Our ability to devise safe, reliable multi-sensor systems is critical in an age where sensor platforms are increasingly complex.
The difficulties inherent in the design and veri.cation of complex systems necessitates intelligent recon.guration of sensors
and their signal processing tools online. This paper introduces a self-adaptive (i.e. deliberative) approach to flexible model-based
image understanding. The methods described in this paper use qualitative reasoning to maintain probabilistic models of qualitative
sensor-centred scene descriptors.