Congestion Detection of Urban Intersections Based on Surveillance Video
Fei-fei Xun, Xinghai Yang, Yu Chan Xie, Lingyin Wang · 2018
This paper proposes a congestion detection algorithm of urban intersection surveillance video. The algorithm is divided into two parts: global vehicle speed detection algorithm and traffic state identification algorithm. The global vehicle speed detection algorithm firstly selects Region of Interest (ROI) according to the scene of intersection, using Shi-Tomasi corner detection algorithm to find the pixels to be tracked. Then take the detected corner points as inputs of Lucas-Kanade optical flow algorithm to obtain the vehicle motion vector. After that filter and obtain global vehicle speed according to calculate the direction of vector. The traffic state identification algorithm analyses global vehicle speed of one traffic light period. With multiple judgments the algorithm final discriminates traffic state. Experiments show that the proposed algorithms have strong anti-interference ability and good vehicle detection accuracy, meet the needs of practical application.