Annual Report 2010 - Fachgruppe Informatik an der RWTH Aachen ...

Overview
The Chair of Computer Graphics, Computer Vision, and Multimedia at RWTH Aachen
University has three major focus areas within the field of Visual Computing: Geometry
Processing, Mobile Multimedia, and Computer Vision.
The Geometry Processing Unit addresses research questions in the context of the generation,
reconstruction, optimization, and modification of digital 3D models most often represented by
polygonal meshes. The typical application domains for the algorithms developed in our group
are CAD/CAM in architecture and engineering, simulation sciences, rapid prototyping, as
well as medical applications. Our recent activities in Geometry Processing were concerned
with the automatic generation of high-quality quad meshes and the robust, efficient, and exact
implementation of geometric operators like Minkowski sums and sweep surfaces.
The Mobile Multimedia Unit emerged from our activities in various third party funded
projects, including the excellence cluster UMIC. Here we are investigating fundamental
functionalities that make high performance graphics and vision capabilities useable on mobile
devices. In particular we are looking into ways how to overcome the restrictions implied by
limited resources (e.g. compute performance, memory, transmission bit-rate) and how to
recover and exploit the location context of the mobile user for novel application scenarios.
The Computer Vision Unit is concerned with the reconstruction of faithful 3D models from
images and video as well as with the detection and recognition of objects. For 3D
reconstruction we distinguish different classes of objects, i.e. human faces in motion, static
medium sized 3D objects, and entire urban environments. On the detection, tracking, and
recognition side, real-time performance and the combination of different detection modules
are our major interest. Some topical overlap with the Mobile Multimedia Unit exists in the
area of image-based self-localization, where the location of a mobile user is determined by
feature-based comparison of his view with a large database of street-level photographs.
Our research is applied to basic research in the sense that we develop fundamental algorithms
and apply them to practical real-world problems. This is why our results are on the one hand
published in academic papers at the major international conferences and journals, while on the
other hand they are being used in practical applications by our industrial partners.
A fully integrated Geometry Processing Pipeline is one of the long term goals on our research
agenda. It requires methods for capturing the shape and texture of real objects, turning them
into efficient computer models and enhancing the quality of the raw data with respect to
surface roughness, sharp feature preservation and topological consistency. More sophisticated
algorithms are able to even detect and extract structural information (e.g. in technical objects
composed by shape primitives such as cylinders and spheres) or they turn unstructured
polygon meshes into CAD-type quad meshes where the orientation and alignment of the quad
faces reflects the intrinsic geometric structure of the given surface.
Eventually we also need algorithms for the interactive visualization of the potentially massive
datasets. To achieve this goal we are investigating new techniques to measure three
dimensional data with digital cameras and laser range scanners. Once this data is obtained we
have to integrate measurements from several sources into a single 3D computer model and
thereby remove erroneous and redundant data.
Besides the reconstruction of existing objects, we are also investigating new algorithms and
concepts for the generation, modelling, and manufacturing of new designs. This is supported
by novel algorithms for the efficient, robust and precise computation of geometric operators
like Boolean operations, Offsets, Minkowski sums and sweep surfaces.
273