Abstract
This paper describes a new procedure for merging several partial models of a given 3D object with the purpose of reconstructing the complete object model. Just 3D models and no 3D data are fused in this approach. The method uses previously defined partial modeling structures, in the context of spherical representations. Special attention is paid to show how these structures are combined, and what problems arise when reconstructing the full object. The work points out how to detect overlapping nodes and how to put them into correspondence and then to fuse them. Occluded views and how they are modeled and coupled with the rest of the model is also discussed. The goodness of the method is demonstrated with experimental results. Partial views range data of a sample object, their corresponding partial models, and the results of fusion of these submodels with the closing object submodel are shown throughout the paper.