Mobile Robot Positioning Target Object Based On Binocular Stereo Vision

During recent decades, many universities and research institution have set off a wave of mobile robot research, accompanied by mobile robot increasingly widely used in production, live and other fields, and the mobile robot technology based on binocular stereo vision has become a hat spot.The subject aims at achieving three-dimensional positioning of target object in scene, on the research platform of AS-R mobile robot's binocular vision system.Working principles and mechanisms of each research module in binocular stereo vision system, besides exploratory research is done in the two modules of feature extraction and feature matching obtaining good result.The main work is summarized as follows:Firstly, the tructural characteristics of the AS-R robot's binocular stereo vision system as the platform of the subject are in research and analysis, grasping the working mechanism of the platform on the whole and understanding the way to expand in robot hardware also expanding a binocular vision system as research platform; I learns the basic theory of Marr machine vision to guide the research and grasps the principle and structure form of binocular vision system; The current popular camera calibration methods are studied and presented in principle analysis.This subject takes Zhang's plane calibration to get the parameters of binocular stereo vision.Secondly, this subject proposes a method to attain the color threshold adaptively.The football target in image is segmented using HSI space threshold and then to estimate the center and radius of football based on "four key points" and the area method;The principle of Canny edge detection operator is studied which is employed to extract the single pixel football edge; "Even and sparse edge feature extraction"methed is put forward based on the imaging characteristics of soccer in extracting matching feature,extracting eight sparse edge points almost evenly in the edge having been detected; " Corresponding partial traverse match"method is put forward in the stereo match method corresponding to feature extraction method,finding the eight match points in right (left) image corresponding to the eight waiting-match points in left(right)image.The match of sparse feature points in both images is completed,getting sixteen match point pairs.Finally, both parallel and intersectant optical axis binocular vision system's principle of three-dimensional positioning target object are analyzed.The subject carries out an experiment of three-dimensional positioning upon sixteen match point pairs by binocular vision system of parallel optical axis,gaining three-dimensional coordinates of sixteen points on the football edge.The average of these coordinates is the three-dimensional coordinate of football core,that is to actualize positioning football target. Positioning results are verified and the causes of error are analysed.