| In the application environment of the robot, industrial robot can only execute a predetermined command in a structured environment strictly defined, lacking of perception and strain capacity for the application environment, which limits the application of robot seriously. The robot loaded the vision system does not need the motion trajectory of the robot off-line programming or teach, saving a lot of programming time and improving production efficiency and quality of work.In the vision system of the robot, robot based on the around two-dimensional image can get the 3D information. The principle of binocular stereo vision based on the stereoscopic imaging is according to the principle of human eye, through the use of calibration of two identical camera shoot on the target object from different angles, the image correction, image preprocessing, image feature extraction,3D image stereo matching technique to obtain the target object and the 3D reconstruction.This paper is based on the principle of binocular stereo vision, makes a deep research o n the robot visual system.. This paper introduces the robot manipulator and the configuration of the camera mode, and select the binocular camera configuration of eye-to-hand, and the i mage preprocessing, image feature extraction are researched, aiming at the calibration precis ion of binocular camera, using RANSAC (Random Sample Consensus) algorithm is the core idea of random checkerboard calibration experiment, secondly aiming at efficiency of the al gorithm, proposes an image traversal method based on Harris corner detection, this paper als o discusses the binocular stereo vision image correction and depth information extraction tec hnology, and has carried on the related experiment, the experimental results show that, RAN SAC algorithm based on Harris image corner point detection algorithm by traversal method t he algorithm has the reference value in the application of 3D coordinate measurement of bin ocular vision system in robot. |
PDF Full Text Request