Modelling of rigid objects by bidimensional moments. Applications to the estimation of 3D rotations

In this paper the moment method has been extended to characterize an object moving in three-dimensional (3D) space. The model evolved is a function of the photometrical characteristics and variations of the projected 2D moments of the object in relation to its various positions. Here, the object is characterized by a set of appropriate moments describing it in relation to image translation and rotation. By using five separate camera images of this object in various positions, the zero-, first- and second-order reference moments could be determined. Experimental results applied to the determination of parameters of a 3D rotation have confirmed the validity of such a method.