Study On Feature Mining Technology Based On Point Cloud In Reverse Engineering

Parametric reverse modeling technology is the focus in the field of reverse engineering. In recent years, large amount of points, acquired by non-contact measuring machine, make the reverse modeling become an inefficient procedure. To solve the problem, feature mining technology is proposed and deep researched in the thesis.The basic theories including the definition, classification and representation of geometric feature are presented firstly. Based on the knowledge of feature, feature mining technology is defined. By researching the method of mining feature, three key steps: data preprocess, segmentation and feature analysis are taken as the main research subjects.Point cloud, composed by dense and error-filled measured points, affects the efficiency and quality of CAD model reconstruction. To solve the problems, a new three parameters Shepard surface is proposed. The surface can be built from point cloud without complicated procedure such as boundary definition, parametrization, surface approximation and etc. Based on the surface, measured points can be cleaned and analyzed effectively. For example, outliers in point cloud can be removed while preserving the feature; Curvature of each measured points can be effective estimated without constructing thousands of local surface patches.An automatic segmentation algorithm based on geometric attribute analysis is proposed. The algorithm subdivides point cloud into cubic grids and then maps the geometric attribute value of points in each grid to normal curvature coordinate system and Gaussian sphere. By testing hypothesis, the patterns of the normal curvature image and the Gaussian image are recognized. Based on grid structure, the image points clustering, and goodness-of-fit testing, point cloud is segmented into several regions and characterized as natural quadrics, extruded surfaces or ruled surfaces, respectively. Applications show that the proposed algorithm deals with large amount of measured points stably and effectively. It can be applied to many other fields including visual reality and computer vision.Deviation estimation between point cloud and given CAD model plays an important role in the reverse engineering. In this paper, a fast estimation method is proposed. The method samples surfaces of the model firstly; based on the sampled points, a projection of each measured point is computed; using the projections, distance from points to surfaces is calculated. Error of the distance can be kept within...