Fast and Accurate Computation of 3D Charlier Moment Invariants for 3D Image Classification

The problem of 3D digital object invariability is encountered in image processing, especially in pattern classification/recognition. The 3D object should be correctly recognized regardless of its particular position and orientation in the scene. This paper proposes a new method to extract 3D Charlier moment invariants to translation and scaling (3DCMITS). These descriptors are extracted directly from discrete orthogonal Charlier polynomials without using 3D geometric moment invariants. This method is fast and does not require any numerical approximation compared to the indirect method based on 3D geometric moment invariants. The results show the proposed method's effectiveness in terms of speed with an improvement exceeding 99,97%. For validation purposes and as an illustration of the interest of 3DCMITS, this paper offers a classification system for 3D objects based on the proposed 3DCMITS and Support Vector Machine (SVM) classifier. The obtained results are verified with K-Nearest Neighbor (KNN) classifier and other existing works in the literature.