Bayesian hypothesis generation and verification

Regarding computer vision as optimal decision making under uncertainty, a new optimization paradigm is introduced, namely, maximizing the product of the likelihood function and the posterior distribution on scene hypotheses given the results of feature extraction. Essentially this approach is a Bayesian formulation of hypothesis generation and verification. The approach is illustrated for model-based object recognition in range imagery, showing how segmentation results can optimally be incorporated into model matching. Several new match criteria for model based object recognition in range imagery are deduced from the theory. The text was submitted by the author in English. Walter Armbruster (born 1948 in Linz, Austria) graduated with a degree in Mathematics from the University of Heidelberg in 1975, where he also received his Ph.D. degree (Dr. rer. nat) in 1980. Subsequently, he worked in the Dept. of Mathematics, publishing research work in several journals including Econometrica. Since 1985 he has been a research scientist at the FOM, where he presently heads the project group “target recognition with laser radar.” A member of the NATO RTO, he has published several dozen articles in the fields of target tracking, helicopter obstacle avoidance, autonomous navigation, and 3D object recognition; some of these articles, however, are not publicly distributed.