Automatic recognition algorithm of vehicle type based on video-content analysis

Vehicle type recognition in traffic video data is a hot problem of the field of computer vision. By analyzing the content of traffic video, we proposed an algorithm of vehicle type automatic recognition. First of all, do moving foreground detection, shadow elimination, and morphologic processing by hybrid inter-frame subtraction algorithm combined with dynamic background model. Secondly, to establish vehicle relationship between adjacent frames, use the method of improved tree tracing diagram for vehicle tracking, and then extract the geometric feature and moment feature of the vehicle. Finally, take vehicle feature with scale invariant feature as parameters, then finishes classification and identification of the vehicle type by support vector machine. The simulation experiment results show that this algorithm can extract the information of vehicle type effectively, with a high rate of vehicle type recognition.