Object recognition using relational clique and cycle mappings

A procedure is adopted for object recognition that uses the general properties of relations, in particular, their clique and cycle invariance, as a mechanism for pruning the corresponding search trees prior to searching for relational monomorphisms. Relational cliques and cycles represent physical properties of the physical objects are being matched, and hence by judicious choice of the relations utilized, the clique and cycle structures can be extracted concurrently with the features by low-level processing, especially those properties of line segments which can be readily obtained from the Hough space.<>