Solving the problem of super-resolution using a model of a neural network of direct propagation

The problem of increasing the efficiency of control of moving objects using new algorithms that improve the quality of images obtained during the survey of space is considered and solved. A significant improvement in quality is based on the achievement of angular resolution, tens of times higher than the Rayleigh criterion. Angular super-resolution provides separate observation of several objects that are not resolved by direct observation, and the accompanying increase in image clarity makes it possible to capture previously unnoticed details of images of complex objects. On this basis, the probability of correct solutions to recognition and identification problems increases. To provide angular super-resolution, the problem of creating a neural network has been solved. For multi-element receiving and transmitting measurement systems, an extrapolation method for achieving angular super-resolution is proposed and justified. The basis of the method is the extrapolation of the values of the complex amplitudes of the received signal by individual elements of the receiving devices outside the measuring system. Thus, a larger virtual system is created, for which its own Rayleigh criterion is fulfilled. As a result, the effective angular resolution increases in proportion to the increase in the size of the virtual system. Comparative results of mathematical modeling of the neural network and other extrapolation methods are investigated and presented, the limits of applicability of the method are determined.