RU2694848C1 - Method of forming a scalable system for detecting and classifying sea targets with artificial intelligence elements - Google Patents

Abstract

FIELD: hydro acoustics.

SUBSTANCE: invention relates to hydroacoustics and can be used to form a scalable target classification system, transmitting data from the marine environment to the atmosphere and back. Main system is formed, for which one radiator (3) and two receiving transducers (4) and (5) are placed on opposite boundaries of medium and working zone of nonlinear interaction and parametric conversion of waves (7) is formed between them from two horizontally spaced in receiving zone and combined in radiation zone of parametric antennae. Nonlinearly transformed waves are received and transmitted to receiving channel (2), where signals are amplified in the band of parametric wave conversion, their phase difference is measured, frequency-time scale of signals is converted to high-frequency region, their narrow-band spectra are measured and recorded. Signals from the spectrum analyzer output are transmitted via a radio communication channel to an information-analytical center (IAC) (18), where identification of information waves is carried out, according to the results of which necessary correction is introduced and signals are transmitted back to radiating path (1) for controlling the pumping waves of the sea medium. At the same time the main system is scaled. For this purpose, n additional systems are introduced, which are spatially distributed within adjacent and remote water areas. Additionally, a single information-analytical center (SIAC) (25) is installed, which controls operation of the main and n additional systems. For paging of medium pumping signals to each additional system include emitting (1.1...1.n) and receiving (2.1...2.n) channels, radiators (3.1...3.n) and two receiving transducers (4.1...4.n), which are placed in marine environment and connected by cable to emitting (1) and receiving (4) paths respectively. Working zones of nonlinear interaction and parametric conversion of waves (7.1...7.n) are formed similarly to the main system. Signals are received and by means of cables are transmitted to receiving paths (2.1...2.n), where signals are amplified, converted, analyzed and transmitted to input of recorder and inputs of units of system analyzers IAC (18.1...18.n), mathematical processing of images of spectrograms of objects and comparing degree of belonging of analyzed spectral region to basic data, recording, accumulating and updating data libraries of portraits of classification objects. Signals from the outputs of the IAC of main (18) and n additional systems (18.1...18.n) via radio communication channels are transmitted to SIAC (25), through which the library of portals of the IAC classification objects of the main and n additional systems is refilled and signals are generated which, over radio communication channels is transmitted from SIAC (25) outputs to main (1) emitting channels and n additional systems (1.1...1.n) to input of corresponding pumping signals of sea medium for correction of their parameters taking into account hydro-acoustic conditions of wave propagation in marine environment and detected objects.

EFFECT: enabling generation of a scalable system for detection and classification of marine targets with elements of artificial intelligence, through which a long parametric reception of interacting waves is provided on adjacent and remote water bodies using data transmission from the marine environment into the atmosphere and back, detected objects are classified based on computational operations of artificial neural networks and libraries of mathematically processed images of sea target spectrograms.

1 cl, 9 dwg

Description

The invention relates to the field of underwater acoustics and can be used to form a target classification system that is scalable within adjacent and / or remote waters, detected during long-range parametric wave reception using data transfer capabilities from the marine environment to the atmosphere and back.

The proposed method is implemented on the basis of the means of marine instrumentation and the technology of long-range parametric reception of interacting waves using data transmission from the marine environment to the atmosphere and back (see Mironenko MV, Malashenko A.E., Karachun L.E., Vasilenko A. M. Low-frequency luminal method of long-range hydrolocation of the hydrophysical fields of the marine environment: monograph - Vladivostok: SKB SAMI FEB RAS, 2006. - 173 pp., Vasilenko AM, Mironenko MV, Pyatakovich VA, etc. Sistema monitoring the fields of the sources of the atmosphere, the ocean and the earth’s crust on on the technology of nonlinear luminal hydroacoustics: a monograph. - Vladivostok: T.O. Makarov TOVVMU, 2015. - 320 p.).

Neural network recognition of target classes detected by the amplitude-phase modulation signs of waves interacting in the marine environment is based on computational operations of artificial neural networks and libraries of mathematically processed spectrogram images of objects, which speeds up the recognition process and increases the likelihood of classifying both surface and underwater targets. (see Pyatakovich V.A., Vasilenko A.M., Khotinsky OV. Recognition and classification of sources for the formation of fields of different physical nature in the marine environment: monograph. - Vladivostok: Maritime State University. GI Nevelsky, 2017. - 255 p.).

The principle of operation of parametric antennas with both high-frequency (tens to hundreds of kHz) and low-frequency (tens to hundreds of Hz) pumping of the marine environment is based on the use of the natural nonlinear properties of the marine environment.

There are known methods and parametric antennas that implement them using high-frequency pumping of the marine environment (see Novikov B.K., Timoshenko V.I. Parametric antennas in sonar systems. - L .: Shipbuilding. - 1990. - P. 17-40, 203 -225), the shortcomings of which are the small range of parametric wave reception (hundreds of meters and only 1-2 km in some cases) and the limited ability to measure the spatial-temporal characteristics of signals, which is especially evident when receiving waves of different physical nature low, infrasonic and a fractional frequency bands.

There are known methods and parametric antennas that implement them, whose operation is based on low-frequency illumination of the medium with weakly damped pump signals with a frequency of tens or hundreds of hertz, which leads to an increase in the range of parametric wave reception by tens or hundreds of times with respect to high-frequency parametric antennas (see Mironenko MV , Malashenko A.E., Vasilenko A.M. and others. Nonlinear translucent underwater acoustics and means of marine instrumentation in the creation of the Far Eastern radiohydroacoustic system for lighting the atmosphere, ocean and emnoy bark, monitor their fields of different physical nature: monograph - Vladivostok. Publ Dalnevost University Press, 2014)..

A known method of transmitting information waves in the marine environment (see patent 2472236 of the Russian Federation, IPC G10K 11/00, publ. 10.01.2013, bull. No. 1), which includes the formation of a parametric radiating antenna, the emission of pumping waves into the medium with simultaneous transmission information waves, as well as receiving signals generated by the interaction of pump waves and information waves, followed by the restoration of the original information. At the same time, on the propagation path of the radiated low-frequency pump signal, preferably in the far zone of the emitter, a spatial non-linear region is formed, which is irradiated with biharmonic audio signals close to the pump signal and waves of a different physical nature are introduced into it, for example, electromagnetic or hydrodynamic disturbances subjected to time conversion in accordance with the characteristics of the transmitted information.

The considered method has the following disadvantages. Low effect of non-linear interaction and parametric transformation of waves and, as a result, a small transmission-reception range of waves of different physical nature in the marine environment, limited information capacity of the transmitted and received signals, as well as the impossibility of transmitting data from the marine environment to the atmosphere and back.

By its physical essence, the closest to the claimed invention is a method of transmitting information waves from the marine environment to the atmosphere and back (see patent RF 2593625, IPC G10K 11/00, publ. 10.08.2016, bull. No. 22), which is chosen as prototype.

The prototype method consists in that one emitter and two receiving transducers are placed at opposite boundaries of the medium and form between them a working zone of non-linear interaction and parametric wave conversion from two horizontally separated parametric antennas in the reception zone; nonlinearly converted waves are received and transmitted to the receiving path, where the signals are amplified in a parametric wave conversion band, their phase difference is measured, the time-frequency scale of the signals is converted into a high-frequency region, their narrow-band spectra are measured and recorded; the signals from the output of the spectrum analyzer are transmitted over the radio link to the information-analytical center, where they identify information waves, introduce corrections and transmit them back to the radiating path to control the process of radiation of pumping waves of the marine environment.

The disadvantages of the prototype method include the following:

- the lack of technology that allows the system to be scaled by additional systems spatially separated within adjacent and (or) remote water areas;

- the lack of technology that provides the neural network classification of sources detected by the signs of amplitude-phase modulation of waves interacting in the marine environment, based on the computational operations of artificial neural networks and libraries of mathematically processed images of spectrograms of marine targets;

- lack of technology that provides replenishment and updating of libraries of portraits of objects of classification.

The main distinctive feature of the proposed method from the prototype method is that the system implemented by this method is formed as scalable, which includes the main system, n additional systems and a single information-analytical center (EIAC).

The problem to which the invention is directed is to develop a method for forming a scalable system for detecting and classifying marine targets with elements of artificial intelligence, with which they provide far-off parametric reception of interacting waves on adjacent and / or remote water areas using data transmission from the marine environment into the atmosphere and back, as well as carry out the classification of detected objects based on the computational operations of artificial neural networks and bi liotek mathematically processed images spectrograms sea targets.

The scaling of the main system with additional systems on adjacent and (or) remote waters provides a long-range parametric reception of waves in the sound and infrasonic frequency ranges, as well as an increase in the capacity of information about sources of fields of various physical nature and hydrologic-acoustic conditions of signal propagation in the marine environment.

The adjustment of the coordinated mode of operation of the radiating and receiving paths of the main and additional systems to the hydrologic-acoustic conditions of wave propagation and manifestations of the signs of information wave sources is solved by introducing into the system structure a single information and analytical center that manages the operation of the scalable detection and classification system for marine targets elements of artificial intelligence in accordance with the objectives and conditions of long-term maritime monitoring.

Replenishing and updating, through the EIAC, libraries of portraits of classification objects formed in the information and analytical centers of the primary and secondary systems, speeds up the recognition process, increases the likelihood of classifying both surface and underwater targets, as well as expands the range of recognized objects.

To solve this problem, a method has been developed to form a scalable system for detecting and classifying marine targets with elements of artificial intelligence, which consists in forming the main system, for which one emitter and two receiving transducers are placed at opposite environmental boundaries and form between them a working zone of nonlinear interaction and parametric conversion waves from two horizontally separated parametric antennas in the reception area and the radiation emitted in the radiation area; then, the nonlinearly converted waves are received and transmitted to the receiving path, where the signals are amplified in the parametric wave conversion band, their phase difference is measured, the time-frequency scale of the signals is converted to the high-frequency region, their narrow-band spectra are measured and recorded; then the signals from the output of the spectrum analyzer are transmitted over the radio communication channel to the information and analytical center, where they identify information waves by the results of which they introduce the necessary corrections and transmit signals back to the radiating path to control the radiation process of the pump waves in the marine environment.

The principal difference of the proposed method is that the main system is scaled, for which n additional systems are introduced, which are spatially distributed within adjacent and / or remote water areas; in addition, a single information-analytical center (EIAC) is introduced, which manages the work of the main and n additional systems; each additional system includes the radiating and receiving paths, the emitter and two receiving transducers, which are placed in the marine environment and connected to the radiating and receiving paths, respectively, as well as the working area of nonlinear interaction and parametric wave conversion, which is formed in the marine environment in in the form of two horizontally separated parametric antennas at the points of reception and at the point of radiation combined at the point of radiation, for which the medium is sounded by pump signals that are generated in the radiating tract cte; then signals are received and transmitted to the receiving path through cables, where signals are amplified in a parametric wave conversion band, their phase difference is measured, the time-frequency scale of signals is converted into a high-frequency region, narrow-band spectra are measured, discrete components of total or difference frequency are identified as signs of manifestation amplitude-phase modulation of the low-frequency pump wave of the medium by radiation and fields of marine targets; then the signals are transmitted to the recorder input, as well as to the input of the system unit There is a lot of analysis of the information and analytical center, where using computational operations of artificial neural networks they perform mathematical processing of the images of spectrograms of objects and compare the degree of belonging of the analyzed spectral region to basic data, then record, accumulate and update data of the library of portraits of classification objects, then signals from the outputs of information the analytical centers of the main and n additional systems are transmitted via radio links to the EIAC, through which the biblical The libraries of portraits of objects of classification of information and analytical centers of the main and n additional systems, as well as form signals that transmit via radio links from the EIAC outputs to the radiating paths of the main and n additional systems to the input of the corresponding marine pump signal generators to adjust their parameters -acoustic conditions of wave propagation in the marine environment and detected objects.

The technical result of the invention is to provide on the adjacent and (or) remote waters far parametric reception of waves in the sound and infrasonic frequency ranges using data transmission from the marine environment to the atmosphere and back, as well as the classification of detected objects based on the computational operations of artificial neural networks and libraries mathematically processed spectrogram images of marine targets.

The physical essence of long-range parametric wave reception in the marine environment

It is known that the influence of fields of different physical nature, formed by sources in the marine environment, on the low-frequency signals of the pumping of the marine environment is carried out through a change in the density and coefficient of elasticity of the medium (see Mironenko MV, Malashenko AE, Karachun L.E., Vasilenko, AM, Low Frequency Translucent Method for the Long-Range Sonar of Hydrophysical Fields of the Marine Environment: a monograph - Vladivostok: SKB SAMI FEB RAS, 2006; Malashenko AE, Mironenko MV, Chudakov MV, Pyatakovich VA Far parametric reception of electromagnetic waves generated by technical sources in the marine environment. Sensors and systems - Moscow: 2016. No. 8-9 (206). P. 14-18).

In its physical essence, the far parametric reception of signals provides for a special change in the density and (or) temperature of the aquatic environment. The change of these parameters can be done in various ways, but the main one is the formation of an extended non-linear region in a given direction of radiation-reception of waves (parametric antenna).

Parametric wave reception manifests itself as an amplitude-phase modulation of a low-frequency pump wave in the marine environment by a wave formed by an object, with their joint propagation. In the path of receiving and processing signals using spectral analysis, there are signs of amplitude-phase modulation of waves.

The process of parametric wave reception can be explained by the usual system of hydrodynamic equations for a viscous fluid when applying to the equation of state the corresponding changes in the phase velocity of sound in time and space. To calculate the propagation velocity of an elastic (acoustic) wave, you can apply the well-known formula

,

,

- коэффициент адиабатической сжимаемости жидкости. where P is the pressure;

- coefficient of adiabatic compressibility of the liquid.

Using the relation between adiabatic and isothermal compressibility, we can obtain the following expression for the phase velocity

,

,

- удельный объем.where ρ is the density;

- specific volume.

. Таким образом, если в морской среде распространяется электромагнитная волна гармонической частоты, то фазовая скорость упругой волны (низкочастотной волны накачки) будет меняться с той же частотой.From the above expression it follows that changes in the density ρ and pressure P at a constant temperature lead to a change in the phase velocity of sound over time.

. Thus, if an electromagnetic wave of a harmonic frequency propagates in the marine environment, then the phase velocity of the elastic wave (low-frequency pump wave) will change with the same frequency.

Testing the performance of ideas that are the basis of the present invention, was carried out using electromagnetic waves. It is obvious that the laws of nonlinear interaction and parametric conversion for other waves, as well as in the case of a positive effect with electromagnetic waves, also exist.

The spectrum of interacting waves consists of an infinite number of lateral components, whose frequency and amplitude can be found from the well-known expression

,

,

- результирующее и мгновенное значения давления модулированной волны, соответственно;

- удвоенная частота модулированной волны;

- волна, генерируемая объектом;

- время;

- функции Бесселя n-го порядка;

- амплитуда модулированной волны;

- коэффициент модуляции.Where

- the resulting and instantaneous values of the pressure of the modulated wave, respectively;

- twice the frequency of the modulated wave;

- wave generated by the object;

- time;

- n-th order Bessel functions;

- modulated wave amplitude;

- modulation factor.

.As can be seen from the expression, the frequencies of the lateral components differ from the doubled center frequency 2ω (equal to the sum of the frequencies of the interacting waves) by ± n величинуΩ, where n is any integer. The amplitudes of the lateral components for the respective frequencies
(2ω ± nΩ) are determined by the magnitude of the multiplier

.

спектр взаимодействующих волн приближенно состоит из удвоенной центральной частоты 2ω и ее боковых частот 2ω+Ω и 2ω-Ω.At small values of the modulation factor

The spectrum of the interacting waves approximately consists of the doubled central frequency 2ω and its side frequencies 2ω + Ω and 2ω-Ω.

The claimed invention is illustrated in the drawings. FIG. 1 shows a block diagram reflecting the method of forming the main scalable system for detecting and classifying marine targets with elements of artificial intelligence, where:

1. The radiating path.

2. Reception path.

3. Emitter.

4, 5. Receiving transducers.

6. Sea target (source of waves).

6a. Nonlinear wake area.

6b. Outboard emitter of acoustic and (or) electromagnetic signals.

7. Working area of nonlinear interaction and parametric transformation of pump and information waves (information waves are waves generated by natural or artificial sources in the marine environment, for example, acoustic, electromagnetic, hydrodynamic).

8. Generator of pump signals (linear-frequency-modulated and / or phase-modulated).

9. The pump signal generator (stabilized frequency in the range of tens to hundreds of hertz).

10. Power amplifier.

11. Block approval.

12. Two-channel broadband amplifier.

13. Phase meter.

14. The converter of time-frequency scale of signals in the high-frequency region.

15. Narrowband spectrum analyzer.

16. Registrar.

17. Transmitting radio unit.

18. Information Analytical Center (IAC).

19. Receiving radio unit.

20. System analysis unit.

21. Transmitting radio unit.

22. Receiving radio unit.

23. Controlled marine environment.

24. Sea surface.

FIG. 2 shows a block diagram with its functional links, reflecting the method of forming a scalable system for detecting and classifying marine targets with elements of artificial intelligence, which combines the main and n additional systems, where:

1. The radiating path of the main system.

1.1 ... 1.n. Radiant paths of additional systems.

2. The receiving path of the main system.

2.1 ... 2.n. Reception paths of additional systems.

3. The emitter of the main system.

3.1 ... 3.n. Emitters of additional systems.

4, 5. Receiving transducers of the main system.

4.1 ... 4.n, 5.1 ... 5.n. Receiving transducers of additional systems.

18. Information and analytical center of the main system.

18.1 ... 18.n. Information and analytical centers of additional systems.

25. Integrated Information and Analytical Center (EIAC).

FIG. 3 shows the level of the information wave of the difference frequency formed by the nonlinear area of the wake of the boat wake. The frequency of the pumping signals of the marine environment was 1040 Hz and 960 Hz. The frequency of the difference frequency information signal was 80 Hz. The length of the route for receiving and transmitting signals was 25 km.

FIG. 4 shows the radiation spectrum of a vessel, measured by a low-frequency translucent method, which implements a long-range parametric reception of acoustic and electromagnetic waves interacting in the marine environment. The frequency of medium illumination is Fa = 390 Hz, the length of the route is 45 km. In the signal spectrum, parametric components of the sum and difference frequencies from the original frequencies of the acoustic medium and electromagnetic radiation of the vessel are observed.

FIG. 5 shows the spectrogram of the noise field of the marine vessel, on which the hydrodynamic field of the wake trace and the discrete component of the resonant oscillations of the ship hull are observed. The frequency of the translucent signals was 400 Hz, the length of the translucent trail was 30 km.

The general structure of the recognition network is shown in FIG. 6. The neurons that make up the network are the same and have an activation function of a known type.

where x _2n ⁽ⁱ⁾ , y _n ⁽ⁱ⁾ and I _n ⁽ⁱ⁾ are the values of the r - th input signal, the output signal and the external bias of the n - th neuron of the i - th layer; N _i - the number of neurons in the i - th layer; i = 1, 2, 3.

FIG. 7 and FIG. 8 presents the results of a computational experiment to determine the recognition coefficient (classification), defined as the ratio of the number of recognized objects to the total number of tests in percent, for surface and underwater objects in a noisy signal in the range from -10 to 20 dB. As can be seen from the figures, the recognition and classification of sea targets with the help of computational operations of the perceptron network allows to increase the probability of classification of both surface and underwater targets by 5-7%.

FIG. 9 shows a table of the interpretation of the elements of the output vector of recognition of sonar signals according to the amplitude-frequency characteristic.

The method of forming a scalable system for detecting and classifying marine targets with elements of artificial intelligence is implemented as follows.

The operation of the primary and secondary systems that make up the scalable system for detecting and classifying marine targets with elements of artificial intelligence is performed uniformly. Underwater sound guidance beacons of the PZM-400 type can be used as low-frequency emitters. Radiating and receiving paths can be formed from existing radio equipment.

As shown in FIG. 1 emitter 3 and receiving transducers 4, 5 of the main system are placed in a controlled marine environment 23 taking into account the laws of multipath wave propagation. Between the emitter and two receiving transducers they form a working zone of nonlinear interaction and parametric conversion of pump and information waves 7, as two horizontally separated parametric antennas in the reception area (see Certificate of state registration of the computer program "Calculation of the ray pattern" №2016616822 from 06/21/2016; Certificate of state registration of the computer program "Program of simulation modeling of the propagation of sonar ignalov »№2017664296 from 12.20.2017; Certificate of state registration of the computer software program" Software-computing system of marine information situation simulation for identification purposes »№2018612944 from 01.03.2018).

Next, in the radiating path 1 generate signals pumping the marine environment. The emitter 3 sounds the medium with pumped signals of a stabilized frequency in the range of tens to hundreds of hertz or linear-frequency modulated and / or phase-modulated signals, which ensures long-range parametric reception of interacting waves in the sound and infrasonic frequency ranges.

In various modes of motion, marine targets 6 generate emissions that lead to a change in the magnitude of the characteristics of the conductive fluid (density and / or temperature and / or heat capacity, etc.), which, depending on their physical nature, modulate the signals from the pumping of the marine environment. Being an inseparably connected component of the information wave, modulation components are transported over long distances (tens to hundreds of km). Low-frequency and high-frequency components appear in the spectrum of an information wave, as a result of amplitude-phase modulation of a low-frequency wave pumping the marine environment by radiation and fields of wave sources.

The signals from the receiving transducers 4 and 5 through cable lines arrive at the input of a two-channel wideband amplifier 12 of the receiving path 2, where they are amplified in the parametric conversion band, and then the phase difference is measured (block 13), the time-frequency scale of signals is transferred to the high-frequency region (block 14), conduct narrowband spectral analysis (block 15) and highlight the signs of information waves, namely the amplitude-phase modulation of the low-frequency wave of the medium pumping by radiation and fields of marine targets.

The operation of the time-frequency scale conversion of the signal in block 14 increases the energy concentration of non-linear converted signals and the efficiency of extracting from them the signs of fields generated by marine targets.

The operations of spectral analysis in block 15 emit discrete components of the total or difference frequency in the narrow-band spectra of nonlinearly converted signals that are indications of the appearance of information waves, namely, amplitude-phase modulation of the low-frequency wave pumping the marine environment by radiation and fields of wave sources.

The amplitude-frequency characteristics of the object signals (detected marine target), obtained using narrowband spectral analysis in the receiving path 2, are fed to the input of the recorder 16, as well as through the transmitting radio block 17 of the receiving path 2 and the receiving radio block 19 of the IAC 18 to the input of the system analysis block 20 IAC 18, where they run computational operations of an artificial neural network, with which they perform mathematical processing of the images of spectrograms of objects and compare the degree of belonging of the analyzed spectral region to the base y data. Then perform the following operations with data - they record, accumulate and update libraries of portraits of objects of classification.

Next, a control signal is formed and transmitted from the output of the IAC 18 unit 20 through the transmitting radio unit 21 of the IAC 18 and the receiving radio unit 22 of the radiating path 1 to the input of the corresponding generator of pump signals 8 or 9 of the radiating path 1, which allows generating signals for pumping the marine environment taking into account the hydrologic and acoustic conditions of wave propagation in the marine environment and detected objects.

The task of classifying marine targets is solved using a three-layer neural network that recognizes seven objects and allows you to select one unknown class, which in the future will significantly expand the range of classified marine targets. The setting of the weighting coefficients of the recognition network is determined by the error back-propagation algorithm. The basic idea of the algorithm is to propagate error signals from the network outputs to its inputs, in the opposite direction to direct signal propagation in normal operation. In order to be able to apply the method of back propagation of an error, it is necessary that the transfer function of neurons be differentiable (see Pyatakovich VA, Vasilenko AM Preliminary information processing by a neuron-like categorizer when recognizing images of marine objects. Underwater naval weapon. - St. Petersburg: 2017. - Issue 1 (32) .- pp. 31-34; Pyatakovich VA, Vasilenko AM Prospects and limitations of using geometric methods for recognizing acoustic images of marine objects as applied to the control problem of a neural network expert basic system. Fundamental research. - M: 2017. - № 7. - p. 65-70).

На каждый нейрон второго слоя через синапсы с весами {T_ij ⁽²⁾}, i = 1, 2, 3; j = 1, 2, 3 подаются выходные сигналы первого слоя. На каждый нейрон третьего слоя через синапсы с весами {T_ij ⁽³⁾}, i = 1, 2, 3; j = 1, 2, 3 подаются выходные сигналы второго слоя. Значения выходного сигнала третьего слоя распознающей сети образуют вектор решений

, элементы которого представлены в табл. 1 на фиг. 9.The analysis of the low-frequency, mid-frequency and high-frequency components of the amplitude-frequency characteristic is performed separately, since the general features for different types of objects can be in different frequency ranges. As shown in FIG. 6, for each neuron of the first layer through synapses with weights {T _ij ⁽¹⁾ }, i = 1, 2, 3; j = 1, 2, 3 all components of the input vector are fed

For each neuron of the second layer through the synapses with weights {T _ij ⁽²⁾ }, i = 1, 2, 3; j = 1, 2, 3 output signals of the first layer are given. For each neuron of the third layer through the synapses with weights {T _ij ⁽³⁾ }, i = 1, 2, 3; j = 1, 2, 3 output signals of the second layer are given. The output values of the third recognition network layer form a vector of solutions.

elements of which are presented in Table. 1 in FIG. 9.

The scaling of the main system with additional systems on adjacent and (or) remote waters provides a far parametric reception of waves in the sound and infrasonic frequency ranges using data transmission from the marine environment to the atmosphere and back, as well as increasing the information capacity on the sources of fields of various physical nature and hydrological acoustic conditions of signal propagation in the marine environment.

For this purpose, n additional systems are spatially distributed within adjacent and (or) remote water areas and a single information-analytical center (EIAC) is introduced that manages the operation of the main and n additional systems.

FIG. 2 shows that additional systems supply 1.1 ... 1.n and emitting 2.1 ... 2.n paths, the blocks of which are connected and functioning as in the radiating 1 and receiving paths 2 of the main system, as well as information and analytical centers 18.1 ... 18.n, blocks of which are connected and function as in IAC 18 of the main system.

Additional systems also supply radiators 3.1 ... 3.n and receiving converters 4.1 ... 4.n, 5.1 ... 5.n, which are placed in the marine environment and connected to the radiating 1.1 ... 1.n and receiving 2.1 ... 2.n paths, respectively, by working zones 7.1 ... 7.n non-linear interaction and parametric transformation of waves, which form in the form of two horizontally-spaced at the points of reception and combined at the point of radiation parametric antennas.

Outputs of the receive paths of the main and auxiliary systems
2, 2.1 ... 2.n through the transmitting radio units are connected to the inputs of the IAC
18, 18.1 ... 18.n respectively, the outputs of which by means of radio links are connected to the inputs of EIAC 25, through which libraries of portraits of classification objects are replenished, which are formed in EAC 18, 18.1 ... 18.n., And also manage the operation of the scalable detection system and classification of marine targets with elements of artificial intelligence.

In EIAC 25, control signals are generated, which are transmitted from the EIAC 25 outputs to the radiating paths 1, 1.1 ... 1.n of the main and n additional systems to the input of the corresponding signal generators to adjust their parameters taking into account the hydrologic-acoustic conditions of wave propagation environment and detected objects.

Thus, using the method of forming a scalable system for detecting and classifying marine targets with elements of artificial intelligence, it is possible to provide in adjacent and (or) remote waters far-ranging parametric reception of waves in the sound and infrasonic frequency ranges using data transmission from the marine environment to the atmosphere and back, and also classify detected objects based on the computational operations of artificial neural networks and libraries of mathematically processed spectrogram images purposes.

The method and the system implementing it are industrially applicable, since for its creation they use common components and products of the radio engineering industry and computer technology.

Claims (1)

The method of forming a scalable detection system and classification of marine targets with elements of artificial intelligence, which consists in the formation of the main system, for which one emitter and two receiving transducers are placed on opposite boundaries of the environment and form between them a working zone of nonlinear interaction and parametric conversion of waves from two horizontally separated into reception area and parametric antennas combined in the radiation area; then, the nonlinearly converted waves are received and transmitted to the receiving path, where the signals are amplified in the parametric wave conversion band, their phase difference is measured, the time-frequency scale of the signals is converted to the high-frequency region, their narrow-band spectra are measured and recorded; then the signals from the output of the spectrum analyzer are transmitted over the radio communication channel to the information-analytical center, where they carry out the identification of information waves, the results of which introduce the necessary corrections and transmit signals back to the radiating path to control the process of radiation of pumping waves of the marine environment, different the fact that the main system is scaled, for which n additional systems are introduced, which are spatially distributed within adjacent and / or remote water areas; in addition, they additionally introduce a single information-analytical center (EIAC) that manages the work of the main and n additional systems; each additional system includes radiating and receiving paths, an emitter and two receiving transducers, which are placed in the marine environment and connected to the radiating and receiving paths, respectively, as well as working areas of nonlinear interaction and parametric wave conversion, which form in the marine environment in in the form of two horizontally separated parametric antennas at the points of reception and parametric antennas combined at the points of radiation, for which the medium is sounded by pumping signals acts; further, signals are received and transmitted to the receiving paths by cables, where signals are amplified in a parametric wave conversion band, their phase difference is measured, the time-frequency scale of signals is converted into a high-frequency region, narrow-band spectra are measured, discrete components of total or difference frequency are identified as signs of manifestation amplitude -phase modulation of the low-frequency pump wave of the medium by radiation and fields of marine targets; then the signals are transmitted to the recorder input, as well as to the input of the system blocks analysis of information and analytical centers, where, using computational operations of artificial neural networks, mathematical images of the spectrograms of objects are matched and the degree of belonging of the analyzed spectral region to basic data is compared, then the data of the library of portraits of classification objects are recorded, accumulated and updated; analytical centers of the main and n additional systems are transmitted via radio links to the EIAC, through which they replenish bi libraries of portraits of objects of classification of information and analytical centers of the main and n additional systems, as well as form signals that are transmitted from the EIAC outputs to the radiating paths of the main and n additional systems to the input of the corresponding marine pump signal generators to adjust their parameters to the hydrological -acoustic conditions of wave propagation in the marine environment and detected objects.

Images