Feature correspondence and localization of 3D objects using range images

In this paper, we have described an approach for segmentation, recognition and pose estimation of 3D objects using range images. The surfaces are represented using second degree implicit polynomials from which we obtain view-invariant feature attributes and orientation of the patch. For recognition we use of invariant geometric constraints between the different surface patches, which helps in reducing the search space. We have also described an approach to obtain Euclidean invariants and orientation of surfaces described using higher degree implicit polynomials.