Research On 3d Reconstruction System And Algorithm Of Heterogeneous Workpieces Line Laser Scanning

The contour and surface features of heterogeneous workpieces are relatively complex,and affected by the casting process,the individual differences of rough embryos are large.If the robot processes heterogeneous workpieces according to the unified teaching path will lead to problems such as grinding empty stroke,excessive removal and so on,it also accelerates the wear of abrasive belt and seriously affects the processing efficiency.Therefore,it is necessary to carry out 3D measurement on heterogeneous workpieces to obtain the contour information,so as to provide basic conditions for the research of Robot Adaptive grinding.At present,most3 D measurement research of heterogeneous workpieces can only detect a certain surface feature of the workpiece,or complete the splicing of images from different perspectives with the help of a rotating platform,pasting marked points and other methods,which has great limitations.In view of the above problems,this paper proposes a method of 3D measurement of heterogeneous workpieces by the cooperation of robot and line laser scanner.Firstly,the hardware platform is built to collect the surface contour information of the workpiece,and then the research is focused on the problems of point cloud data processing,registration and reconstruction,to complete the restoration of the 3D model of heterogeneous workpieces.The main research contents are as follows:(1)According to the detection requirements of heterogeneous workpieces,the line laser scanner is selected to build a 3D measurement system for the scanning equipment,and the ceramic standard ball is used as the calibration target to calibrate the hand-eye of the robot in the Eye-in-Hand mode,which is used to solve the conversion matrix between the line laser scanner coordinate system and the robot end flange coordinate system,and finally convert the2 D information collected by the line laser scanner is into 3D point cloud data.(2)The spatial topological structure of the point cloud is established to realize the fast index of the nearest neighbor.On this basis,the noise points of the point cloud are eliminated.Then,the collected multi view point cloud images are registered based on the FPFH feature histogram,and the coarse registration position is used as the initial correspondence for the fine registration of the point cloud images based on the ICP algorithm,so as to obtain the complete point cloud model of the workpiece.(3)The greedy projection algorithm based on Delaunay triangulation principle reconstructs the contour surface of the spliced 3D point cloud model.Aiming at the problems of surface cavity and large roughness in the registration overlapping area of the reconstruction model,based on the KD-tree duplicate point deletion method,the duplicate redundant points are deleted by calculating the distance between the corresponding points,so as to improve the quality of surface reconstruction.(4)Taking the blade as the measurement experiment object.Two ceramic standard spheres with different radii are used to solve and verify the robot hand eye calibration conversion matrix,and the calibration root mean square error is 0.036 mm.The multi view point cloud images are spliced,and the registration deviation is calculated.The registration method in this paper is compared with the single ICP fine registration and NDT + ICP registration,and the accuracy is improved by 71.9% and 31.5% respectively.Finally,the 3D surface features of the point cloud model are reconstructed,and the profile size information of the reconstructed model is extracted.Compared with the actual workpiece size,the size deviation of the model is less than 0.15 mm.