Research On Target Location And Grab System For Robot Based On Binocular Vision

Binocular vision is one of the hot topics in the field of computer vision,and can actualize the acquisition and reconstruction of target object's environmental info rmation,so it is widely used in target positioning.In order to make the robot repl ace the workers to complete the work in the complex and dangerous working envi ronment and reduce the amount of workers' labor,this paper studies the robot targ et location and grab automatically based on binocular vision.(1)Research some image preprocessing algorithms which are used to reduce the noise interference and enhance the detail features of target image.Through the analysis of image feature extraction,edge extraction modeling and template matchin g,this paper propose edge extraction based on grey level transformation.(2)The algorithm is developed in the Halcon vision development environme nt and finally converted into C code by analyzing the binocular calibration of bino cular vision system and the mathematical model of binocular vision.Binocular cali bration is carried out using "binocular?calibration" operator,and "calibrate?hand?eye" executes hand-eye calibration and determines the relative position relationship bet ween the left camera and the robot base.(3)In order to make the data simple for processing,the binocular vision sy stem is changed from non-standard polar geometry structure to standard polar geom etry structure.This paper proposes using sub-pixel contour based on image pyramid as shape template for target positioning,and the effect of the algorithm are verifi ed by experiments.(4)Proposing grey correlation stereo matchings based on region to achieve s tereo reconstruction,and the "eye-in-hand" hand-eye model is used to design a co mplete system based on JLRB20 robot.The stability of system is verified by grab bing experiment,and the grab error is analyzed to verify reliability of the system.