Description, Matching and Retrieval by Content of 3D Objects

In this work, we report on three research results achieved during the first three years of activities carried out under the task 3.8 of the DELOS Network of Excellence. First, two approaches for 3D objects description and matching for the purpose of 3D objects retrieval have been defined. An approach based on curvature correlograms is used to globally represent and compare 3D objects according to the local similarity of their surface curvature. Differently, a view based approach using spin image signatures is used to capture local and global information of 3D models by using a large number of views of the object which are then grouped according to their similarities. These approaches have been integrated in task prototypes and are now under integration into the DELOS DLMS. To open the way to 3D objects retrieval based on similarity of object parts, a method for the automatic decomposition of 3D objects has been defined. This approach exploits Reeb-graphs in order to capture topological information identifying the main protrusions of a 3D object.