Research On Robot's Disorderly Grabbing Based On Binocular Vision

Drived by the era of Industry 4.0,intelligent manufacturing technology continues to develop,requiring robots to have intelligent level.And traditional industrial robots Offline programming is generally adopted,and it is suitable for pipeline working mode,so it is difficult to adapt to complex scene capture task.In the face of the present disorderly stacked workpiece products,express package sorting,manual operation is difficult to face the new demand.And visual technology has brought about a new revolution,it changed the grasping way of traditional robot.Therefore,this paper combined binocular vision and six-axis manipulator as the research object,to achieve the disordered environment grab the task.The main work is as follows:Introduce the background and significance of the subject research,draw the current problems,and put forward the research purpose.Select a suitable binocular camera,build a six-axis manipulator,set up the capture scene,and design the experimental process.In the experiment,combining with Zhang Zhengyou's plane calibration method to solve the internal and external parameters of the camera,Camera calibration toolbox using Open CV image library camera calibration function and MATLAB software,the calibration plate images were collected for calibration.The experimental results were analyzed.Then,the knowledge of image preprocessing is analyzed.The image enhancement,transformation and filtering are discussed,and the best preprocessing method is obtained.The improved SIFT algorithm is used for local stereo matching.The parallax diagram is obtained by analyzing the principle of binocular parallax.The SAD algorithm,BM algorithm and SGBM stereo matching algorithm are verified by experiments.Finally,the hand eye calibration experiment was carried out.The kinematics analysis of the six-axis manipulator is carried out by using D-H parameter method.Verify the correctness of hand-eye calibration.Using the robot toolbox of MATLAB software kinematics simulation and path planning research,obtains the effect diagram.In this thesis,the binocular camera is used as the perceptron of the six-axis manipulator to capture the target of the set disordered scene.For multi-target recognition,target segmentation algorithm based on feature points is used to extract features.SGBM algorithm is used to convert the depth map to obtain the three-dimensional coordinates of the target.Finally,the task of capturing the target in a disordered environment is realized,and the target object can be accurately located.