Research On Ranging Technology Based On Computer Vision

Computer vision has been widely used in various fields and promoted the development of society.As an important research direction of computer vision,vision ranging has played a huge role in the fields of UAV vision positioning,robot autonomous navigation,non-contact measurement and so on.Binocular stereo vision ranging,with its advantages of high accuracy,low investment,good concealment and non-destructive,can realize the measurement of the size,shape and position of irregular processing objects,and meet the increasing demand for measurement accuracy and detection efficiency.Its application prospect is broad and has certain research value.In this paper,a method of combining MATLAB calibration with OpenCV matching is proposed to realize binocular stereo vision ranging.The main work includes the following aspects:In the introduction of camera calibration,three main camera calibration methods are studied,and the advantages and disadvantages of the three methods as well as the use scenarios are analyzed.Aiming at the disadvantage of traditional calibration method which needs high precision calibrator and self-calibration method,this paper adopts a calibration method which does not need high precision calibrator and is accurate and convenient.Based on this method,binocular stereo calibration is completed after single target calibration of left and right lenses in MATLAB respectively.Accurate calibration data of internal and external parameters are obtained,and data parameters are verified.For the most important step of stereo matching before measurement,the basic theory of stereo matching is introduced in detail from three aspects: matching principle,matching criterion and matching classification.The regional stereo matching algorithm is analyzed in detail.The influence of different window sizes on the accuracy of stereo matching has been analyzed.SGBM and BM algorithms are simulated in depth.By comparing disparity maps,their advantages and disadvantages are analyzed.Considering the timeliness and choosing the appropriate matching algorithm..The hardware system platform of binocular stereo vision is built and the actual ranging experiment is carried out.A total of 165 groups of measured data were recorded,and the average relative error of each group of measured distances was calculated.The result of error analysis shows that the ranging accuracy of this system is less than 5% in the range of 200 mm-4000 mm.Finally,it is compared with the ranging system based on ORB matching and the ranging system based on traditional algorithm.The system has a wider measuring range and higher measurement accuracy.The proposed visual ranging method can meet the requirements of high accuracy,and has certain practical significance in the application scenarios of obstacle avoidance and robot navigation.