Study Of Inspection Technologies Based On Point Cloud For Complex Surfaces And Its System Development

With the development of product's aesthetic feeling and customer's individuation, more and more free-form surfaces and their combination, complex surfaces, are used in the product design of automobile, home appliances, commodity. This made the product design diversification and made industry design, reverse engineering, 3D CAD technology being an efficient design technique. It brings the digital manufacture and digital control at the same time. Advanced digital control tool equipment degrades the difficulty of manufacture. How to inspect free-from surfaces of product and control the quality of machining is becoming more important. It is a problem not to be solved completely. The inspection system of complex surfaces is the most efficiency way to solve this problem based on point cloud. The main steps include point cloud's preprocessing, registration between point cloud and CAD model, assignment of point cloud to CAD model's surfaces, deviation between point cloud and CAD model, visualization of the result of deviation. This dissertation study digital inspection system and key technology of complex surfaces parts based on point cloud. The main tops of research and creative point in the dissertation are as follows:The dilemma of removing outliers and keeping feature information in point clouds is being analyzed strictly. The principle reason is found when filter method is used to smooth point cloud. A math model of heat transfer is built and proved that they have the same essence in math principle. So a methodology for smoothing of point cloud data that based on anisotropy heat conduction theory is proposed and penalty function of the method is also analyzed. The method successfully resolves the problem of removing outlier and keeping feature information. Some example is used to prove the rightness and effectiveness of the algorithm.It is necessary to register point cloud to 3D CAD model in free-from surface inspection. The efficiency and the establishment of corresponding points between point cloud and 3D CAD model of registration method is the key problem. A new method that based on extend Gaussian sphere is used to define the corresponding point of coarse registration point cloud to 3D CAD model. In order to solve the efficiency of the external feature registration method such as iterative closest point, firstly the complex surface of 3D CAD model is scattered as an aim lead point cloud, then using the searching method to get the closest point in point cloud replace the optimization method between the point cloud and NURBS. It is clear that the optimization always cost more time. Some cases are done to prove the method's correction.