Pedestrian-vehicle conflict can reflect potential hazard in mixed traffic. Automatic pedestrian-vehicle conflict detection algorithms can provide both real-time warning information to traffic users and collect data for further analysis. This paper proposes a method to automatically classify different traffic users and to detect pedestrian-vehicle conflict using video data. A mixed Gauss-based background differencing algorithm is used to detect a foreground target. The speed and area parameters are used to classify traffic objects. A tracking algorithm based on multi-feature fusion is designed to track classified objects. The notion of pedestrian-vehicle conflict is adjusted for video-based detection, and a pedestrian-vehicle conflict video detection model is designed. In total, 117 pedestrian-vehicle conflicts are collected. The results show that the proposed method can correctly detect about 80.3% of pedestrian-vehicle conflicts. Further research will focus on the improvement of the object extraction algorithm, and more mixed traffic scenes will be tested based on this proposed method.
Real-Time Pedestrian-Vehicle Conflict Detection Algorithm Using Video Data
14th COTA International Conference of Transportation Professionals ; 2014 ; Changsha, China
CICTP 2014 ; 535-546
2014-06-24
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
REAL-TIME PEDESTRIAN AND VEHICLE DETECTION FOR AUTONOMOUS DRIVING
British Library Conference Proceedings | 2018
|