Online Three-dimensional Measurement Of Wheel Tread Based On Point Cloud Registration

With the rapid development of high-speed railway in our country,the passenger transport capacity of railway transportation has been improved.In order to ensure the traffic safety and the safety of passengers' lives and property,it is necessary to pay attention to the dimensions of the wheel and the parameters of the tread.Three dimensional measurement technology has been widely used in many fields,such as automatic detection,quality control,machine vision,biomedicine and so on.In this thesis,the type CRH2A of EMUs sample wheel as the research object,the three dimensional measurement technology and point cloud registration technique are applied to the 360°measurement of wheel tread profile.The in-depth research from two aspects of theory and practice,and the research contents are as follows:First,in the aspect of wheel tread cloud acquisition,the principle of 3D measurement technology based on grating projection is studied.Focusing on the phase unwrapping technique,a phase unwrapping algorithm is introduced,which based on quality guided map algorithm and the relative reliability of adjacent pixels as the expansion path,phase is unwrapped by setting region partition and merging.Finally,the simulation and experimental results are compared with the branch-cut method,the quality map guidance algorithm and the least square method,and the mean square error is reduced by about 35%compared with the quality map guidance algorithm,the speed is increased by about 10 times.What's more,the noise diffusion in the path dependent algorithm is effectively suppressed.Secondly,this thesis introduces the basic process and principle of 3D point cloud registration technology,and understands the data registration mechanism of traditional ICP algorithm.The covariance matrix is used to initial registration,and the simulation results are compared with the point cloud feature histogram method(PFH).An improved ICP algorithm is introduced,which takes the point to tangent plane as the objective function,and then uses the kdtree algorithm to accelerate the search and constraint by the consistency of the rigid transformation consistency.The simulation results show that the time is only 0.1 s based on the covariance matrix method,which is more reliable and more suitable for the coarse registration of wheel tread cloud data.The iterative precision of the improved ICP algorithm is about 10%higher than the other algorithms,and the registration time is not more than 3s.Finally,the three-dimensional profile measurement platform of train wheel tread is designed and built.The 360° online measurement of wheel tread profile is realized by using optical three-dimensional measurement technology and point cloud registration technology,the total registration time and average registration precision are 148.48s and 0.0638mm respectively based on the improved ICP algorithm to complete the measurement of the whole wheel,and it can improve the accuracy and efficiency of online detection.