Research On Video Vehicle Tracking And Traffic Incident Detection Algorithm

With the rapid development of economy in our country,the number of motor vehicles has a trend of increased year by year.The road traffic accident economic loss and the death toll has increased dramatically.All kinds of traffic accidents has become a severe problem which endanger the safety of the urban traffic.The leading cause of traffic accident is motor vehicle traffic violations.Currently the most serious violations is including: abnormal vehicle parking,change lanes,ultra-high speed and super-low speed driving,illegal retrograde,etc.These illegal behavior seriously affect the urban traffic order,and cause traffic accidents easily.The traffic incident detection system is based on video.It can monitor all kinds of traffic violations on the road with real-time,and automatically get the evidence of vehicle violations.It also makes drivers consciously abide by traffic rules.In turn,it can effectively reduce the number of traffic accidents,and maintain road traffic safety.Therefore,road traffic abnormal incident detection is very important for the study.This article mainly studied on vehicle detection,tracking and all kinds of traffic incident detection.The main work includes the following four aspects:We discussed features of several kinds of target detection methods,and used the combing method of background difference and three frame difference to detect targets.The experimental results show that the moving target detection algorithm is able to positioning vehicle targets with high accuracy,speed.This paper proposed a target tracking method.It based on dynamic search window and joint characteristics.The characteristics are the color of the target,the center of mass and the height and width of external rectangular.The method had good real-time and overcame the problem of poor effect of single feature tracking.We proposed a tracking compensation algorithm by searching for connected domain based on the public pixel of two frames of target.The algorithm has robustness.It often lead to tracking failure because of targets adhesion,background covering goals or targets moving too fast.So this paper proposed a motion estimation approach.It predicts the location of the target in thenext frame,and realizes the stability of target tracking at last.The experimental results show that the tracking algorithm is simple,and stable,and can track moving targets with real-time.We analyzed the abnormal parking,change lanes,ultra-high speed and super-low speed,and vehicle retrograde traffic events.And we put forward a detection method.It is on the basis of the history of the target tracking trajectory.The experimental results show that the traffic incident detection algorithm can judge all kinds of traffic abnormal events with high accuracy.At first,we analyzed several important function modules,and then designed the detection system and software platform.It was constructed on VC6.0 platform,and used object-oriented method.The visual interface is used for interaction and DirectShow is used for image display.