Research On Automatic Vehicle Model Recognition Technology In Intelligent Transportation System

Intelligent Transportation System (ITS) is the frontier of traffic technology throughout the world. The service sectors acknowledged internationally include advanced traffic management system, advanced traffic information service system, commercial vehicle transportation and management system, electronic toll system, public transportation management system, emergency management system, and advanced vehicle monitoring system. Automatic vehicle model recognition is the key technology in ITS, which has been widely applied in the fields, such as transportation monitor system, automatic toll system.This dissertation mainly research the automatic vehicle model recognition system, which is operated based on the vehicle image acquired from video cameras by employing moving object detection algorithm, feature extraction and pattern recognition algorithm. The main works and contributions of this dissertation are summarized as follows: (1) Study the existing digital image processing technology and algorithms, analyze the popular algorithms in vehicle detetion ,and present a novel doule inter-frame sum method based on the method of inter-frame difference used in vehicle detection. This method is simple and has less calculation complexity and strong anti-jamming ability; (2) Considering the probable phenomena of parallel vehicles in vehicle recognition based on video, this dissertation puts forward a parallel vehicle detection and division algorithm based on mathematic morphologic labeling connected components; (3) In view of the fact that the pseudo-Zernike moments has the advantages of rotation invariant and insensitive to noise properties, which make it represent image pattern better, in addition to the features of vehicle length, vehicle width and the ratio of length to width, the vehicle pseudo Zernike moments feature is added in vehicle model features to classification, and the results from the experience shows that this feature can improve the recognition effect; (4) Study basic theories and main algorithms of pattern recognition and present a vehicle model recognition method based on LSSVM ( Least Squares Support Vector Machines ).The results from the experience shows that the method of vehicle model recognition presents in this dissertation can improve the recognition effect.