Automatic Traffic Event Detection Based on Dynamic Time Warping Algorithm
Yi Shi, Xiaobo Lu, Huang Wei, Tao Liu, Yun-ting Xian, Kun Hua Zhong · 2007
In order to detect traffic event automatically and to handle timely, a new traffic event detection algorithm is presented. Firstly, vehicle tracking algorithm is designed based on dynamic image analysis technique. Secondly, trajectory modeling method is constructed to extract motion trajectories of vehicles. Finally, a dynamic time warping algorithm is introduced. The trajectory data are firstly trained. After computing the distance function, warping path can be obtained. As a result, the match between testing data and typical trajectory pattern can be realized. In order to validate the algorithm, three measures are presented and concrete experiment is designed. Groups of experiments have been done on U-turn, illegal changing-lane, illegal left-turns. The results show that the event detection algorithm has valid real-time performance.