3d Reconstruction And Positioning Capture Of Heterogeneous Components Based On Structured Light

Aiming at the problem of how to achieve precise grasping in complex scenes such as weak textures and reflections,and how to achieve precise grasping to meet the requirements of the next robot welding processing,this paper uses dual The key technology of three-dimensional information of different components and the pose recognition of different components are studied by the eye structured light system.The main research contents include the following three points:(1)Research on construction and 3D reconstruction of structured light systemAiming at the shortcomings of traditional binocular vision that is easily affected by ambient light,a set of binocular structured light system is built to complete hardware,software design,camera calibration,etc.The multi-frequency heterodyne method is selected as the method for dissolving the projected pattern of the structured light,and the 3D reconstruction of the binocular structured light is studied to complete the 3D reconstruction of the dissimilar components.Aiming at some abnormal phase jump points that cannot be removed by phase correction,an improved monotonic method is proposed.On the basis of the original method,the slope fitting of the phase of the entire line is added as a constraint to restore the phase that cannot be corrected normally.Finally,the three-dimensional reconstruction of different components is completed according to the matching principle of binocular vision.(2)Pose estimation based on point cloudFor a large number of redundant points and noise points around different components after3 D reconstruction,the point cloud is preprocessed,and for the existence of multiple heterogeneous components in the same scene,Euclidean clustering is used to achieve segmentation.The ISS algorithm is used to complete the extraction of key points on the grasping plane of different components,which improves the efficiency of rough and fine registration of point clouds.Finally,the target pose is obtained through the rotation and translation transformation of the coordinate system,and the pose estimation of heterogeneous components is realized.(3)Research on positioning and graspingThe principle of hand eye calibration is studied.The eye-to-hand method is used to carry out calibration experiments,and the conversion matrix between the camera coordinate system and the robot coordinate system is solved to prepare for the grasp work.Both point cloud registration and hand-eye calibration will calculate their respective rotation and translation matrices.Combined with the two matrices,the position and orientation information of different components in the workpiece coordinate system can be solved and written into the robot control cabinet,so that the robot can move to the specified position for calculation to achieve positioning and grasping.Taking a variety of heterogeneous components as the research object,the camera calibration accuracy and hand eye calibration accuracy are verified by using the calibration board.The pixel errors of left and right camera calibration are 0.05 and 0.04,and the rotation root mean square error of hand-eye calibration is 0.15 °.Aiming at the problem of reconstruction accuracy,three-dimensional reconstruction of the standard blocks stacked on top of each other and the calculation of the average distance between the two point clouds on the surface of the standard block are carried out to verify the accuracy of the experiments.The results show that the accuracy can reach 0.04 mm.Grasp different components with different poses,and the experimental results show that the grasping success rate reaches 98%,which verifies the effectiveness of the target grasping.