The Research On Detection And Tracking Techniques For Video’s Moving Vehicles In Intelligent Transportation System

Intelligent transportation system (ITS) is a system for applying computer visiontechnology and image processing technology in traffic domain. Also, it is currentlybeing studied and worldwide focused on in the area of world traffic transport. The aimof ITS is to make full use of available road facilities，improve reciprocity of vehicle，road and people, and then enhance the economical efficiency of the whole system.The main research aim of detection and tracking for moving vehicles in ITS ishow to raise the accuracy of detection and tracking algorithm for moving vehicles, theresponse speed and the anti-interference ability. This paper mainly discusses thefundamental theories and key technologies of vehicle detection and tracking systembased on video sequences in ITS. After summarizing and analyzing the existingtechnologies of vehicle detection and tracking system, a new method for vehicledetection and tracking in municipal transportation was presented. Experiments aredone to demonstrate the validity of the proposed method．The main contents of thestudy are summarized as following:1. Detection of moving objective vehicles. Captured vehicle images are firstpreprocessed to diminish noise and interference. Then, the paper focuses on theadaptive background updating algorithm based on multi-information fusion which canadaptively adjust the updating speed according to the background disturbance and theintensity of illumination; and segments the moving vehicles on binary image bychoosing the OTSU. Then, it, with morphological filtering, makes theclose-before-open operation on binary image in order to eliminate the image noise and’holes’; Finally, it proposes a new improved algorithm based on the HSV color spacemodel, that effectively reduces the shadow detection error rate. The effectiveness andreliability of the proposed method is verified by experimental simulation.2. Tracking of moving objective vehicles. The paper studies the characteristics ofthe SIFT descriptor, and proposes the simplified way of SIFT in practical trafficsituation. Then concludes a Mean Shift object tracking algorithm based on SIFTdescriptor (SIFT-Mean Shift). The algorithm can calculate the position and scale oftracking targets by simplified SIFT. And then extract the key points’ direction featurevectors in the tracking area, to describe the moving targets by using the directiondistribution histogram. In the end, with Bhattacharyya coefficient, assess the similarity of feature vectors distribution for target tracking. Experimental results show that thealgorithm can obtain the good tracking effect for moving vehicles in which scales arechanged and rotated, and have a strong adaptability for noise, illumination or occlusion.It also has good robustness.