Automatic CAD model reconstruction from measurement data

Abstract

This paper will present a new method for automatic CAD model reconstruction from discrete measurement point cloud. Due to the enormous amount of data from a digitizer, a data reduction algorithm will first be implemented to reduce data on planar or near planar areas. A general quadric surface representation will then be used to fit pre-defined data subsets by a least square algorithm. The evolution of the surface representation is achieved by repetitive applications of crossover and mutation operations that are characteristic of generic algorithms (GAs). When a surface is extracted, it is classified into a specific quadric surface according to a classification table. CAD models are reconstructed by quadric surface intersection evaluation. This method is for solid models composed only of planar and quadric surfaces.