Deviation Analysis Of Three-dimensional Model Based On Laser Scanning

It is an important issue for three-dimensional computer graphics research to make a quantitative analysis or evaluation of an entity model. The use of three-dimensional laser scanning and computer for quantitative analysis and evaluation of the models, it is not only fast, reliable, more accurate but also fair and objective for the uniform evaluation criteria. So this paper presents a quantitative deviation analysis of three-dimensional model method based on laser scanning.This method makes use of three-dimensional laser scanner to pick the point sets of models, and matches the standard model with the model to be evaluated, to make coordinate system of the two models same; and then, reconstructs continuous surfaces from the point sets, and discrete the continuous surface; and finally, generates offsets of discrete surfaces based on the standard model, so that you can calculate the quantitative deviation of the evaluated model in all directions and levels. In this paper, the main issues to be resolved in derivation analysis of three-dimensional model are: scanning three-dimensional models; registration of models; quantitative representation of three-dimensional model, and so on. To solve the above problems, this paper presents the following methods:1. Scanning models. Laser scanner is fast and accurate, and the scanned point set is very convenient to be represented and processed in the computer. Therefore, three-dimensional laser scanner is suitable as a scanning device, in order to obtain the 3D point set. The point set is saved in STL file.2. Improve the fast ICP algorithm. After the pre-processing of 3D point sets, the point set is mapped into a 2D binary image, and matching the 2D image can get a rough registration result, and then use the fast ICP algorithm proposed by Szymon for accurate alignment.3. Reconstruct continuous curve from point set and discrete the curve. To accurately quantify the deviation between the two models, in addition to the registration of models, but also we need to reconstruct the surface of model, and then discrete surface for accurate calculation. In this paper, the Delaunay triangulation algorithm is used to reconstruct the surface, and the models are discretized with Octree data structure. 4. Generate offsets of the discrete surface, and calculate the deviation with discrete voxels. The algorithm calculates discrete voxels on the envelope surface of discrete model to generate offsets, and it effectively solves topology, self-intersection problems. Voxels are convenient to calculate the deviation of model in the volume or length.