Moment Invariants (MIs) are widely used for object recognition. Most of the moment based object recognition work reported is for 2-Dimensional (2-D) objects only. There have been considerable attempts made in extending the 2-D MIs to 3-D space. However, using moments for 3-D motion parameter estimation is relatively neglected. In this paper we present two iterative schemes for motion estimation of planar objects using moments as features. One, using the Levinberg-Marquardt method, performs better compared with the other. Only the pure rotational case is considered. By using moments, the correspondence problem is completely eliminated. We show from simulation experiments that this method is a feasible one and the error performance is reasonable. As motion of a planar patch is considered, the algorithm estimates both the rotational parameters and the planar coefficients.
Explicit Separable two dimensional Moment Invariants for object recognition