METHOD FOR RECOGNIZING TYPICAL COMPOSITION OF A CLUSTERED AIR TARGET OF VARIOUS CLASSES UNDER VARIOUS CONDITIONS OF THEIR FLIGHT BASED ON KALMAN FILTERING AND A NEURAL NETWORK Russian patent published in 2023 - IPC G01S13/52 G01S7/41 G06N3/02 G06T7/20 

FIELD: radar location.

SUBSTANCE: invention can be used for recognition in a pulse-Doppler radar location station (RLS) of a typical composition of a clustered air target (CAT) of various classes under various conditions of its flight based on Kalman filtering and a neural network. The claimed method consists of the fact that the radar location (RL) signal reflected from a group of aircrafts (AC), from the output of the pulse-Doppler radar receiver at an intermediate frequency, is subjected to narrow-band Doppler filtering and converted into an amplitude-frequency spectrum (AFS), the spectral components of which are caused by signal reflections from AC airframes. Next, threshold processing of the ASF signal is carried out with formation of counts F_ijq of Doppler frequencies, where i=1,…,I; I - the number of ACs in the group; j=1,…, J; J is the number of AC types of different classes; q=1,…Q; Q is the number of AC flight conditions. Then Doppler frequencies are sequentially formed in discrete time, according to which, in K cycles, the values of the corresponding ijq-th ACF and the value σ²_F*ijq of variances of the Doppler frequency derivative for each i-th AC of the j-th type with q-th flight conditions are calculated. Preliminary training of the neural network for each i-th AC of the group is carried out. In the process of recognizing the typical composition of the CAT by the RL signal reflected from its elements, using each ijq-th Kalman filter with the corresponding dynamic model, Doppler frequency readings are filtered, as a result of which, at the output of each ijq-th Kalman filter, an estimate of the ΔF_ijq(k+1) Doppler frequency is formed, from which the ACF estimate and the estimate of the σ²_F*ijq variance of the Doppler frequency fluctuation derivative estimate for each i-th aircraft group are calculated, which is fed to the corresponding input of the neural network to make a preliminary decision about the presence in the group of the i-th aircraft group of the j-th type with q-the flight conditions with a probability Р_ijq, which is compared with the threshold value Р_thr, if the condition is met for each probability value Р_ijq ≥ P_thr the final decision is made that the i-th AC in the group has the j-th type with q-th flight conditions, otherwise the decision is made that this type of AC is absent in the group.

EFFECT: creating a method that makes it possible to recognize in a pulse-Doppler radar with a probability not lower than a given standard a composition of CATs of various classes under various conditions of their flight.

1 cl, 1 dwg, 1 tbl