The Research On Vehicle Recongnition Technique Based On Support Vector Machine

Taking up the task of designing and realizing video detection of traffic flow and vehicle recognition system based on the technology of video and image processing, this thesis researches into the technology of monitoring traffic flow and traffic condition by means of video detection and recognition. The paper is recommending the traffic flow of the video to measure the research background in the vehicle recognition system, and on the foundation of the application current situation both at home and abroad ,show the significane and prospect of this research Through image filtering in the vehicle recognition system and noise reduction, image segmentation, feature pick-up and object recognition. establish the basic method of calculation that will be adopted in the design and realization of the system. Finally, it presents the design of video detection of traffic flow and vehicle recognition system. In this article, some new algorithms and methods are introduced in such stages as preprocessing of images, classification of characteristics and character recognition and vehicle style recognition. For example, Firstly, In this paper,we make an research on image process method of SVM image. Making use of the characters of SVM image and all kinds of traditional edge detection methods, the methods of edge detection based SVM image have been proposed, which differ from operator methods and are easier to understand and realize. Satisfactory results were obtained.. Secondly, The wavelet transformation method is adopted to withdraw vehicle features and then to assort vehicles automatically. Thirdly,This paper proposes a vehicle automatic-classification method based on SVM theory.trained Support Vector Machine(SVM) off-line,and the tested vehicle characteristic data as tested samples then the vehicle type can be judged by means of the trained SVM according to the method proposed in this paper. Oftentimes multi-class classification problem have been treated as a series of binary problems in the SVM paradigm, but it is computationally more expensive to solve multi-class problems. In this paper, a new multi-class classification method is proposed based on one-class support vector machine (1-SVM),and then applies the method to vehicle recognition. The results of simulation experiments at vehicle database show that the proposed method is effective and feasible.