CONCEPT OF TRIANGULAR SPATIAL RELATIONSHIP AND B-TREE FOR PARTIALLY OCCLUDED OBJECT RECOGNITION: AN EFFICIENT AND ROBUST APPROACH

Abstract

In this paper we propose a novel method for recognizing partially occluded objects. The proposed method uses corner points and their spatial relationships perceived through the application of Triangular Spatial Relationship (TSR)1 by considering three successive corner points at a time. The perceived TSR among corner points are used to create a model object data base using B-tree, an efficient multilevel indexing structure. Since the TSR is preserved by the use of quadruples and B-tree does not support multidimensional data (4-dimensional in our case), a distinct and unique key is computed for each distinct quadruple and then the computed keys are stored in B-tree as representatives of the corresponding quadruples. Each key value stored in B-tree is then attached with a list of objects, which have the corresponding quadruple as one of their associated quadruples. During matching, B-tree is accessed through in search of each key of the scene image to retrieve the list of objects associated with that key. The matched sequence is preserved in a two-dimensional matrix called status matrix. The object for which there is a longest substring of key match in the status matrix is declared to be the identified object. Experiments conducted on different sets of objects revealed the superiority of the proposed method over others and also its consistency with human perception.