Dec. 19 2005 2:00PM

Dec. 19 2005 2:30PM

Computer vision and computer graphics both deal with the visual appearance of real world objects. A major issue of computer vision is the recognition of objects from images. A means to this end are internal object representations. An important problem of computer graphics is the generation of computational representations of objects from images, for example for the purpose of creative further processing (geometric modelling) or graphical rendering. The current situation is widely characterized by a separation of the acquisition of object data from their subsequent processing. This often leads to the fact that the recorded data do not meet the demands of further processing. An approach to tackle this problem is the active visual acquisition of objects. This means that the processing part of an active vision system exerts influence on the acquisition part. For this purpose appropriate methods
have to be developed. To augment the efficiency of the discovery of suitable methods the system can learn acquisition techniques dynamically depending on the aim of an application (analogously to a human learning
paradigm). In this talk I introduce existent methods for object acquisition and propose a dynamic approach which aims in the development of an intelligent, vison-based 3d-scanner which learns autonomously strategies for object acquisition.

International Workshop on Data - Algorithms - Decision Making (Contact person: Martin Janžura)