Consider The Angle Of Inclination Of Svm-based Vehicle Classification

Vehicle automated classification is the important component of Intelligent Transportation Systems (ITS) as well as the Image processing and pattern recognition application field of research hotspot. In recent years, with the rapid development of economy and rising rates of vehicle ownership, the method of vehicle classification is increasingly being widely used in today's traffic monitoring, urban emergency command, safe urban construction and automatic toll collection system. However, due to our large number and variety of vehicles, vehicle classification is still a problem (to be solved).Therefore, the technology and the method of vehicle classification has been widespread concerned in recent years.The thesis presented a Vehicle Model Classification System of being tilt angle based on the image acquired from cameras by image preprocess, tilt angle correction, feature extraction, feature selection and pattern recognition algorithm. In order to achieve the purpose of classification and recognition of vehicles, the main work and innovation points in this thesis are summarized as follows:(1) Image preprocessing of vehicles from the CCD camera, effectively eliminated a variety of image noises. Focus on the vehicle image of tilt angle to correct and use different methods. Compared with the both methods I chose the better one.(2) Feature extraction. In this paper, Freeman chain code approaches to contour tracking of vehicles, we get the external profile and calculate the top of the vehicle length, vehicle length,vehicle height, perimeter, area, bounding rectangle and other characteristics based on Freeman chain code. On this basis, we also proposed the other important parameters, such as center distance of two tires, front suspension, rear suspension, rear height and several other features as identified in this paper models. Among them, we should detect the two round wheels before the wheelbase extract.(3) The basic theory of SVM and their main method. In this paper, we present a phased strategy for object classification based on comparative analysis, the first roughly divided into small car; bus and truck, and then for each kind of vehicle type further subdivided. That is, for different models based on directed acyclic graph decision-making approach to classify, the same model used the binary tree approach to classify., To some extent,we can solve the inter-domain object cross-cutting issues in this way.(4) Finally, we carried the experiment on muti-classification model of SVM.We select the appropriate kernel function and the penal factors by comparing, and then use trained vehicle classification to classify.The experimental study on the vehicle image video show that the method used has some effects and achieve the common vehicle classification target.