Describing 3D Geometric Primitives Using the Gaussian Sphere and the Gaussian Accumulator

Abstract

Most complex object models are composed of basic parts or primitives. Being able to decompose a complex 3D model into such basic primitives is an important step in reverse engineering. Even when an algorithm can segment a complex model into its primitives, a description technique is still needed in order to identify the type of each primitive. Most feature extraction methods fail to describe these basic primitives or need a trained classifier on a database of prepared data to perform this identification. In this paper, we propose a method that can describe basic primitives such as planes, cones, cylinders, spheres, and tori as well as partial models of the latter four primitives. To achieve this task, we combine the concept of Gaussian sphere to a new concept introduced in this paper: the Gaussian accumulator. Comparison of the results of our method with other feature extractors reveals that our approach can distinguish all of these primitives from each other including partial models. Our method was also tested on real scanned data with noise and missing areas. The results show that our method is able to distinguish all of these models as well.