A framework for uncertainty reasoning in hierarchical visual evidence space

A computational framework is presented to show how Dempster-Shafer (D-S) evidence theory can be applied to a hierarchically structured hypothesis space in a computer vision system. It is shown how to make use of partial and locally ambiguous information at different levels of abstraction to achieve a reliable interpretation. It is also shown how the reasoning process can make use of spatial relationships among pieces of visual evidence to strengthen the reasoning results. Because a frame of discernment consists of a set of mutually exclusive visual events, the reasoning process is visual evidence driven, and the goal is to assign labels to visual events. However, at each level of the event hierarchy it is possible to pose additional D-S problems involving spatial relationships among hypotheses at that level and other evidence. The results of these subsidiary D-S problems contribute to the belief functions for the original hypotheses at that level. Thus, a framework for incorporating a relational model into an event-driven reasoning process is formed. An implementation example is given.<<ETX>>