Fast Reconstruction And Parameter Calibration Method Of High-precision Point Cloud Data

High precision device surface digitization plays an important role in detection and assembly.According to the precision capacity of the device,the collected point cloud data increases rapidly.The accuracy and speed of 3d reconstruction of point cloud data are key considerations in virtual assembly,and these indicators are closely related to system construction and point cloud reconstruction.In this paper,3D reconstruction and key parameter calculation of industrial spherical parts and mass data are deeply studied from three aspects: 3D reconstruction algorithm,point cloud model parameter detection and 3D system design.Rapid reconstruction of mass data and 3D parameter measurement of model surface are realized.The main work of this paper is as follows:(1)Using the parallel features of GPU,we proposed a method to process the point cloud data in blocks and parallelize,and implement multi-thread parallel reconstruction.Starting from the data relationship of ordered point cloud,a data processing method including point cloud processing and smoothing is designed.The GPU is used to quickly reconstruct the triangulated surface of the block data,and the Laplace algorithm is used to smooth the reconstructed surface.By performing reconstruction model tests on 5 different sets of massive data,compared with the traditional reconstruction method using only CPU,the model reconstruction speed has been improved.(2)The existing three classical parameter calculation methods,including centroid method,Hough transform method and curve fitting,are analyzed,and the RANSAC algorithm is proposed to perform surface fitting on the reconstructed model,and surface fitting on the fitted model.The relative parameters are calibrated,and the height parameter value is calculated,which can be effectively applied to the pre-judgment of device assembly inspection and improve the assembly quality.(3)The design process of the platform is described in detail from the analysis of the functional and non-functional requirements of the system,the overall design of 3D visualization,and the realization of the point cloud 3D visualization platform.The visual operation platform can realize the rapid display of multi-scale zooming of 3D models,automatically and quickly construct three-dimensional digital models of parts,and support functions such as smooth zooming and display interaction of users at different scales.