Study And Application Of Moving Vehicle Image Matching Methods

In recent years, intelligent transportation system have been developed rapidly and used more widely. Due to the complexity of real traffic, such as vehicles is changing lanes, the presence of vehicle interference, etc., which lead to the inefficiencies in application. In the system of measuring speed of a motor vehicle, if the object changed the lane or exceeded by the interfere vechicle that can lead to mistake, so it need to judge whether the object in the previous frame and the updating frame is the same vechicle that using the image match, which can privde the basis of the final traffic enforcement.This requires monitoring in the video stream processing to add a feature to identify the vehicle to determine whether it is the same vehicle in the previous frame and the update frame, to achieve the accuracy of the monitoring. In intelligent transportation systems, surveillance equipment, location and work environment often are not ideal, which makes the traditional matching becomes difficult in the vehicle identification.Firstly, SIFT feature point extracted, structural SIFT feature vectors, using Euclidean distance to achieve feature points that finish the initial matching, the final match by RANSAC algorithm to remove the errors. This paper designed a detection step to solve the identification between the target vehicle in the previous frame and the update frame which are caused by the existence of vehicles changing lanes, the body too large, interference between the adjacent lane and other factors in Intelligent Traffic System.The actual experiment shows that the detection of steps designed to match the recognition for the movement of vehicles with good results.Paper first introduces the background of the application,the traditional method of classification and the advantages and disadvantages, and the development of SIFT algorithm, highlighting and analyzing the Scale-invariant feature transform algorithm and RANSAC algorithm, and designed several experiments to prove as the matching algorithm. Secondly, this paper introduces the detection step software platform OpenCV, in the final it design the experiments which proved the detection step for identification of moving vehicles with good result.