Application Of Video Processing In Several Key Algorithms Of Intelligent Transportation System

With the development of China's economy, transportation is becoming more and more important. The establishment of a modern intelligent transportation system can effectively solve traffic congestion, blocking and violations. License Plate Recognition system and the detection of the illegal tax evasion in highway Toll Collection System are two important applications in intelligent transportation system. Some valid feature extraction and recognition methods are proposed for the key issue of license plate recognition system--license plate chinese character recognition. Meanwhile, the detection algorithm for the illegal tax evasion in highway Toll Collection System was discussed and the embedded technology in Intelligent Transportation System was preliminary introduced.In the research on license plate chinese character recognition, the features of Chinese character are extracted using the combining of structural methods and statistical methods. And the minimum distance classifier is used for identification. A database for license plate chinese character is established firstly, and then some pre-treatment including graying, smoothing noise removal, binarization, cutting and normalizing are conducted. After the refine or contour extraction processing, the 4-direction features are extracted. The last minimum distance classifier is used for classification and identification. Experimental results show that the algorithms can meet the real-time performance and accuracy requirements for a license plate recognition system.In the study on the detection of the illegal tax evasion in highway Toll Collection System, the detection algorithm based on optical flow technology is proposed for illegal jumping pounds behavior. The algorithm is mainly composed of image preprocessing, the vehicle motion detection and behavior analysis based on optical flow. In which inter-frame difference method is used for moving target detection, and obtain the appropriate motion vectors by calculating optical flow of the motion region. Finally the illegal behavior is determined by analyzing the vertical motion vectors of the traffic video images. Experimental results show that the proposed algorithm can achieve good real-time performance and accuracy requirements.