The Algorithm Of Vehicle Detection Based On Support Vector Machine (SVM)

With the development of society and technology, one feature of theArtificial systems and products in the future is intelligentization, that is using thecomputers to insteading of some aspects of the human activities. The first step isdetecting test of vehicle to analysis all kinds of intelligent traffic behavior inthe field of intelligent traffic. So the accuracy and speed of vehicle detection have avery large influence to intelligent transportation system. In this paper we have astudy of vehicle detection method and have a experimental comparison of targetdetection algorithm of mainstream based on the target detection technology of supportvector machine (SVM) method and we improve the traditional algorithm from theangle of the speed of detection. The main contents are as follows:(1): Introducing the background and significance of the research of vehicledetection and the research status at home and abroad.(2) we have analyzed the thoughts, Algorithm and the application of supportvector machine (SVM), Image feature extraction, HOG feature extraction, Enhancesexual HOG feature extraction technology in the vehicle detection technology fromthe experiment.(3) Under the environment of Visual Studio2008programming we have donethe experiment based on the OPENCV and basic characteristics of HOG, theexperiment based on OPENCV and improve the HOG feature, the experiment basedon the OPENCV and an improved algorithm based on support vector machine (SVM)experiments and the experiment based on the root filter of vehicle detectionexperiments. These experiments evaluate the Algorithm performance from the aspectof both accuracy and speed of detection(4) With nonlinear support vector machine (SVM) model for training and thedetection of linear support vector machine (SVM), we have improved the traditionalalgorithm. The experimental results showed that the improved algorithm not onlyimproved the detection accuracy, and reduced the test time consuming.(5) Combined with the traditional support vector machine (SVM) testing ideasand DPM (deformable component model) detection method we have onlytaken root in the DPM filter algorithm for the vehicle for testing. And the resultsshowed that although this method’s accuracy was worse than the traditionalsupport vector machine (SVM) detection but still be able to detect the vehicleaccuratel--y. So we can apply this method to the occasion which therequire--ments for accuracy is not high, while requiring higher detect-ion.