Research On The Technology Of Moving Vehicle Detecting And Tracking In Video Image

With the accelerated pace of urbanization, the number of motor vehicles continued growth, the city's transport system faces enormous pressure. Computer vision-based intelligent transportation system is an effective way to solve the problem, while the motion detection and tracking of vehicles as one of its key core technology will become very important. In this paper, detection and tracking of moving vehicles has been studied, analyzed and summarized the existing detection and tracking technology, focusing on aspects of sports vehicle tracking algorithm, and the corresponding algorithm improvement. In this paper, the research work are as follows:1) The detection of moving vehicles, when the case of static camera, through the inter-frame difference method, the average background reconstruction based on time difference method, classification based on pixel gray background reconstruction difference, based on Gaussian mixture Difference of Background Modeling results of these four methods of comparison, the last use of the Gaussian mixture model-based background modeling difference method to detect moving vehicles. Pre-test and post-test and the images were pre-, and HSV color space based on the shadow removal algorithm to remove the shadow of the image, making detection more complete vehicle area.2) Presented to the SIFT algorithm. This tracking method is based on SIFT feature points between vehicles match, but the SIFT algorithm to generate the feature vector is 128-dimensional computation of large and can not meet the requirements of real-time. On this basis, with the PCA algorithm is proposed to reduce the dimensionality of the feature vector, the SIFT-PCA algorithm. By comparing the experimental data, SIFT-PCA algorithm to real time has increased.3) The proposed color feature as a secondary feature. In the tracking process, there could be a vehicle of SIFT feature points and multiple vehicles to match the SIFT feature points, so there could be false match. In order to accurately match the color of the proposed vehicle moments (lower third moment) features to further determine whether the match between vehicles, a simple measure of Euclidean distance formula between them and compare the color characteristics, color characteristics of selected distance value least two vehicles to the correct match.