Research On The Key Technologies Of Point Model Process With Primitive Geometry

Point model adopts discrete point as surface presentation. In comparison with triangle mesh model, Point model has following properties:convenient data acquisition, simple data structure, and no need to maintaining topological structure. In consequence, digital geometry processing in point model has been an emerging research focus in computer graphics.The paper serves point model as reseach object, and serves geometry primitives as basic geometry presentation. It aims to develop efficient and robust methods for detection of geometry primitives and several key applications.The main researches are as follows:1. A RANSAC based detection method has been introduced, discussed and analyzed. A theoretical analysis of the employed local sampling strategy has been given. The local sampling allows robust extraction of geometry primitives with high probability even in large point model. In conjunction with the lazy score evaluation scheme this leads to an efficient and effective algorithm.2. An algorithm for compression of the decomposed point model has been presented. Due to the good approximation quality of the geometry primitives it is possible to efficiently compress displacement maps using image-based techniques.This even allows for fast decompression on the GPU during interactive rendering.3. The segmentation and matching provided by the geometry primitives has been exploited to automatically detect user specified entities in the point model.Detection is reduced to a graph matching problem:Entities are described by their comprising geometry primitives, represented as graph nodes, and their geometrical relations, represented by graph edges. The matching is efficient even on large point model since the number of geometry primitives is much lower than the number of points and invalid matches can be quickly pruned if geometric constraints associated with the edges are not met.4. Completion of unobserved parts of geometry has been approached by extending detected geometry primitives in the empty areas. In order to allow the completion of even very complex holes with possibly multiple boundaries the proposed method minimizes a novel surface energy. The energy prefers surfaces that follow the geometry primitives such that the completed parts are effectively closed by extended geometry primitives. However, in case no suitable geometry primitives exist the method automatically and gracefully resorts to a completion by minimal surfaces. In a word, this paper proposes new ideas and methods to key procedures in efficient point model processing with geometry primitives, and presented new means and techniques for reverse engineering. It indicated that they have wide application in computer aided design and computer graphics.