Research On Vehicle Object Extraction And Speed Measurement In Traffic Vide

With the rapid development of China’s economic level and people’s living standard,as the national public infrastructure,highway traffic has also been rapid developing,which makes people’s travel very convenience and brings great challenges to the work of the traffic management department,in this case,the intelligent transportation systems is emerging as the times require.Intelligent transportation system is the development direction of traffic management,which can combine traffic control with computer science effective,and will improve the efficiency of traffic management and save manpower resources greatly.Intelligent traffic system mainly includes these key technologies extracting,tracking and behavior analysis of moving objects in traffic video image sequence intelligent.The research content of this paper is the part of the basic research work of national fund "the key technology research of distributed video vehicle unified spatiotemporal Association retrieval ",and also is the foundation part of intelligent transportation system.The main research work of this paper are as follows:The moving object detection in traffic video.For the road environment of traffic video is complex and changeable(such as the changes of weather,the changes of light illumination,tree branches sway and so on),based on the analysis of various classical algorithms,this paper presents a moving target detection algorithm based on local texture feature of the pixel Gaussian mixture background modeling method,It can effectively improve the anti-jamming performance of moving target detection and reduce the false detection rate due to the similar background.In the end,the foreground area is processed by closed operation and open operation,which can remove the noise and the "internal hole" of prospect effectively and further improve the prospects of the integrity.The vehicle shadow elimination in traffic video.Due to the transformation of intensity,size,direction and other properties of the vehicle shadow in the traffic video,the shadow elimination becomes a very difficult task.Which ask that the shadow elimination algorithm has a certain generality.Aiming at the above problems,this paper put forward a kind of based on dynamic texture window improved operator shadow elimination algorithm on the basis of LBP Operator,the algorithm propose an improved texture operator based on dynamic window,window is dynamic only in shadow region in the calculation of the texture and modified the calculation method of texture,the algorithm is not sensitive of the shadow’s strength,size,and direction of characteristics,so you can get better effect in different environment.The vehicle speed measurement based on high speed traffic video.In order to obtain the speed information of the vehicle in the high-speed road section without the hardware speed measuring equipment,this paper mainly uses video velocimetry based on virtual coil,according to the high-speed road lane standard,in the video for the manual calibration virtual velocity region 30M length,In view of the inaccuracy of the vehicle collision detection,In this paper,we propose a video speed measurement algorithm,which is based on vehicle corner tracking in the virtual area,when the vehicle passes through the virtual area of velocity measurement by using the LucasKanade of Pyramid method based on the Shi-Tomasi corner to track and record the characteristic points of the vehicle area,then,PAM clustering algorithm gets the time parameters after the vehicle go through the virtual velocity region,and then obtain information about the speed of the vehicle,this algorithm can improve the video velocity measurement accuracy effective.