Workpiece Surface Contour Detection Based On 3D Point Cloud And Robort Grinding Trajectory Planning

With the transformation and upgrading of the manufacturing industry,intelligent manufacturing is gradually becoming the pillar force of the transformation and upgrading of the manufacturing industry with its characteristics of high intelligence,manufacturing personalization and integration.The development of intelligent manufacturing technology promotes the combination of machine vision technology and robot processing technology.Robot machining technology based on traditional teaching method realizes automation to a certain extent,but it is not competent when machining a variety of parts with complex surfaces.The real-time detection technology based on machine vision provides a solution to this problem.Taking the grinding robot as an example,firstly,the contour position information to be machined on the workpiece is extracted by visual detection technology,and then the detected contour information is transformed into trajectory points for the grinding robot to track,which can guide the robot to grind the workpiece contour along the preset path.Because machine vision detection technology has the characteristics of high precision and high convenience,combining it with robot processing technology can complete the complex workpiece processing task with high precision and high efficiency.In this paper,the contour detection technology of workpiece complex surface based on three-dimensional point cloud is deeply studied,the contour detection system for workpiece with complex surface is developed,and the trajectory planning in robot grinding is studied.The main research contents of this paper are as follows:Firstly,the principle of 3D measurement of structured light projection is studied,and a measurement platform is built to collect the three-dimensional point cloud information on the workpiece surface,and the quality evaluation method for point cloud data collected is introduced.Based on establishing the topological relationship between points,a variety of filtering algorithms are used to filter and denoise the original point cloud and remove the useless redundant point information.Starting from the neighborhood point set density of sampling points and the point product between sampling points and normal vector,the down sampling method based on traditional voxel filtering is improved to realize the adaptive compression of point cloud data.The experimental results show that through the construction of three-dimensional measurement system and the function of point cloud preprocessing algorithm,the panoramic point cloud data with high quality and integrity can be obtained.Secondly,to express the panoramic features of the measured workpiece,this paper proposes a hierarchical point cloud registration algorithm,which converts single point cloud data in different coordinate systems to a coordinate system through rigid body transformation matrix.From the point clouds to be registered,the key points of the Internal Shape Signature(ISS)which can represent the characteristics of the global point cloud are extracted and described.On this basis,rough registration is completed.The traditional Iterative Closest Point(ICP)registration algorithm is improved from three aspects: corresponding point search,error point pair removal and reasonable selection of error function,and the accurate registration between point clouds is completed.To restore the surface features of the measured workpiece,two surface reconstruction algorithms,Greedy Projection Triangulation algorithm and Poisson Reconstruction algorithm,are studied in detail,and the original surface features of the measured workpiece are restored.Thirdly,to accurately extract the surface contour of the measured workpiece,a hierarchical point cloud boundary extraction algorithm based on multi threshold is proposed in this paper.The neighborhood centroid deviation of sampling points and the sum of neighborhood point vectors are fused to roughly extract the characteristic boundary points with high computational efficiency.Then,the characteristic boundary points are determined based on the maximum included angle threshold between neighborhood point vectors on the micro tangent plane,and the accurate extraction of boundary point cloud is realized.The extracted workpiece contour points are smoothed and sorted,and they are derived as path points for grinding robot system to recognize.The kinematics modeling and analysis of the six degree of freedom grinding robot are carried out,and the forward and inverse solutions are completed.Then,the robot trajectory planning is deeply studied in joint space and Cartesian space respectively,so as to ensure that the robot tracks the contour trajectory points of the workpiece with the correct orientation.Finally,taking the surface contour detection of complex workpiece as the research object,a point cloud processing and detection system is developed based on Visual Studio2013 platform,QT framework and open source Point Cloud Library(PCL)in Windows10.The point cloud processing algorithm proposed in this paper is integrated into the detection system and verified by instantiation.The contour points of the workpiece surface extracted by the detection system are imported into the grinding robot controller to verify the tracking effect of the robot end on the path points.