Research On Vehicle Detection And Tracking Method Based On Machine Vision

Machine vision-based moving object detection and tracking technology is one of the hot spots and difficulties in the field of image processing and is used in various fields such as intelligent transportation,human-computer interaction,and artificial intelligence.Through the detection and tracking of the moving vehicles in the traffic surveillance video,the traffic information of the vehicle is provided for the intelligent traffic system.The research of this thesis mainly focuses on the accurate detection of the moving vehicle in the surveillance video,and then the vehicle is effectively tracked on the basis of the detection.Therefore,this thesis starts with the most commonly methods for the detection and tracking of moving targets by domestic and foreign researchers.The main tasks are as follows:Firstly,in the detection of moving vehicles,based on the frame difference method and the edge detection method,a moving vehicle detection algorithm based on the improved three-frame difference method and edge detection method is proposed to improve the problem of three-frame difference algorithm.Then the improved three-frame difference algorithm combined Gaussian mixture model is used to improve the result,which solves the problem that the vehicle interior information detected by the three-frame difference method is not complete and the Gaussian mixture model is sensitive to light.The improved algorithm improves the accuracy of the algorithm for vehicle detection and adaptability to changes in lighting.Secondly,in the tracking of moving vehicles,based on the accurate detection of vehicles,the contour center coordinates of the vehicle are taken as the research object.Firstly,the center coordinates of the vehicle contour in the current frame are extracted,then the vehicle center coordinates in the next frame image are predicted by Kalman filter,and then the coordinates of the detected vehicle contour in the frame are extracted as the Kalman filter when the next frame image arrives.The observation value is used to correct the prediction value at the previous moment through this observation value.Finally,this corrected value is used as the predicted value at the next calculation moment,and the loop is continuously looped to match the detection result with the tracking result until the end.Finally,in order to facilitate the maintenance and improvement of the algorithm in the later period,according to the research route and results of the video image preprocessing,modular design is used for vehicle detection and tracking in this thesis,and it includes four modules: video preprocessing,vehicle detection,vehicle tracking and result display.Results showed that the system can effectively detect and track the vehicles in the surveillance video.Through the analysis of the tracking results,relevant information about the vehicle can be calculated.