Contact point: Alberto Del Bimbo ( ) University of Florence, Italy

Technical Contact Points: Stefano Berretti ( ) and Pietro Pala ( )

Type of Software: PROTOTYPE

Descriptive Keywords: 3D content based retrieval

Potential Use and Applications: 3D models are being created and employed in a wide range of application domains, including medicine, computer aided design and engineering, and cultural heritage. The development of techniques to enable retrieval by content of 3D models assumes an ever increasing relevance in all those area where it is necessary to find 3D objects according to their shape similarity.

General Description: At the Media Integration and Communication Center of the University of Florence, a prototype system for content based retrieval of 3D objects has been developed. The basic idea of 3D content based retrieval, is to allow users to find objects in a large repository by providing to the system an object as example of the the main characteristics the user is looking for. The query object is described according to a set of representative features which capture salient information of the object surface. The goal for the retrieval system is to find 3D objects similar to the query model in terms of the extracted features. In the proposed prototype, the spatial distribution of curvature information extracted from the object surface is used to represented 3D objects by curvature correlograms. This 3D description and retrieval module is under integration into the DELOS DLMS.

Technical Description: The 3D retrieval system, supports feature extraction and content based retrieval of three-dimensional (3D) objects. Functionalities that are currently supported by the system are: § 3D feature extraction – Given a 3D model as a mesh of triangles, this component extracts a feature vector representation of the 3D object. An object is described by a set of feature vectors capturing its main shape characteristics. § Description storage – Given a 3D object descriptor, this module stores it in a local archive of 3D object descriptors. In the current implementation, feature vectors are stored in a relational database (MySQL). § 3D object retrieval – Given an input 3D object, this module match the query model descriptor against the set of reference descriptors stored in the local archive. The match is performed by determining the best correspondence between features vectors representing the query and each of the database models. The module is devised to perform matching between 3D object representations in the format generated by the “3D Feature Extraction” module. The module returns a list of (name of) 3D models and their distances, ranked in order of increasing distance from the query. § Query interface – A user interface is available to query the retrieval system. The interface is developed as a web application using the JSP/servlet technology.

Required User Skills: No particular skills are required in using the prototype. Just a normal level of practice in interacting with software applications is needed.

Pre-Requisites for Installation: Two 3D retrieval applications are available. The first one is based on a web interface which allows remote querying of an archive of 3D objects. This application is completely written in Java so, in principle, it is platform independent. The requirement to run the server side part of the system is the installation of the Java platform. Running the client side only requires a web browser. The second prototype is a standalone application written in C++ developed for the Windows operating system.

Conditions of Use: Contact the authors.