Font Size: a A A

Research Of Binocular Stereo Vision Distance And Speed Measurement Based On OpenCV

Posted on:2013-05-02Degree:MasterType:Thesis
Country:ChinaCandidate:F B LiuFull Text:PDF
GTID:2248330371485543Subject:Software engineering
Abstract/Summary:PDF Full Text Request
Vision system is the most important organ human beings used to perceive theworld. Human get most of the information of the outside world though sense of sight.Computer as the core of the machine vision is to make computer applications have thecognitive ability of the three-dimensional world of environmental information throughthe two-dimensional image information. Machine vision system simulates humanvision system which is becoming one of the hottest researches. It has more and moreapplications in the military, aerospace, medical and other fields.This article focuses on binocular stereo vision system. Based on parallaxprinciple, we achieve: capture of images, camera calibration, the feature pointsdetection and extracting, image’s stereo matching and3D reconstruction gradually.We finally accomplish distance measurement and speed measurement of the movingtarget.In the experiments of this topic, we firstly program the capture of video framefrom two cameras simultaneously with the help of OpenCV in VS2005. Consideredwith the complexity and feasibility of the process, we adopt the Zhengyou Zheng’sChessboard camera calibration method, a method between the traditional calibrationand self-calibration, to get the inside and outside parameters of the cameras byexperiment. Then, we respectively use Harris corner detection method, Shi-Tomasicorner detection method (GoodFeatureToTrack) and KLT corner tracking method todetect the two images captured by two cameras simultaneously, then extract thefeature points. According to the stereo matching of image, we accomplish SIFT stereomatching algorithm experiment and block matching under epipolar constraintalgorithm experiment. We contain each matching conjugate points, and we can gettheir coordinates. Then, with the matching points by stereo matching, we can achieve3D reconstruction based on parallax principle. According to the corner detectionmethod and stereo matching algorithm we have achieved, we combine three differentranging programs: Harris+SAD,GFT+KLT+SAD,SIFT+KLT. We use this threeprograms to accomplish distance measurement,compare and analyze the results.Finally, we study Kalman filter theory and then use this theory to measure the velocityof moving targets.Using two HD-USB cameras, we compile each parts of this experiments’ codewith C++based on OpenCV in VS2005.
Keywords/Search Tags:Binocular stereo vision, Camera calibration, Feature extracting and stereomatching, 3D reconstruction, Kalman filter, OpenCV
PDF Full Text Request
Related items