Efficient object recognition method based on hierarchical representation

It is a very challenging problem to make robot vision autonomously identify and recognize objects in the real world, and the recent research in this field has been moving forward mostly through designing smart shape descriptors for providing better similarity measure. In this paper, we propose a novel shape descriptor called hierarchical representation to describe the object shape efficiently and perfectly. Firstly, the contour can be divided into several segments according to corners and the distribution of local curvature. Secondly, we evaluate the importance of each contour segment by hierarchical description, and then the multi-level contour segment set combined algorithm is carried out to combine the useless and redundant contour segments. Finally, a set of contour feature segments, completely representing local features of objects, are obtained. The experimental results of MPEG-7 database indicate that this algorithm has great advantage over recently published algorithms, especially for the objects with partial occlusion.