The structure of Bayes nets for vision recognition

This paperl is part of a study whose goal is to show the efficiency of a Bayes net to carry out model based vision calculations. [Binford et al. 1987] The network model is drawn up from the object's geometric and functional description to predict the appearance of an object. Then this net can be used to find the object within a photographic image of a scene. Many existing and proposed vision recognition techniques resemble the uncertainty calculations of a Bayes net. In contrast, though, they lack a derivation from first principles, and tend to rely on arbitrary parameters. that we hope to avoid by our network model. The connectedness of the net depends on what independence considerations can be identified in the vision problem. Greater independence leads to easier calculations, at the expense of the net's expressiveness. Once this tradeoff is made and the structure of the net is determined, it should be possible to tailor a solution technique for it. This paper explores the use of a net with multiply connected paths, drawing on both techniques of belief networks [Pearl 86] and influence diagrams. We then demonstrate how one formulation of a multiply connected net can be solved.