Research On The Generation And Segmentation Method Of Workpiece Point Cloud Based On Wrapped Phase Matching

Coded structured light 3D point cloud generation technology is an important method to realize high-precision measurement of objects in space.It has a wide range of applications in many fields,such as machine vision,reverse engineering and medical imaging.However,the existing structured light point cloud generation technologies generally have shortcomings such as poor anti-disturbance,long time-consuming and low accuracy.In view of the above deficiencies,this thesis compensates the error of the traditional structured light system based on phase shift codec,aiming to generate the 3D model of independent workpiece point cloud with repeated or weak morphological features efficiently and accurately.The following research contents are mainly carried out in this thesis:Firstly,this thesis analyzes and studies the binocular imaging system.Aided by the plane calibration board,the binocular system is calibrated and the parameters required to calculate the three-dimensional coordinates are obtained.In addition,the horizontal alignment correction and irrelevant background region filtering of binocular image are realized to improve the efficiency and accuracy of matching search.Aiming at the causes and solutions of errors,this thesis conducts a systematic and in-depth study.For effectively avoiding jumping cumulative error,a stereo matching method based on variable multiple frequency wrapped phase is designed.Faced with the problem that the surface features of the workpiece are not significant,the matching search range is reduced and the reliable candidate points are retained after multiple screening.According to the matching constraints,the weighted window template and jump point filtering criterion are constructed to compensate the phase errors.For the disparity map generated by matching,the disparity fitting method is used to get the surface which is more close to the real information of the workpiece.During the point cloud post-processing phase,most of the error point clouds is filtered out,and the point cloud is simplified to further improve the rapidity of the system.In order to obtain independent point cloud models of different workpieces,the depth threshold is reliably solved by combining the global iterative threshold method and the maximum inter-class variance method.The complete segmentation and extraction of the target workpiece point cloud is achieved through the depth histogram and the three views of the point cloud.The experimental results show that the proposed research on point cloud generation method of workpiece based on coded structured light binocular vision has certain efficiency advantages over the traditional methods,and can obtain the independent point cloud of simple feature workpiece with less error and faster.