AN ALGORITHM OF RECOGNITION SYSTEMS OF THE PERCEPTRON TYPE WITH A CORRELATOR AT THE INPUT,
FOREIGN TECHNOLOGY DIV WRIGHT-PATTERSON AFB OHIO
Pagination or Media Count:
The paper deals with a recognition system, consisting of a correlator and perceptron, designed for recognition of patterns on a uniformly illuminated background. To reduce the size of the recognition system, a two-dimensional autocorrelation function is employed as the input description. This function is invariant to some isomorphous transformations. A new autocorrelation function, obtained by positive and negative images, is proposed, which permits the reduction of the dimensions of the apparatus. The algorithm for calculating the autocorrelation function is adapted for digital computers. An analog television variation of the autocorrelator is also described. The perception part of the system choice of properties, mask-random prototypes, etc. is calculated.