Model-Based Pose Proposal for 2-D Object Recognition

TL;DR: This work considers the problem of finding a known two- dimensional object in an image, or verifying that it does not appear in the image, by adopting the strategy of doing a fast scan for potential places in theimage where the object could be, and proposes a scan pose proposal, which works by finding U- shaped segments of object boundaries.

Abstract: We consider the problem of finding a known two- dimensional object in an image, or verifying that it does not appear in the image. We adopt the strategy of doing a fast scan for potential places in the image where the object could be; we call this scan pose proposal. Each pose hypothesis is a set of edges that correspond to a subset of the transformed object boundary. Our algorithm works by finding U- shaped segments of object boundaries, doing a quick match process between U- shaped segments in the image and the model, and combining the matches into overall pose hypotheses. Analysis and experiments show that the algorithm runs efficiently, and does a good job of discarding all but a few spots in the image as possible pose hypotheses.