Optical pattern recognition using bayesian classification

摘要

Two recognition systems which classify optical correlation data are described and experimentally compared. Both require a priori estimates of multivariate distributions generated from their training images (TIs). One system uses a quadratic classifier to partition the signal space into regions associated with single TIs and then defines object regions by forming a union of the appropriate TI regions. The other system uses a composite Bayes' classifier to partition the signal space directly into regions associated with single objects. Accordingly, it requires object class distributions which it approximates with composite algebraic functions constructed from the TI distribution estimates. Experimental results demonstrate that the composite Bayes' classifier consistently outperforms the modified quadratic classifier, albeit marginally, when non-TIs are processed.