Manipulator Grabbing Connecting Rod System Based On Binocular Stereo Vision

With the vigorous advancement of the intelligent process,especially in the context of the Industry 4.0 era with high-end equipment as the core,the machine vision industry is developing rapidly.In recent years,industrial robots have developed rapidly,and the market demand has grown significantly.Machine vision technology,which plays the role of the "eye of fire" of industrial robots,has contributed.This article combines current market needs,a set of gripping systems for mechanically placed mechanical links was designed using industrial robot systems combined with binocular stereo vision technology.The main research contents are as follows:According to the needs of the project,choose the appropriate hardware equipment including cameras,lenses,light sources,etc.Depending on the installation environment and institutional sports field trip to choose the right robot model.The industrial camera imaging model,the installation method of the binocular camera,the dual-target calibration,and the algorithm principle of hand-eye calibration are introduced in detail.Write a program to implement the camera and robot calibration algorithms.The final calibration results are analyzed and compared.In this paper,by analyzing the surface features of the connecting rod parts,we decided to use speckle lasers to enhance the surface information of the parts.The disparity map of the scene where the target is located is obtained through the study of multiple binocular stereo matching algorithms,and the 3D point cloud map of the link part is obtained through 3D reconstruction.Aiming at the problem that too much point cloud information of the scene where the target part is located leads to the slow registration of subsequent point clouds,this paper proposes a threshold segmentation method based on the Z-axis coordinate values to remove the background point cloud scene.The problem of 3D pose estimation for multiple target objects placed randomly is studied.Finally,in this paper,a model-based point cloudregistration method is used to estimate the pose of target parts in the scene point cloud.First,a large amount of point cloud data is used.Calculate the normal vector of each point,then extract the feature points,characterize the feature points,perform preliminary registration of the template point cloud and the scene point cloud,and then use the ICP algorithm to accurately register the point cloud and calculate the point pair Rotation transformation relationship.Finally,the 3D pose information of the target in the manipulator base coordinate system can be obtained by combining the results of previous hand-eye calibration.The obtained posture information is sent to the manipulator,and the manipulator is guided to grasp the link parts.Finally,combined with software and hardware,the design of the link grabbing system based on binocular stereo vision was completed,and the manipulator guided grabbing of any placed link parts was successfully implemented,and the final result data was analyzed and compared.Confirm that the actual requirements of the project have been met.The experimental results prove that this project can accurately and quickly grasp the target workpiece,and has a certain stability,which meets the requirements of the actual production and manufacturing environment.