3-D mesh-classification method based on angular histograms

Abstract

Methods for analyzing of 3-D mesh topology play an important role in computer graphics problems such as segmentation, correspondence, shape recognition, editing, and animation. We present a new approach for recognition, analysis, and classification of 3-D mesh-based objects. Here, we show how histogram descriptors based on the angles between vertices can be used for grouping similar shapes and for shape classification. Our method can be customized to classify objects of varying mesh complexity (i.e., level of detail) by changing the descriptor’s resolution parameter. We demonstrate the effectiveness of our method on a 3-D object database generated using shape grammars and a hybrid morphing algorithm.