Research On The Method Of Vehicle Illegal Jumping Feature Extraction And Tracking Based On Vehicle Monitor Video

With the increase of the number of car in China,traffic problems have emerged more frequently,especially in the morning and evening peaks period.It is one of the main reasons for urban traffic congestion,which not only seriously reduces the traffic efficiency.disrupts traffic order,but also easily causes accidents.It is difficult for the fixed cameras to satisfy supervision and management of traffic order by administration department in the whole area due to the increasing number of cars and the mileage of the highway.Meanwhile,the vehicles have widely installed the driving recorders that can not only record and save the environment around the vehicle,but also capture and upload the monitoring video,which effectively expand the range of monitoring.However,the vehicle mounted camera itself is in a fast moving state compared with traditional surveillance cameras such as electronic police,which brings new research for target detection and tracking of vehicles jumping.Therefore.this paper puts forward a method of extracting and tracking of video vehicle monitoring based on the characteristics of vehicle jumping.It aims to provide legal basis for the effective supervision of the media and the fast handling of the traffic police department by quickly and automatically extracting and saving the forensics images of vehicle jumping from received monitoring video.The main works are as follows:Firstly,the detection and fitting of lane in the region of interest with the feature constraint are completed based on image pre-processing,edge detection and Hough transform.Moreover,the width of lane in region of interest is calculated by line equation,providing data for analysis of the jumping.Secondly,an adaptive method based on license plate is proposed to enhance efficiency of the existing vehicle tracking method.It contributes to the results that it reduces search and detection of invalid information and improve the efficiency by adjusting the detection area of the next frame according to the current frame of the license plate detection and monocular ranging.Thirdly,an image forensics method for jumping is proposed in this paper.The method utilizes the relationship between the lateral displacement of the vehicle in the vehicle-mounted monitoring image and the detected lane width to analyze and extract the forensic image of the process of the jumping.Finally,the experiment have been carried out to verify the accuracy and robustness of the method.