Design And Implementation Of Traffic Flow Statistics System

In recent years, with the rapid development of computer vision technology, intelligentvideo processing technology is applied to real life more and more. Intelligenttransportation system based on video surveillance is a hot research field of computervision technology. It can analyze the video image data to obtain real-time informationabout the road traffic situation. This paper makes a deep research on vehicle targetdetection and vehicle tracking algorithm in Intelligent Traffic Monitoring System. Putforward the video vehicle automatic detection identification method based on AdaBoostClassifier discriminant and Vehicle tracking method based on Lucas-Kanade pyramidoptical fow computation.In order to Controls and constraints vehicles, Road traffic monitoring system requiresall camera installation perspectives can clearly see the vehicle license plate area. There islarge difference between different vehicles in models and colors, but the partialcharacteristics of the plate area having similarity. This article use this feature, puttingforward vehicle detection method based on license plate information, and trainingdetecting plate area AdaBoost classifier through offline Learning. First of all, theMonitoring video vehicle automatic detection method extract Interested region of themovement by Motion Modeling, then, detect the license plate area targets by slidingAdaBoost Classifier within the region of interest, and use Clustering method to merge theinformation of the license plate area became the vehicle target location information in theend.Vehicle in the course of driving is rigid motion. This feature meets the constraints ofOptical flow feature matching algorithm. Vehicle tracking method based on Lucas-Kanadepyramid optical fow computation select a set of feature points in the target area of thevehicle at first, use pyramid optical flow algorithm to strike light flow of this set of featurepoints in front and rear of two frames. And then use the Matching Errors Criterion and Forward error detection method excludes the Error larger feature points, get the motionvector of the target thereby. To track processing Continuous video sequence for obtainingthe vehicle target trajectory.This paper has analyzed the vehicle trajectory obtained by the algorithm, performederror correction is by using the motion characteristics of the vehicle, and proposed analgorithm to make the system adapt to the long-stay parking situation. We finally realizethe function of traffic statistics by analyzing and getting the traveling locus of the vehicletarget in surveillance video sequences, and counting. In order to validate the performanceof the algorithm proposed in this paper, experiments of a large number of surveillancevideo were completed. The experimental results show that the algorithm system proposedin this paper is a high-precision, adaptability, and efficient traffic flow statistics system.