Position, rotation, and scale-invariant recognition of 2-dimensional objects using a gradient coding scheme

Abstract

The two-dimensional representation and recognition of objects, based on external space-domain descriptors, is addressed. A new shape signature is proposed for two-dimensional shape recognition, namely the gradient-perimeter plot. The recognition method is position, rotation, and scale invariance. For the matching process a cost function, based on mean square error, is minimized. It is shown that, as the basis of an alternative method, the angle-perimeter plot can be used to provide a feature set for two-dimensional shapes. This method can be applied to polygonal shapes, or shapes for which polygonal approximations are available. Both methods were successfully applied to a set of standard shapes.<>