Vehicle Violation Based On Video Image Research On Behavior Detection And Recognition Technology

Due to the continuous increase of urban population,the demand for motor vehicles is on the rise.The popularity of vehicles has brought a series of traffic problems,the most obvious is the increase of traffic accidents year by year.The main cause of traffic accidents is that drivers do not abide by the traffic rules and can not ensure safe driving.In order to better manage and plan the vehicles on the road,the development of intelligent transportation system is widely concerned by the domestic and foreign governments and relevant research institutions.As an integral part of intelligent transportation system,vehicle violation detection technology has important research value.With the rapid development of image processing,pattern recognition and other technical means,more and more researches have been made on various intelligent control systems based on video analysis.In this paper,the knowledge of image preprocessing is studied,and the algorithms of image gray,image filtering,edge extraction,expansion corrosion,cavity filling and so on are introduced in detail.In the process of moving object detection,firstly,some commonly used moving object detection algorithms are introduced,and the advantages and disadvantages of these algorithms are analyzed.The principle and experimental process of vibe algorithm are introduced in detail.Based on the traditional vibe algorithm,it is improved.In the background modeling stage of vibe algorithm,the method of expanding the sample neighborhood range is used for selection;the method of dynamic background measurement is used Methods: the threshold value of vibe algorithm was adjusted adaptively to adapt to the impact of dynamic background on the detection results;the shadow removal method based on edge information was used to eliminate the interference of shadow on the detection results.Compared with the experimental results of the traditional vibe algorithm,the improved vibe algorithm can detect the moving target better in the complex environment.After that,the multi feature cascade classifier is used to recognize the vehicle by extracting the hog and Haar like features of the vehicle.According to the experimental data,the algorithm can achieve the expected effect on the accuracy of vehicle recognition.At last,we detect and recognize the illegal behaviors of the moving vehicles,specifically study the two kinds of illegal behaviors: illegal line pressing and illegal parking.The experiment shows that the algorithm adopted in this paper is practical and feasible,and meets the expected design requirements.