FIELD: hydro acoustics.
SUBSTANCE: invention relates to underwater acoustics and can be used to implement neural network recognition of target classes (surface or underwater object) detected by the signs of amplitude-phase modulation of low-frequency signals pumping the marine environment by radiation and fields of objects. Method of detecting and classifying marine targets using a mathematical model for determining the type of target, consisting in that, first, forming in the marine environment a zone of nonlinear interaction and parametric conversion of pumping waves with object signals, for which the radiator and the receiving antenna are placed on the opposite boundaries of the controlled portion of the marine medium, then the pump waves, modulated by the object signals, are received and amplified in the parametric conversion band, transferring their frequency-time scale to a high-frequency region, performing narrow-band spectral analysis, selecting parametric components of the sum or difference frequency, from which, based on the time and parametric conversion of waves, restoring the characteristics of the object signals, which are supplied to neural network recognition and classification circuit, consisting of unit of target class identification by amplitude-frequency characteristics, which realizes computational operations of artificial neural network and covered with feedback to training unit, in memory of which data of mathematically processed images of sea target spectrograms are recorded, wherein the first input of the target class recognition unit by amplitude-frequency characteristics receives data from the spectral analyzer output of the signal receiving, processing and recording channel, and its second input receives data from neural network recognition and classification path learning unit. Principal difference from prototype is that data are supplied to an additional introduced path of adaptive neuro-fuzzy correction to input of unit of productive rules and functions, where the new product rule number signal and the new type of the membership function are generated, and then supplied to the neuro-fuzzy network adapter input, the function of which is performed by the neural-fuzzy network (ANFIS), in which the membership of the new type of rules is compensated, the number of the rule required for replacement in the main rules base is determined, as well as a new membership type function for the given rule, and further to a differentiator, the output of which is connected to the input of the fuzzy controller, in which the rule base is automatically adjusted based on a sampling of mathematical models of marine targets, formation and reduction of sampling of reference samples of mathematical models of sea targets and correction of data of the operatively updated library of mathematically processed images of sea target spectrograms for the neural network recognition and classification channel learning unit, thereafter, an artificial neural network is tuned and a conclusion on the degree of belonging of the analyzed spectral region to the classification object (surface or underwater object) is formed. Said technical result is achieved by forming and reducing a sample of reference samples of mathematical models of sea targets by an adaptive neuro-fuzzy correction channel, independently performing automatic adjustment of its rules base and its neuro-fuzzy correction, using computational operations of adaptive neuro-fuzzy network (ANFIS), for quickly updated library of spectrograms of sea targets unit learning network neural network recognition and classification, providing final classification of detected marine targets and high probability of correct classification of marine targets by 5–7 %.
EFFECT: technical result of proposed invention is automation of process of recognition of classes of sea targets, detected by signs of amplitude-phase modulation of low-frequency pumping signals of marine environment by radiation and fields of objects, complex reduction of data size during automatic adjustment of rules base due to generation and reduction of sampling of reference samples of mathematical models of sea targets, carried out by means of tract of adaptive neuro-fuzzy correction.
1 cl, 7 dwg
