Research On Robot Positioning Technology Based On Binocular Vision

With the rise of the "Industry 4.0" strategy concept,robot technology has been widely used in the field of intelligent manufacturing.The integration of robot systems and machine vision is an important manifestation of intelligent work.Today,machine vision has evolved from the original monocular vision to binocular vision and multi-eye vision.Among them,binocular stereo vision technology simulates the function of human binocular vision,and can understand the depth information of space through parallax.Compared with multi-eye vision,binocular vision has better real-time.This article builds a binocular vision system based on the ABB robot body.Under the environment of Robot Studio software and HALCON software,the robot positioning system based on binocular vision is studied.Main tasks as follows:1.In the binocular calibration part,the calibration method and camera calibration principle are researched.Camera parameters are solved based on Zhang Zhengyou calibration method.The calibration plate image is collected,and the binocular system calibration is finally realized through operations such as dividing the calibration plate area,extracting sub-pixel coordinates of the mark points,and the internal and external parameters and correction map of the left and right cameras are obtained.2.Aiming at the hand-eye system calibration,the Eye-in-Hand calibration model is researched,the robot tool end is manipulated by a teaching device to acquire calibration plate images,generate robot pose files,perform hand-eye calibration,and obtain pose conversion matrix between the left camera and the tool end.3.For the binocular images collected after calibration,the spatial and frequency domain filtering algorithms are studied,and Gaussian filtering algorithm based on FFT realizes image enhancement.The area of interest is selected based on the area features of the area.Feature points are extracted using the centroid algorithm and Hessian matrix attributes.4.The gray-correlation region matching algorithm combining the polar constraint criterion and NCC(Normalized Cross-Correlation)is used to complete the stereo matching of the feature points.The feature points are reconstructed in combination with the three-dimensional positioning principle to obtain the three-dimensional coordinates in the left camera as the reference coordinate system.5.The pose transformation matrix obtained by hand-eye calibration is used to convert the pose in the reference camera coordinate system to the pose in the robot base coordinate system,and realize data transmission of pose through Socket communication.Finally,a virtual workstation for automatic handling of cylindrical parts is established in the Robot Studio system,and the RAPID language is written to implement the positioning simulation test of the robot.