3D object search engine and exploitation to e-business (2006: PAVET 2005 (Subcontract))

The aim of the altaB23D project is to research and develop the basic building blocks (methods, algorithms) for the creation of the first 3D object search engine. The engine will be based on the content, the context of the 3D objects and on the user’s subjectivity concerning “what s/he thinks is similar”. Moreover, traditional methods such as the name of the object and/or keywords will also be integrated.

All the Internet users are familiar with the wide-known text-based search engines (Yahoo, Google, etc.). By providing some keywords, the search engine is capable of retrieving the websites which contain these keywords, within a few seconds. Concerning the 3D objects, such a procedure is not adequate since the object owners often use names irrelevant to the object’s functionality. Further, concerning the present e-commerce applications, the companies use only simple 2D images of their products, thus, no significant visual details of them can be provided to the potential customers.

From the aforementioned cases it becomes clear that in order to create a reliable 3D object search engine the main algorithms should be effectively combine information related to objects’ geometry (content), context (knowledge) and user’s subjectivity.

oncerning the “context”, the main research topics which are expected to be expanded during the project’s lifecycle include:

• 3D object annotation,
• annotation propagation to every 3D object of the database, and
• relevance feedback in the sense of user’s subjectivity.

In this way the human factor is expected to be an active member of the whole process and not be just “a user”.