3D Object Categorization Based on Histogram of Distance and Normal Vector Angles on Surface Points

This paper propose a method to 3D models categorization based on geometric features from face and vertex of any 3D model using probabilistic neural network. For 3D model classification, we use histogram of two variables, i.e., the angle between normal vector on the object surface point and vector that connect shape origin to that point; and distance of object surface point to shape origin. Also, for better separability of different models, Euclidean distance histogram for pairs of surface points is used. The most advantage of using histogram to present the features is that it leads to reduce the feature vector dimension and consequently computational cost in classification process. Performance of the proposed method is investigated using McGill database. The final result shows desired classification rate.