På svenska

Presentation Information     2014-03-20 (14:15)   •  The seminar room at Vi2

Speaker Prof. Walter Kropatsch
Comment Vienna University of Technology
Type External presentation
Title Presentation of the Pattern Recognition and Image Processing (PRIP) Group
Abstract The PRIP group is part of the institute of Computer Graphics and Algorithms in the faculty of Informatics of Vienna University of Technology. After shortly presenting the members of the group I will address the current five research projects. The first project is aiming at phenotyping of plants, both above and below the ground. It is a collaboration with Dr. Busch of the Gregor Mendel Institute in Vienna. The project works with the plant Arabidopsis thaliana of which we aim to extract geometrical and topological traits of the upper part and use a Reeb graph to study the structure of the roots. Graph centralities defind the importance of a node in a graph. They are wellknown from social network analysis and have interesting applications in shape matching, MaxCut-based segmentation and for the registration of point clouds. We have shown in the past how to construct pyramids with combinatorial maps. The newly proposed "canonical encoding" allows the encoding of a complete pyramid by simply ordering the darts of the map in the order they have been contracted and removed during the bottom-up construction. It allows the complete reconstruction of all the pyramid levels without any extra data. If the higher levels reflect a higher importance of the map entities it should be suited for efficient top-down matching similar to the combinatorial maps used by Solnon and Damiand who showed recently a polynomial time algorithm for matching (in general it is NP) with a canonical map representation. Tracking related multiple targets in videos was the PhD topic of Nicole Artner. She studied the potential of graph-based representations and methods in the temporal domain. Her method is able to automatically extract a graph-model from a moving target object and to use it to reliably track it under severe occlusions. Finally we helped a group of artists that want to make people realize that "new media changes people's reality towards the unreal shadows of Plato's Cave Allegory". In the exhibitions in Washington DC and later in Vienna a quadcopter takes pictures of participants faces which will be displayed on a video screen with some very personal information to make them aware of the omnipresent surveillance that our society is currently facing. In the future we plan a cooperation with the group of Antoine Vacavant in Le Puy in France on the segmentation and tracking of Liver Lesions. The goal is to avoid ionizing radiation and to use contrast agents and track their propagation wave in the images. This project is still in a planning phase. Together with the Veterinary University of Vienna we plan to study the phenotyping of Lipizzan horses with the aim to extract traits allowing objectivity and repeatability of judgement. The difficulty is that the required landmarks are mostly not the visible features, i.e. the specific properties of the bones. Unfortunately the skeleton of the bones is not the same as the digital skeleton of the horse's shape. After this overview I would like to discuss more informally related research issues.