Research And Development Of Surface Reconstructioin Of Unorganized Points

With the ever-enlarging domain of 3D data gathering and application, data from super computer, earth satellite, all kinds of scanners and various capture devices needs instant treatment. Since the calculating speed, store capability, display technology and web capability enhanced constantly, direct treatment of the data become a possibility. Surface reconstruction technology is to manage 3D information which is recorded in certain ways into data that can be reorganized and treated by computer. And the geometry information contained in 3D data can be restored into graphics therefore make quantitive analysis, display and treatment of the objects more convenient and fastMost of the research in this paper is about surface reconstruction from unorganized points which is one kind of surface reconstruction technology. Normally, we can divide this algorithm into three classes: the first is based on distance field and contour tracing, the second is based on three-dimensional Delaunay triangulation and the third is based on local area incremental algorithm. After went deep into the thought, the computing flow and the reconstruction result of these three kinds of reconstruction algorithm, and did analysis on their time complexity, advantages and disadvantages, I integrate all the excellent thought of the three kinds and bring forward a new surface reconstruction method named surface reconstruction based on bread-first-search.This method brings the breadth-first-search algorithm in graph search area into our surface reconstruction field. From the incremental computing idea, we make full use of the state expanding characteristic of search algorithm. Recurring to octree space division, searching constraint,optimum vertex estimation and the thinking of Delaunay triangulation, we use initialized triangle as searching base and orientation edges as searching elements to reconstruct model surface gradually and symmetrically. The proposed algorithm supports grid simplified, parallel computing and visualization and do not depend much on parameters. In addition, holes and gaps can be filled optionally. Experimental results show that our algorithm is effective, robust and works well for models with arbitrary topology.