Research On Tracking And Classification Algorithm And Realization Of Video Based On Vehicle Detection Technology

As the key of the intelligent transportation, vehicle detection system provides reliable data for traffic control and management. Based on the summary and analysis of the existing video vehicle extraction theory and key technology, this paper focuses on detection, recognition and tracking of video vehicle from stationary perspective. The article mainly studies the background image extracting which is based on the background modeling, video vehicle extraction, video vehicle identification and classification of video and vehicle tracking. The main research work is as follows:( 1) Extracting the background image based on the background modeling: by implementing the mean background modeling, single gaussian background modeling and gaussian mixture background modeling algorithm. After analyzing the characters of various algorithms to extract the background image. this paper selects the average method, which is real-time and simple, to extract the background.(2)Video vehicle extraction: the edge detection, thresholding segmentation, frame difference and background subtraction method are used to extract moving vehicle as well as to discusses the influence of the initial frame video image information for the moving target extraction. The experimental results show that under the condition of the initial frames which is close to the background image in video image, the background reduction method can obtain clearer moving target, which almost has no problems of noise and vehicle shadow.(3)Classification of video vehicle identification: this paper puts forward an idea that vehicle feature identification can be based on the vehicle aspect ratio and video vehicle binary duty cycle. After vehicle feature extraction, the article respectively uses the template matching and the support vector machine(SVM) to classify four vechicle types of buses, minibuses, car and tricycle. The results show that the support vector machine(SVM) has higher accuracy of recognition.(4)Vehicle video tracking: the paper analyzes and studies the MeanShift tracking algorithm, Kalman filtering algorithm, which is combined with MeanShift tracking, and multiple target tracking algorithm. In multiple target tracking, Using the Interested in region extraction and external rectangular box of the improved algorithm of the vehicle, this paper proposes a multiple target tracking based on the center of mass of the algorithm. The results show that the multiple targets tracking algorithm that is based on centroid increases the speed and accuracy.The experiments, which are realized by OpenCV and python under Windows system, show that the improved algorithms can quickly detect the real-time video moving vehicle and then accurately identify and track the vehicles. Therefore the algorithms can be applied to the intelligent transportation.