The Surface Reconstruction Algorithm Based On RBF

With the rapid development of scanning devices, especially non-contact measuring devices, people can get vast Data Points instantly. Whether it can truly reappear model's outline or not is the most important criteria to measure surface reconstruction. The more data points, the more details can be reappeared. However, overmuch data points always lead to increased computation and make it impossible for surface reconstruction on small and medium size computers. In this connection, based on the thought of divide and conquer, this article have studied related theory of surface reconstruction. The major works involved are as follows:First of all, I make an overall narrative and review of surface reconstruction. This article mainly introduces reconstruction methods that based on messy point set, voxel, neural networks and triangle mesh. Meanwhile, this article have given the evaluation criterion of surface reconstructionSecondly, an improved the multi-level geometrical image presentation of point cloud surface was presented. Aiming at the defects of multi-level geometrical image representation in the spherical coordinates, this article put forward to represent it in cylindrical coordinates. In this way it can precisely represent complex point cloud surface and make it easier to compressing, editing, simplifying, model retrieval and progressive transmission.At last, this article improved the surface reconstruction algorithm based on radical basis function. Aiming at data dimension which cannot reconstruct on small and medium size computers, this article improved the surface reconstruction algorithm of radical basis function based on the bounding box, all style bounding boxes were introduced in this article as well. Compared with the advantages and disadvantages, interpolation theory of radical basis function in detail were introduced, this can rationalize the choice of bounding box. Meanwhile, experiments showed that improved bounding box can improve the reconstruction effect obviously.This thesis is supported by grant No.10YSY02 from Northwest University Graduate Innovative Education Project.