Using a spatio-temporal reasoning system to improve object models on the fly
Martin Antenreiter, Johann Prankl, Markus Vincze and Peter Auer
33rd Workshop of the Austrian Association for Pattern Recognition - Visual Learning
Österreichische Computer Gesellschaft
, Vienna, Austria
We present a system, which is able to track multiple objects under partial and total occlusion. The reasoning system builds up a graph based spatio-temporal representation of object hypotheses and thus is able to explain the scene even if objects are totally occluded. Furthermore it triggers learning new appearances in case of plausible object hypotheses. We represent objects in a star-shaped geometrical model of local descriptors using a codebook. The novelty of our system is to combine a spatio-temporal reasoning system and a local descriptor based object detector for on-line improving object models in terms of adding new, and deleting unreliable interest points. We propose this system for learning object models on the fly, to keep them smart and manageable.