40

Michael Hunter (PI), Angshuman Guin (Co-PI), Wonho Suh, James Anderson

Feasibility Study for Using Video Detection System Data to Supplement Automatic Traffic Recorder DataFeasibility Study for Using Video Detection System Data to Supplement Automatic Traffic Recorder Data

Conclusion• Comparing vehicle counts from ATR, VDS, and

RTMS to manual counts showed that all of these systems’ counts were generally within 10% of the manual counts.

• It was also discovered that VDS count accuracy varied slightly between different mounting styles and offsets.

• Furthermore, VDS camera viewing angle and number of lanes covered can significantly impact the vehicle count accuracy. Therefore, candidate VDS units must be checked for vehicle count accuracy before incorporation into the Office of Transportation Data’s data streams for Federal reporting.

• Certain maintenance activities, such as roadway repaving, can leave the OTD blind on these sections as ATR systems may be damaged or removed during these activities. The Office of Traffic Operation’s detectors use non-intrusive technology and can provide an alternative data source for OTD under such conditions

This research was sponsored by the Georgia Department of Transportation under Project RP 11-13. Opinions expressed here are those of the authors and not necessarily those of the Georgia Department of Transportation.

OverviewThe Georgia Department of Transportation (GDOT) continuously collects traffic data using Automated Traffic Recorder (ATR) stations, Video Detection System (VDS) cameras and Remote Traffic Microwave Sensors (RTMS). ATR stations are utilized by GDOT’s Office of Transportation Data (OTD) while the VDS and RTMS systems are used for the Georgia NaviGAtor Intelligent Transportation System by GDOT’s Office of Traffic Operations (OTO). This research compared the automated traffic counts from these three technologies to manual traffic counts, counted using a proprietary Android application, to determine their accuracy and the feasibility of incorporating the VDS and RTMS counts into the OTD data streams to enhance Federal reporting. This research focused mainly on VDS, as it is one of the more widely used vehicle sensors in the Atlanta area. Furthermore, many VDS locations were tested to find the effects of the various mounting styles and offsets on the accuracy of vehicle counts.

Site Selection1st Round ATR: on I-285 near Orchard Road VDS: Pole Mounted with 24 feet offset on I-285 near Orchard Road VDS: Gantry Mounted on 285 near Orchard Road2nd Round VDS: Gantry Mounted on I-285 near US-78 VDS: Pole Mounted with 36 feet offset on I-285 near US-78 RTMS: on US-78 near Idlewood Road VDS: Pole Mounted on I-75/I-85 Connector

ATR, VDS, and RTMS Sensor LocationsSource: Google Maps

Data CollectionTablet Counting Application Allows data collectors to count vehicle by tapping specified detection

zones as vehicles pass under it. Video playback can be controlled by data collectors. When reviewing counted video, the detection zones light up to indicate

previously recorded vehicles. Videos can be reviewed by a second data collector to increase manual

count accuracy.

Lane Hours Counted1st Round 380 Lane hours of manual counts collected2nd Round 1096 Lane hours of manual counts collected

Student Collecting Data (left), Data Collection Application (Right)Images Courtesy: Wonho Suh

Results

Lane 1 Lane 2 Lane 3 Lane 4-40

-30

-20

-10

0

10

20

30

40

%

ATR vs Manual Count (Southbound, Hourly counts, n=24)

ATR vs. Manual counts box plots (left), lane-by-lane ATR vs. Manual counts y-y plot (Right), I-285 near Orchard Road.

Video Detection System

0 500 1000 1500 2000 25000

500

1000

1500

2000

2500

Manual Count

AT

R C

ount

ATR vs Manual Count (Southbound, Hourly counts, n=96)

Lane 1 Lane 2 Lane 3 Lane 4-40

-30

-20

-10

0

10

20

30

40

%

VDS vs Manual Count (Southbound, Pole mounted, Hourly counts, n=23)

0 500 1000 1500 2000 25000

500

1000

1500

2000

2500

Manual Count

VD

S C

ount

VDS vs Manual Count (Southbound, Hourly counts, n=92)

VDS vs. Manual counts box plots (left), lane-by-lane VDS vs. Manual counts y-y plot (Right), I-285 near Orchard Road.

WB Lane 1 WB Lane 2 WB Lane 3-40

-30

-20

-10

0

10

20

30

40

%

RTMS vs Manual Count (Hourly counts, n=34)

0 500 1000 1500 2000 25000

500

1000

1500

2000

2500

Manual Count

RT

MS

Cou

nt

RTMS vs Manual Count (Westbound, Hourly counts, n=78)

Remote Traffic Microwave Sensor

RTMS vs. Manual counts box plots (left), lane-by-lane RTMS vs. Manual counts y-y plot (Right), US-78 near Idlewood Road.

Vehicle Count Comparisons Each detection technology’s counts were compared to manual counts Box plots show that the average difference of the counts for each

technology from the baseline manual counts were within 10% of the manual counts.

Wider variation of deviations from the baseline were observed VDS and RTMS counts than ATR counts.

Automated Traffic Recorder

0 500 1000 1500 2000 25000

500

1000

1500

2000

2500

Manual Count

VD

S C

ount

VDS vs Manual Count (Southbound AM, Hourly counts, n=112)

RTMS vs. Manual counts y-y plot (left), VDS camera view of I-285 at US-78 (Top Right), VDS camera view of I-75/85 at 14th Street (Bottom Right)Image Courtesy: GDOT

14th Street VDS Camera Issues Wider variation occurred with the 14th Street VDS camera. Multiple lanes and low viewing angle created increased problems with

vehicle occlusion. Shows that candidate VDS camera stations must be checked for

accuracy before inclusion in Federal reporting.

Selection Criteria 1st Round Site was chosen based on proximity of ATR and VDS to each

other as well as two nearby PTZ cameras. 2nd Round sites were selected based on VDS mounting styles, sensor

type (RTMS), and proximity to a PTZ camera.

I-285 near US-78

I-75/85 near 14th Street