E.A. Samoylin, Yu.A. Kuznetsov, A.S. Pomysov

The method of separation of the useful signals of radar-tracking images against the Gaussian modal pulse interferences, based on use of neuron network algorithm and statistically optimum procedure of decision-making is offered. The purpose of operation is lowering of errors of separation of the useful signals of radar-tracking images against noises of impulse structure on the basis of sharing of statistical rules and neuron network algorithms. The offered method of separation of the useful signals of radar-tracking images against pulse interferences is based on sharing of neuron network algorithm and the modified Bayesian procedure. Thus on the first processing stage the trained two-layer neural network, capable to function in the conditions of nonparametric prior uncertainty and allowing to receive some rough estimates of parameters of pulse interferences, and on the second processing stage - the optimum algorithm modified statistically using neuron network estimates of parameters of noises will be involved. In this case sharing of two different principles of processing - trained neuron network, not always giving in to mathematical reasons, and strict statistically optimum, allows to reduce probabilities of errors of separation of the useful signals of radar-tracking images against pulse interferences. Results of numerical researches of the neural statistical technique, probabilities of errors of separation of signals of radar-tracking images testifying to lowering against noises in comparison with a traditional method are provided. It is given an example processing of the digital radar-tracking image by an offered and traditional method.