CISIM 2010

Włodzimierz Kasprzak, Associate Professor

Włodzimierz Kasprzak
Prof. Włodzimierz Kasprzak received his Ph.D. in Computer Science in 1987 and the D.Sc. in Automation Control in 2001, both from Warsaw University of Technology, Faculty of Electronic Engineering and Information Technology. He also holds a German Dr.-Ing. in Computer Science - with specialization in Pattern Recognition - obtained in 1996 at the University of Erlangen-Nuremberg.
He joined the Warsaw University of Technology lately in 1997, where he is currently Associate Professor in the Institute of Control and Computation Engineering. His main research interests are in Pattern Recognition, speech analysis and computer vision.
Before 1997 he worked as a researcher in Bavarian Research Center for knowledge-based systems, Erlangen, Germany (1990-1995) and in RIKEN Institute, Wako-shi, Japan (1995-1996). W. Kasprzak has made fundamental contributions to, at that time new emerging technologies, computer vision systems for driver support and Independent Component Analysis in blind signal processing.
Besides research activities has also has experience in commercial and private sector in Poland - working for 2 years as a division manager for the FESTO company and 6 years as a lecturer in the private Warsaw School of Computer Science.
Keynote Title:
Integration of different computational models in a computer vision framework

A general (application independent) computer vision framework is proposed. It follows the methodology of knowledge-base systems – dividing a system into knowledge base and control. We choose procedural semantic networks for object-oriented modeling of the world. It is basically a non-monotonic logical language. Several inference rules are proposed within the knowledge base that allow to create instances of model concepts. In order to activate an inference rule a model-to-image data matching process need to be performed. We view this matching as a solution to constraint satisfaction problem (CSP), supported by Bayesian net-based evaluation of partial variable assignments. A modified incremental search for CSP is designed that allows partial solutions and calls for stochastic inference in order to provide judgments of partial states. Hence the detection of partial occlusion of objects is handled consistently with Bayesian inference over evidence and hidden variables. Our specific application field is the recognition of partially hidden 3-D objects. We apply the framework to image analysis of scenes with objects passed to a machine by a human hand. Different types of objects are modeled in proposed framework.