Efficient 3D object classification by using direct Krawtchouk moment invariants

In this paper, we present an efficient set of moment invariants, named Direct Krawtchouk Moment Invariants (DKMI), for 3D objects recognition. This new set of invariants can be directly derived from the Krawtchouk moments, based on algebraic properties of Krawtchouk polynomials. The proposed computation approach is effectively compared with the classical method, which rely on the indirect computation of moment invariants by using the corresponding geometric moment invariants. Several experiments are carried out so as to evaluate the performance of the newly introduced invariants. Invariability property and noise robustness are firstly investigated. Secondly, the numerical stability is discussed. Then, the performance of the proposed moment invariants as pattern features for 3D object classification is compared with the existing Geometric, Krawtchouk, Tchebichef and Hahn Moment Invariants. Finally, a comparative analysis of computational time of these moment invariants is illustrated. The obtained results demonstrate the efficiency and the superiority of the proposed method.