Research On Vehicle Detection And Tracking System

In recent years, because of the rapid development of domestic’s economy, thevehicle has entered into thousands of households, but also brought a variety of trafficproblem, especially traffic congestion phenomenon exists in large and medium-sizedcity, giving people a lot of inconvenience at daily travel. In order to solve the presentnumerous vehicles cause traffic problems, intelligent transportation technology hasattracted more attention in each country. Because traffic congestion phenomenon hasgradually spread to the small city, to solve this problem has become more and moreurgent and important. Video-based vehicle detection and tracking technology is the fieldof intelligent transportation category, can detect and track vehicles with statistics onroad, is the effective addressing way to solve vehicle congestion problem.Vehicle detection and tracking system mainly includes three technology of imageprocessing, including moving vehicle detection, moving vehicle tracking and movingvehicle shadow removal. In this paper, according to these three issues in-depth study,through the experimental comparison to choose a suitable method, to finish vehicledetection and tracking system. First of all, research on image preprocessing, and thenstudy the real time high of the frame subtraction and background subtraction in movingvehicle detection technology. Moving vehicle tracking has been more mature methods.Here mainly target the center point and the color combination of multiple features fortracking experiment. Because shadow existence causes error tracking,So to researchshadow and make corresponding improvement. Finally to test the integration of thewhole system.In this paper, based on the two kinds of shadow removal algorithms including thecolor space detection and edge detection are studied and improved. Based on the colorspace detection of shadow removal, proposed uses the single brightness scale separationof shadow and vehicle. After the algorithm is improved, the experiment shows that thecolor contrast are not bright between the shadow and vehicle color, and the situationalso can be very good for shadow removal. Based on the edge detection, the four edgeoperators are filtered to extract foreground contour, finally experimental result showsbetter than the commonly used Canny operator, Sobel operator and Log operatorshadow removal effect.