Utility-based control for computer vision

Several key issues arise in implementing recognition in terms of Bayesian networks. Computational al effiCiency is a driving force. Perceptual networks are very deep, typically fifteen levels of structure. Images are very wide, e.g. an unspecified number of edges may appear anywhere in an image 512 x 512 or larger. For efficiency, we dynamically instantiate hypotheses of observed objects. The network is not fixed, it is only partially instantiated. Hypothesis generation and indexing are important, but they are not considered here [Nevatia 73}, [Ettinger 88}. This work is aimed at near-term implementation with parallel computation in a radar surveillance system, ADRIES [Levitt 88], (Franklin 88] and a system for industrial part recognition, SUCCESSOR [Bin­ ford 88]. For many applications, VlBion must be !aster to be practical and so efficiently controlling the machine vision process is critical. Perceptual operators may scan megapixels and may require minutes of computation time. It is necessary to avoid unnecessary sensor actions and computa­ tion. Parallel computation is available at several levels of processor capability. The potential for parallel, distributed computation for high-level vision means distributing non-homogeneous compu­ tations. This paper addresses the problem of control in machine vision systems based on Bayesian probability models. We separate control and inference to extend the previous work [Binford 87] to maximize utility instead of probability. Maximizing utility allows adopting perceptual strategies for efficient information gathering with sensors and analysis of sensor data. Results of controlling machine vision via utility to recognize military situations are presented in this paper. Future work extends this to industrial part recognition for SUCCESSOR.