Research On Optimization Of Matching Algorithm For Binocular Stereo Vision Vehicle Speed Measurement System

Economic development and social progress have made cars a means of transportation for more people when traveling.With this,the pressure on road traffic has increased,and the demand for intelligent transportation networks has become more urgent.The measurement of vehicle speed is a basic and important part of intelligent transportation network.The vehicle speed measurement system based on binocular stereo vision vehicles can overcome the shortcomings of traditional speed measurement methods such as high installation cost,complex construction,and measurement limitations.The speed measurement system consists of four parts: target detection,image matching,binocular camera calibration and speed measurement.The research content of this paper is the image matching module in the system.Image matching is a basic and critical problem in the field of computer vision.In this speed measurement system,image matching is to obtain the correct matching points between the two images.However,the method of combining SURF and the Homography matrix in the solution process does not involve the interference of noise/outliers and the amount of calculation.Reach the trade-off.This article mainly focuses on the deficiencies of the above-mentioned matching methods.The image matching module is the follow-up work of the target detection module.In the target detection,the coordinate position information of the target area can be obtained.The position information is used to constrain the position of the feature points extracted from SURF,and the relationship between the pixel proportion of the license plate area and the distance of the camera license plate can be fitted by the characteristics of license plate system.This relation is used to roughly eliminate the matching point pairs whose distance to the camera is obviously abnormal in the matching point pairs set obtained by SURF algorithm matching the license plate area.Then,the local neighborhood consistency constraint was used to eliminate the mismatched point pairs in the matching point pair set,and the size of the matching point pair set was further reduced.Using the estimation of the camera pose,the mismatched points in the car logo,car lights,and mirror areas are eliminated;In the speed measurement process,in order to measure the same point in the two frames before and after the speed measurement,combined with the continuity of the video sequence,a vehicle speed measurement scheme based on the video time and space domain is proposed.This paper studies the deficiencies of the matching algorithm in the vehicle speed measurement system based on binocular stereo vision vehicles,and proves the feasibility of this method in theory and practice.After six sets of experimental tests,the speed measurement results are compared with the speed obtained by GPS.The speed measurement error of a single feature meets the application requirements of the Chinese national standard GB/T 21555-2007 with an error of less than 6%.In order to reduce the error of single-feature velocity measurement,the four characteristics of velocity measurement errors are averaged,the maximum velocity measurement error is 2.79%,which is 1.01% less than the original velocity measurement system.The speed test results prove that the matching algorithm has high accuracy and reliability.