Evidence Accumulation & Flow of Control in a Hierarchical Spatial Reasoning System

To elaborate, suppose a helicopter-based computer vision system is looking at a snow-covered terrain; this terrain knowledge must then be explicitly taken into account in a target recognition procedure. Clearly, the processing required for a snow-covered background is different from that for, say, a wooded area in spring. As a simpler example of knowledgebased processing, consider the problem of self-location for a vehiclemounted vision system (Kak et al. 1987). Let's say the vehicle's whereabouts are approximately known from the position encoders mounted on the wheels, the precision of this information limited by the extent of slippage in the wheels, and so on. Given this approximate information, is it possible to make a more precise fix on the location of the vehicle by integrating the vision data with the map knowledge while the two are out of registration?