No abstract

No abstract

Object recognition in natural surroundings is one of the most ambitious goals of digital image processing. In our work, strategies for the analysis of traffic scenes are investigated. The Hierarchical Structure Code (HSC) is applied to this problem and it has been shown that fast object recognition is possible. The HSC provides a unified approach for contour-based and region-based descriptions of structures. Hardware is being built to compute the HSC in video real time and a library of HSC-operations has been developed to extract invariant features like structure types, shape descriptions, or relations between structures. As part of the PROMETHEUS-project, we examined a set of 141 traffic scenes in order to develop strategies for an automatic analysis. Some of them are already implemented by using the set of domain independent HSC-operations together with some newly created ones. Hereby, we could show that fast analysis of traffic scenes is possible with the HSC by using parallel hardware architectures. 1

No abstract

No abstract