Region based on object recognition in 3D scenes

In this paper, we present a novel object matching approach using the method considering both the similarity on regions and structure in its feature space. The previous works [1], [2] and [3] show that it's possible to formulate the object matching problem as a linear programming problem. However, it remains an open problem how to better use the feature similarity and structure similarity at a same time. In our approach, the contour of the regions of the objects as well as the local structures of these regions in its feature space are considered as two parts of the object matching problem. Hongsheng, Li et al proposed an efficient way to solve these object matching problems by reducing the problem to a linear programming problem.