文摘
The Dempster-Shafer theory and the convex Bayesian theory have recently been proposed as alternatives to the (strict) Bayesian theory in the field of reasoning with uncertainty. These relatively new formalisms claim that missing information in the probabilistic model of a process not necessarily disables uncertainty reasoning. However, this paper shows that this does not apply to processes where the reasoning is part of a decision-making process, such as object recognition. In these cases, a complete probabilistic model is required and can be obtained by estimating missing probabilistic information. An examplary approach towards the estimation of uncertain probabilistic information is described in this paper for a multi-sensor system for recognition of electronic components on printed circuit boards.