3D Reconstruction And Measurement Based On Machine Vision

With the development of modern industry,many metal structural parts with very complex features require 3D physical digital measurement.Efficient,real-time and accurate 3D measurement in the production or assembly process is a difficult problem that modern manufacturing must solve.This paper mainly focuses on the key components such as servo mechanism,antenna in special structure,and scans them in all directions using 3D laser scanning equipment.According to the multi-view point cloud data obtained from scanning,using 3D registration technology to realize the 3D reconstruction of point cloud model.The feature point cloud to be tested is extracted by the point cloud segmentation algorithm,and the key parameters are accurately measured.The major contents of the thesis are shown as below:1.3D point cloud filtering algorithm is studied and analyzed.For the point cloud data obtained by 3D laser scanning,the filtering method combined with large-scale noise filtering and small-scale noise filtering is used to remove noise by step which is introduced during scanning.Finally,carried out the experimental simulation of the combined filtering algorithm using the workpiece point cloud model,and the effectiveness of the filtering algorithm is proved.2.The technology of rough registration and accurate registration in 3D reconstruction of point cloud model is studied.In the rough registration,the rotation registration algorithm is mainly introduced,and used the cone-based calibration method to calibrate the turntable quickly.In the accurate registration,the ICP accurate registration algorithm is mainly studied,and improved the ICP algorithm for multi-view point cloud registration.Finally,the experimental simulation of the improved ICP algorithm is carried out using the multi-view statue point cloud model.It is proved that the improved ICP algorithm effectively improves the registered speed and accuracy.3.The existing 3D point cloud segmentation methods are analyzed,and the region growth segmentation algorithm and the RANSAC algorithm are studied.For the complex point cloud model such as servo mechanism,a set of complex point cloud model segmentation process using region growth algorithm and RANSAC algorithm is proposed.4.According to the measurement requirements of key components in special structure,designed the corresponding measurement plan.Carried out detailed measurement of the servo mechanism and T/R components of antenna,and verified the final measurement results.5.Aiming at the measurement of key components in special structure,an on-line measurement software is designed and developed.The software mainly realizes the online measurement function of the servo mechanism,antenna,and other key components,which can provide real-time and accurate key parameters for the assembly of the special structure.