Moment Methods for the Identification of Three Dimensional Objects from Optical Images

The approach presented here makes use of the theory of two- dimensional moment invariants for planar geometric figures developed by Ming-kue Hu. Complete systems of moment invariants under translation, similitude and orthogonal transformations are derived. By carefully utilizing these properties, a sample set is constructed in which each sample is represented by a vector which characterizes the image for a certain orientation of some object from the given group. A pattern recognition technique is then described in which a parametric representation of the input signal is employed. The decision process using typical samples partitions the space into regions that envelop the chosen samples of a class. A simulation program based on the above outline is successfully developed which not only identifies objects, but also determines their orientation and position in space.