Research On Vision-based Detection Technology Of Vehicles Ahead

The research of intelligent transportation system is very urgent to reduce traffic accidents.As the most important part of intelligent vehicles,the research on vehicle detection and tracking technology is getting more and more attention from researchers.In this paper,the problem of frequent traffic accidents during vehicle driving is studied in the aspect of vehicle detection and tracking.By using binocular vision sensor to analyze image preprocessing,vehicle detection and tracking,camera calibration and stereo vision system,a Camshift algorithm with improved features and Kalman filtering is proposed to solve the problem of inaccurate tracking during vehicle driving.details as follows:Firstly,further research on the research status and development trend of intelligent transportation systems at home and abroad,and briefly introduce the current mainstream vehicle detection technology.The image preprocessing related techniques are analyzed.The median filtering is used as the image filtering method.The image is binarized by the OSTU method,and the Canny edge detection operator is selected to extract the image edges.Secondly,the improvement of the Camshift algorithm is realized by combining vehicle characteristics and Kalman filter tracking.Based on the vehicle detection technology in front,the paper uses the improved Hough transform to extract the lane line and accurately locate the vehicle according to the shadow of the bottom edge of the bottom of the vehicle.The Camshift algorithm is used to track the vehicle in front,and combined with Kalman filter to update and predict the position of the next moment,the accuracy of tracking is improved.Finally,the imaging principle of the binocular camera and the conversion relationship between the coordinate systems were analyzed,and the binocular stereo vision system ranging was studied.It mainly includes stereo correction,polar geometry principle and stereo matching algorithm.The hardware and software platform of stereo vision system is built.the distance measurement experiment is verified and analyzed,and the ranging error is kept within 5% in the end.