Multiple Images, Hypotheses, and Wavelengths, Image Analysis in Space, Time, and Energy
Christian Wöhler
TU Dortmund/Germany

Abstract:

This talk gives an overview of the research activities in the Image Analysis Group at TU Dortmund. In the first part, approaches to image-based 3D reconstruction of dynamic scenes and methods for the cognitive interpretation of the reconstruction results in terms of motion trajectories, actions, intentions, and situations are presented. Applications in the fields of driver assistance systems and safe human-robot interaction are discussed. The second part describes a framework for the 3D reconstruction of surfaces based on photometric and also polarimetric information, focusing on natural materials with non-Lambertian reflectance properties, where variations of the observed surface brightness due to 3D shape are separated from those due to non-uniform albedo. These techniques are applied to industrial quality inspection tasks as well as to the construction of digital elevation maps in the context of lunar remote sensing. In the remote sensing scenario, it is shown how photometric surface reconstruction is favourably combined with the analysis of hyperspectral imagery in order to calibrate and normalise spectral data and compensate topographic effects. Results are shown for recent hyperspectral orbital imagery of the Moon.

back to the list of talks