Research Of Mobile Robots Binocular Stereo Vision

With the rapid booming advance of social economy and science and technology,Intelligent mobile robots will become an important part of the future of society.duringoperation process of the intelligent mobile robot, they use their system to detect andperceive the external environment,and guide the robot beforethe next step of work.And because of it gets large amount of informationand feels the intuitive, stereovision systemgradually become intelligent mobile robot to researchhot spot of bafflediscovery, location, baffle avoidance etc.Therefore, it is vital significance toresearchstereo vision system.So, this paper mainly include three content：camera calibrationof Binocularvision,images matching,3-d restoration of obstacles in space.Firstly, this paper presents the research platform---uAgent intelligent mobilerobot. The paper introduces working theory of the comprehensive terrace,andrecommend hardware extension methods for mobile robot; Then thepaper introducesthe Marr machine vision of the basic theory, and use it to instruct this paper research.In detail, this paper introduces the basic knowledge of camera calibration andanalysisthe traditional, self-calibration, and some new calibration methods, this paper adopts apractical, accurate calibration method,---zhang’s calibration method，and the cameracalibration experiment was carried on.Secondly, in order to use binocular stereo vision system to recover obstacleinformation in space, we studied a lot of data, found the SIFT algorithm of imagescaling and rotation to Keep the same features. The paperresearches SIFT incombination with KD tree+BBF algorithm to Analysis of the left and right as thecorresponding feature points in the process of searching, using128dvector expressing a feature point SIFT descriptor for registration. It reduces the error rate, improve the efficiency.Finally, the paper researchthe binocular stereo vision system for stereo matchingthe characteristics of the key points for the recovery of3-d information, Based on theprojection matrix, we calculate the3-d space coordinates so that reduces calculationof complicated degree. In addition, it can quickly restore the obstacles ofthree-dimensional spatial information. Based on VS2010+OPENVC experimentsshow that, this paper puts forward the method has high accurateness and strongreal-time, so this method can satisfy the demand.