Abstract
A rule based expert system is described that distinguishes targets from background clutter. The expert system uses a Bayesian inference system and has the unique feature of allowing graduated or fuzzy responses for the different “events.” This allows the system to handle non-exact or uncertain data from the input seeker. For instance, one can specify (on a scale of –1 to 1) if something is “pretty large” or “relatively flat.” The optical implementation is based on the diagnostic expert system design described by McAulay.1 It uses a 1-D SLM to handle the input (fuzzy) responses and a 2-D SLM to store the a priori probability matrices for the Bayesian inference system. The 2-D SLM also provides interconnection between the event outcomes and the various hypotheses.