To improve the performance of freeway Wrong-Way driver detection systems, verify that articulating camera systems can consistently return to exact pre-set positions needed for analysis and automatically reset positioning if needed using reference points.
Florida DOT study evaluated the performance of Wrong-Way driver detection systems using video analytics.
Made Public Date
05/27/2020

949

Tampa Bay
Florida
United States
TwitterLinkedInFacebook
Identifier
2020-00966

Testing and Evaluation of Freeway Wrong-way Driving Detection Systems

Background

Florida DOT (FDOT) designed and conducted an evaluation of Wrong-Way detection systems offered by three different vendors. Systems evaluated primarily utilized existing FDOT camera video feeds and analyzed the images to identify Wrong-Way vehicles and notify the Traffic Management Center.

This study selected six testing locations on I-275 in the Tampa Bay area, and were assigned into four testing scenarios:

  • Testing with normal daily traffic conditions
  • Testing consecutive Wrong-Way Drivers in both directions
  • Testing under normal light nighttime traffic conditions
  • Testing under low light nighttime traffic conditions.

The following measures were developed for use in the performance evaluation:

  • Wrong-Way Driving detection system accuracy
  • Percentage of False calls
  • Actual Wrong-Way Driving detection accuracy
  • Percentage of Missed Calls.

Data collection occurred after vendors were able to set-up and configure and optimize their systems for testing. Duration of data collection was set for one full week with an additional week of data when needed. Collected Wrong-Way driving detections included date, time and location information. Recorded video data was used as ground truth for the assessment.

Comparison of different testing scenarios showed varied detection performance, but all were able to send notifications without issues. The performance indicators of detection system accuracy and actual detection accuracy were excellent indicators of vendor system performance, but did not necessarily reflect true detection in actual implementation.

Lessons Learned

  • Performance measures used were found to be excellent indicators of vendor freeway Wrong-Way driver detection, however real system performance should be further investigated based on actual Wrong-Way driving detection over a sufficiently long time, or under a controlled environment.
  • Existing Pan-Tilt-Zoom (PTZ) cameras were found to sometimes not return consistently to the exact pre-set position needed for analysis. As a result, fixed cameras were used for this study. The research team offered two recommendations: (1) Check PTZ cameras to determine if they can return to the exact pre-set position for similar studies, and (2) Improve freeway Wrong-Way driver detection systems to automatically adjust settings using reference points for minor movements of PTZ cameras and provide alerts for large movements.
Goal Areas