Research On Disk Components Identification And Location Technology Based On 3D Point Cloud

In recent years,China’s traditional manufacturing industry has entered a bottleneck period with the increasing domestic labor costs,deepening social aging and the impact of the epidemic at home and abroad.In view of the long-term goal of the 14 th Five-Year Plan,China’s manufacturing industry will achieve highly automated and intelligent industrial manufacturing in the new round of industrial revolution,which meets the needs of The Times.In this paper,an identification and positioning method for footwall components in unstructured environment is proposed to solve the problem.A robot grasping system based on 3D vision is built to verify the effectiveness of the proposed method.The main research work is as follows:Firstly,the recognition and location system of disordered sorting scene is designed.Through comparative analysis,the selection scheme of the camera and the industrial robot was determined.Aiming at the point cloud acquisition device,the monocular structured light system was designed.The transformation relationship between the camera imaging principle and the relevant coordinate system was studied,and the mathematical model of the system was established.On the basis of component depth image acquisition,point cloud data is recovered by mapping method.Secondly,a point cloud data preprocessing method based on octree and an improved region growing image segmentation method are presented.The redundant background was removed based on the straight-through filtering algorithm,the parts and the platform plane were separated by the random sampling consistency algorithm,and the point cloud space topological relationship was established by the octree method.Based on the characteristics of disk elements,an improved region growing image segmentation algorithm was proposed.The algorithm calculated the curvature by selecting the local radial basis function method,took the minimum point of curvature as the initial seed point,and carried out region growing according to the set spatial threshold range.Experimental results verified the efficiency and accuracy of point cloud segmentation.Thirdly,an identification and location method of disk elements combining point cloud feature description,initial registration and precise registration is proposed.By obtaining the surface normal of point cloud data,the influence area and time complexity of query points in point feature histogram(PFH)and fast point feature histogram(FPFH)are calculated,and the FPFH descriptor that is more consistent with real-time effect is selected.The corresponding relation estimation optimization algorithm is used to eliminate the false matching points and provide an appropriate initial transformation matrix for the subsequent matching process.The iterative nearest point(ICP)algorithm is used for accurate point cloud registration,and then the pose information of point cloud is obtained.The experimental results verify the validity of the above method.Finally,the identification and positioning system of disordered sorting scene disk components is built,the compilation platform based on Visual Studio 2017 configuration PCL library and Open CV library is built,and the MFC plug-in is configured to create humancomputer interaction interface.Eye-to-hand model was selected for hand-eye calibration experiment,and the transformation relationship between robot base coordinate system and camera coordinate system was obtained.The effectiveness of the proposed algorithm is verified by experiments.