Research On Robot Automatic Grasping Technology For Stacking Objects

As an important representative of intelligent manufacturing equipment,industrial robots are widely used in many industrial fields.Industrial robot grasping combined with machine vision is the premise of many operations,which plays an important role in the process automation of manufacturing workshop.Traditional planar visual detection can be used to detect two-dimensional images,but the accuracy of workpiece recognition is low in the case of stacking,which makes it impossible to grasp the target workpiece.Therefore,based on binocular vision system based on surface structured light,research on robot grasping technology under stacked conditions was carried out,experimental platform was built,algorithm design and experimental verification were carried out,and main research contents were as follows:In order to obtain the pose information of the target,an NDT point cloud registration algorithm based on ISS-3DSC algorithm was proposed to reconstruct the target point cloud to obtain the pose information.The algorithm firstly preprocesses the target point cloud collected by the visual system and the template point cloud transformed by the CAD model,such as voxel filtering and RANSAC fitting segmentation,and then builds the description of the 3DSC feature point pair based on ISS key points for the two point clouds.After rough registration of two piece point clouds was completed by SAC-IA algorithm according to ISS-3DSC feature point pair description to obtain the initial transformation relationship,fine registration was completed by NDT algorithm to obtain the final transformation relationship of two pieces,complete three-dimensional reconstruction and obtain the target pose.In order to select the priority of grabbing target and grab point,a priority evaluation criterion based on Euclidean distance was established for grabbing target workpiece,and a cluster of points with small Euclidean distance between the camera image plane and the point cloud was selected as the priority grabbing target.After the three-dimensional reconstruction of the target is completed,the OBB bounding box is established,and the center of the bounding box,namely the point cloud center of mass,is taken as the grasping point,and the positive direction of the Z-axis of the bounding box is taken as the grasping posture of the claw in reverse.For the calibration study of the selected Eye-to-hand hand-eye system,the rapid hand-eye calibration method based on the robot workpiece calibration is selected.The robot workpiece coordinate system is taken as the reference coordinate system,and the binocular structured light vision system coordinate system is established in the position where the workpiece coordinate system coincides,so as to indirectly complete the calibration of the hand-eye system.Finally,the hardware platform of industrial robot grasping is built,the binocular structured light vision system and the robot grasping system are designed respectively,and the robot grasping experiments are carried out in a variety of scenarios.The experimental results show that the robot visual grasping system can meet the requirements and complete the grasping of stacked workpieces with a high success rate,which has a certain engineering application value.