The National Academies

NCHRP IDEA 20-30/IDEA 217 [Completed (IDEA)]

A Real-Time Proactive Intersection Safety Monitoring and Visualization System Based on Radar Sensor Data

  Project Data
Funds: $137,000
Staff Responsibility: Inam Jwed
Research Agency: University of Louisville
Principal Investigator: Zhixia (Richard) Li
Fiscal Year: 2019

This IDEA project developed a proactive intersection safety monitoring and visualization system (IP SV) implementable at any type of intersection for any type of safety and operation analysis under a long-term data collection period. The IPSV, in contrast to other traffic detection techniques, such as using video cameras, drones, or Lidar sensors, employs a 24GHz Microwave Doppler radar sensor, which can track all detected approaching vehicles and pedestrians’ trajectories with update frequency up to 0.3s/object. Based on the vehicle trajectory data, the IPSV calculates time-to-collision (TTC) and detects all possible traffic conflicts at the intersection. Finally, two field data collections were conducted with assistance from the Kentucky Transportation Cabinet to validate the accuracy of the IPSV. An application interface was developed along with the IPSV for users to configure IPSV and run queries and visualizations. IPSV was validated at both signalized and non-signalized intersections. The improved IPSV system was seen to achieve a high traffic conflict severity detection accuracy with an average 4.8% error rate. The system also achieved an 80% true positive rate and a100% true negative rate for conflict quantity detection. The results were validated through manual comparisons of ground truth found in videos based on field-collected traffic data. The IPSV provides a cost-effective method to quickly evaluate safety treatment effectiveness for an intersection without the need of waiting crashes to happen. It  complements the crash data to help transportation agencies and local governments better understand the safety issues at an intersection with traffic conflicts data collected. The system recognizes all types of road users, including vehicles, bicyclists, and pedestrians, via an explainable feature-based algorithm and detects traffic anomalies such as illegal left-turns, jaywalkers, and red-light-runners in any desired observation period. It also visualizes conflicts severity and quantity straightforwardly with the identifiers of TTC, vehicle speed, movements or conflict types. The automated and integrated system applies to any type/number/geometry shape of lanes of an approach at an intersection, providing real-time feedback on target intersection safety issues with 24/7 traffic monitoring.

The final report is available.

To create a link to this page, use this URL: http://apps.trb.org/cmsfeed/TRBNetProjectDisplay.asp?ProjectID=4878