State of the art evaluation of traffic detection and monitoring systems

              State of the Art Evaluation of Traffic
Detection and Monitoring Systems
Volume I – Phases A & B: Design
Final Report 627 ( 1)
Prepared by:
Dan Middleton, Ph. D., P. E.
Ryan Longmire
Shawn Turner, P. E.
Texas Transportation Institute
Texas A& M University
College Station, TX 77843- 3135
October 2007
Prepared for:
Arizona Department of Transportation
206 South 17th Avenue
Phoenix, Arizona 85007
in cooperation with
U. S. Department of Transportation
Federal Highway Administration
The contents of this report reflect the views of the authors who are responsible for the
facts and accuracy of the data presented herein. The contents do not necessarily reflect
official views or policies of the Arizona Department of Transportation or the Federal
Highway Administration. The report does not constitute a standard, specification, or
regulation. Trade or manufacturers’ names that appear herein are cited only because
they are considered essential to the objectives of the report. The U. S. Government and
the State of Arizona do not endorse products or manufacturers.
Technical Report Documentation Page
1. Report No.
FHWA- AZ- 07- 627( 1) 3. Recipient's Catalog No.
5. Report Date
4. Title and Subtitle October 2007
State of the Art Evaluation of Traffic Detection and Monitoring Systems
Volume I – Phases A & B: Design 6. Performing Organization Code
7. Author
Dan Middleton, Ryan Longmire, and Shawn Turner
8. Performing Organization Report No.
10. Work Unit No.
9. Performing Organization Name and Address
Texas Transportation Institute, Texas A& M University
2929 Research Parkway
3135 TAMU
College Station, TX 77843- 3135
11. Contract or Grant No.
SPR- PL1-( 71) 627
JPA 07- 006T
13. Type of Report & Period Covered
INTERIM REPORT
February 2007- October 2007
12. Sponsoring Agency Name and Address
Arizona Department of Transportation
206 S. 17th Avenue
Phoenix, Arizona 85007
Project Manager: Stephen R. Owen, P. E.
14. Sponsoring Agency Code
R0627- 19P
15. Supplementary Notes
Prepared in cooperation with the U. S. Department of Transportation, Federal Highway Administration
16. Abstract
This report covers the Phase A and B activities of Research Project SPR 627 for the Arizona Department of
Transportation ( ADOT). Phase C is planned as a separate research activity and is anticipated to begin in the near
term, following the completion of Phases A and B. The need for a better evaluation program for new traffic detection
systems came in part from a lack of confidence in existing detectors, as well as the need for non- intrusive detectors to
replace failing embedded inductive loops. The primary objectives of this research were to identify the most promising
vehicle detection technologies to meet ADOT needs, to identify candidate test sites, to develop a field test evaluation
plan, and to develop and deliver a detailed design of the detection testbed on the selected segment of freeway. The
Texas Transportation Institute ( TTI) met these objectives through an Internet and literature search, a state- of- the-practice
review, a search of relevant new detector systems, and through meetings with the Technical Advisory
Committee ( TAC). Relying on TAC input, TTI developed first a conceptual design, followed by a detailed design and
budget for a proposed test facility located on I- 10 in Phoenix just west of the 16th Street interchange.
Detectors selected for test in the initial period of 12- plus months during Phase C ( and the technology used) are as
follows: Wavetronix SS- 125 ( microwave radar), Sensys Networks ( magnetic), Global Traffic Technologies microloops
( magnetic) and Autoscope Solo Pro ( video imaging). The baseline system selected for providing ground truth data is
the Peek ADR- 6000 using inductive signatures as its basis of detection. It is anticipated that this Phase C testing will
include two summer seasons to expose selected detectors to the extreme heat and related environmental conditions
found in the Phoenix area. The initial cost of the testbed will include detectors sufficient to ultimately cover eight lanes
in the westbound direction ( currently seven lanes) and six lanes in the eastbound direction. Besides the detectors, the
total cost estimate includes a 12 ft by 12 ft node building, three equipment cabinets, inductive loops for the baseline
system, conduit, and boring. The total cost of the facility is estimated to be approximately $ 566,000.
17. Key Words
Vehicle detection, non- intrusive detector,
inductive loop, magnetometer, video imaging,
microwave radar.
18. Distribution Statement
Document is available to the U. S. Public
through the National Technical Information
Service, Springfield, Virginia, 22161
19. Security
Classification
Unclassified
20. Security Classification
Unclassified
21. No. of
Pages
124
22. Price
23. Registrant's Seal
SI* ( MODERN METRIC) CONVERSION FACTORS
APPROXIMATE CONVERSIONS TO SI UNITS APPROXIMATE CONVERSIONS FROM SI UNITS
Symbol When You Know Multiply By To Find Symbol Symbol When You Know Multiply By To Find Symbol
LENGTH LENGTH
in inches 25.4 millimeters Mm mm millimeters 0.039 inches in
ft feet 0.305 meters M m meters 3.28 feet ft
yd yards 0.914 meters M m meters 1.09 yards yd
mi miles 1.61 kilometers Km km kilometers 0.621 miles mi
AREA AREA
in2 Square inches 645.2 square millimeters mm2 mm2 square millimeters 0.0016 square inches in2
ft2 square feet 0.093 square meters m2 m2 square meters 10.764 square feet ft2
yd2 Square yards 0.836 square meters m2 m2 square meters 1.195 square yards yd2
ac acres 0.405 hectares Ha ha hectares 2.47 acres ac
mi2 square miles 2.59 square kilometers km2 km2 square kilometers 0.386 square miles mi2
VOLUME VOLUME
fl oz fluid ounces 29.57 milliliters mL mL milliliters 0.034 fluid ounces fl oz
gal gallons 3.785 liters L L liters 0.264 gallons gal
ft3 Cubic feet 0.028 cubic meters m3 m3 cubic meters 35.315 cubic feet ft3
yd3 Cubic yards 0.765 cubic meters m3 m3 cubic meters 1.308 cubic yards yd3
NOTE: Volumes greater than 1000L shall be shown in m3.
MASS MASS
oz ounces 28.35 grams G g grams 0.035 ounces oz
lb pounds 0.454 kilograms Kg kg kilograms 2.205 pounds lb
T short tons
( 2000lb) 0.907 megagrams
( or “ metric ton”)
mg
( or “ t”)
mg
( or “ t”)
megagrams
( or “ metric ton”) 1.102 short tons ( 2000lb) T
TEMPERATURE ( exact) TEMPERATURE ( exact)
º F Fahrenheit
temperature
5( F- 32)/ 9
or ( F- 32)/ 1.8 Celsius temperature º C º C Celsius temperature 1.8C + 32 Fahrenheit temperature º F
ILLUMINATION ILLUMINATION
fc foot- candles 10.76 lux Lx lx lux 0.0929 foot- candles fc
fl foot- Lamberts 3.426 candela/ m2 cd/ m2 cd/ m2 candela/ m2 0.2919 foot- Lamberts fl
FORCE AND PRESSURE OR STRESS FORCE AND PRESSURE OR STRESS
lbf poundforce 4.45 Newtons N N Newtons 0.225 poundforce lbf
lbf/ in2 poundforce per
square inch 6.89 kilopascals kPa kPa kilopascals 0.145 poundforce per
square inch lbf/ in2
SI is the symbol for the International System of Units. Appropriate rounding should be made to comply with Section 4 of ASTM E380
TABLE OF CONTENTS
Page
EXECUTIVE SUMMARY ................................................................................................. 1
PART ONE: STATE OF THE PRACTICE
1. INTRODUCTION ........................................................................................................... 5
BACKGROUND..................................................................................................... 5
SCOPE OF THE PROJECT .................................................................................... 6
ORGANIZATION OF THE CONTENTS .............................................................. 7
2. PHOENIX AREA DETECTORS.................................................................................... 9
INTRODUCTION................................................................................................... 9
FREEWAY MANAGEMENT SYSTEM ( FMS) DETECTORS............................ 9
3. LITERATURE AND INTERNET SEARCH; STATE CONTACTS ........................... 15
INTRODUCTION................................................................................................. 15
SINGLE TECHNOLOGY SOURCES.................................................................. 15
SINGLE AREA TECHNOLOGIES...................................................................... 28
RECENT MULTIPLE TECHNOLOGY SOURCES............................................ 29
FOLLOW- UP PHONE CALLS TO SELECTED STATES ................................. 39
4. VEHICLE DETECTOR TESTBEDS............................................................................ 43
INTRODUCTION................................................................................................. 43
CALTRANS I- 405 TESTBED IN IRVINE, CALIFORNIA ................................ 43
MNDOT NON- INTRUSIVE TEST FACILITY ON I- 394................................... 45
TTI TESTBED IN AUSTIN.................................................................................. 49
TTI TESTBED IN COLLEGE STATION ............................................................ 51
PART TWO: DESIGN DEVELOPMENT
5. TESTBED SITE SELECTION AND CONCEPTUAL DESIGN................................. 55
INTRODUCTION................................................................................................. 55
TESTBED SITE SELECTION.............................................................................. 55
TESTBED CONCEPTUAL DESIGN................................................................... 58
PRIMARY DETECTION SYSTEMS................................................................... 60
SPECIALIZED DETECTOR ( OPTIONAL)......................................................... 63
SPEEDINFO: A UNIQUE BUSINESS PLAN ..................................................... 63
PROGRAM BUDGET........................................................................................... 63
6. DETECTOR EVALUATION PROGRAM................................................................... 67
INTRODUCTION................................................................................................. 67
SELECTION, SCREENING, AND TESTING PROCEDURES ......................... 67
PROJECT MANAGEEMNT / DESIGN .............................................................. 69
TABLE OF CONTENTS ( Continued)
Page
CONSTRUCTION ................................................................................................ 70
MAINTENANCE ................................................................................................. 70
SYSTEMS ISSUE MANAGEMENT ................................................................... 70
EVALUATION ..................................................................................................... 70
APPROVED PRODUCT LISTS........................................................................... 74
7. DETAILED DESIGN .................................................................................................... 75
INTRODUCTION................................................................................................. 75
BASELINE SYSTEM ........................................................................................... 75
EQUIPMENT CABINETS.................................................................................... 78
CONDUIT AND PULL BOXES........................................................................... 78
EQUIPMENT POLE ............................................................................................. 79
NODE BUILDING................................................................................................ 79
DETECTOR EQUIPMENT................................................................................... 79
OTHER ON- SITE EQUIPMENT.......................................................................... 81
PROGRAM BUDGET.......................................................................................... 82
APPENDICES ................................................................................................................... 91
APPENDIX A: OVERVIEW OF SELECTED DETECTION SYSTEMS........... 91
APPENDIX B: CONDUIT CHART ..................................................................... 97
APPENDIX C: INDUCTIVE SIGNATURE TECHNOLOGIES, INC .............. 101
APPENDIX D: SPEEDINFO .............................................................................. 105
LIST OF REFERENCES................................................................................................. 113
LIST OF FIGURES
Figure Page
1 FMS Detector Locations .......................................................................................... 10
2 FMS Cabinet with Inductive Loop Equipment ........................................................ 12
3 SmarTek SAS- 1 Contact- closure Cards................................................................... 12
4 Phoenix Metro FMS Existing and Planned Projects ................................................ 13
5 Speed Comparison of ADR- 6000, Doppler Radar, and Video Imaging System ..... 18
6 PATH Traffic Detector and Surveillance Sub- Testbed............................................ 44
7 MnDOT NIT Site Layout......................................................................................... 45
8 Catwalk for Mounting Detectors Overhead ............................................................. 46
9 Aluminum Tower for Sidefire Mounting ................................................................. 47
10 View of NIT Building from the Catwalk ................................................................. 47
11 MnDOT Shelter Schematic Layout.......................................................................... 48
12 Layout of TxDOT’s I- 35 Site................................................................................... 50
13 Views of I- 35 Testbed .............................................................................................. 51
14 Layout of S. H. 6 College Station Testbed................................................................ 52
15 View of S. H. 6 Testbed Looking South ................................................................... 53
16 View of Equipment Cabinets and Weather Station.................................................. 53
17 Aerial View of I- 10 at 13th Street ............................................................................. 57
18 Ground Level View of I- 10 at 13th Street Facing Northward .................................. 57
19 Conceptual Testbed Design...................................................................................... 59
20 Aerial View of 16th Street Interchange Ramps ( View to North).............................. 61
21 ADR- 6000 Loop Layout .......................................................................................... 77
22 Arizona DOT Testbed – General Arrangement & Details ( 22a & b) ...................... 83
23 DVSS- 100 Doppler Vehicle Speed Sensor ............................................................ 110
LIST OF TABLES
Table Page
1 Peek ADR- 6000 Classification Accuracy Comparison for I- 35 .............................. 18
2 SAS- 1 Count Error Rates on S. H. 6 during Dry Weather........................................ 25
3 SAS- 1 Count Error Rates on S. H. 6 during Wet Weather ....................................... 25
4 Detector Cost Comparison ....................................................................................... 30
5 Detector Error Rates................................................................................................. 30
6 Detector Ease of Installation and Reliability............................................................ 31
7 Estimated Life- Cycle Costs for a Typical Freeway Application ............................. 31
8 Sensor Performance Descriptions ( Hawaii) ............................................................. 32
9 Site Descriptions ( Pennsylvania) ............................................................................. 33
10 Results of Portable Setup Field Testing ( Pennsylvania) .......................................... 34
11 Comparisons of Mean Percentage Volume Errors ( Nebraska) ................................ 36
12 Overall Results for Volume and Speed Detections ( Minnesota) ............................. 37
13 Result Summary for Length- Based Class Detection ( Minnesota) ........................... 38
14 FY 2006 TTI Detector Test Plan ( Texas)................................................................. 38
15 Annualized Life- Cycle Cost of Detectors on Freeways........................................... 62
16 Qualitative Assessment of Detectors........................................................................ 62
17 Responsibilities for Detector Evaluation.................................................................. 67
18 Anticipated Testbed Initial Costs for Phase 1 and Phase 2 ...................................... 86
19 Sources of Cost Estimates ........................................................................................ 88
20 Conduit Sizes and Types for Detailed Design ......................................................... 99
21 Estimated Cost of the IST System Using Existing Loops...................................... 102
TERMS, ACRONYMS, AND ABBREVIATIONS
AADT Average Annual Daily Traffic
AC Alternating Current
ADOT Arizona Department of Transportation
ADR Automatic Data Recorder
AP Access Point Unit ( Sensys)
APD Absolute Percent Difference
ASIM ASIM Technologies, Ltd, Detector manufacturer in Switzerland
ASTM American Society for Testing and Materials
ATR Automatic Traffic Recorder
ATRC Arizona Transportation Research Center ( ADOT)
AWG American Wire Gauge
Caltrans California Department of Transportation
CCD Charge- Coupled Device ( camera)
DC Direct Current
DETT Detector Evaluation and Testing Team ( Caltrans)
DRI ( Caltrans) Division of Research and Innovation
DSL Digital Subscriber Line ( communication)
EIS Electronic Integrated Systems, Inc.
FHWA Federal Highway Administration
F. M. Farm- to- Market ( Texas)
FMS Freeway Management System
FTP File Transfer Protocol
GTT ( 3M) Global Traffic Technologies, LLC ( formerly a 3M business unit)
HD High Definition ( Wavetronix SS- 125)
IST Inductive Signature Technologies, Inc.
ITD Intermodal Transportation Division ( ADOT)
ITS Intelligent Transportation Systems
JPEG Joint Photographic Experts Group ( graphic image format)
LAN Local Area Network
MAG Maricopa Association of Governments
MAPE Mean Absolute Percent Error
MnDOT Minnesota Department of Transportation
NB Northbound
NDOR Nebraska Department of Roads
NIST National Institute of Standards and Technology
NIT Non- Intrusive Tests ( by MnDOT)
NTP Network Time Protocol
NTSC National Television Standards Committee
ORADS Off- Road Axle Detection Sensor, from Spectra Research
ORITE Ohio Research Institute for Transportation and the Environment
PAD Passive Acoustic Detector
PATH Partners for Advanced Transit and Highways ( at UC Berkeley)
PCD Phoenix Construction District ( ADOT)
PeMS ( Caltrans) Performance Measurement System
PennDOT Pennsylvania Department of Transportation
PMD Phoenix Maintenance District ( ADOT)
PMD ITS Phoenix Maintenance District Intelligent Transportation Systems
PNITD Portable Non- Intrusive Traffic Detection System
PVR Per Vehicle Records
RMSE Root Mean Squared Error
RTMS Remote Traffic Microwave Sensor
S. H. State Highway ( Texas)
SR State Route ( Arizona and Pennsylvania)
SAS- 1 SmarTek Acoustic Sensor
SEO Schwartz Electro- Optics
SRF Consulting Group in Minnesota
STIP Short- Term In- Pavement
SUV Sport Utility Vehicle
TAC Technical Advisory Committee
TCPEC Traffic Control Product Evaluation Committee
TDS2 Traffic Detector Surveillance Sub- Testbed ( Caltrans)
TIRTL The InfraRed Traffic Logger
TOC Traffic Operations Center
TOG Transportation Operations Group ( of TTI)
TPD Transportation Planning Division ( ADOT)
TTG Transportation Technology Group ( ADOT)
TTI Texas Transportation Institute ( of Texas A& M University System)
TxDOT Texas Department of Transportation
VIP Video Image Processor
VISION PCD Field Office ( ADOT)
VSV Video Signature Vector ( Caltrans testbed)
1
EXECUTIVE SUMMARY
INTRODUCTION
Accurate, complete, and timely traffic data is critical to the effective management of
Arizona’s highway system. Limitations in current traffic monitoring abilities are an ongoing
challenge for the Arizona Department of Transportation ( ADOT) and for its customers in
both urban and rural areas. Many technologies exist for detecting vehicles and determining
traffic volume, speed, and lane occupancy.
This research project, as performed by the Texas Transportation Institute ( TTI), included a
state- of- the- practice review, intended to assist ADOT and other Arizona transportation
agencies in identifying the most appropriate detection technologies to meet local needs. The
scope of this project also required the development of a design for a detector testbed facility,
which ADOT proposed to develop in the near future, in the Phoenix area.
CURRENT PHOENIX AREA DETECTION SYSTEMS
Embedded inductive loops and pole- mounted passive acoustic detectors ( PADs) are currently
the primary detectors used on the roughly 250 centerline miles of freeways in the Phoenix
metro area. Inductive loops are a mature and accurate technology but lane closures for their
installation and maintenance have become less feasible in recent years due to traffic volumes,
safety issues, and costs.
PADs have not been as accurate as originally anticipated, so viable alternative detectors are
needed. ADOT’s preference, due to the historical problems noted with embedded roadway
sensor designs, was to focus the study initially on new non- intrusive detection concepts.
LITERATURE AND INTERNET SEARCH
A thorough literature and Internet search produced a number of useful documents on the
topic of vehicle detection on freeways. Due to the dynamic nature of this topic and the ever
changing nature of the detector market, the most useful information came recently – from the
late 1990s until the most recent. None of the newer non- intrusive detectors appear to be as
accurate in vehicle presence detection as loops under all environmental conditions, but many
agencies have determined that other positive features outweigh the modest reduction in
accuracy.
Close scrutiny of literature and Internet sources reveals that comparing each research
project’s findings is difficult at best due to the use of different metrics, different traffic
conditions, different models of the same detector ( or at least different firmware), and
different positioning of detectors. In general, however, a few of the newer detectors can
count traffic consistently within 5 percent of true counts. Speed estimates from these devices
indicate similar accuracy, and in limited cases exceed the speed estimates of standard
inductive loops ( e. g., Doppler radar). The non- intrusive technologies that have withstood the
scrutiny of several installations and have the most promise for replacing inductive loops are
2
microwave radar, video imaging, and magnetic detectors. Of the two prominent magnetic
detectors, one is intrusive but is still deemed worthy of consideration.
VEHICLE DETECTOR TESTBEDS
Existing field test facilities in California, Minnesota, and Texas make use of components that
were already in place along existing freeways such as overhead structures, poles, and
conduit. For example, the Caltrans facility on I- 405 in Irvine, California utilizes a unique
camera system to establish ground truth, substantially reducing the need for manual viewing
of recorded video. It also has inductive loops in the pavement to be used as needed. The
Minnesota test facility on I- 394 near downtown Minneapolis used standard inductive loops
for ground truth and mounted test systems either on an overhead bridge, on movable
telescoping poles ( sidefire), or underneath the pavement. The Texas facility on I- 35 near
downtown Austin used an upscale vehicle classifier, the Peek ADR- 6000, for ground truth,
existing luminaire poles and conduit, and an overhead sign bridge. Most of the test detectors
were mounted on the luminaire pole. Lessons learned from these vehicle detector testbeds
could provide useful information as this project moves forward.
STATE CONTACTS
The information gathered in this section complements the other sections. The literature and
Internet search helped identify states that had tested and installed new technology detectors,
and the state contacts brought some of the information up to date. In summary, all three states
– California, Minnesota, and Texas – have conducted research on some of the same detectors
but there are differences in how the research is being implemented. The various districts in
California and Texas are installing a wide variety of non- intrusive detectors to replace failing
loops, but Minnesota continues to install and rely almost exclusively on loops.
TESTBED SITE SELECTION AND CONCEPTUAL DESIGN
The site selected for the future ADOT testbed consists of a mainline element, which is on
Interstate 10 at 13th Street ( see map page 3), and a westbound entrance ramp element from
the nearby 16th Street interchange. Some strengths of this site include its central location,
high traffic volume ( including trucks), number of lanes, proximity to the I- 10 tunnels ( related
to lane closures), space availability for testbed equipment and parking, and unobstructed
view of approaching traffic. The conceptual testbed design includes many of the existing and
proposed components of the testbed such as equipment cabinets, underground boring, pull
boxes, ground truth devices, and safety barriers. The offset sign bridges offer locations for
mounting detectors and surveillance cameras overhead, and proposed poles could be used for
mounting sidefire detectors and surveillance cameras. Horizontal 3- inch bores spaced 18 feet
apart can accommodate microloops from Global Traffic Technologies ( GTT, formerly 3M).
Peek ADR- 6000 inductive loops in the roadway on the upstream side of each structure are
intended to be used for ground truth. It is anticipated that if the testbed is built in two stages
that initial equipment installation will emphasize the westbound direction of traffic flow on
the north side of the freeway.
3
DETAILED DESIGN AND TESTING PROGRAM DEVELOPMENT
Building on what had already been accomplished in the conceptual design and the site
selection process, the detailed design provided more accurate information on quantities and
sizes of components. Given the uncertainties of installing shallow inductive loops in the
rubberized asphalt pavement layer, there was also an investigation of the Inductive Signature
Technologies ( IST) system as an alternative baseline system instead of the Peek ADR- 6000.
The Technical Advisory Committee ( TAC) decided that the IST system did not have the
maturity needed, electing to stay with the Peek unit. Components of the Detailed Design that
evolved or were modified from the Conceptual Design included the use of more ADOT-furnished
components such as pull boxes, poles, and cabinets. The poles ( two on the north
side of I- 10) will now be standard ADOT poles without mast arms. Also, the existing on- site
camera will provide surveillance coverage, supplemented by two proposed Autoscope video
imaging systems, one per direction of traffic flow.
The overall detector evaluation process is currently envisioned as a seven- step process, and it
is anticipated that the Transportation Technology Group ( TTG) will be primarily responsible
for selecting detectors for test and the exact procedures to be used. The total cost of the
ADOT testbed facility on I- 10 is estimated to be about $ 566,000. Some components will be
provided by ADOT at no cost to the project; however, all proposed traffic detector systems to
be evaluated are included in the budget estimate. Other major cost elements include
infrastructure, communications, detector mountings, node building, design package,
construction management, and contingency.
Future I- 10 Testbed Site
4
5
PART I: STATE OF THE PRACTICE
1. INTRODUCTION
BACKGROUND
The Arizona Department of Transportation spends about $ 50,000 at each location to install
Freeway Management System ( FMS) detection, and another $ 1,000 per year to maintain each
site. For roughly 250 centerline miles of freeway in the greater Phoenix area ( termed “ Metro
Phoenix”), ADOT anticipates spending about $ 25 million to install traffic sensors at some
500 sites, and another $ 10 million to maintain them, over the 20- year life of the Regional
Transportation Program. Added together, these costs represent a substantial investment, but
not making a serious commitment to traffic monitoring equipment would undoubtedly result
in even greater cost in terms of reduced safety, lost federal funding, greater motorist delay,
and heightened motorist frustration.
Accurate, complete, and timely traffic data are critical to the effective management of
Arizona’s highway system. Limitations in current traffic monitoring abilities are an ongoing
negative issue for the Arizona Department of Transportation and for its customers as well in
both urban and rural areas. Many technologies exist for detecting vehicles and determining
traffic volume, type, lane occupancy, and speed. The key is to invest wisely in the
appropriate traffic monitoring technology, and hence, the need for this research project.
The Maricopa Association of Governments ( MAG), ADOT’s Transportation Planning
Division ( TPD), and others use data from the Freeway Management System ( FMS) traffic
detectors for planning and budgeting purposes. It is critical for ADOT to be aware of the
state- of- the- art in traffic monitoring technology, and to implement the latest and best
technology for both planning and operations applications. The metropolitan freeway ramp
metering program is another critical focus area for this study of detection options. ADOT
needs to identify the best of the new standard detector designs for its basic long- term needs,
and to also consider specialized systems for various localized data issues. To this end, the
Texas Transportation Institute was requested to conduct this study, to include the key facility
design deliverables for ADOT.
Transportation Technology Group
ADOT’s Transportation Technology Group, established in July 1996, has the responsibility
for the planning, development, deployment, management, and operation of new technologies
related to the transportation industry. Prior to the establishment of TTG, a variety of different
organizations managed technology- related activities. However, TTG is not responsible for
the Department’s information resources; this effort is handled by the Information Technology
Group ( ITG).
Intelligent Transportation Systems ( ITS) is the application of computers, electronics, control
systems, communications technologies, and management strategies to transportation systems
6
in an integrated manner, providing travel information to increase the safety and efficiency of
the surface transportation systems.
Since the creation of TTG, all ITS activities throughout the State have been consolidated
within one dedicated group, interfacing with one another, and proceeding toward the same
goals and objectives. With a vision focused primarily on the ITS activities throughout the
State and close coordination and interaction, this Group plans, develops, deploys, manages,
and operates ITS projects to better serve its customers.
Texas Transportation Institute
ADOT project stakeholders recommended the selection of TTI to conduct this research,
recognizing the unique abilities and qualifications of TTI to perform the work. The brief
overview below describes the organization, and the specific units involved in the research.
Established in 1950, TTI is the world's largest university- based transportation research and
education institute. TTI’s objective is to solve transportation problems through research, to
transfer technology and to develop diverse human resources to meet the transportation
challenges of tomorrow. The Texas Transportation Institute is part of the Texas A& M
University System and maintains research divisions, regional divisions, research centers, and
field offices. Within TTI, the Transportation Operations Group ( TOG) is a national leader in
transportation operations. TOG’s mission is “ to facilitate innovations in transportation
system operations through leadership in research, education and technology transfer.” TOG
has access to all the equipment and resources within TTI and actively coordinates work
within TTI and with partners to provide effective services to sponsors.
TOG facilities provide outstanding capabilities for conducting leading edge research in all
aspects of the transportation operations research program. Examples of TOG facilities
include the TransLink ® Laboratory, a multi- modal, multi- agency public/ private program of
research, development and professional education; freeway and intersection field test
laboratories; and a Traffic Control Device Outdoor Laboratory and Demonstration Facility.
Additional information about the TOG research program, staff, and facilities can be found at
http:// transops. tamu. edu/. It also contains links to other TOG/ TTI websites.
SCOPE OF THE PROJECT
The research objectives are:
• Gather information on the current vehicle detection systems used in Arizona.
• Identify the most promising vehicle detection technologies to meet the needs of
ADOT and other Arizona agencies.
• Identify candidate test sites, develop preliminary site design, and develop field test
evaluation plan.
• Develop a detailed design of the detection testbed on the selected freeway segment.
7
ADOT Research Project SPR- 627 has a total of three phases, but this report only covers the
work by TTI on the first two – Phase A and Phase B. Phase A is the “ Global ‘ State- of- the-
Practice’ Review and Conceptual Design,” and Phase B is the “ Detailed Design.”
Phase C, the initial long- term detector evaluation study, is expected to be done as part of a
separate agreement to be developed at a later date.
The work plan consisted of eight tasks. In Task 1, information was collected on detection
elements currently used in the Phoenix area. This information came primarily from the
project’s Technical Advisory Committee ( TAC) members. Task 2 was a state of the practice
review of embedded and non- intrusive traffic detection technologies that have the potential to
more accurately monitor key data of volume, lane occupancy and speed. TTI gathered this
information through a global Internet survey and literature search of relevant new products
and innovations. Task 3 involved a search of relevant new detector systems including both
breadth of detector types and success stories that could apply to ADOT and local Arizona
agencies. Task 4 involved development and presentation of an interim state- of- the- practice
report to the TAC.
Task 5 involved identifying the testbed site, calibration equipment, a benchmark “ truthing
system” for comparison of results, and the detection systems that are proposed to be tested.
Task 6 involved planning and conducting TAC meetings for Phases A and B. Task 7 was the
detailed design and included development of the testing program. Task 8 involved
development of deliverables.
ORGANIZATION OF THE REPORT
This report consists of seven chapters. Following this introductory chapter is Chapter 2,
which covers Phoenix area detectors that are currently being used. Following that is Chapter
3, which details findings of the literature and Internet search. Chapter 4 covers vehicle
detector testbeds in California, Minnesota, and Texas, and information gathered in phone
calls to the same three states. Chapter 5 covers the testbed site selection and the conceptual
design of the testbed. Chapter 6 discusses a proposed seven- step evaluation program for
selected detectors. Chapter 7 describes the final detailed design, including the budget for
implementing the testbed.
8
9
2. PHOENIX AREA DETECTORS
INTRODUCTION
With the exception of a few vehicle detectors installed in the Phoenix area for test and
evaluation, the detectors currently used by the Arizona Department of Transportation are
inductive loops and passive acoustic detectors ( PADs) from SmarTek ( SAS- 1).
FREEWAY MANAGEMENT SYSTEM ( FMS) DETECTORS
Original ADOT Freeway Management System detector locations were every 1/ 3- mile along
most of the major highways in the Phoenix area. Following FMS deployment, approximately
two- thirds of the locations were decommissioned, which leaves operating detector stations
about every mile. Figure 1 on page 10 shows ADOT Freeway Management System ( FMS)
detector locations in the Phoenix area, indicating all 219 locations that are currently
maintained ( not decommissioned) for collecting traffic data for operations or planning
purposes. Red triangles represent PAD locations and green circles are loop sites.
Inductive Loops
As a general rule, loops are no longer being installed on Phoenix freeways due to the
disruption to traffic imposed by their installation and the fact that if lanes shift laterally, the
existing loops are no longer useful. Loops are a mature technology and, in most cases, are
sufficiently accurate for most purposes if properly installed and maintained. However,
installation and replacement of loops represent a significant cost when traffic control,
motorist delay, and increased crash risk during installation/ maintenance are considered. Also,
when loops are installed using a sawcutting procedure, a weakening of the pavement occurs.
Some of the recent installations on Phoenix freeways have utilized preformed loops placed
on top of an existing pavement in advance of a hot- mix asphalt overlay, overcoming the
negative aspects of sawcutting. One exception to the general moratorium on freeway loop
installation is where an accurate truth system is needed such as for a detector test facility.
The existing inductive loop sites in the Phoenix area have different numbers of wire turns,
different loop depths, and different wire gauges. The loops are either 18 gauge wire with five
turns ( below the hot mix asphalt overlay) or 14 gauge wire with four turns ( in the upper
portion of the hot mix asphalt overlay). Some of the preformed loops were probably forced
out of position by the paving operation. The nominal spacing of the loops is 18 feet, but the
actual spacing varies from 17 feet to 24 feet. Checks of speeds based on assumed loop
spacing indicates that speeds are off by as much as 15 mph.
Investigating data discrepancies requires an awareness of not only the detector accuracy but
other aspects of data processing in the cabinet and communication elements providing
information to ADOT’s Traffic Operations Center ( TOC). Discrepancies in data received at
the TOC could be caused by detector calibration issues or by communication errors. Potential
issues with inductive loops include: incorrect loop placement ( loop shift during overlay),
Figure 1: FMS Detector Locations
Loop Detectors
Acoustic Detectors
10
11
Broken and/ or shorted loops, false actuations due to similar frequencies, sensitivity settings,
and malfunctioning detector cards. Potential issues with the acoustic detectors include:
pavement texture, echoes/ reflections, detector alignment, and various software settings.
Unless data are buffered locally at the cabinet, there is a chance of incorrect and/ or missing
data when it is received by the TOC. Detector actuations that are sent to the controller as
contact- closure events are potentially lost when there are communication errors between the
cabinet and the TOC.
Passive Acoustic Detectors
There seems to be general agreement among local agencies that the PADs are a significant
contributor to data errors, but there may also be errors introduced either in the cabinet or
beyond the cabinet. Input based on other Texas Transportation Institute ( TTI) research
suggests that calibration of the PADs may be an issue at some sites. One calibration issue
requires that PADs be recalibrated following a “ quiet pavement” rubberized asphalt overlay.
Figure 2 on page 12 is a typical cabinet, and Figure 3 shows a SAS- 1 contact closure card for
PAD sites. These cabinets house Model 179 controllers to process data provided by vehicle
detectors ( PADs and loops), detector cards ( SAS Relay Interfaces or loop detector cards),
and a fiber- optic data transceiver. FMS data are transmitted from the cabinets to the Traffic
Operations Center ( TOC) on a dual- ring fiber- optic network, or via analog modem over
twisted pair.
FMS Improvements
Figure 4 on page 13 shows the FMS improvements planned. These corridors planned for
improvement will be useful as detector test site identification moves forward since it may be
desirable to locate the test facility within a construction project and perhaps even share
funding with the construction project.
ADOT Transportation Planning Division ( TPD) Test Site
The ADOT Transportation Planning Division Data Bureau operates a test site at the
interchange of Northern Avenue and Loop 101. The sensors installed there are GTT ( 3M)
magnetometers, microwave radar ( RTMS and Wavetronix), and inductive loops. This
research project scope includes identifying other candidate sites that ADOT could choose
from to implement a full- scale, long- term freeway vehicle detector test facility. It also
includes developing a conceptual plan for the proposed test facility. In related non- ADOT
tests, Sensys Networks magnetometers were installed in Scottsdale in early 2006 but not for
long enough to fully explore longevity.
12
Figure 2: FMS Cabinet with Inductive Loop Equipment
Figure 3: SmarTek SAS- 1 Contact- closure Cards
Figure 4: Phoenix Metro FMS Existing and Planned Projects
13
PHASE DESCRIPTION MILE FY STATUS
1 I- 10 ( 83rd Ave.- Southern Ave.)
I- 17 ( Thomas Rd.- Maricopa T. I.)
21
8 94 Completed
2
SR51 ( I- 10- Glendale Ave.)
SR143 ( I- 10- L202)
L202 ( I- 10- SR143)
5.5
4
3
96 Completed
3A I- 17 ( Thomas Rd.- Peoria Ave.) 7 99 Completed
4 L101 ( US60- L202)
L202 ( SR143- L101)
3
6 98 Completed
5 SR51 ( Glendale Ave.- Bell Rd.) 8 99 Completed
7A I- 10 ( Southern Ave.- Chandler Blvd.) 6 99 Completed
6A L101 ( Southern Ave.- Guadalupe Rd.) 2 00 Completed
7B US60 ( I- 10- Val Vista Dr.) 13 01 Completed
Subtotal, completed 86.5
6B, 6C L202 ( L101- Gilbert Ave.) N
L101 ( Princess- L202.)
8
14
09
07 Bid Advertise
8 US60 ( Val Vista Dr.- Power Rd.) 4 06 Under construction
9 L101 ( Guadalupe Rd.- L202) S 5.5 06 Under construction
10 L101 ( Grand Ave.- I- 17) 13 07 Under construction
11 I- 10 ( 99th Ave.- 83rd Ave.)
L101 ( I- 10- Grand Ave.)
2
9.5 06 Under construction
12A L101 ( I17.- Princess DR.) 7 14 Under design
13 US60 ( Power Rd- Crismon.) 4 06 Under construction
Subtotal, design and construction 67
3B I- 17 ( Peoria Ave.- Happy Valley Rd.) 9 13 Programmed
7C I- 10 ( Chandler Blvd.- Queen Ck. Rd.)
I- 10 ( Dysart- 99th) 4,4 18 Programmed
12B L101 ( I- 17- Scottsdale Rd.)
SR51 ( Bell Rd.- L101)
7
2.5 13 Programmed
14 L202 ( Gilbert Rd.- I- 10) S 11 13 Programmed
15 L202 ( SR87- Power Rd.) N 10 19 Programmed
16 L202 ( Power Rd.- Gilbert Rd.) NE- SE 25 22 Programmed
Subtotal, programmed 72.5
Freeway Management
System
Phoenix Metro Area
Gilbert Rd.
Glendale Ave.
Shea Blvd.
Scottsdale Rd.
Thomas Rd.
10
1
10
20
1
Happy Valley Rd.
14
TOC
1
Northern Ave.
Peoria Ave.
Thunderbird Rd.
1
Pima Rd.
8
83rd Ave.
5
19th Ave.
University Dr.
McKellips Rd.
Southern Ave.
Ray Rd.
McDowell Rd.
Power Rd.
Pecos Rd.
34
Queen Creek Rd.
99th Ave.
Chandler Blvd.
6
56th St.
Grand
Ave.
Bell Rd.
20
20
20
Crismon Rd.
Elliot Rd.
Guadalupe Rd.
Warner Rd.
Indian School Rd.
Camelback Rd.
Bethany Home Rd.
Dunlap Ave.
Cactus Rd.
Greenway Rd.
Union Hills Dr.
Beardsley Rd.
Deer Valley Rd.
Pinnacle Peak Rd.
Broadway Rd.
Brown Rd.
6
90th St.
1
Red Mtn. T. I.
Maricopa T. I.
Baseline Rd.
Val Vista Dr.
Dobson Rd.
10
6/ 25/ 07 DB
Carefree Hwy
14
15
3. LITERATURE AND INTERNET SEARCH; STATE CONTACTS
INTRODUCTION
TTI researchers conducted a comprehensive literature search covering the past five years
and an Internet search to determine emerging and promising vehicle detector systems that
are worthy of further consideration by the Arizona Department of Transportation
( ADOT) and other Arizona agencies. Previous TTI research covered the time period prior
to this five year period. A keyword search used a variety of combinations of words such
as freeways, non- intrusive detectors, performance, reliability, cost, video image
detection, machine vision, microwave, radar, Doppler radar, passive acoustic, inductive
loop detectors, installation methods, count accuracy, speed accuracy, occupancy
accuracy, non- intrusive vehicle classification, FHWA classification, length- based
classification, field test lab, freeway test site, testbed, and testbed conceptual plan. The
search yielded over 150 total records. Of these, some were duplicates of each other or
reports found elsewhere, and many others were not as useful as first thought. The
remainder is included in this document.
Beyond using the key word search to develop the list of document summaries, some of
the criteria used for selecting detectors for further evaluation included:
• Technologies known or found to be reliable and reasonably accurate.
• Emphasis on detectors that cover multiple lanes.
• Technologies of reasonable cost.
• Emphasis on vehicle detection and not pedestrian or bicycle detection.
• Emphasis on freeways and not signalized intersections.
• Emphasis on non- intrusive detectors but not to the total exclusion of others.
The organization of the literature and Internet findings in this section begins with
findings pertaining to single technologies – providing the reader with detailed
performance on each technology separately. Findings from research projects that used
similar environmental and traffic conditions to compare technologies are toward the end
of the section ( e. g., comparing PADs with microwave radar). Within this structure, the
more sophisticated multiple- lane detection systems come first followed by single
detection area devices. Findings are also organized chronologically with the most recent
research results placed last.
SINGLE TECHNOLOGY SOURCES
Detection technologies discussed below are primarily non- intrusive, although there is
also information on loops because they remain the most prominent detection system used
16
around the country. Two of the early series of detector research projects whose findings
still have current relevance in this context were conducted by the Minnesota Department
of Transportation ( MnDOT) along with consultant SRF Consulting, Inc. ( 1, 2, 3, 4) and by
the Texas Transportation Institute. ( 5, 6, 7, 8, 9, 10, 11) The more recent study findings will be
emphasized but some of the earlier findings are included as well.
The MnDOT Phase I Non- Intrusive Tests ( NIT) ran from November 1995 to January
1997, involved 17 devices representing eight technologies, and used standard inductive
loops for ground truth in both phases. Volume and speed data were the primary
parameters tested, with classification also included on some devices. Phase II included
building a permanent test shelter at the site, which was completed in April 2001,
purchasing the detectors to be tested, and pre- testing the detectors through the summer of
2001. The official freeway data collection lasted from October 2001 to early March 2002.
The most recent round of detector tests was the Portable Non- Intrusive Traffic Detection
System ( PNITDS), which began in 2003 and concluded in 2005. ( 12)
Recent TTI research projects on this topic span a time period from 1995 to 2007. These
projects relied on two freeway test facilities, one in College Station on S. H. 6 and the
other on I- 35 near downtown Austin. The College Station testbed maintains free- flow
conditions while the Austin site is usually free- flow during off- peak periods but becomes
congested with stop- and- go traffic during peak times or due to incidents. Cross- sections
of the two roadways are different as well – the College Station site has four lanes with
two in each direction with a 60- ft depressed grass median while the Austin site has five
contiguous lanes in the southbound direction with a center median barrier.
TTI developed and equipped the two freeway testbeds with equipment such as equipment
cabinets, computers, baseline inductive loops, charged couple display ( CCD) cameras,
Digital Subscriber Line ( DSL) communication, and baseline inductive loops. More
information on both of these facilities is provided later in this document. Detectors
investigated in this TTI research included: Autoscope Solo Pro ( video imaging),
Accuwave ( microwave), Remote Traffic Microwave Sensor ( RTMS) ( microwave radar),
non- invasive microloops by GTT ( 3M) ( magnetic), SAS- 1 by SmarTek ( acoustic)
Traficon VIP ( video imaging), Iteris Vantage ( video imaging). Other detectors tested by
TTI with less viability were not included in this document.
Inductive Loop Detectors
It is appropriate that a comparison be made of newer detectors with the most commonly
used detector in current practice— the inductive loop. If non- intrusive detector accuracy
compares favorably with loops and they are otherwise similar but non- intrusive, there are
many agencies that would choose the non- loop option. ( 1,2,3) The Minnesota research used
6 ft by 6 ft loops installed in previous tests for baseline comparison of counts and speed
accuracy. Therefore, the inductive loops were only approximately four years old when
this Minnesota research began. Initial loop accuracy tests showed that the loops in lanes 1
and 2 on the freeway undercounted by 0.1 percent, while the high- occupancy vehicle
( HOV) lane loops undercounted by 0.9 percent. Speed tests indicated that lane 1 loops
17
underestimated true speed by 6.1 percent, and lane 2 loops underestimated speed by 1.9
percent.
Peek ADR- 6000
The Peek ADR- 6000 is a high- end vehicle classification system ( using Idris technology)
which exclusively uses inductive loops as its in- pavement sensors. It is unique in the fact
that it uses inductive loops as axle sensors, using vehicle signatures generated from loop
actuations for its classification algorithm. Therefore, its speed, count, and classification
results exceeded previous experience from the more typical classifiers using loops and
axle sensors ( e. g., piezoelectric sensors). The ADR stored classification data internally to
be downloaded later to a site computer or to other computers via the Internet using file
transfer protocol ( FTP). ( 7)
TTI findings ( from I- 35 only) indicated that the ADR- 6000 was accurate for
classification, counts, speeds, and lane occupancy ( although TTI had to develop software
to monitor occupancy). Table 1 on page 18 shows the classification result for a dataset of
1923 vehicles, indicating only 21 errors and resulting in a classification accuracy of 99
percent ( ignoring Class 2 and 3 discrepancies). This data sample occurred during the
morning peak and included some stop- and- go traffic. For count accuracy, the Peek in this
same dataset only missed one vehicle ( it accurately accounted for vehicles changing
lanes). Figure 5 on page 18 shows the close agreement of the ADR with two other test
systems – an overhead Doppler radar system, and an Autoscope Solo Pro – using one-minute
speed bins from the Peek. The graphic indicates discrepancies only at slow speeds
( below about 15 mph) where the Doppler radar is known to drop out and the Autoscope
speed accuracy decreases slightly. ( 7)
Researchers expect the future of the ADR- 6000 in similar applications to be a function of
its cost, willingness of agencies to continue installing inductive loops, and willingness of
multiple agencies to develop agreements to share maintenance responsibilities ( e. g., for
shared data) due to its high cost. The fact that it can serve a dual role is expected to be a
positive factor, especially at more demanding locations with extremely high volumes and
where it can serve both the traffic operations and traditional data needs.
Video Image Vehicle Detection Systems
Autoscope Solo
The Autoscope Solo is a video imaging system whose cameras can be mounted either
overhead or to the side of the road. MnDOT tests of the Autoscope 30 ft over the center
of three lanes indicated excellent performance. The absolute percent volume difference
between the sensor data and loop data was under 5 percent for all three lanes. The
detector also performed well for speed detection. The absolute average percent difference
was 7 percent in lane one, 3.1 percent in lane two, and 2.5 percent in lane three. For
other mounting locations beside the roadway, the detector performed best when mounted
high and closest to the roadway. ( 4)
18
Table 1: Peek ADR- 6000 Classification Accuracy Comparison for I- 35
Vehicle Classification
1 2 3 4 5 6 7 8 9 10 11 12 Total Errors
Lane 1 Count 0 330 118 1 9 0 0 2 15 0 1 0 476
Errors 0 0 0 0 1 0 0 0 2 0 0 0 3
Lane 2 Count 0 299 84 0 16 3 1 11 23 0 1 0 438
Errors 2 1 3 1 1 8
Lane 3 Count 2 306 96 1 11 3 0 7 6 0 0 0 432
Errors 1 2 1 1 5
Lane 4 Count 0 312 88 1 14 1 0 4 2 0 0 0 422
Errors 1 1 1 1 4
Lane 5 Count 0 106 36 0 5 3 0 0 5 0 0 0 155
Errors 1 1
Totals 4 1356 423 7 60 12 1 24 55 0 2 0 1923
Total Errors 2 3 1 4 5 2 0 0 4 0 0 0 21
Source: Reference 7.
Source: Reference 7.
Figure 5: Speed Comparison of ADR- 6000, Doppler Radar, and Video Imaging System
Lane 5 Afternoon Peak Speeds I- 35 ( 7/ 3/ 02)
0
5
10
15
20
25
30
35
40
45
50
16: 45
16: 48
16: 51
16: 54
16: 57
17: 00
17: 03
17: 06
17: 09
17: 12
17: 15
17: 18
17: 21
17: 24
17: 27
17: 30
17: 33
17: 36
17: 39
17: 42
17: 45
Time
1 Minute Average Speed
ADR6000 RTMS Solo Pro Luminaire
19
Autoscope Solo Pro
At the time of this TTI research, the Autoscope Solo Pro was the latest version of the
integrated camera and processor. TTI tested this detector at both testbeds, but the results
reported in this section come from the I- 35 testbed and are based on five- minute samples
of count and speed data. The I- 35 site has five southbound lanes with lane 1 ( the median
lane) being furthest from the detector. For these tests, the Solo Pro was 35 ft above the
pavement and 6 ft from the nearest lane. ( 7)
In TTI research which ended in 2002, the Autoscope Solo Pro count accuracy was within
5 to 10 percent of the baseline counts during free- flow conditions, but it generally
diminished in all lanes when 5- minute interval speeds dropped below 40 mph and
especially during stop- and- go conditions. On all four of the monitored lanes, it
overcounted during free flow, but almost always within 10 percent of baseline counts.
During the peak periods, however, it undercounted. On lane 1, its error was always within
10 percent. On lane 2, its undercounts were about half within 10 percent and half between
10 and 20 percent. On lane 3 ( closer to the camera), its undercounts were two- thirds
within 10 percent and one- third between 10 to 20 percent of baseline counts. On lane 4,
the Autoscope had 9 out of 10 within 10 percent and one out of 10 between 10 and 20
percent. Speed and occupancy of the Solo Pro were the best of any non- intrusive devices
tested by TTI in this research project. Speeds were almost always within 0 to 3 mph of
the baseline system. Its 15- minute cumulative occupancy values differed from loops by
as much as 4 percent, but during most intervals its difference was less than 1 percent. ( 7)
Iteris Vantage
TTI tested the Iteris Vantage on I- 35 immediately following its initial release for freeway
applications. It had the highest standard deviation during free flow of all test devices on
both lanes 1 and 3. Overall, the Iteris count accuracy was not as dependent on prevailing
freeway speeds as some other devices. It did not have a significant bias toward
overcounting or undercounting. Its lane 1 morning peak counts were between - 1 and - 22
percent during slow speeds ( 20 to 30 mph); then it overcounted by as much as 10 percent
when speeds increased. It mostly overcounted in lane 1 during the afternoon peak with a
range from - 4 to + 10 percent. Lane 2 Iteris morning peak counts were all within the range
of 0 to - 10 percent except one and that one was at + 5 percent. In the afternoon, its range
was - 5 to + 10 percent, and all but four of its intervals were within ± 5 percent. Lane 3
Iteris morning peak counts were all within the range of + 2 to - 7 percent. In the afternoon
peak, the Iteris was + 5 to - 10 percent. Lane 4 counts were not available. ( 7)
For speed accuracy, the Iteris standard deviation was among the lowest of the devices
tested on both lanes 1 and 3. Its mean values of speed differences were lowest on lane 3,
perhaps indicating better calibration than on lane 1. The Iteris Vantage speed estimates
were both higher and lower than the baseline speeds but usually within 5 mph in lane 1
during the morning peak. During the afternoon peak, it was always within 5 mph on lane
1. On lane 2, its morning peak speed estimates exceeded the baseline by as much as 15
mph. During the afternoon peak, it was always within 5 mph on lane 2. On lane 3 during
20
the morning peak, its speeds were excellent in all intervals showing speeds within 0 to 2
mph of the baseline. During the afternoon peak, it was within 5 mph of the baseline. On
lane 4, the Iteris was consistently within 5 mph of baseline during the morning peak.
Speeds during the afternoon peak were not available. ( 7)
Of the three non- intrusive devices tested for occupancy output in lanes 3 and 4, the Iteris
Vantage was the second most accurate. Its 15- minute cumulative occupancy values
differed from loops by as much as 8.1 percent, but during most intervals its difference
was less than 6 percent.
Traficon NV
MnDOT Phase II tests mounted the Traficon camera directly over the lanes at heights of
21 ft and 30 ft facing downstream. The preferred orientation was facing oncoming
vehicles, but site features precluded this orientation. At the 21- ft height, the absolute
percent difference between the sensor data and loop volume data was under 5 percent for
all three lanes. At the 30- ft height, its off- peak performance was similar but it
undercounted during congested flow showing an absolute percent difference of some
15- minute intervals from 10 percent to as high as 50 percent. Reasons suspected for the
reduced accuracy were snow flurries and sub- optimal calibration. Its speed accuracy at 21
ft indicated good performance. Its absolute average percent difference was 3 percent in
lane 1, 5.8 percent in lane 2, and 7.2 percent in lane 3. During the snowfall, its speed
accuracy declined to a range of 8.9 percent to 13 percent. ( 4)
A critical finding of the Phase I NIT research was that mounting video detection devices
is a more complex procedure than that required for other types of devices. Camera
placement is crucial to the success and optimal performance of this detection device.
Lighting variations were the most significant weather- related condition that impacted the
video devices. Shadows from vehicles and other sources and day/ night transitions also
impacted count accuracy. ( 3)
Microwave Radar Detectors
Minnesota Phase I researchers tested one radar device, the RTMS X2 by Electronic
Integrated Systems, Inc. ( EIS). This device can be mounted either overhead or in a
sidefire position aimed perpendicular to traffic. The RTMS is easily mounted but requires
a moderate amount of calibration to achieve optimal performance. MnDOT researchers
found that rain affected the performance of the RTMS, although they attributed this
degradation to water entering the device and not to limitations of the technology. When
the RTMS was mounted overhead, it undercounted vehicles by 2 percent or less at the
freeway site. When it was in a sidefire orientation, it undercounted traffic by
approximately 5 percent. ( 3)
Results of research ending in 2002 at TTI’s I- 35 testbed in Austin indicated that the
RTMS X2 is more accurate in both counts and speeds in the overhead position although it
covers only one lane in that orientation. The more popular orientation is sidefire, so the
following discussion focuses on its sidefire accuracy. In sidefire, the RTMS can generate
21
speeds and counts for five lanes with reasonable accuracy. ( The tests at the I- 35 site used
five lanes.) Its advantages also include being mounted only 17 ft above the roadway, and
good user interface. Its coverage and initial cost make the RTMS an economical means
of monitoring several lanes. ( 7)
More specifically, the TTI research found that the RTMS undercounted in all lanes
during both peak and off- peak intervals. Its five- minute counts in lane 1 were all in the -
10 to - 25 percent range. ( The detector location was nearest lane 5, so lane 1 was farthest
away.) Researchers did not evaluate lane 2. In lane 3, 95 percent of the time intervals
were within 5 percent of baseline. In lane 4, 98 percent of the time intervals were within
15 percent of baseline counts. These findings indicate that distance from the detector and
occlusion affected count accuracy. Lane 1 was slightly worse than lane 3, and lane 4 was
slightly worse than lane 3, suggesting either calibration differences or middle lanes
naturally being better than either extreme. Aggregated speed estimates by the sidefire
RTMS differed from baseline speeds by as much as 15 mph during peak periods, but it
was usually within 5 to 10 mph of baseline speeds during the off- peak. This research did
not include occupancy tests on the RTMS. ( 10)
In the overhead position, the RTMS was more accurate in counting vehicles, but it only
covers one lane. In TTI tests, the overhead RTMS ( Doppler mode) generated excellent
speeds until prevailing traffic speeds dropped below about 15 mph. It is a mature product
and is not significantly affected by weather or lighting conditions. ( 7)
The Detector Evaluation and Testing Team ( DETT) of the California Department of
Transportation ( Caltrans) tested two radar detectors, the RTMS X3 and the Wavetronix
SmartSensor. ( 13) The test site was the Caltrans test facility on I- 405 near the University
of California at Irvine, which uses the seven northbound lanes for tests. Traffic volume at
this site is about 3 million vehicles per week. Another technology tested at this site was
the Inductive Signature Technologies ( IST) product that has the capability of tracking
vehicles using inductive loop signatures. The team collected both 30- second and 5-
minute aggregate data at this site. Results indicate that the ground truth inductive loops
overcounted by 1.0 to 1.5 percent. This overcounting is due at least in part to lane
changers that cross sensors in two adjacent lanes.
The California tests indicated that with proper installation and calibration either detector
can deliver better than 95 percent overall vehicle count accuracy at 5- minute and
30- second intervals and 95 percent speed accuracy at 5- minute intervals. However, due to
a very strict Caltrans specification, neither detector was found to be suitable for
determining occupancy. One of the comments from researchers was that the RTMS
requires considerable effort to achieve acceptable data accuracy, requiring expert know-how
and a lot of time to set up and calibrate. The Wavetronix only required 15 to 20
minutes total to set up, whereas a factory representative took about one hour per lane for
the RTMS. Also, the technology can be very accurate in the center of a roadway but the
presence of trucks and heavy traffic can cause the detectors to miss counting of some
vehicles as well as to create some false readings in side lanes. ( 13)
22
Since a number of previous studies had compared aggregate data from one or more
detectors to concurrent measurements from another device ( perhaps a ground truth
device), Coifman ( 14) chose to compare actuations of individual vehicles at one detector to
concurrent measurements of the same vehicle at another detector. He used four inductive
loop sensor models and the RTMS. More specifically, the research used the following
loop detection units: Peek GP6 and Reno A& E Model 222 inductive loop detectors, along
with the reportedly higher performing GTT ( 3M) and IST Model 222 detectors. The
research used the Berkeley Highway Laboratory to collect data from all five of the
detectors using Videosync, the software package developed by Caltrans Division of
Research and Innovation ( DRI), as the primary tool for data reduction. This software
allows the direct comparison between concurrent detector and video data.
Each of the sensors exhibited problems. Study conclusions stated that agencies could
identify and correct most of the problems with additional fine- tuning in the data
processing by the controller or data aggregator, but most operating agencies did not
attempt to accomplish the correction. Therefore, the study findings should replicate
conventional practice. Some of the errors could be corrected by improved controller
logic, but some would require a trip to the field to correct. The Reno detector tended to
flicker on for short periods in absence of a vehicle in the detection zone, which could be
correctable in the controller software. Other errors resulted from lane- changing
maneuvers over the detection zone. IST and Peek tended to detect such vehicles in both
lanes, while the Reno and GTT ( 3M) sensors tended to underestimate the on- time of
vehicles changing lanes in one lane while not detecting them in the other. ( 14)
The RTMS showed systemic errors in its performance— manifested as differences
between nearest ( small detection zone) and farthest lanes ( occlusion). This systematic
change in on- time would be an important consideration for applications that rely on
occupancy. Also, the RTMS count and on- time are typically noisier than loops, although
pluses and minuses tend to cancel each other. Detection zone sizes varied across all
detectors, from the larger detection zones of the RTMS and even across the four models
of loop detectors whose in- pavement dimensions were the same. These variations will
impact the occupancy values. Of course, the sizes of loop detection areas are a function
of sensitivity settings, but perhaps equally important are site- specific factors. ( 14)
Another relevant report by Coifman ( 15) also used the Berkeley Highway Laboratory to
investigate aggregate data from the RTMS sensor. The study evaluated the performance
of the RTMS in sidefire mode relative to loop detectors in the freeway setting. The
documented results first reported the aggregated data by the RTMS using its internal
controller emulation and compared these results with data from nearby dual- loop
detectors. The RTMS measures of flow and occupancy are noisier than loop detectors,
although the RTMS estimates for speeds are almost as good as those from single- loop
detectors. The second aspect of the study considered aggregate measurements from
contact closure data and compared RTMS results against the dual- loop detectors. For
reference, the research also compared one loop against the adjacent loop in the same lane
in a trap loop configuration. In the flow measurements, the RTMS was within 10 percent
of values generated by the loops with the loops being within 3 percent of each other.
23
Occupancies were not as accurate, ranging from 13 percent to 40 percent, again
compared to the inductive loops.
A research project conducted by the Ohio Research Institute for Transportation and the
Environment ( ORITE) investigated the use of a custom- built trailer fitted with two
microwave radar detectors to monitor traffic along selected segments of roadway. The
trailer consisted of a steel frame with a solar panel plus battery box containing four deep-cycle
gel batteries and a power controller. The solar unit was rated at 225 watts and
outputs 12V DC; it was also equipped with a charge controller capable of regulating up to
15 amps of current. As equipped, the system can run for 8 days on batteries without
sunlight. The trailer had two telescoping poles capable of reaching heights of 20 ft; it had
four sockets so that the poles could be erected on either side of the trailer. It also had anti-theft
devices such as a removable hitch, and special lock- nuts on the wheels. ( 16)
During a traffic monitoring session, the ORITE trailer used one Wavetronix SmartSensor
model SS105 attached to each pole, with each detector pointed in the same direction and
operating in parallel. The available information did not specify the separation distance,
but photos indicated a separation of about 8 ft. The available information also did not
discuss the possibility of interference between the two detectors, which would likely
occur at that spacing and orientation. ORITE typically operated both detectors in the
sidefire orientation. The trailer also housed a controller, which was a small computer
used to collect the data from each sensor and combine the data into a single text file.
Storage of the text file is on the computer’s hard drive and on a 256 MB flash memory
card. Setup of the entire operation takes about 45 minutes— 30 minutes for the trailer and
about 15 minutes for the sensors. ( 16)
The data collected by the system include: timestamp, lane number, and moving average
speed ( based on the last 16 vehicles) from the first sensor followed by a similar dataset
from the second sensor. Next came an average of the two running speeds, vehicle length,
and speeds for each sensor. Vehicle classes for this research were: Class 0 ( 0 to 20 ft),
Class 1 ( 21 to 40 ft), and Class 2 ( at least 41 ft). A portion of the ground truth came from
videotape with time- stamped video synchronized with the same laptop the radar units
were synchronized with. The baseline vehicle speeds came from a Kustom Signals TR- 6
radar unit. ( 16)
Results indicate that the Wavetronix system misses some vehicles due to occlusion and it
sometimes registers phantom vehicles from extraneous radar echoes ( e. g., from a truck in
an adjacent lane). On one of the test days, the number of phantoms was 7.03 percent, but
on other days, the number of phantoms and misses was always less than 5 percent, and
often under 1 percent. Speeds measured by the Wavetronix system ( based on the moving
average technique) usually correlated well with true speeds. These moving average
speeds were a combination of the speeds detected by both sensors. The largest difference
was 3 mph. The standard deviations in data measured by the trailer were always higher
than those from the hand- held radar unit, generally by a factor of 2 to 3. The smallest
difference found was 0.1 mph ( 2.0 mph Wavetronix vs. 1.9 mph radar), and the largest
difference was 3.8 mph ( 4.2 mph Wavetronix vs. 1.1 mph radar). ( 16)
24
Other results based on vehicle length ( or classification) were not as accurate. For
example, one dataset had 8 percent true vehicles with lengths over 40 ft while the
Wavetronix data indicated 21.4 percent with lengths over 40 ft. Some results were better
and some were worse, but the authors conclude that the system does not reliably estimate
the number of trucks in the traffic stream. Weather was not a factor in any of the tests. ( 16)
Passive Acoustic Detectors
In MnDOT phase II tests, the sensor was bench- tested in the lab in March 2001, and then
was mounted on the sidefire tower in May 2001. These tests used a total of five heights
and three offsets during the actual field tests on the freeway between October 2001 and
January 2002. Results indicated that at the first base ( 15 ft from the first lane), the
detector provided better results for lanes 2 and 3 than for lane 1. The 24- hour data show
that the absolute percent differences for lanes 2 and 3 were under 8 percent at all heights,
and between 12 percent and 16 percent for lane 1 with heights less than 30 ft. Results
were good for free- flow traffic conditions, but the detector undercounted during
congested flow when speeds dropped.
Test data showed that 15- minute absolute percent differences were between 0 and 5
percent during off- peak, and varied from 10 percent to 50 percent during congested
periods, depending on site geometry. In speed detection, the detector performed well at
base one. The absolute average percent differences were under 8 percent for most
mounting heights and between 12 percent and 16 percent for lane 1 at heights less than
30 ft. Overall test results show that the detector performs best when mounted with equal
height and horizontal offset between the detector and the centerline of multiple lanes ( 45-
degree angle). ( 4)
The first full test of the SAS- 1 by TTI was at its S. H. 6 testbed in research ending in
2000. ( 6) The only factor found to compromise the SAS- 1 count accuracy in this series of
tests was rainfall. The detector’s performance declined during wet weather, as indicated
by a comparison of Tables 2 and 3 ( page 25). The vendor, who was involved on- site in
the initial setup, discovered an error in the lane sensitivity setting that might have
accounted for the undercounting that occurred during rain. Unfortunately, there was no
other wet weather during these tests to verify the assumed improvement.
The second project at TTI to test the SmarTek SAS- 1 detector ended in 2002. The SAS- 1
height above the freeway was 35 ft and its offset from the nearest lane ( lane 5) was 6 ft.
Its count accuracy for lane 1 ( farthest) dropped during congested flow compared to free
flow, but on lane 3 the accuracy was similar for the two conditions. The SAS- 1 generally
undercounted almost all intervals. In lane 1 during the a. m. peak and while speeds were
over 40 mph its count range was 0 to - 10 percent. During slower speeds, its range was -
12 to - 32 percent. Its range for lane 1 afternoon peak intervals was + 2 to - 20 percent with
all but two intervals between 0 and - 10 percent. The SAS- 1 lane 2 ranges for the morning
and afternoon peaks were between + 5 to - 18 percent and 0 to - 10 percent, respectively.
Lane 3 counts fell in the range of + 6 to - 12 percent during the morning peak and - 2 to - 14
25
percent during the afternoon peak. In lane 4, it undercounted during both the morning and
afternoon peak by the range of - 3 to - 15 percent and 0 to - 12 percent, respectively. ( 7)
Table 2: SAS- 1 Count Error Rates on S. H. 6 during Dry Weather
Lane
Error Range (%) Left Right
0 to 10 % 353 of 378 ( 93.4 %) 376 of 378 ( 99.5 %)
10 to 20 % 25 of 378 ( 6.6 %) 2 of 378 ( 0.5 %)
20 to 30 % 0 0
Source: Reference 6.
Table 3: SAS- 1 Count Error Rates on S. H. 6 during Wet Weather
Lane
Error Range (%) Left Right
0 to 10 % 4 of 20 ( 20.0 %) 4 of 20 ( 20.0 %)
10 to 20 % 12 of 20 ( 60.0 %) 3 of 20 ( 15.0 %)
20 to 30 % 4 of 20 ( 20.0 %) 13 of 20 ( 65.0 %)
Source: Reference 6.
The speed accuracy of the SAS- 1 was similar in congested flow and free flow on
lane 1. For lane 3, its mean and standard deviations indicate that its accuracy was more
consistent in free flow than in congested flow. The SAS- 1 consistently overestimated
speeds in lane 1 during the morning peak by 5 to 10 mph. During the afternoon peak, it
overestimated speed by as much as 20 to 25 mph during very slow speeds then improved
to within 5 mph as speeds reached free- flow conditions. On lane 2 during both the
morning and afternoon peaks, the SAS- 1 was almost always over the baseline system by
0 to 5 mph with a maximum of 10 mph. On lane 3 this detector was consistently within 2
to 5 mph of the baseline system. On lane 4, its morning peak speed estimates were
consistently within 5 mph and its afternoon peak speed estimates were less consistent but
still within ± 5 mph. ( 7)
This research also compared the lane occupancy output of the SAS- 1 with the baseline
loop system in lanes 3 and 4. Its 15- minute cumulative occupancy values differed from
loops by as much as 14.7 percent, but during most intervals the difference was less than 4
percent. ( 7)
Active Infrared Detectors
Preliminary testing of active infrared detectors by public agencies indicates promising
results for monitoring vehicle speeds and classifications. Active infrared systems appear
to be immune to lighting changes but appear to be affected by heavy fog and heavy dust.
Disadvantages of infrared sensors include: cost; inconsistent beam patterns caused by
changes in infrared energy levels due to passing clouds, shadows, fog, and precipitation;
26
lenses used in some devices may be sensitive to moisture, dust, or other contaminants;
and the system may not be reliable under high- volume conditions. ( 5)
An active infrared device detects vehicle presence by emitting a laser beam toward the
road surface and measuring the time required for the reflected signal to return. The
presence of a vehicle reduces the return time for the reflected signal to the detection unit.
Phase I of the Minnesota project evaluated one active infrared device, the Schwartz
Electro- Optics ( SEO) Autosense I, and only on the freeway. In addition to detecting
stationary and moving vehicles by presence, the Autosense I can obtain vehicle speed and
vehicle profile ( which researchers can use for classification). Heavy snowfall, as well as
rain and freezing rain, caused the detector to both overcount and undercount vehicles.
During snow, the undercounting was attributed to vehicles traveling out of the detection
zone, while overcounting was probably the result of falling snow reflecting the laser
beams causing false detections. These discrepancies were attributed to the change in
reflectivity properties of the pavement. ( 3)
In research aimed at reducing the number of trucks stopping at isolated signalized
intersections, TTI tested the SEO Autosense II as one of the options for detecting and
classifying vehicles. Weaknesses included its cost ($ 10,000 per lane in 1995), lack of
ruggedness for field applications, inconsistent speed accuracy, and requirement for
mounting directly over the lane. For a sample of 160 vehicles, it missed 3 percent and
misclassified 7.5 percent. ( 9)
Magnetic Detectors
Passive magnetic devices measure the change in the earth’s magnetic flux created when a
vehicle passes through the detection zone. The only two systems deemed appropriate for
this research are the GTT ( 3M) non- invasive microloop and Sensys Networks
magnetometers. Both systems are passive sensing systems that are based on the earth’s
magnetic field. When a vehicle passes through the detection zone, it temporarily distorts
the earth’s magnetic field enough to create a measurable change in the field. ( 4) A passive
magnetic device must be relatively close to the vehicles it is detecting; therefore most
applications require installation below the pavement. The Sensys Networks
magnetometers require a 4- inch diameter core to a depth of 3 inches, mount flush with
the pavement surface, use wireless communication to the roadside, and require a short
lane closure for installation or replacement. The GTT ( 3M) detectors use copper wire
connections, are more time- consuming to install requiring a horizontal bore in most
cases, but can be installed at depths up to about 36 inches without significant loss in
accuracy. Both systems can be used for vehicle speed, length, and occupancy if two
detectors are available per lane at a known spacing.
GTT ( 3M) Microloops
MnDOT results indicated that GTT ( 3M) probe performance was compromised by water
in the conduit. During periods of heavy rain, erratic performance could have been due to
27
intermittent grounding problems. Vehicles straying from the normal lanes resulted in
overcounting during periods of snow. ( 3)
Besides installation in a horizontal bore underneath the roadway, GTT ( 3M) detectors
have also performed successfully under bridge decks. Installers must first use a
magnetometer to determine proper placement of the probes; otherwise optimum
performance requires a trial- and- error process. One of the requirements of this system is
that the probes remain relatively vertical, so keeping the horizontal bores straight is
critical. Probes placed in a non- vertical orientation can lead to speed errors. MnDOT tests
under pavement indicated excellent volume and speed results. The absolute percent
volume difference between sensor and baseline was under 2.5 percent, which is within
the accuracy capability of the baseline loop system. For speeds, the test system generated
24- hour test data with absolute percent difference of average speed between baseline and
test system from 1.4 to 4.8 percent for all three lanes. ( 4)
TTI tested GTT ( 3M) microloops at its S. H. 6 testbed in College Station. At this
relatively low- to moderate- volume site over a six- day count period, TTI found that
microloop counts were within 5 percent of baseline counts 99.4 percent of the time in the
right lane ( dual probes). In the left lane ( single probes), 94.5 percent of the 15- minute
intervals were within 5 percent, 4.5 percent were between 5 and 10 percent, and in 1.0
percent there was a more than 10 percent difference from baseline. ( 6)
Sensys Networks
Cheung et al. investigated the use of single wireless magnetic detectors as an alternative
to inductive loops for traffic monitoring on freeways as well as at intersections. Their
advantages appear to include cost, ease of deployment and maintenance, and enhanced
measurement capabilities. Components of this magnetic detector include “ sensor nodes,”
which communicate with an Access Point ( AP) unit installed at the roadside. A sensor
node is comprised of a magnetic sensor, a microprocessor, a radio, and a battery. ( 17)
The paper covers two experiments, with the first and longer one being a two- hour
monitoring session on Hearst Avenue in Berkeley, California, downstream of a signalized
intersection. During this two- hour session, 332 vehicles passed through the detection
zone. The single magnetic sensor achieved a detection accuracy of 99 percent ( 100
percent if motorcycles are excluded), and average vehicle length and speed estimates that
appear to exceed 90 percent. ( 17)
For vehicle classification, a single dual- axis magnetic sensor measures the earth’s
magnetic field in both the vertical direction and along the direction of the lane, each
sampled at 64 Hz. A simple algorithm uses the information to classify the vehicle into six
types: passenger vehicles, SUV, van, bus, mini- truck, and truck. Of the sampled vehicles,
the detector correctly classified 24 out of 37 vehicles ( 63 percent). Combining classified
vehicles into the FHWA classification scheme suggests an 83- percent accuracy rate for
the FHWA scheme. The sample size was small, but the results appear to be promising.
The sensor correctly classified all buses, vans, and passenger vehicles, but it had
28
problems with SUVs and mini- trucks. Further experiments are needed to determine its
accuracy with trucks. It is important to note that adding length as a measured feature of
the single magnetic sensor would probably improve the classification accuracy. ( 17)
This research compared this single magnetic detector and its capabilities with inductive
loops. In comparison, measuring accurate lengths with loops requires two loops
compared to only one magnetic detector. The magnetic detector is easily installed and
measures the earth’s magnetic field, which is a three- dimensional vector. This detector
records the changes in the field caused by different parts of the vehicle and that is how it
can classify the vehicle. The size of an inductive loop, on the other hand, is larger,
causing it to lose some of the distinctive features from the inductive signature. In other
words, magnetic signatures provide more detail on the vehicle to improve its use as a
classifier.
Other advantages of magnetic detectors include the ability to install them on bridges,
where sawcuts ( for loops) would weaken the structure. Finally, wireless magnetic sensor
networks should be much less expensive to maintain than inductive loops while providing
more of the needed information. ( 17)
The authors suggest that both the speed and classification accuracy could be improved
significantly by using two magnetic detectors spaced a known distance apart. They
predict vehicle classification accuracies in the 80- percent range would be likely. The
authors planned on additional tests to further develop the classification accuracy. ( 17)
SINGLE AREA TECHNOLOGIES
The next three sub- sections cover single area detector systems using the following three
technologies: passive infrared, microwave, and pulse ultrasonic. These detectors are
simpler than the detectors covered above in that they do not distinguish vehicles by lane.
The application of these detectors in a freeway environment might be most appropriate
for ramps ( e. g., ramp metering or queue length monitoring).
Passive Infrared Detectors
MnDOT researchers found that passive infrared devices were not impacted by weather
conditions and were very easy to mount, aim, and calibrate. However, there were
significant differences in performances of the devices tested. Eltec Models 833 and 842
showed significant fluctuations in count accuracy. The ASIM IR 224 was easy to mount
and calibrate, and repeatability was good. Results for these detectors were as good as
within 1 percent of baseline data for freeway tests. ( 3) Two other research projects that
included passive infrared detectors found mediocre results. ( 18, 19)
Microwave Detectors
The MnDOT research tested four different Doppler microwave devices, but the research
team presented detailed data for only two. All four devices were easily mounted and
29
calibrated, and none of the devices seemed to be affected by weather conditions. The
devices tested revealed differences in performance. Both the Peek PODD and the Whelen
TDN- 30 required mounting overhead or slightly to the side of the roadway.
Under optimal conditions, the Peek PODD was able to count vehicles at the freeway site
within 1 percent of the baseline, provided that the device was properly aimed. During one
of the procedures, it detected vehicles in the adjacent lane. The Whelen TDN- 30
undercounted at the freeway site by approximately 3 percent. ( 20)
Pulse Ultrasonic Detectors
The Minnesota research team tested two pulse ultrasonic devices, the Microwave Sensors
TC- 30 and the Novax Lane King. Overhead mounting of the devices provided optimal
signal return and vehicle detection; however, sidefire mounting is possible for some
devices. Pulse ultrasonic devices are relatively easy to mount and weather does not affect
performance; however, the ease of calibration varies with devices. Calibration for the
Lane King was extensive for optimum performance.
Both detectors were accurate in counting vehicles at the freeway site, but the two pulse
ultrasonic devices interfered with one another when mounted side by side for tests, which
would not be done in practice. ( 3)
RECENT MULTIPLE TECHNOLOGY SOURCES
University of Utah
In a report published in 2003, Martin and Feng developed a traffic detector selection
procedure which included the following technologies: inductive loops, magnetic, active
infrared, passive infrared, microwave radar, ultrasonic, passive acoustic, and video image
vehicle detection systems.
The selection criteria included: general installation conditions, cost, data accuracy,
reliability, and ease of installation and maintenance.
Tables 4 and 5 ( page 30) and Tables 6 and 7 ( page 31) summarize this information. ( 21)
30
Table 4: Detector Cost Comparison
Technology/ Sensor Device Cost Lanes d Mounting d
3M Microloop 2 ch. Canoga Detector $ 546
4 ch. Canoga Detector $ 704
702 Microloop Probe $ 160
701 Microloop Probe $ 138
Installation kit $ 114
Carriers $ 355/ pkg
Cable: $ 0.39/ ft
S 3- inch conduit
placed under
roadway
SmarTek SAS- 1 $ 3500/ unit M ( 5 lanes) S ( 25- 40 ft)
Autoscope Solo a
Single direction: $ 4900
Autoscope
Autoscope 2020
Single direction: $ 4820
M ( 32) b O/ S
Traficon $ 4000 per camera ( camera,
processor, housing, lens, cables,
surge protection, setup & training) c
M ( 24) b O/ S ( 25- 45 ft)
Source: Adapted from Martin and Feng: Reference 21.
a Autoscope solo has integrated camera and processor.
b Maximum number of detection zones per camera.
c A high resolution CCD black/ white or color camera. The video camera should provide detailed video
without lag, image retention, or geometric distortion.
d S – Single lane detector, M – Multiple lane detector, O – Overhead, S – Sidefire.
Table 5: Detector Error Rates
Sensor
Mounting
Location
Count
Speed a
Evaluation Organization
( Year)
3M Microloop Pavement 2.5% 1.4%- 4.8% MnDOT ( 2002)
3M Microloop Bridge 1.2% 1.8% MnDOT ( 2002)
3M Microloop Pavement 5% μ= - 0.25 mph
σ = 3.6 mph
TTI ( 2000)
SAS- 1 Sidefire 8%- 16% 4.8%- 6.3% MnDOT ( 2002)
SAS- 1 Sidefire 4.0%- 6.8% 3.4%- 6.8% TTI ( 2000)
SAS- 1 Sidefire 10% μ = - 0.5 mph
σ = 4.8 mph
TTI ( 2000)
Autoscope
Solo
Sidefire 5% 8% MnDOT ( 2002)
Autoscope
Solo
Overhead 5% 2.5%- 7% MnDOT ( 2002)
Autoscope
Solo
Sidefire 2.1%- 3.5% 0.8%- 3.1% TTI ( 2000)
Traficon Sidefire 5% ( 45 ft) 2%- 12% MnDOT ( 2002)
Traficon Overhead 10%- 15%
( 25- 30 ft)
3%- 7.2% MnDOT ( 2002)
Source: Adapted from Martin and Feng: Reference 21.
a μ = mean, σ = standard deviation.
31
Table 6: Detector Ease of Installation and Reliability
Technology/ Sensor Ease of Installation b Ease of Calibration b Reliability a
3M Microloop 0 1 2
RTMS 2 1 1
SmarTek SAS- 1 2 2 2
Autoscope Solo 2 1 2
Traficon 2 1 2
Source: Adapted from Martin and Feng: Reference 21.
a Reliability level is based on the performance shown in tests.
b 2: Performs satisfactorily in the stated condition; 1: Meets some but not all criteria for
satisfactory performance; 0: Does not perform satisfactorily in the stated condition.
Table 7: Estimated 2003 Life- Cycle Costs for a Typical Freeway Application
Detector Initial
Cost
Mounting
Type c
Install.
Cost
Ann. Mtce.
Cost
System
Life ( yrs)
Life- Cycle
Cost/ system a
3M Microloops $ 13,125 b $ 200 15 $ 1380
RTMS $ 6600 O $ 2400 $ 1700
S $ 400
$ 200 7
$ 1370
SmarTek SAS- 1 $ 7000 S $ 800 $ 400 7 $ 1700
Autoscope O $ 3000 $ 1980
Solo Pro
$ 9800
S $ 1000
$ 400 10
$ 1730
Traficon $ 8000 O $ 3000 $ 1760
S $ 1000
$ 400 10
$ 1510
Source: Adapted from Martin and Feng: Reference 21.
a Costs are for a total of six freeway lanes, three per direction.
b Total of 16 lanes and 32 probes.
c O - Overhead, S – Sidefire.
University of Hawaii
The Department of Civil and Environmental Engineering at the University of Hawaii at
Manoa evaluated eight vehicle detectors ( five non- intrusive) at several locations in
portable and permanent installations. These systems are: GTT ( 3M) microloops, Spectra
Research ORADS portable laser sensor, RTMS model X2, SmarTek SAS- 1, and
Wavetronix SmartSensor SS105. The research retrieved data from these detectors using
TrafInfo’s Trafmate satellite modem, TrafficWerks cellular system, and conventional
9600 baud modems. ( 22)
The GTT ( 3M) microloops and ORADS portable laser sensor require installations below
or very near the road surface, respectively. Lane closures are not required but personnel
are still exposed to traffic. The GTT ( 3M) microloops and Canoga 702 detector card
provided excellent volume and speed results. They are expected to have a long life cycle
but their initial cost and installation were expensive. The ORADS laser sensor from
Spectra Research did well with volume counts but performed poorly in classification. It
was found that the ORADS laser sensor did not perform well on uneven pavement. ( 22)
32
The research included RTMS X2, SmarTek SAS- 1, and SmartSensor SS105 in sidefire
mode at various heights and distances from the roadway, and in inclement weather
conditions. Researchers found that these three sensors could provide high- quality data at
a low cost, with low energy consumption, and simple calibration. Installation required a
pole at least 20 ft tall, and offset at least 20 ft from the first lane. Researchers retrofitted a
trailer- based “ light plant” with the sensors and deep- cycle batteries for power. They
deployed this portable, stand- alone unit at various locations where sensor mounting
options were limited. ( 22)
The baseline system for determining the volume, speed, and classification performance
aspects of the sensors was usually pre- existing loops, supplemented in some cases with
manual counts. Table 8 on page 32 summarizes the values associated with some
qualitative ratings of the detectors. Based on comparisons of volume counts and speeds,
the non- intrusive detectors rated as follows:
• The count rating for the RTMS X2 and RTC data unit by EIS was good to very
good. The speed rating for the X2 and RT data unit was very good to excellent.
• The count rating for the SAS- 1 acoustic sensor and SAS- CT board by SmarTek
was good to very good. The speed rating for the SAS- 1 and SAS- CT board was
very good to excellent.
• The research did not rate the SmartSensor SS105 by Wavetronix by values from
the table for either speeds or counts, but described its performance as similar to
the RTMS unit.
Table 8: Sensor Performance Descriptions ( Hawaii)
Rating Volume/ Classification
(% error)
Speed ( mph)
Excellent ± 1 ± 3
Very Good ± 3 ± 6
Good ± 5 ± 10
Possibly Adequate ± 10 ± 15
Inadequate > ± 10 > ± 15
Source: Prevedouros: Reference 22.
Researchers recommended the SmartSensor as the top sensor based on ease of setup,
lower height requirement, and exceptional feedback and assistance from the vendor. The
SmartSensor’s auto- ranging and calibration features made it the quickest sensor to install.
The researchers recommended both the SmartSensor and the RTMS for quiet rural
locations with power already available. They recommended the SAS- 1 for battery or
solar power operations due to its minimal power consumption. The SAS- 1 was able to
detect bicycles in quiet locations, but loud music and other background noises caused the
sensor to register a count. ( 22)
33
Pennsylvania Department of Transportation
In a research article published in 2006 sponsored by the Pennsylvania Department of
Transportation ( PennDOT) and the Federal Highway Administration, a research team
field- tested selected non- intrusive traffic data collection equipment. The site selection
process provided a cross- section of roadside environments for the equipment setup and
utilized sites near in- pavement traffic counting stations operated by PennDOT. These
ground truth data sources were either Automated Traffic Recorder ( ATR) sites or Short-
Term In- Pavement ( STIP) sites. ( 23)
Two of the detectors tested were RTMS microwave radar sensors by Electronic
Integrated Systems and two were SAS- 1 acoustic sensors by SmarTek. The RTMS model
used in this research was the X2. Mobility Technologies, a local data provider,
participated in the selection of sites and equipment to ensure that test detectors were
operating properly. This process involved a quality control process that was not covered
in detail in the report. Of the six sites selected by Mobility Technologies, two were in the
Pittsburgh area and four were in the Philadelphia area. The other two test detectors were
the Wavetronix SS105, and “ The InfraRed Traffic Logger” ( TIRTL). Some tests were set
up in a portable scenario and some were permanent; the difference was in how the
detectors were mounted. ( 23)
One limitation of the study was its short duration – in its entirety lasting only two days
( September 14 and 15, 2005). Vendor representatives were allowed a maximum of two
hours for setup of the equipment, and the data collection at each site was only four hours
in duration. The authors calculated absolute percent difference ( APD) as follows:
×100%
−
=
M
S M
V
V V
APD
where:
APD = Absolute Percent Difference (%)
Vs = Volume from the sensor or ATR/ STIP ( vehicles)
Vm = Volume from the manual count ( vehicles)
Table 9 provides summary information on the four sites. Site number 1 and site number 3
are in the same location; however site 1used only the northbound lanes, while site 3 used
both directions simultaneously with a single sensor. ( 23)
Table 9: Site Descriptions ( Pennsylvania)
Site No. Description
1 S. R. 0119, NB only, freeway – single direction
2 S. R. 0040, both directions – two- lane highway
3 S. R. 0119, both directions – freeway
4 S. R. 0040, both directions – 5- lane suburban arterial
Source: French and French: Reference 23.
34
Table 10 summarizes the results of the portable setup tests for data collected in 2005. It
provides the average APD data across the 16 15- minute time intervals along with the
APD between the sensor and manual count in the total four- hour volume. In these tests,
the sensor that matched the manual counts most closely was the SmartSensor. For the
permanent setup comparisons, there was no statistically significant difference at the 95
percent confidence level between the best performing microwave sensor and the best
performing acoustic sensor. ( 23)
Table 10: Results of Portable Setup Field Testing ( Pennsylvania)
Site No. SmartSensor SAS- 1 TIRTL RTMS
( Dir.) APD% 4- Hr% APD% 4- Hr% APD% 4- Hr% APD% 4- Hr%
1 ( NB) 6 5.8 1 1.0 1 0.2 6 4.1
2 ( EB) 4 3.7 2 0.7 1 0.0 3 1.3
2( WB) 2 0.5 20 20.4 1 0.2 25 24.9
3( NB) 2 1.4 1 0.8 19 18.5 2 1.5
3( SB) 2 0.9 3 1.0 26 26.0 1 0.1
4( EB) 2 1.2 3 1.1 25 25.7 2 0.2
4( WB) 3 1.1 5 5.0 15 14.9 43 40.6
Average 3 2.1 5 4.3 13 12.2 12 10.4
Source: French and French: Reference 23.
Nebraska Department of Transportation
Like other states, Nebraska has used inductive loop detectors as its dominant means of
detection for many years. However, the installation and maintenance of loops has become
an issue due to the necessary lane closures that cause excessive delays to motorists and
unacceptable risks for construction workers. These factors, along with the increased
availability of emerging non- intrusive detectors that could be used to replace loops led to
this research project to investigate the pertinent aspects of two of the newer detectors –
the Autoscope RackVision by Econolite and the SmartSensor SS105 by Wavetronix.
Even though previous research had evaluated these detectors, the Nebraska Department
of Roads ( NDOR) believed it was important to test them under local conditions because
of the differences in detection systems, applications, and operating conditions. Study
goals were to compare count accuracy of the two detection systems and compare them to
inductive loops and also to compare speed and vehicle classification characteristics. ( 24)
The test location for this research was on I- 80 at the 36th Street overpass in Omaha,
Nebraska. The paper did not provide traffic volume data, but it indicated that it was
adjacent to an interchange which experienced over 3 million hours of delay in the year
2002, ranking it among the top 60 bottlenecks in the United States. ( 25) Three major
elements needed for this test were: 1) a data collection trailer equipped with the
SmartSensor and positioned beside the roadway, 2) a data collection van equipped with
the Autoscope camera and video recording equipment, and 3) an existing NDOR loop
detector station with a single inductive loop in each freeway lane. ( 24)
35
The van was positioned so cameras were centered over the eastbound lanes of I- 80 and
cameras were 63.1 feet above the roadway. Recorded video from the van was post-processed
using the Autoscope RackVision in a laboratory setting. Lane numbering used
the convention of lane 1 being the outside lane and lane 5 is closest to the median. The
SmartSensor was installed in sidefire configuration at a mounting height of 18 feet and
was 19 feet away from lane 1. The SmartSensor system had a Click! 100TM module which
provided contact closure data to a Peek Automated Data Recorder ( ADR) model ADR-
1000 Plus, where data were stored. The data were subsequently downloaded to a laptop
computer at the end of each data collection period. ( 24)
Data collection for this research occurred on three weekday morning peaks ( Tuesday,
Thursday, and Friday) during the week of August 8 to 12, 2005 from 6: 00 a. m. to 10: 00
a. m. and two afternoon peak periods of the same week ( Wednesday and Thursday) from
3: 00 p. m. to 7: 00 p. m. A critical element in all tests of this type is clock synchronization
on all systems. In this case, researchers used the clock associated with the loop detectors
as the standard clock for all other detectors and they used an accurate digital watch to
synchronize other systems to the loop detectors. They checked and synchronized clocks
at the beginning of each four hour data collection period and found that clock drift within
that four hour window was insignificant. The ground truth for this research came from
manual counts of the recorded videotapes in 15 minute intervals. ( 24)
One of the statistics used to compare the three systems ( inductive loops, SS- 105, and
Autoscope RackVision) was percentage volume errors. The null hypothesis was that the
mean percentage volume errors of the three sensors were statistically equal. The analysis
used the F statistic to investigate the null hypothesis. The larger the F statistic, the more
likely it is that the null hypothesis will be rejected. The research first developed a
comparison of aggregate data ( all lanes, all conditions) and then developed a disaggregate
comparison by lane, by speed, by weather condition, and by lighting conditions for the
Autoscope ( weather does not affect the other two systems). ( 24)
The results of the aggregate comparison indicated that the Autoscope had the lowest
mean percentage error of 0.67 percent and the SmartSensor had the highest mean
percentage error of 1.39 percent. All three technologies had positive mean percentage
errors, indicating overcounts. The SmartSensor had the highest standard deviation ( S. D. =
4.290) of the three technologies. On the basis of the F statistic at ά = 0.05, the null
hypothesis was rejected, concluding that the mean percentage volume errors of the
technologies were not all statistically equal. However, the mean percentage errors of all
three systems were within 2 percent, so their average accuracies were greater than 98
percent. The authors conclude that, even though there were statistically significant
differences between the systems, they were practically very similar and would be suitable
for most volume counting applications. ( 24)
The next computation was disaggregated by lane. For the SS- 105, lane 1 was closest and
lane 5 was the lane furthest from the sensor. The Autoscope was centered over lane 3 so
lanes 1 and 5 were furthest from the camera. In this case, the null hypothesis was that the
mean percentage errors of the three sensors were statistically equal ( at ά = 0.05) across all
36
lanes. The corresponding F statistic indicated that the mean percentage errors of the three
sensors were not statistically equal for any of the five lanes. Lane 1 data had the largest F
statistic ( F = 111.52) while lane 4 had the smallest ( F = 8.79) indicating that each
individual detector’s performance was a function of lane location. The SS- 105 had mean
percentage volume errors in lanes 2, 3, and 4 in the range of 1 to 2 percent, but lane 5
results indicated larger overcounting errors ( mean = 5.7 percent) and larger
undercounting errors in lane 1 ( mean = - 3.5 percent).
For the Autoscope, the mean percentage errors in lanes 1 through 4 were approximately
zero but the mean percentage error in lane 5 was higher at about 3 percent. Researchers
hypothesized that the presence of trucks in lane 4 might have occasionally activated the
detector in lane 5 and caused the overcounting errors in lane 5. ( 24)
In another analysis, results were disaggregated by speed where “ normal” traffic involved
average 15- minute speed greater than 50 mph and “ slow” traffic involved average 15-
minute speed less than or equal to 50 mph. Of the three systems, the SS- 105 was most
affected by speed – overcounting in normal speeds and undercounting during slow
speeds. The Autoscope displayed a similar pattern although the effect was not as
pronounced. Inductive loops showed little variation by condition. ( 24)
When results were disaggregated by weather ( rain and dry), the mean percentage errors
of the three sensors were not statistically different. Across clear and rainy conditions, the
mean percentage errors for all three sensors were very close with differences within 0.3
percent. This finding indicates that rainy weather did not seem to affect performance of
any of the sensors. ( 24)
The research also investigated the effects of light on the Autoscope. The mean percentage
errors of the Autoscope data collected under dark and daylight conditions were found to
be 0.797 percent ( S. D. = 2.116 percent) and 0.767 percent ( S. D. = 2.503 percent),
respectively. Both mean percentages fall within plus or minus 1 percent, which indicates
Autoscope performs well under both conditions. Table 11 ( page 37) summarizes the
results of this research. ( 24)
Finally, the research investigated the influence that levels of aggregation had on results.
Findings indicate that the SS105 is more sensitive to the data aggregation type chosen.
The analysis using combined lane and cumulative approaches produce lower percentage
errors as compared to the by lane approach. The reason for this finding is that
undercounting on lane 1 is balanced by overcounting in lane 5.
From a practical perspective, all three technologies performed adequately for most
volume counting applications faced by transportation agencies – particularly with respect
to total counts. The body of the paper ends with a discussion of speed and vehicle length,
but this section is inconclusive because ground truth was not available for comparison.( 24)
37
Table 11: Comparison of Mean Percentage Volume Errors ( Nebraska)
Loop SmartSensor
SS105
Autoscope
RackVision
% Error
N
Mean Std.
Deviation
Mean Std.
Deviation
Mean Std.
Deviation
F
Sig.
Overall 320 0.911 1.229 1.390 4.290 0.669 2.507 4.939 0.007*
“ Lane” based analysis:
Lane 1
Lane 2
Lane 3
Lane 4
Lane 5
64
64
64
64
64
1.066
0.956
1.022
1.010
0.501
1.266
1.077
1.002
1.141
1.538
- 3.479
1.695
1.066
1.974
5.698
2.183
2.499
1.580
4.214
4.294
0.350
- 0.006
0.078
- 0.096
3.021
1.978
1.574
1.565
2.093
3.386
111.523
14.135
10.058
8.787
40.140
0.000*
0.000*
0.000*
0.000*
0.000*
“ Traffic Condition” based analysis:
Normal
Slow
293
27
0.971
0.266
1.231
1.022
1.609
- 0.981
4.176
4.857
0.739
- 0.083
2.543
1.962
7.029
1.178
0.000*
0.313
“ Weather Condition” based analysis:
Clear
Rainy
220
100
0.946
0.834
1.239
1.211
1.481
1.193
4.375
4.111
0.675
0.657
2.454
2.632
4.156
0.883
0.016*
0.415
* indicates a statistical difference at ά = 0.05 level of significance.
Source: Zhang et al.: Reference 24.
Minnesota Department of Transportation
In 2005, MnDOT published results from its most recent Pooled Fund study on non-intrusive
detectors, entitled “ Evaluation of Portable Non- Intrusive Traffic Detection
System” ( PNITDS). While a substantial part of the report addressed mounting
techniques, it also covered detector accuracy for three systems – the Wavetronix SS105,
RTMS by EIS, and SAS- 1 by SmarTek. It also provides some limited results on the
TIRTL classifier. Table 12 is a brief overall summary of results for the first three. The
report also provided results by individual site where testing occurred. These sites were on
the following cross- sections: an eight- lane freeway, a four- lane freeway, a four- lane
arterial, a two- lane roadway with a side wall, an unparallel lane at a freeway exit ramp,
and a narrow local street without center striping. ( 12)
Table 12: Overall Result for Volume and Speed Detections ( Minnesota)
SS105 RTMS SAS- 1
Volume Detection 1.4% - 4.9% 2.4% - 8.6% 9.9% - 11.8%
Speed Detection 3.0% - 9.7% 4.4% - 9.0% 5.6% - 6.8%
Heavy Traffic Impact No No Yes, Undercount
Weather Impact No No No
Barrier Impact Minimal Moderate Not tested
Source: MnDOT: Reference 12.
Table 13 on page 38 summarizes results of vehicle length data collected from the RTMS
and SS105 microwave radar detectors. Researchers collected two hours of data at each of
two sites, monitoring three lanes at each site.
38
Table 13: Result Summary for Length- Based Class Detection ( Minnesota)
Lane SS105 RTMS
1 0.8% to 5.6% 1.2% to 4.4%
2 0.6% to 4.7% 0.2% to 1.2%
3 0.4% to 1.5% 0.4% to 1.4%
Source: MnDOT: Reference 12.
Texas Department of Transportation
TTI’s most recent vehicle detector research project used both the I- 35 testbed and the
S. H. 6 testbed. Table 14 summarizes the detectors that were included in the final year of
the project. The original intent was to include GTT ( 3M) microloops as well but an
adequate site that met all the criteria was not found.
Table 14: FY 2006 TTI Detector Test Plan ( Texas)
Test Location ( No. Lanes)
Detector
Technology Austin College Station
Autoscope Solo Pro Video Imaging I- 35 ( 5 lanes) a S. H. 6 ( 2 lanes)
Iteris Vantage Video Imaging I- 35 ( 5 lanes) S. H. 6 ( 2 lanes)
SmarTek SAS- 1 Passive Acoustic I- 35 ( 5 lanes) S. H. 6 ( 2 lanes)
Sensys Networks Magnetometer I- 35 ( 2 lanes) S. H. 6 ( 2 lanes)
SmartSensor SS105 Microwave Radar I- 35 ( 4 lanes) b S. H. 6 ( 4 lanes)
Traficon VIP Video Imaging I- 35 ( 5 lanes) S. H. 6 ( 2 lanes)
a The I- 35 site has a total of five southbound lanes, but lane 1 has a failed inductive loop.
b The mounting pole was too close to lane 5, so the SmartSensor could only monitor three lanes.
Detector performance in this TTI research was compared against the Peek ADR- 6000, a
high- end vehicle classifier using inductive loop signatures for baseline count, speed, and
occupancy data. Sidefired microwave radar detectors in this research exhibited consistent
speed accuracy, although limited tests of an overhead- mounted SmartSensor SS105 in its
Doppler mode was even better ( it can only cover one lane in Doppler mode). Therefore,
the SmartSensor SS105 should be considered as an accurate speed detector for replacing
loops with its orientation depending on site- specific accuracy needs. For a three- color
urban speed map display, most of the detectors tested in this research have the needed
speed accuracy.
All non- intrusive technologies that are mounted beside and above the roadway appear to
be affected by side- to- side occlusion and even more so in congested conditions. When
congestion reaches a point where the prevailing speed begins to drop, the accuracy of
most non- intrusive detectors typically declines significantly. Especially for video
imaging, it is critical that installers place cameras high enough to minimize these effects.
The height is not as critical for measuring speeds as it is for achieving the desired
accuracy in counts and occupancies. To investigate camera placement needs, TTI utilized
three- dimensional visualization software and a two- dimensional occlusion calculator
39
using Microsoft Excel to develop camera placement criteria. For example, a typical
camera mounting height of 35 feet at an offset of 10 feet ( from the nearest monitored
lane) would only cover two lanes adequately.
From a performance standpoint, microwave radar, magnetometers, and certain video
image vehicle detection systems are probably all suitable for freeway applications but
certain caveats apply. Beyond camera mounting as noted above, video imaging is more
complex, requires periodic lens cleaning, and is usually more expensive, but a positive
attribute is that it offers a view of the traffic stream. Some limited weather and lighting
conditions may affect the latest video imaging systems although the manufacturers have
reduced those impacts in recent models.
Sensys Networks magnetometers warrant continued evaluation over a longer period of
time. Accuracy levels were certainly acceptable but its battery life needs to be verified in
high- volume traffic and in extreme temperatures. One negative attribute is that it is an
intrusive device, requiring interference with traffic for installation. It is a promising
replacement for loops since installation would take less time than loops. Finally, the
SmartSensor SS105 ( and its newer version, the HD) is a rugged device that can be
mounted on an existing pole, automatically calibrates speed and configures lane positions
for each lane monitored, can cover up to eight lanes ( 10 lanes for the HD) in sidefire
orientation, and is apparently not affected by any weather or lighting conditions.
FOLLOW- UP PHONE CALLS TO SELECTED STATES
TTI contacted states to follow up on information gathered from the written literature and
from the Internet. States that were helpful in this regard were California, Minnesota, and
Texas. All three states have excellent detector test facilities.
California
TTI contacted the engineer with the Caltrans Detector Evaluation and Test Team ( DETT)
who was responsible for one research project which used the I- 405 test facility in Irvine
to evaluate two microwave radar detectors, the RTMS and the Wavetronix SS105. ( 13)
His comments were mostly about the test results, but they provide insight on Caltrans
expectations on these detectors at the time. This and other research of vehicle detectors
was at least in part a result of aggressive promotion and sales by manufacturers without
proper guidance on setup.
The basic conclusions of Caltrans based on the DETT tests ( published in 2004) are that
inductive loops will continue to be the detection method of choice. Caltrans will probably
continue to use the RTMS and SmartSensor SS105 microwave detectors but only for
counts in locations where loops are not available or perhaps where a short- term count is
needed. Where microwave detectors are used, Caltrans will most likely use contact
closures to emulate loops and use them only for traffic counts ( presence detection). The
Caltrans spokesman doubts that his agency has the technical expertise to properly set up
these microwave detectors and get performance that is similar to factory installers.
40
Another phone interview was with a seasoned research engineer with the Caltrans
Division of Research and Innovation ( DRI). Caltrans received $ 10 million in 2006 for
districts to improve vehicle detection across the state. The various districts have elected
to spend their share for different types of detectors, resulting in a wide variety of freeway
detector types. Some districts fixed existing inductive loops or installed new loops while
other districts chose to purchase non- intrusive detectors to replace failed loops. For
example, District 8 ( San Bernadino) purchased Wavetronix radar detectors and Sensys
Networks magnetometers. These new magnetometers are getting the attention of Caltrans
engineers because their accuracy appears to be commendable, especially under free- flow
conditions. Based on early results of testing in California, their accuracy appears to be
better than radar but they require lane closures for installation or replacement.
Caltrans tracks the performance of its detectors through a database called the California
Freeway Performance Measurement System ( PeMS) database, which provides input on
“ detector health.” Not long ago, vehicle detectors in this database ( mostly loops) were no
better than 75 percent functional overall. However, Caltrans has determined that properly
installed and maintained loops were 99 percent accurate. Their conclusion was that other
components, perhaps within the cabinet, are also sources of error.
Of the newer detectors, San Diego State University is currently evaluating the TIRTL.
One of the applications initially envisioned for this sensor was in Yosemite and perhaps
other remote locations. However, there is substantial snowfall three to four months out of
the year at the Yosemite site, which would undoubtedly degrade its accuracy. One of the
recent additions to the lineup of test detectors at the I- 405 test facility in Irvine is the
Wavetronix SS125 High Definition ( HD). Based on a short observation period during
free- flow traffic, it appeared to be working quite well, but its behavior during more
congested or even stop- and- go conditions was not as well known.
The camera system developed by a Cal Poly professor and installed at the I- 405 test
facility was a subject of interest. The camera system works very well during daylight
hours but not at night. The intent in developing the camera system was in its use as a
ground truth system and for re- identification of vehicles. Both require adequate lighting
for success. Besides lighting, another of its requirements is an overhead structure to
mount the cameras. There is one camera mounted to the bridge structure pointing
downward ( near vertical) over each lane. Installation personnel initially had concerns
about vandalism or even kids playing on the structure leading to serious injury but none
of that has happened to date.
Minnesota
The Twin Cities Traffic Management Center has cameras at most interchanges on the
twin cities freeway network as well as ramp meters. MnDOT uses inductive loops at
almost all of these locations, both on the mainline and on the ramps for ramp metering.
MnDOT uses loops for its AADT counts as well. There are 77 ATR sites statewide but
many are being converted to classification sites by adding an axle sensor. The state is
using piezoelectric sensors for axle detectors along with inductive loops.
41
MnDOT is trying only a few non- intrusive detection systems as possible replacements for
inductive loops in the near future or for gathering data where loops are not appropriate.
Two systems that MnDOT is currently testing are the TIRTL and the Wavetronix
SmartSensor SS125 ( or High Definition). MnDOT reported on initial results of the
TIRTL at the 2006 North American Travel Monitoring and Exhibition Conference
( NATMEC). This system shows promise for some limited applications, but MnDOT does
not plan on purchasing any more of these classifiers. The cost of the detector is one
negative at $ 25,000 each. Another factor is the time to set it up at the data collection site,
requiring two persons at each location.
MnDOT purchased two of the Wavetronix detectors and is seriously considering
collecting length data using a length- based classification scheme. This unit has, or soon
will have, eight length bins for this classification mode. MnDOT has not used length-based
classification before so there are some uncertainties, since FHWA still wants
reports to use the standard Scheme F with 13 classes.
Another detector that MnDOT is considering but is not committed to is Nu- Metrics
“ Groundhog” magnetometers, which will be used at Road Weather Information Systems
( RWIS) sites. Using two of these per site will facilitate speed data collection. Another
detector which was recently demonstrated to MnDOT was the magneto- resistive sensor
by Sensys Networks. However, MnDOT is currently not serious about buying any of
these detectors. One final detector that has been demonstrated recently is the Peek
AxleLightTM and is similar in concept to the TIRTL.
One of the reasons inductive loops continue to be the most popular form of detection is
that they last as long as 10 or even 15 years. A few of the non- intrusive detectors are used
to fill in gaps. One example is the RTMS by EIS, which the state uses to conduct high-volume
counts. These appear to work satisfactorily for this application. MnDOT also
does some classification counts manually.
Texas
The Texas Department of Transportation ( TxDOT) districts use a variety of detection
technologies, with inductive loops continuing to be used in many cases until they fail.
While the Traffic Operations Division in Austin provides support and guidance to
districts in choosing among the technologies, each district decides what detectors to use.
In a few cases, especially smaller districts still rely on loops so they replace failed loops
with more loops. District decision- makers are practically unanimous in believing that
none of the newer technologies are as accurate as properly installed and properly
maintained loops. However, most districts replace failed loops with another technology
due to some legitimate reasons.
The costs of some of the newer technologies such as video imaging have declined over
the past five or more years, especially at signalized intersections, where each camera
replaces several loop detectors. At the same time, districts are facing much higher lane
closure costs and they realize that closing lanes increases safety risks for both installation
42
crews and motorists along with increasing motorist delay. One other important
consideration is that newer technologies have improved over the past 10 or so years for
both intersections and freeways.
The larger urban districts have been installing newer technologies longer than most of the
smaller, less urbanized districts. Most districts make incremental changes by installing a
few units for observation before making a wholesale changeover in the types of detectors
used. When a purchase is made, it is often on a low- bid basis, leading to unsatisfactory
results and also resulting in multiple technologies being deployed across each district. It
is anticipated that recent TxDOT- sponsored research on vehicle detectors will make a
difference, but it will take time. Following are a few examples of the larger urban
districts that may be helpful.
The Austin District allows contractors on construction projects to determine the detection
technology as long as district requirements are met. The district still uses loops far more
than any other technology even with this freedom of choice, mostly because contractors
know how to install loops and they are not as familiar with newer technologies. In the
Dallas District, inductive loops on freeways are being replaced by Autoscope Solo Pro
and SmartSensor SS105. The Corpus Christi District has had many problems with loops
and has reverted to video imaging systems but with disappointing results with that
technology as well. The El Paso District still largely relies on inductive loops but has
replaced some of them with RTMS and a few video imaging systems at intersections.
The Fort Worth District has discontinued the installation of inductive loops in lieu of
Wavetronix SmartSensors, RTMS, and video imaging systems. The Houston District has
had more than its share of inductive loop problems for many years and was one of the
first to install video imaging systems on freeways. In fact, the district began using that
technology before it had matured enough to be reliable. Problems led the district to
abandon many of the video units and continue its search for an adequate technology.
The most recent interest has been in the Wavetronix SmartSensor 105. The San Antonio
District has had success with inductive loops on freeways and at signalized intersections
and these units are anticipated to continue to serve the needs of the district. For signalized
intersections, the district primarily installs video imaging systems.
TIRTL vendors have contacted TxDOT’s planning division to demonstrate the detector
but the TxDOT representatives do not plan to purchase them in the immed