APPLICATION OF PERCEPTRONS TO PHOTOINTERPRETATION.

Abstract : The paper summarizes the accomplishments of three phases of a continuing research program that is directed towards the development of techniques for automatic recognition of features in photographs. This research places strong emphasis on the pre-processing of original imagery in order to require only a present state-of-the-art capability in automatic pattern recognition. Efforts generally confirmed the validity of the system philosophy being pursued, and evaluated a number of techniques for accomplishing the various steps (object detection, object isolation, object standardization, and object recognition) in the automatic system. Volume quantities of actual photographic material were processed within a general purpose digital computer for detection, isolation, and standardization. Excellent object recognition results were obtained with a simple perceptron, using quite badly distorted synthesized object material. (Author)