Design Of Vehicle Video Retrieval System Based On License Plate Recognition

As the traffic monitoring system has been built and improved all over the world inrecent years, there are a lot of new traffic suveillance video produced every day. To therapidly increasing video data, the traditional video retrieval method based on the manualannotation has obvious defect, because valuable clues are often neglected when people arescatterbrained. In contrast, the video retrieval technology based on content is more efficient,objective and has a broader application prospect.This paper designed a vehicle video retrieval system based on license plate recognition,by which the vehicle in a video or an image can be retrieved among the stored videos.There are two main parts of this system: license plate recognition and video retrieval.On the license plate recognition research, a new license location algorithm combiningvertical edge detection, color information of the license plate and mathematical morphologyis presented in this paper. For balance of computing load and recognition accuracy, a“200-d” character feature rule is designed, and the “200-d” feature is used as the input ofBP neural network to recognize the characters. Based on the above-mentioned methods, alicense plate recognition system is set up, which can locate and recognize the license plateeffectively, even when the resolution of pictures and the position of vehicles in the picturesare not fixed.On the video retrieval research, the entire video is treated as a same shot so that thedetection operation can be simplified. To the key frame extraction, a moving targetdetection method fusioning three frame difference medthod and background subtractionmethod is presented.With the shadows removed by the method of weighted top-hattransform, vehicle images are extracted in the interested area successfully. Comparing thehistogram of the vechiles, vehicles which have the similar color can be found. Consideringthe license plate information, the vehicle can be retrieved accurately.The experimental results show that the license plate location method, license platerecognition method and video retrieval method reach a good accuracy, and the system hasapplied value to the engineering practice.