A new point cloud reconstruction algorithm based-on geometrical features

With the device of reverse engineering development, the point cloud data as a common and important form existed in surface reconstruction especially in non-contacting measurement. The 3D scanner usually is used to measure the point data, though, the row data is so large, scatter and unordered. So, the representation of point cloud and how to change it to surface is a key content in reverse engineering system. This paper provides a new method to describe the point cloud, relative to it; we define the import relationship of the point data according to set theory. Based on the definition of near region of data, the rule of neighbour data can be reduced, and the different method to search neighbour or partition data can get. All of the algorithm, this system choose the KD-tree and improved the search way with the definition of neighbour. In addition, the GeoSurface algorithm is put forward in this paper. In this algorithm, the curvature and normal are used to estimate computer geometrical features. And based on geometrical features, GeoSurface uses iterate triangle to mesh the surface and uses the depth of KDtree to control the parameters algorithm. At last, by using the existing experimental equipment, we verify the GeoSurface algorithm. The experiment adopts two data sets from the Stanford 3D scanning repository; they are "Stanford Buuny" and "Happy Buddha". The experimental results show the GeoSurface is an effective algorithm and achieved good results in running time and quality of surface reconstruction compared to crust and Poisson reconstruction algorithm.