Research On Point Clound And Grid Model Construction And Shape Distribution Retrieval

With the develop of 3D data acquisition technology such as 3D laser scanner, the study of 3D modeling technology is more and more interested.3D model has varietys of representation ways, especially,3D model based on point cloud and grid is widely used in computer-aided design, virtual reality, visualization and other fields.While, the 3D model feature extraction is basic problems and key technologies which to solve 3D retrieve technology,and the 3D model feature extraction have a direct impact on 3D model retrieval results directly.This thesis focuses on point cloud and mesh of 3D models, which inculde the representation and modeling technology, the grid model of 3D model feature extraction. The point cloud simplification and shape distribution algorithm is improved in this paper.This paper mainly include follow aspects:(1)Construction of point cloud and grid model. First,we import the data of sampling points of real objects surface by using 3D laser scanner, and put point cloud data preprocess, such as data registration, noise removal, repair pretreatment and so on; Laterly,it achieve point cloud model base on the representation point cloud; Then, it construct the grid model based on point cloud data which reconstruction the surface mesh model, and simplified the size of grid model in order to reduce the memory.(2)We propose a new simplification algorithm for point cloud models,which based on K neighborhood density simplification algorithm. First the algorithm judge the distance about point of the K neighborhood to the center, if the distance larger than the average distance,the point is retained. Otherwise, vector error method is used to measure whether delete the point. This method pre-control the error which is suitable for non-uniform point cloud data, especially it is not easy to change the shape of the model in large parts of the curvature change. Experiment shows that this method can keep the details of model's information when is simplified to in the curvature of the model.(3)A new shape distribution method was proposed it is based on enhancing 3D model detail. First, the preprocessed model was segmented into a set of blocks, Second, these blocks were processed to feature extraction separately and the feature histograms were constructed for each block.Finally, according to calculating the similarity between blocks, the similarity of 3D models was achieved. This method is more accurately to compute the similarity between models than shape distribution. And the distinguishment of details is increased. It resolves the problem of the similar shape but dissimilar detail.(4)Besides theory research and the practical needs of 3D modeling,we bulit a point cloud and grid of 3D model construction system, by Visual C++ 6.0 and OpenGL 3D graphics library for the establishment.This system achieved many functions,such as importing data, preprocessing point cloud data, displaying 3D model, reconstructioning point data, simplifying mesh of 3D model, feature extraction functions based on mesh of 3D model and so on, which have simple operation and built better model.This paper is supported by National High Technology Research and Development Program(863 Program No.2008AA01Z301) of China.