Data-driven indexed hypotheses for distorted shape matching

The problem of identification of multiple flat objects in a cluttered environment is dealt with. The objects become distorted due to occlusion, noise and other spurious effects. Unlike the previous model-based approaches involving exhaustive searching, this paper utilizes a data-driven indexing mechanism for model retrieval. Groups of consecutive segments, called super segments, are used as boundary descriptors. The whole approach is based on attributed string matching. With this technique model library can be modified as per requirement