Detection of vehicle beating trajectories using color license plate location

In order to prevent the phenomenon that the lorries take some illegal operations, such as jumping and taking the s-shaped line so as to reduce the vehicle weight when they are passing the highway toll station, and also in order to curtail fees disputes in the meantime, this article studied the algorithm which can detect the vehicle beating trajectory automatically. Firstly, license plates are located by HSI color space and texture features. Secondly, it achieved the position of license plates more exactly with the K-means clustering. Furthermore, plate centroid was calculated by image threshold, morphology and connected component labeling algorithm. Because there is hardly distortion for plate centroid in the image sequence, which can be used for calculating the centroid trajectory between the first image frame and the following image frames, and characterizing the vehicle´s movement pattern. In order to utilize the hard disk space which is used for saving the image sequence efficiently, improved track detection method was proposed. Through experiments, trajectories of vehicles can be reflected. And it provided the basis for follow-up study.