Research On Data Processing Of Point Cloud Based On Curvature Feature

3D optical scanner plays an important role in digital design. With the popularity and application of3D optical scanner, the processing of scanning measurement data is more and more important. Point cloud data fairing and reduction are vital elements of scanning measurement data processing, which have been the popular and difficult points in the study of discrete data reverse reconstruction.The thesis analyzes all kinds of errors contained in3D scanning measurement data, which include the random errors caused by the vibration of Semiconductor laser, CCD camera, the image data acquisition card, thermal noise of electron components like computer motherboard and mechanical platform. The errors, caused by width of laser line, resolution ratio of CCD camera and the image data acquisition board, assembly accuracy of mechanical platform and roughness, color, texture of the measured object surface, are included as well. This thesis makes a literature review of point cloud denoising and reduction abroad and at home and mainly explores the method of point cloud denoising and reduction based on the curvature information.The thesis regards the calculation formula of differential geometric curvature as theoretical foundation, systematically introduces the present estimation method of discrete point cloud curvature. Nira Dyn’s and Mark Mayer’s estimation methods of discrete point cloud curvature are studied in detail. By comparing the curvature calculation results of spherical, cylindrical, parabolic, saddle surface of two algorithms, the thesis improves Mark Mayer algorithm on the basis of Voronoi area. The algorithm is applied in spherical, cylindrical, parabolic, saddle surface. The calculation results shows that this algorithm enhances the estimation accuracy and stability of discrete point cloud curvature.The improved estimation method of Mark Mayer’s discrete point cloud curvature is exploited in the curvature calculation of practical scanning measurement data. The noise, concave, convex, data points of flat area and curvature feature distribution of edge contour points and height information and normal vector information are all analyzed. The influences on object shape by height, normal vector and curvature are summarized. The fairing Algorithm for point cloud data based on curvature feature information is designed to be used for fairing processing of practical data. The results indicates that this algorithm can recognize the edge contour and the characteristics of the shape well and then reach the goal of denoising on the premise of keeping shape and features. The statistical results of the curvature features before and after point cloud data fairing processing present that the sudden change of curvature caused by noise data is effectively inhibited. Except edge contour, curvature of point cloud data transmits smoothly, which not only increases the quality of object surface, but also benefits the follow-up processing. With the help of improved estimation method of Mark Mayer’s discrete point cloud curvature, the Mean curvature of each point of practical measurement data is calculated. Then they are sorted and statically analyzed. The point cloud data reduction algorithm based on curvature feature information is designed. According to the requirement of streamline points, this algorithm can calculate the distribution density of point cloud. The priority is given to the reduction of data point of small curvature in flat area, which keeps edge contour points well, and decreases influence on contour details along with data reduction.The research is implemented on the basis of3D laser scanning measurement software independently improved by our labs. It penetrates into two links, namely denoising and reduction of scanning measurement data processing. Some achievements have been made on estimation of discrete point cloud curvature and point cloud data denoising and reduction based on curvature feature information, which reinforces subsequent data processing function of scanning measurement software.