Research On Key Technologies Of High Precision Measurement Of Point Cloud Surface Geometry

With the deepening of computer stereo vision technology research,3D scanning technology has been continuously developed,and its related point cloud data processing technology has gradually become a research hotspot.The point cloud edge is a collection of measurement points used to describe the contour boundary of the target object.It is the basis for understanding and expressing the geometric characteristics of the threedimensional object.The high-precision extraction of the point cloud edge is also a prerequisite for achieving high-precision measurement of the point cloud target object parameters.Starting from the complete extraction and precise positioning of the edge of the point cloud,this paper proposes a point cloud edge extraction algorithm based on the mean value of the normal vector angle and the iteration of the weighted normal.At the same time,a point cloud optimization method based on non-edge structure information is used to optimize the point cloud edge de-redundancy,and finally achieve high-precision measurement based on the precise point cloud edge.Its main work is as follows:(1)In view of the situation that the point cloud normal vector description will fail to locate the edge when extracting the edge when the handover angle is extremely small or extremely large,this article first proposes a new division standard for the area division of the point cloud,and constructs the neighborhood normal vector Feature model,calculate the mean value of the normal vector included angle of each point in different neighborhoods,analyze the change rule of the mean normal vector included angle of the points in different regions,and screen out the edge region and the mixed region as edge candidate points.In this process,the edges can be preliminarily screened but some nonedge points will be mixed.Therefore,candidate points are used to design a weighted projection plane objective function and iteratively determine the edge points,and finally complete extraction and precise positioning of the edge points of the point cloud are realized on various forms of data sets.(2)For the edge points in the point cloud edge extraction process,there is a certain degree of redundancy,while the feature description of the neighborhood type fails at complex junctions and the information of non-edge points is relatively stable.In this paper,combining the prior information of non-edge points,the non-edge points are divided into regions and the surface equations are fitted to each region,and the boundary line equations of the structure a priori are obtained by using the connection relationship between the surfaces to realize the optimization of the edge points of the point cloud.Through experiments on a variety of complex structure point cloud data sets,it is proved that the algorithm can adapt to various complex geometric structure data and provide optimized edge extraction with de-redundancy.After obtaining the optimized edge for de-redundancy,curve fitting of three-dimensional space points is performed to form an edge line,and the corner point of the point cloud is determined and measured by the method of corner point repair.Experiments show that the accurate extraction process of the edge points of the point cloud in this paper can effectively improve the measurement accuracy.(3)In industrial measurement,the geometric parameters of the workpiece are often an important criterion for evaluating whether the manufacturing process is qualified.In the process of extracting contact spots in the wheel-rail contact state,the geometric measurement parameters of the contact spots can indirectly reflect the change of the contact posture,so as to effectively monitor the wheel-rail disease problems.Relying on the research needs of the wheel-rail contact attitude relationship in scientific research projects,the wheel-rail contact data set under different postures is designed and the point cloud data is formed according to the wheel-rail contact scene.The technical route of this paper is used to extract the contact spots and the geometric parameters of the point cloud wheel-rail contact process.