Research On Tracking And Recognition Of Moving Object Based On SVM In Intelligent Traffic

Intelligent Transportation Systems (ITS) has become the development of the futuretransportation system in recent years, and it has an extremely important scientific significanceand broad application prospects. The core of the research in this area is the moving targetdetection, tracking and recognition based on video. It deals with the issues of high-tech inmany fields such as advanced computer technology, image processing, machine vision,artificial intelligence and pattern recognition and so on, thus creating a real-time, accurate andefficient integrated management system.This paper describes the basic theories and key technologies of moving target detection,tracking and recognition in video sequence images. It focuses on study of the real-timetracking and moving target recognition in traffic video based on SVM.For moving target detection, first, It provides an introduction and summary for the threemajor detection methods (ie, background subtraction, inter-frame difference, optical flowmethod), which are currently used, and gives the homologous experimental results. Then, Itmakes a brief overview of the target detection methods based on adaptive backgroundmodeling. In order to extract the moving target more accurately, it has done some operationssuch as median filtering and mathematical morphology processing later in the experiment.For target tracking, we proposed a method by combing CamShift tracking algorithm andKalman Filter, we have used the class of functions, data structures and some basic frameworkby OpenCV to establish a system of moving object tracking. This system includes thefollowing modules such as image frame preprocessing module, the prospect of moving targetdetection module and moving object tracking module. It has achieved the desired results.At last, for moving target recognition, we introduced a theory of image processing basedon SVM. It analyzed shape features, area features, and moment features of the moving target compositely, trained Support Vector Machine (SVM) off-line through proposed feature vector,and the tested characteristic data as tested samples, and then applied the method to therecognition between people and vehicles. The results of simulation experiments show that theproposed method is effective and feasible.