Abstract:

The spatial motion pattern recognition method and the abnormal traffic hehavior detection method of the moving object were proposed. The spatial distance similarity matrix of the trajectory dataset was constructed using the mearurement method in the trajectory space with the improved Hausdorff distance. The typical motion pattern of the moving object was extracted by the spatial distributions of the trajectory dataset learnt by the spectral clustering algorithm. A motion pattern matching algorithm in the trajectory space was presented for the abnormal traffic behavior detection based on Bayes classifier.The vehicle lane-changing behavior in the realistic scenario was recognized as an example, and the result was encouraging. The proposed method provides a technical support for the studies of the traffic behavior of moving object and the intelligent traffic management and control.

Key words: transportation engineering, video processing, motion pattern recognition, Bayes classifier