Intelligent electromagnetic detection method and system for underwater targets

Abstract

The present application relates to an underwater target intelligent electromagnetic detection method and system, belonging to the geophysical exploration and object detection technical field. The electromagnetic detection subsystem collects underwater electromagnetic signals; the electromagnetic data preprocessing unit decomposes the underwater electromagnetic data samples into time domain signal sequences, frequency domain power spectrum signal sequences and time-frequency power spectrum diagrams; the time domain signal depth learning network identifies whether there is an underwater target according to the time domain signal sequences; the frequency domain power spectrum signal depth learning network identifies whether there is an underwater target according to the frequency domain power spectrum signal sequences; the time-frequency power spectrum signal depth learning network identifies whether there is an underwater target according to the time-frequency power spectrum diagrams; according to the identification results of the underwater target by the time domain signal depth learning network, the identification results of the underwater target by the frequency domain power spectrum signal depth learning network and the identification results of the underwater target by the time-frequency power spectrum signal depth learning network, the final time-frequency power spectrum diagram labeled underwater target identification result is generated by comprehensive judgment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of geophysical exploration and object detection, and particularly relates to an underwater target intelligent electromagnetic detection method and system. BACKGROUND

[0002] Acoustic detection is the main detection method for underwater targets for a long time. However, with the continuous improvement of noise reduction technology, the acoustic radiation signal of the underwater target is significantly weakened and gradually approaches the level of the ocean background noise. Purely using acoustic detection cannot meet the needs of underwater target detection, and non-acoustic detection technology needs to be developed. The main non-acoustic detection technologies include optical imaging detection, laser radar detection, infrared detection and electromagnetic detection. Optical imaging detection is suitable for sea areas with clear water quality, and its performance is limited by light conditions and water transparency, and its long-distance detection application is limited. Laser radar detection has significant signal attenuation and scattering effects in turbid seawater, limiting its long-distance detection application. Infrared detection uses temperature difference for detection, and is a passive detection method, which has a certain concealment, but its detection accuracy is easily affected by seawater stratification. Electromagnetic detection realizes target feature and position recognition by detecting the electromagnetic field change caused by the underwater target, and has the advantages of not being easy to expose, strong anti-interference ability, high hydrological adaptability and high positioning accuracy. Compared with other non-acoustic detection technologies, electromagnetic detection has stronger applicability and technical advantages in complex marine environments, and can be widely applied to underwater target detection, identification and positioning.

[0003] At present, underwater target electromagnetic detection technology has made great progress, mainly developing into fixed electromagnetic detection system based on wired network, mobile electromagnetic detection system based on underwater mobile platform and shipborne electromagnetic detection system based on water surface. The fixed electromagnetic detection system based on wired network is mainly developed for the demand of nearshore underwater target detection, which contains multiple fixed nodes of electric field, magnetic field and other physical field on the seabed, and the nodes are connected with the shore-based monitoring center through wired network to realize real-time detection of electromagnetic characteristics of underwater targets. The system is based on high-speed wired network and can realize real-time transmission of electromagnetic detection data, but the deployment cost is high, the difficulty is great, the mobility is poor and the detection range is limited. The mobile electromagnetic detection system based on underwater mobile platform is mainly developed for the demand of underwater target mobility and large-scale detection. The system takes AUV, underwater glider and other underwater mobile platforms as carriers to carry electric field and magnetic field detection system to realize the mobility and large-scale detection of underwater target electromagnetic characteristics. The advantage of the system is flexible deployment, good mobility and large detection range, but it is difficult to realize real-time transmission of underwater target electromagnetic detection data due to the limitation of low-bandwidth wireless communication mode such as underwater acoustic communication. The shipborne electromagnetic detection system based on high-speed wired network is built through the towing of the water surface ship to realize the large-scale real-time mobile detection of underwater targets, which combines the advantages of the fixed electromagnetic detection system based on wired network and the mobile electromagnetic detection system based on underwater mobile platform.

[0004] From the perspective of underwater target electromagnetic signal recognition method, the marine electromagnetic environment is complex, there are many natural electromagnetic field signals such as sea waves, sea currents, tides and seabed medium induced in the geomagnetic field, and there are also many artificial electromagnetic field signals such as marine engineering activities, submarine cables and nearshore power systems. These electromagnetic field signals are coupled with the weak electromagnetic field signals of underwater targets, making it difficult to identify underwater target electromagnetic signals. Artificial intelligence technologies such as deep learning and transfer learning have made great progress in image recognition, speech recognition and natural language processing, providing technical reference for intelligent detection and recognition of underwater target electromagnetic signals. At present, there is no intelligent recognition method for underwater target electromagnetic signals based on deep learning and other artificial intelligence algorithms. SUMMARY

[0005] The present application at least partly solves one of the problems in the related art, and provides an underwater target intelligent electromagnetic detection method and system based on artificial intelligence algorithm, which solves the problem of poor detection accuracy of underwater target electromagnetic signals in the prior art.

[0006] To achieve the above-mentioned purpose, the first aspect of the present application provides a kind of, including:

[0007] An underwater target intelligent electromagnetic detection method, comprising the following steps:

[0008] S1: collecting underwater electromagnetic signals to obtain underwater electromagnetic data and electromagnetic data samples;

[0009] S2: decomposing the electromagnetic data samples into time-domain signal sequences, frequency-domain power spectrum signal sequences, and time-frequency power spectrum maps;

[0010] S3: inputting the time-domain signal sequences into a time-domain signal deep learning network to identify whether the time-domain signal sequences contain underwater targets, inputting the frequency-domain power spectrum signal sequences into a frequency-domain power spectrum signal deep learning network to identify whether the frequency-domain power spectrum signal sequences contain underwater targets, and inputting the time-frequency power spectrum maps into a time-frequency power spectrum signal deep learning network to identify whether the time-frequency power spectrum maps contain underwater targets;

[0011] S4: comprehensively determining and generating a final underwater target recognition result labeled by the time-frequency power spectrum map according to the underwater target recognition results of the time-domain signal deep learning network, the underwater target recognition results of the frequency-domain power spectrum signal deep learning network, and the underwater target recognition results of the time-frequency power spectrum signal deep learning network.

[0012] In some embodiments of the present application, step S1 further includes:

[0013] The collected electromagnetic data is intercepted by a fixed-length time sliding window;

[0014] The fixed-length time sliding window is moved at a fixed time interval, and the electromagnetic data intercepted by the time sliding window at all time intervals is taken as the electromagnetic data samples.

[0015] In some embodiments of the present application, the electromagnetic detection method further includes:

[0016] The electromagnetic data intercepted by the fixed-length time sliding window is taken as the time-domain signal sequences and input into the time-domain signal deep learning network;

[0017] The electromagnetic data intercepted by the fixed-length time sliding window is calculated by power spectrum estimation to obtain frequency-domain power spectrum data, which is taken as the frequency-domain power spectrum signal sequences and input into the frequency-domain power spectrum signal deep learning network;

[0018] The electromagnetic data intercepted by the fixed-length time sliding window is calculated by power spectrum estimation and sliding time window method to obtain time-frequency power spectrum data of the fixed-length time sliding window, which is taken as the time-frequency power spectrum maps and input into the time-frequency power spectrum signal deep learning network.

[0019] In some embodiments of the present application, the comprehensive determination of step S4 to generate the final underwater target recognition result labeled by the time-frequency power spectrum map further includes:

[0020] If the time-domain signal deep learning network, the frequency-domain power spectrum signal deep learning network, and the time-frequency power spectrum signal deep learning network all identify the underwater target, the first intensity level is determined as the target;

[0021] If the time-domain signal deep learning network and the frequency-domain power spectrum signal deep learning network only one of them identifies the underwater target, and the time-frequency power spectrum signal deep learning network identifies the underwater target, the second intensity level is determined as the target;

[0022] If the time-domain signal deep learning network and the frequency-domain power spectrum signal deep learning network both identify the underwater target, and the time-frequency power spectrum signal deep learning network identifies the underwater target, the third intensity level is determined as the target;

[0023] If the time-domain signal deep learning network and the frequency-domain power spectrum signal deep learning network both identify the underwater target, and the time-frequency power spectrum signal deep learning network identifies the underwater target, the fourth intensity level is determined as the target;

[0024] If the time-domain signal deep learning network and the frequency-domain power spectrum signal deep learning network only one of them identifies the underwater target or both identify the underwater target, and the time-frequency power spectrum signal deep learning network identifies the underwater target, no target is determined as the target;

[0025] In the order of the first intensity level, the second intensity level, the third intensity level, the fourth intensity level, and no target, the accuracy of target identification gradually decreases.

[0026] In some embodiments of the present application, the step S1 further includes: collecting underwater electromagnetic signals through electromagnetic detection nodes arranged at different positions connected in sequence through a wired network, to obtain underwater electromagnetic data and electromagnetic data samples;

[0027] The underwater target intelligent electromagnetic detection method further includes: comprehensively processing the underwater electromagnetic data samples of each electromagnetic detection node to identify the underwater target and the motion trend of the underwater target.

[0028] In some embodiments of the present application, the step S1 further includes: collecting underwater electromagnetic signals to obtain underwater electromagnetic data, and performing denoising processing on the underwater electromagnetic data, and taking the denoised underwater electromagnetic data as electromagnetic data samples.

[0029] In some embodiments of the present application, the network model training step further includes:

[0030] According to the recognition result of the underwater target by the time domain signal deep learning network, the time domain signal sequence is labeled as underwater target time domain signal feature and no underwater target time domain signal feature, and the labeled time domain signal sequence is constructed as a time sequence signal deep learning network training set for training the time sequence signal deep learning network.

[0031] According to the recognition result of the underwater target by the frequency domain power spectrum signal deep learning network, the frequency domain power spectrum signal sequence is labeled as underwater target frequency domain power spectrum signal feature and no underwater target frequency domain power spectrum signal feature, and the labeled frequency domain power spectrum signal sequence is constructed as a frequency domain power spectrum signal deep learning network training set for training the frequency domain power spectrum signal deep learning network.

[0032] According to the recognition result of the underwater target by the time-frequency power spectrum signal deep learning network, the time-frequency power spectrum graph is labeled as underwater target time-frequency power spectrum signal feature and no underwater target time-frequency power spectrum signal feature, the underwater target electric field and magnetic field characteristic signals are distinguished, and the time of the time-frequency power spectrum graph is labeled as before the underwater target appears, during the underwater target appears, or after the underwater target appears. The labeled time-frequency power spectrum graph is constructed as a time-frequency power spectrum signal deep learning network training set for training the time-frequency power spectrum signal deep learning network.

[0033] The embodiment of the present application also provides an underwater target intelligent electromagnetic detection system, which can execute the underwater target intelligent electromagnetic detection method provided by the foregoing embodiments, and includes:

[0034] An electromagnetic detection subsystem includes a plurality of electromagnetic detection nodes, each electromagnetic detection node is sequentially connected through a wired network, and is used for collecting underwater electromagnetic signals and obtaining electromagnetic data samples of the underwater electromagnetic signals;

[0035] An electromagnetic monitoring and intelligent processing subsystem includes:

[0036] An electromagnetic monitoring unit communicates with the electromagnetic detection subsystem through a wired network, acquires electromagnetic data samples of the underwater electromagnetic signals, and stores the electromagnetic data samples in an electromagnetic database;

[0037] An electromagnetic data preprocessing unit is used for denoising the electromagnetic data samples of the underwater electromagnetic signals acquired by the electromagnetic monitoring unit, and further decomposing the electromagnetic data samples into time domain signal sequences, frequency domain power spectrum signal sequences and time-frequency power spectrum graphs;

[0038] An electromagnetic data intelligent processing unit includes:

[0039] A time domain signal deep learning network is used for taking the time domain signal sequence as input and identifying whether there is an underwater target;

[0040] A frequency domain power spectrum signal deep learning network is configured to take a frequency domain power spectrum signal sequence as an input and identify whether there is an underwater target;

[0041] A time-frequency power spectrum signal deep learning network is configured to take a time-frequency power spectrum as an input and identify whether there is an underwater target.

[0042] The comprehensive decision unit is connected with the output end of the time domain signal deep learning network, the output end of the frequency domain power spectrum signal deep learning network, and the output end of the time-frequency power spectrum signal deep learning network, and is configured to generate a final underwater target identification result according to the identification result of the underwater target by the time domain signal deep learning network, the identification result of the underwater target by the frequency domain power spectrum signal deep learning network, and the identification result of the underwater target by the time-frequency power spectrum signal deep learning network according to a comprehensive decision rule.

[0043] In some embodiments of the present application, each of the electromagnetic detection nodes comprises an electric field sensor, a magnetic field sensor, an electric field pre-signal processing circuit, a magnetic field pre-signal processing circuit, an acquisition circuit, and a main controller; the acquisition circuit is a multi-channel synchronous sampling analog-digital conversion chip; the main controller comprises an FPGA and an MCU; the FPGA is connected with the acquisition circuit to control multi-channel synchronous sampling; the electric field sensor is connected with the acquisition circuit through the electric field pre-signal processing circuit; and the magnetic field sensor is connected with the acquisition circuit through the magnetic field pre-signal processing circuit.

[0044] In some embodiments of the present application, the electric field pre-signal processing circuit comprises an input impedance matching unit, a chopping modulation unit, an audio transformer isolation amplification unit, a high input impedance pre-amplification unit, a low-noise integrated operational amplifier amplification unit, a program-controlled gain unit, a chopping demodulation unit, a low-pass filter unit, and an output impedance matching unit connected in sequence; the input impedance matching unit is connected with the electric field sensor, and the output impedance matching unit is connected with the acquisition circuit.

[0045] The magnetic field pre-signal processing circuit comprises an input impedance matching unit, a low-noise integrated operational amplifier amplification unit, and an output impedance matching unit; the input impedance matching unit is connected with the magnetic field sensor, and the output impedance matching unit is connected with the acquisition circuit.

[0046] Compared with the prior art, the underwater target intelligent electromagnetic detection system and method provided by the present application has at least the following beneficial effects:

[0047] 1. The underwater target intelligent electromagnetic detection system and method proposed in the present application apply artificial intelligence technology to underwater target electromagnetic detection, realize underwater target intelligent electromagnetic detection, improve the detection capability of underwater targets, and have good detection effect.

[0048] 2、The underwater target intelligent electromagnetic detection system and method provided by the present application processes the collected electromagnetic signals into time domain signals, frequency domain power spectrum signals and time-frequency power spectrum signals, and learns through respective independent deep learning networks, each deep learning network can better adapt to the characteristics of various signals, thereby realizing more targeted training of the deep learning network, and further realizing more accurate underwater target recognition through comprehensive judgment.

[0049] 3、The underwater target intelligent electromagnetic detection system and method provided by the present application adopts a fixed time interval sliding time window method to intercept electromagnetic data for intelligent processing, which meets the calculation time of the intelligent processing algorithm while improving the real-time performance of underwater target detection.

[0050] 4、The underwater target intelligent electromagnetic detection system and method provided by the present application can identify the electromagnetic signal patterns of underwater targets at different time stages such as before, during and after the appearance of underwater targets, and further connect multiple electromagnetic detection nodes in sequence through a wired network to form an underwater target electromagnetic detection array, which can further identify the motion trend of the underwater target while identifying the underwater target.

[0051] 5、The underwater target intelligent electromagnetic detection system and method provided by the present application, the designed electric field pre-signal processing circuit includes sequentially connected input impedance matching unit, chopping modulation unit, audio transformer isolation amplification unit, high input impedance pre-amplification unit, low noise integrated operational amplifier amplification unit, program-controlled gain unit, chopping demodulation unit, low-pass filter unit and output impedance matching unit, which can realize nV-level ultra-low background noise signal amplification processing of extremely low frequency weak electric field signals: the low input impedance matching mode of direct current coupling is adopted to realize the access of extremely low frequency weak electric field signals; the chopping modulation amplification mode of "chopping modulation + chopping demodulation" is adopted to realize the suppression of 1 / F noise of the low noise operational amplifier; the three-stage fixed gain amplification of "audio transformer isolation amplification + high input impedance pre-amplification + low noise integrated operational amplifier amplification" is adopted to realize nV-level ultra-low background noise signal amplification processing of channel signal isolation; program-controlled gain amplification is adopted to realize a larger measurement dynamic range.

[0052] 6、The underwater target intelligent electromagnetic detection system provided by the present application can be applied to various application scenarios such as shore-based fixed electromagnetic detection, ship-borne towed electromagnetic detection and underwater mobile electromagnetic detection, and provides effective technical support for the development of underwater target electromagnetic detection technology and equipment. BRIEF DESCRIPTION OF DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments or prior art description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0054] Figure 1 is a structural schematic diagram of an underwater target intelligent electromagnetic detection system according to an embodiment of the present application;

[0055] Figure 2 is a logic structure diagram of an electromagnetic detection node according to an embodiment of the present application;

[0056] Figure 3 is a multi-channel synchronous acquisition digital logic block diagram of an electromagnetic detection node according to an embodiment of the present application;

[0057] Figure 4 is a composition block diagram of an electric field pre-signal processing circuit of an electromagnetic detection node according to an embodiment of the present application;

[0058] Figure 5 is a composition block diagram of a magnetic field pre-signal processing circuit of an electromagnetic detection node according to an embodiment of the present application;

[0059] Figure 6 is an Ethernet-based system network topology diagram of an electromagnetic detection subsystem according to an embodiment of the present application;

[0060] Figure 7 is a composition block diagram of a shipborne electromagnetic monitoring and intelligent processing subsystem according to an embodiment of the present application;

[0061] Figure 8 is a processing flow block diagram of a time-frequency domain combined underwater target electromagnetic identification intelligent processing method according to an embodiment of the present application;

[0062] Figure 9 is an intelligent electromagnetic detection comprehensive decision logic diagram of an underwater target according to an embodiment of the present application;

[0063] Figure 10 is a training data set construction flow diagram of a deep learning network according to an embodiment of the present application;

[0064] Figure 11 is a time-frequency power spectrum marked as a pre-stage according to an embodiment of the present application;

[0065] Figure 12 is a time-frequency power spectrum marked as a middle stage according to an embodiment of the present application;

[0066] Figure 13is a time-frequency power spectrum marked as a later stage according to the embodiment of the application;

[0067] Figure 14 is a schematic diagram of a joint detection method of underwater targets of a multi-node electromagnetic detection system according to the embodiment of the application;

[0068] Figure 15 is a background noise test result of an electric field and a magnetic field channel of an electromagnetic detection node according to the embodiment of the application;

[0069] Figure 16 is a wired network real-time transmission performance test result of an electromagnetic detection node according to the embodiment of the application;

[0070] Figure 17 is a network structure and parameters of a YOLOv5s lightweight underwater target detection network according to the embodiment of the application;

[0071] Figure 18 is a magnetic field test result of an underwater target electromagnetic identification intelligent processing method at a certain moment according to the embodiment of the application;

[0072] Figure 19 is a magnetic field test result of an underwater target electromagnetic identification intelligent processing method at a certain moment according to the embodiment of the application;

[0073] Figure 20 is a magnetic field test result of an underwater target electromagnetic identification intelligent processing method at a certain moment according to the embodiment of the application;

[0074] Figure 21 is an electric field test result of an underwater target electromagnetic identification intelligent processing method at a certain moment according to the embodiment of the application;

[0075] Figure 22 is an electric field test result of an underwater target electromagnetic identification intelligent processing method at a certain moment according to the embodiment of the application;

[0076] Figure 23 is an electric field test result of an underwater target electromagnetic identification intelligent processing method at a certain moment according to the embodiment of the application;

[0077] Figure 24 is a schematic diagram of a shore-based fixed electromagnetic detection application according to the embodiment of the application;

[0078] Figure 25 is a schematic diagram of a ship-mounted towed electromagnetic detection application according to the embodiment of the application;

[0079] Figure 26 is a schematic diagram of an underwater mobile electromagnetic detection application mounted on an AUV according to the embodiment of the application;

[0080] Figure 27 is a flow chart of an intelligent electromagnetic detection method for underwater targets according to an embodiment of the present application. DETAILED DESCRIPTION

[0081] In order to make the technical problems to be solved by the present application, technical solutions and beneficial effects clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0082] In the embodiments of the present application, the prefix words such as "first", "second" are only used to distinguish different description objects, and have no limiting effect on the position, order, priority, quantity or content of the described objects. The use of ordinal words such as "first", "second" in the embodiments of the present application does not limit the described objects, and the description of the described objects should be referred to the description of the context in the embodiments, and should not be considered as redundant limitation because of the use of such prefix words. In addition, in the description of the embodiments, unless otherwise specified, the meaning of "multiple" is two or more than two.

[0083] The technical solutions in the embodiments of the present application will be described below with reference to the drawings in the embodiments of the present application. In the description of the embodiments of the present application, unless otherwise specified, " / " represents the meaning of or, for example, A / B can represent A or B; "and / or" in this paper is only a description of the association between the associated objects, which means that there can be three kinds of relationships, for example, A and / or B, which can represent three cases of A alone, A and B together, and B alone.

[0084] In several embodiments provided in the embodiments of the present application, it should be understood that the disclosed system and method can be implemented in other ways. For example, the above-described device embodiments are only schematic, for example, the division of the units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the shown or discussed objects can be indirect coupling or communication connection through some interfaces, devices or units, and can be electrical, mechanical or other forms.

[0085] The embodiments of the present application provide an intelligent electromagnetic detection system for underwater targets, and the composition of the electromagnetic detection system is referred to Figure 1 , which includes an electromagnetic detection subsystem and an electromagnetic monitoring and intelligent processing subsystem.

[0086] The electromagnetic detection subsystem comprises a plurality of electromagnetic detection nodes for collecting underwater electromagnetic signals; the electromagnetic detection nodes are sequentially connected with each other, and each electromagnetic detection node comprises an electric field sensor, a magnetic field sensor, a pre-signal processing circuit, a main controller and a collection circuit. For example, the electric field sensor uses an Ag / AgCl electrode or a carbon fiber electrode which can be used in a marine environment; the magnetic field sensor uses a three-axis magnetic flux gate sensor.

[0087] Specifically, the electromagnetic detection nodes are arranged at intervals underwater and can collect electromagnetic signals at different positions underwater. The electromagnetic detection nodes are connected with each other through a wired network, and the logical structure diagram of each electromagnetic detection node is shown in Figure 2 . The sensing unit comprises an electric field sensor and a magnetic field sensor; the main controller comprises an FPGA and an MCU. The MCU is connected with an SD card and a network communication interface, and is responsible for system management, data storage and data transmission; the FPGA is connected with two 24-bit 4-channel synchronous sampling analog-to-digital conversion chips (ADCs), and the two 24-bit 4-channel synchronous sampling ADCs are connected with the electric field sensor and the magnetic field sensor respectively, so as to realize 24-bit 8-channel synchronous signal collection; a high-precision clock provides a high-precision time base signal for synchronous collection, data storage and data transmission of the MCU and the FPGA.

[0088] In order to ensure the accuracy of underwater target identification, the synchronism of collected electromagnetic signals needs to be ensured. Based on this, the present application provides a synchronous collection logical architecture and method of an electromagnetic detection node.

[0089] Referring to Figure 3 , in the present application, the synchronous collection of signals is realized based on an FPGA. The digital logic of the FPGA comprises:

[0090] 4-channel synchronous sampling ADC sampling control logic unit: the 4-channel synchronous sampling ADC sampling control logic function comprises an interface signal timing, register configuration and sampling state timing of the 4-channel synchronous sampling ADC, so as to realize collection data acquisition of the 4-channel synchronous sampling ADC;

[0091] Channel data extraction logic unit: used for data extraction of each channel, so as to realize channel data separation;

[0092] Data time stamping logic unit: adding a high-precision time stamp generated by an RTC clock logic to a group of data of each channel according to a data storage format;

[0093] Pong RAM data cache unit: used for continuous ADC sampling of each channel;

[0094] Data cache interrupt unit: used for generating an interrupt signal to inform the MCU to read cache data when the pong RAM data caches of the 8 channels are full;

[0095] RTC clock logic unit: used for providing high-precision clock for 4-channel synchronous sampling ADC sampling control logic and data time stamping logic to realize 8-channel high-precision synchronous acquisition.

[0096] In order to realize high-precision detection of electromagnetic signals, further pre-processing of the collected electromagnetic signals is realized through the pre-signal processing circuit, and the processed electromagnetic signals are input to the 4-channel synchronous sampling ADC. Due to the different characteristics of electric signals and magnetic signals, in the embodiment of the application, an electric field pre-signal processing circuit and a magnetic field pre-signal processing circuit are further designed.

[0097] The electric field pre-signal processing circuit comprises an input impedance matching unit, a chopping modulation unit, an audio transformer isolation amplification unit, a high input impedance pre-amplification unit, a low-noise integrated operational amplifier amplification unit, a programmable gain unit, a chopping demodulation unit, a low-pass filter unit and an output impedance matching unit connected in sequence; the input impedance matching unit is connected with the electric field sensor, and the output impedance matching unit is connected with the 4-channel synchronous sampling ADC of the main control unit; the magnetic field pre-signal processing circuit comprises an input impedance matching unit, a low-noise integrated operational amplifier amplification unit and an output impedance matching unit, the input impedance matching unit is connected with the magnetic field sensor, and the output impedance matching unit is connected with the 4-channel synchronous sampling ADC of the main control unit.

[0098] Reference Figure 4 The circuit is one of the core technologies for weak electric field signal detection and processing, and realizes nV-level ultra-low noise signal amplification processing of extremely low frequency weak electric field signals. The direct current coupling low input impedance matching mode is adopted to realize the access of extremely low frequency weak electric field signals; the chopping modulation amplification mode of “chopping modulation + chopping demodulation” is adopted to realize the suppression of 1 / F noise of the low-noise operational amplifier; the three-stage fixed gain amplification of “audio transformer isolation amplification + high input impedance pre-amplification + low-noise integrated operational amplifier amplification” is adopted to realize nV-level ultra-low noise signal amplification processing of channel signal isolation; the programmable gain amplification is adopted to realize a larger measurement dynamic range. The low-pass filter at the rear end of the circuit is used to filter out the chopping modulation noise, and the output impedance matching is used to realize the output of the low-output-resistance amplified signal.

[0099] Reference Figure 5 The circuit realizes low-noise amplification processing of the output signal of the magnetic field sensor, and is realized by using a low-noise integrated operational amplifier. The circuit comprises input impedance matching, low-noise integrated operational amplifier amplification and output impedance matching, the input impedance matching realizes the access of high-input-impedance magnetic field signals, the low-noise integrated operational amplifier amplification realizes the amplification processing of the magnetic field signals, and the output impedance matching realizes the output of the low-output-resistance magnetic field amplified signal.

[0100] In some embodiments of the present application, each electromagnetic detection node is connected through a wired network. Each electromagnetic detection node can be connected to the electromagnetic monitoring and intelligent processing subsystem, or can be ultimately connected to the electromagnetic monitoring and intelligent processing subsystem through a master electromagnetic detection node.

[0101] As shown in Figure 6 , the wired network uses Ethernet. Each electromagnetic detection node is configured with a network communication interface. In the embodiments of the present application, each electromagnetic detection node is configured with a three-port Ethernet router. Each three-port Ethernet router includes a WAN interface and two LAN interfaces (LAN1 and LAN2). Among them, the LAN1 of the current electromagnetic detection node is connected to its electromagnetic detection circuit, and the LAN2 interface of the current electromagnetic detection node is connected to the WAN interface of the rear electromagnetic detection node.

[0102] The underwater electromagnetic data collected by the electromagnetic detection subsystem contains the characteristics of underwater targets. The electromagnetic data is transmitted to the electromagnetic monitoring and intelligent processing subsystem, and the underwater target signal recognition is completed by the electromagnetic monitoring and intelligent processing subsystem.

[0103] The structure of the electromagnetic monitoring and intelligent processing subsystem is shown in Figure 7 , which is connected to the electromagnetic detection subsystem through a network communication interface, and includes:

[0104] The electromagnetic monitoring unit is configured with a network communication interface and is connected to an electromagnetic detection node through the network communication interface. Since the electromagnetic detection nodes are connected in series, the electromagnetic monitoring unit can obtain the underwater electromagnetic signals collected by each electromagnetic detection node. The underwater electromagnetic data received through the network communication interface is stored in the electromagnetic database.

[0105] The underwater target recognition intelligent processing unit obtains electromagnetic data from the electromagnetic database for processing, which includes:

[0106] The electromagnetic data preprocessing unit is used to denoise the underwater electromagnetic data obtained by the electromagnetic monitoring unit, and further decompose it into time domain signal sequence, frequency domain power spectrum signal sequence and time-frequency power spectrum diagram.

[0107] The electromagnetic data intelligent processing unit includes:

[0108] The time domain signal deep learning network is used to take the time domain signal sequence as input and identify whether there is an underwater target.

[0109] The frequency domain power spectrum signal deep learning network is used to take the frequency domain power spectrum signal sequence as input and identify whether there is an underwater target.

[0110] The time-frequency power spectrum signal deep learning network is used to take the time-frequency power spectrum diagram as input and identify whether there is an underwater target.

[0111] The comprehensive decision unit is connected with the output end of the time domain signal deep learning network, the output end of the frequency domain power spectrum signal deep learning network, and the output end of the time-frequency power spectrum signal deep learning network, and is configured to generate a final underwater target recognition result according to the recognition result of the underwater target by the time domain signal deep learning network, the recognition result of the underwater target by the frequency domain power spectrum signal deep learning network, and the recognition result of the underwater target by the time-frequency power spectrum signal deep learning network according to a comprehensive decision rule.

[0112] In some embodiments of the present application, the electromagnetic monitoring unit includes an output storage module and a data display module. The data storage module uses a data storage device to store the acquired electromagnetic data. The data display module uses a data display device to display the real-time electromagnetic data of each electromagnetic detection node and to display the underwater target recognition result according to the output information of the electromagnetic data intelligent processing method. The underwater target recognition intelligent processing unit acquires the electromagnetic data of the electromagnetic database, runs the electromagnetic data preprocessing method and the underwater target recognition intelligent processing method, and realizes underwater target recognition.

[0113] It should be noted that the electromagnetic data of each electromagnetic detection node of the underwater target intelligent electromagnetic detection system is processed independently. After the electromagnetic data of each electromagnetic detection node is acquired and preprocessed, the time domain signal sequence, the frequency domain power spectrum signal sequence, and the time-frequency power spectrum graph are used as network inputs, respectively, and three independent deep learning networks are used to recognize underwater target information. The recognition results of the three deep learning networks are comprehensively decided to finally give the underwater target recognition result of each electromagnetic node labeled by the time-frequency power spectrum graph.

[0114] It should be understood that the three deep learning network models can be the same or different, and the three deep learning network models can select a quantized network or a lightweight network. For example, the deep learning network can select a typical LeNet, AlexNet, VGG, ResNet, GoogleNet, or other quantized networks or YOLO, SqueezeNet, MobileNet, ShuffleNet, or other lightweight networks according to different computing platform resources. Figure 17 An example of a deep learning neural network model is given.

[0115] Based on the above underwater target intelligent electromagnetic detection system, the present application further provides an underwater target intelligent electromagnetic detection method.

[0116] S1: Collecting underwater electromagnetic signals to obtain electromagnetic data samples of the underwater electromagnetic signals.

[0117] Specifically, the present application is through the electromagnetic detection subsystem to collect underwater electromagnetic signals, and obtain underwater electromagnetic data. After the electromagnetic detection subsystem obtains the underwater electromagnetic data, the electromagnetic monitoring and intelligent processing subsystem is transmitted, and the electromagnetic monitoring unit stores the underwater electromagnetic data in the electromagnetic database.

[0118] In the process of obtaining electromagnetic data samples, the collected electromagnetic data is intercepted by a fixed time length time sliding window;

[0119] The fixed time length time sliding window is moved at a fixed time interval, and the electromagnetic data intercepted by the time sliding window at all time interval moments is taken as the electromagnetic data sample.

[0120] Specifically, in order to improve the real-time performance of underwater target electromagnetic identification, the present application adopts a sliding time window method to obtain electromagnetic data input into the intelligent processing network each time, intercepts electromagnetic data by a time window of a certain time length (such as 2 hours), and slides the time window at a certain time interval (such as 5 minutes) to obtain electromagnetic data to be processed from the electromagnetic database in real time. The sliding time window method (usually referred to as the sliding window method) is a data processing technology for analyzing local features by moving a fixed size window on a data sequence. The method maintains a fixed size window, which is gradually slid on the data sequence by moving one or more units each time. The window contains the data points currently processed, and the starting and ending positions of the window are dynamically adjusted to efficiently calculate statistics (such as sum, maximum value) or detect patterns.

[0121] In order to improve the identification accuracy, in step S1, after collecting the underwater electromagnetic data, the underwater electromagnetic data is further subjected to denoising processing, and the underwater electromagnetic data after denoising processing is taken as the data sample. The denoising processing method includes but is not limited to: filtering processing, normalization processing, etc.

[0122] S2: decompose the electromagnetic data sample into a time domain signal sequence, a frequency domain power spectrum signal sequence and a time-frequency power spectrum diagram.

[0123] S3: input the time domain signal sequence into the time domain signal deep learning network, identify whether the time domain signal sequence exists underwater target, input the frequency domain power spectrum signal sequence into the frequency domain power spectrum signal deep learning network, identify whether the frequency domain power spectrum signal sequence exists underwater target, and input the time-frequency power spectrum diagram into the time-frequency power spectrum signal deep learning network. Identify whether the time-frequency power spectrum diagram exists underwater target.

[0124] In some embodiments of the present application, after decomposing the electromagnetic data samples into time domain signal sequences, frequency domain power spectrum signal sequences, and time-frequency power spectrum signal, the electromagnetic data intercepted by a fixed time length time sliding window is input into the time domain signal deep learning network as the time domain signal sequence; the electromagnetic data intercepted by the fixed time length time sliding window is calculated by power spectrum estimation method to obtain the frequency domain power spectrum data as the frequency domain power spectrum signal sequence, which is input into the frequency domain power spectrum signal deep learning network; the electromagnetic data intercepted by the fixed time length time sliding window is calculated by power spectrum estimation method and sliding time window method to obtain the time-frequency power spectrum data of the fixed time length time sliding window as the time-frequency power spectrum graph, which is input into the time-frequency power spectrum signal deep learning network.

[0125] For example, the preprocessed electromagnetic data is input into the network as three kinds of signal forms, including time domain signal sequences, frequency domain power spectrum signal sequences, and time-frequency power spectrum graphs. The time domain signal sequence is the electromagnetic data intercepted by a certain time window (such as 2 hours) which is directly input into the intelligent processing network; the frequency domain power spectrum signal sequence is the electromagnetic data intercepted by a certain time window (such as 2 hours) which is converted into frequency domain power spectrum data by Welch method and then input into the intelligent processing network; the time-frequency power spectrum graph is the electromagnetic data intercepted by a certain time window (such as 2 hours) which is converted into time-frequency power spectrum graph by Welch method and sliding time window method and then input into the intelligent processing network. The time window length for time-frequency power spectrum graph calculation is 128 points, 50% overlap rate is used to reduce spectral leakage and improve spectral resolution, and Hanning window function is used to reduce edge effect and improve smoothness.

[0126] S4: generating the final underwater target recognition result according to the recognition result of the underwater target by the time domain signal deep learning network, the recognition result of the underwater target by the frequency domain power spectrum signal deep learning network, and the recognition result of the underwater target by the time-frequency power spectrum signal deep learning network. The underwater target recognition result labeled by the time-frequency spectrum graph is given according to the comprehensive decision result.

[0127] In some embodiments of the present application, according to the characteristics of different signals, the network outputs of the two deep learning networks for time domain signal sequences and frequency domain power spectrum signal sequences are binary outputs of target and no target, for example, target is labeled as 1 and no target is labeled as 0. The network output of the deep learning network for time-frequency power spectrum graph is the underwater target recognition output labeled by the time-frequency power spectrum graph.

[0128] In some embodiments of the present application, the recognition results of the underwater target by each learning network are integrated to generate a final underwater target recognition result, which specifically includes a first intensity level recognition target, a second intensity level recognition target, a third intensity level recognition target, a fourth intensity level recognition target, and a no recognition target. In the order of the first intensity level recognition target, the second intensity level recognition target, the third intensity level recognition target, the fourth intensity level recognition target, and the no recognition target, the accuracy of target recognition gradually decreases. Figure 9 The specific recognition method is as follows.

[0129] If the time domain signal deep learning network, the frequency domain power spectrum signal deep learning network, and the time-frequency power spectrum signal deep learning network all recognize the existence of the underwater target, it is judged as the first intensity level recognition target.

[0130] If only one of the time domain signal deep learning network and the frequency domain power spectrum signal deep learning network recognizes the existence of the underwater target, and the time-frequency power spectrum signal deep learning network recognizes the existence of the underwater target, it is judged as the second intensity level recognition target.

[0131] If the time domain signal deep learning network and the frequency domain power spectrum signal deep learning network both recognize the non-existence of the underwater target, and the time-frequency power spectrum signal deep learning network recognizes the existence of the underwater target, it is judged as the third intensity level recognition target.

[0132] If the time domain signal deep learning network and the frequency domain power spectrum signal deep learning network both recognize the existence of the underwater target, and the power spectrum signal learning network recognizes the non-existence of the underwater target, it is judged as the fourth intensity level recognition target.

[0133] If only one of the time domain signal deep learning network and the frequency domain power spectrum signal deep learning network recognizes the existence of the underwater target or both recognize the non-existence of the underwater target, and the time-frequency power spectrum signal deep learning network recognizes the non-existence of the underwater target, it is judged as the no recognition target.

[0134] In some embodiments of the present application, the decision results are divided into no target, fourth intensity level recognition target, third intensity level recognition target, second intensity level recognition target, and first intensity level recognition target. In the case of the decision results being the first intensity level recognition target, the second intensity level recognition target, and the third intensity level recognition target, the deep learning network with the time-frequency power spectrum as the network input can output a network output result with a target. The target signal form recognized by the time-frequency power spectrum can be labeled as three different target appearance time stages, i.e., target before, target in, and target after. The target before is used to represent that the time of the current data frame is in the first half of the underwater target feature, the target in is used to represent that the underwater target feature is completely present in the current data frame, and the target after is used to represent that the time of the current data frame is in the second half of the underwater target feature.

[0135] Specifically, the deep learning network training of the underwater target electromagnetic identification intelligent processing method based on deep learning network of time-frequency domain joint in the embodiments of the present application needs a data set, and the construction method of the data set is as shown in Figure 10 The electromagnetic data acquisition and preprocessing method and the intelligent processing method are the same, the training data set of the time domain signal sequence and the frequency domain power spectrum signal sequence is marked as corresponding underwater target presence / absence according to the time period information of the acquired electromagnetic data and the underwater target presence / absence information in the time period; the training data set of the time-frequency power spectrum diagram is further marked as electric field and magnetic field characteristic signals according to the generated time-frequency power spectrum diagram, and is divided into three stages of front, middle and rear according to the signal feature form, and an example of the training set marking of the time-frequency power spectrum diagram is as shown in Figures 11 to 13 .

[0136] In some embodiments of the present application, in order to obtain more sufficient electromagnetic detection data, the electromagnetic detection subsystem includes a plurality of electromagnetic detection nodes distributed at different positions underwater, and these electromagnetic detection nodes can collect electromagnetic detection signals at different positions underwater. The underwater target is usually mobile, and the mobility of the underwater target can be reflected according to the electromagnetic detection signals at different positions and different times underwater. Therefore, step S1 further includes: collecting underwater electromagnetic signals by electromagnetic detection nodes arranged at different positions through a wired network, obtaining underwater electromagnetic data and underwater electromagnetic data samples; and step S4 further includes: comprehensively identifying the underwater target and the motion trend of the underwater target by using the underwater electromagnetic data samples obtained by the underwater electromagnetic data collected by each electromagnetic detection node.

[0137] The underwater target intelligent electromagnetic detection system and method provided by the embodiments of the present application are tested, and the test results are as follows.

[0138] (1) Background noise test results of electric field and magnetic field channels

[0139] The background noise test results of the electric field and magnetic field channels of the electromagnetic detection node are as shown in Figure 15 Ex, Ey, and Ez are three electric field channels, using an electric field pre-signal processing circuit, and Mx, My, Mz, Hx, and Hy are five magnetic field channels, using a magnetic field pre-signal processing circuit. As can be seen, the background noise of the electric field channel is 1.585~3nV / rt(Hz)@1Hz; the background noise of the magnetic field channel is 1.579uV / rt(Hz)@1Hz, and the connection sensitivity of the magnetic flux gate sensor is 100mV / uT, and the background noise of the magnetic field channel is 15.79pT / rt(Hz)@1Hz, which is equivalent to the noise level of the magnetic flux gate sensor with a background noise of 10~20 pT / rt(Hz)@1Hz. It can meet the collection requirements of nV-level electric field and pT-level magnetic field signals of underwater targets.

[0140] (2) The test result of the real-time transmission performance of the wired network

[0141] The real-time transmission performance of the wired network guarantees the real-time detection capability of the underwater target electromagnetic detection system of multiple electromagnetic detection nodes, Figure 16 The test result of the network transmission capability of the STM32H7 series high-performance MCU is shown, and it can be seen that the network transmission rate is between 4MB / s and 7MB / s, and the average network transmission rate is 4.9MB / s. According to the network transmission efficiency of 40%, it can theoretically meet the data transmission requirements of about 20 500Hz sampling 8-channel electromagnetic detection nodes.

[0142] (3) The test result of the underwater target electromagnetic identification intelligent processing method

[0143] Figure 17 The network structure and parameters of the YOLOv5s lightweight underwater target detection network are shown in the figure. In the figure, Input is the network input, Backbone performs feature extraction, Neck is responsible for fusing the three scale feature maps extracted from the shallow layer features, and Head performs target prediction. In the specific component module, Conv is a convolutional layer module, C3_1, C3_2, C3_3 are modules containing three convolutional layers and a bottleneck layer, SPPF is a spatial pyramid pooling module, Upsample is an up-sampling module, Concat is a feature fusion module, and Detect is a target detection module. Figures 18 to 23 The time-frequency spectrum labeling identification of the underwater target electromagnetic identification intelligent processing method using the YOLOv5s lightweight target detection network is shown in the figure, Figures 18 to 23 In the figure, c1, c2, and c3 are the front, middle, and back three stages of the underwater target magnetic field signal, Figures 18 to 23 In the figure, d1, d2, and d3 are the front, middle, and back three stages of the underwater target electric field signal. It can be seen that the method of the present application using YOLOv5s can clearly distinguish the time-frequency spectrum information of each different stage of the underwater target electric field and magnetic field, and has good intelligent identification capability.

[0144] The underwater target intelligent electromagnetic detection system provided by the embodiments of the present application mainly faces the shore-based fixed electromagnetic detection and ship-borne towed electromagnetic detection application scenarios based on wired networks. The simplified single electromagnetic detection node system can be applied to mobile electromagnetic detection application scenarios based on underwater mobile platforms.

[0145] (1) Shore-based fixed electromagnetic detection application scenario

[0146] Figure 24The figure shows the application scenario of the shore-based fixed electromagnetic detection of the application, in which the system includes an electromagnetic monitoring and intelligent processing subsystem deployed on the shore and an electromagnetic detection subsystem deployed in the near-shore water.

[0147] (2) Ship-borne towed electromagnetic detection application scenario

[0148] Figure 25 The figure shows the application scenario of the ship-borne towed electromagnetic detection of the application, in which the system includes an electromagnetic monitoring and intelligent processing subsystem deployed on the ship and an electromagnetic detection subsystem towed in the water.

[0149] (3) Electromagnetic detection application scenario of carrying an AUV and other underwater mobile platforms

[0150] Figure 26 The figure shows the application scenario of the electromagnetic detection of the application simplified as a single electromagnetic node applied to the electromagnetic detection application scenario of carrying an AUV and other underwater mobile platforms, in which the electromagnetic monitoring and intelligent processing subsystem and the electromagnetic detection node are both deployed inside the underwater mobile platform.

[0151] The above is merely a specific implementation of the application, but the protection scope of the application is not limited to this, and any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the application, which should be covered.

Claims

1. An intelligent electromagnetic detection method for underwater targets, characterized in that, The method comprises the following steps: S1: collecting underwater electromagnetic signals to obtain underwater electromagnetic data and electromagnetic data samples; S2: decomposing the electromagnetic data samples into time domain signal sequences, frequency domain power spectrum signal sequences, and time-frequency power spectrum diagrams; S3: inputting the time domain signal sequences into a time domain signal deep learning network to identify whether the time domain signal sequences contain underwater targets, inputting the frequency domain power spectrum signal sequences into a frequency domain power spectrum signal deep learning network to identify whether the frequency domain power spectrum signal sequences contain underwater targets, and inputting the time-frequency power spectrum diagrams into a time-frequency power spectrum signal deep learning network to identify whether the time-frequency power spectrum diagrams contain underwater targets; S4: comprehensively determining to generate a final underwater target recognition result labeled by a time-frequency power spectrum diagram according to the recognition results of underwater targets by the time domain signal deep learning network, the recognition results of underwater targets by the frequency domain power spectrum signal deep learning network, and the recognition results of underwater targets by the time-frequency power spectrum signal deep learning network; The comprehensive determination of step S4 to generate a final underwater target recognition result labeled by a time-frequency power spectrum diagram further comprises: if the time domain signal deep learning network, the frequency domain power spectrum signal deep learning network, and the time-frequency power spectrum signal deep learning network all identify that underwater targets exist, the first intensity level recognition target is determined; if only one of the time domain signal deep learning network and the frequency domain power spectrum signal deep learning network identifies that underwater targets exist, and the time-frequency power spectrum signal deep learning network identifies that underwater targets exist, the second intensity level recognition target is determined; if the time domain signal deep learning network and the frequency domain power spectrum signal deep learning network both identify that underwater targets do not exist, and the time-frequency power spectrum signal deep learning network identifies that underwater targets exist, the third intensity level recognition target is determined; if the time domain signal deep learning network and the frequency domain power spectrum signal deep learning network both identify that underwater targets exist, and the time-frequency power spectrum signal deep learning network identifies that underwater targets do not exist, the fourth intensity level recognition target is determined; if only one of the time domain signal deep learning network and the frequency domain power spectrum signal deep learning network identifies that underwater targets exist or both identify that underwater targets do not exist, and the time-frequency power spectrum signal deep learning network identifies that underwater targets do not exist, no recognition target is determined; the first intensity level recognition target, the second intensity level recognition target, the third intensity level recognition target, the fourth intensity level recognition target, and no recognition target are arranged in order of gradually decreasing accuracy of target recognition.

2. The intelligent electromagnetic detection method of underwater targets according to claim 1, characterized in that, Step S1 further comprises: cutting the collected electromagnetic data by using a fixed time length time sliding window; moving the time sliding window at a fixed time interval, and taking the electromagnetic data cut by the time sliding window at all time interval moments as electromagnetic data samples.

3. The intelligent electromagnetic detection method of underwater targets according to claim 2, characterized in that, Further comprising: inputting the electromagnetic data cut by the fixed time length time sliding window into the time domain signal deep learning network as time domain signal sequences; The electromagnetic data intercepted by the fixed time length time sliding window is calculated by power spectrum estimation method to obtain frequency domain power spectrum data as a frequency domain power spectrum signal sequence, which is input into a frequency domain power spectrum signal deep learning network; The electromagnetic data intercepted by the fixed time length time sliding window is calculated by power spectrum estimation method and sliding time window method to obtain time-frequency power spectrum data of the fixed time length time sliding window as a time-frequency power spectrum graph, which is input into a time-frequency power spectrum signal deep learning network.

4. The underwater target intelligent electromagnetic detection method according to claim 1, characterized in that: The step S1 further comprises: collecting underwater electromagnetic signals by electromagnetic detection nodes arranged at different positions and connected in sequence through a wired network to obtain underwater electromagnetic data and electromagnetic data samples; The underwater target intelligent electromagnetic detection method further comprises: comprehensively integrating underwater electromagnetic data samples of each electromagnetic detection node to identify underwater targets and motion trends of the underwater targets.

5. The intelligent electromagnetic detection of underwater targets method of claim 1, wherein, The step S1 further comprises: collecting underwater electromagnetic signals to obtain underwater electromagnetic data, and performing denoising processing on the underwater electromagnetic data, and taking the denoised underwater electromagnetic data as electromagnetic data samples.

6. The intelligent electromagnetic detection of underwater targets method of claim 1, wherein, Further comprising a network model training step, comprising: According to the identification result of the underwater target by the time domain signal deep learning network, the time domain signal sequence is labeled as underwater target time domain signal characteristics and non-underwater target time domain signal characteristics, and the labeled time domain signal sequence is constructed as a time sequence signal deep learning network training set for training the time sequence signal deep learning network; According to the identification result of the underwater target by the frequency domain power spectrum signal deep learning network, the frequency domain power spectrum signal sequence is labeled as underwater target frequency domain power spectrum signal characteristics and non-underwater target frequency domain power spectrum signal characteristics, and the labeled frequency domain power spectrum signal sequence is constructed as a frequency domain power spectrum signal deep learning network training set for training the frequency domain power spectrum signal deep learning network; According to the identification result of the underwater target by the time-frequency power spectrum signal deep learning network, the time-frequency power spectrum graph is labeled as underwater target time-frequency power spectrum signal characteristics and non-underwater target time-frequency power spectrum signal characteristics, the underwater target electric field and magnetic field characteristic signals are distinguished, and the time of the time-frequency power spectrum graph is labeled as a pre-underwater target appearance stage, an underwater target appearance stage or a post-underwater target appearance stage; the labeled time-frequency power spectrum graph is constructed as a time-frequency power spectrum signal deep learning network training set for training the time-frequency power spectrum signal deep learning network.

7. An intelligent electromagnetic detection system for underwater targets, characterized in that, The underwater target intelligent electromagnetic detection method according to any one of claims 1 to 6, comprising: An electromagnetic detection subsystem: comprising a plurality of electromagnetic detection nodes, each electromagnetic detection node being sequentially connected through a wired network, for collecting underwater electromagnetic signals to obtain electromagnetic data samples of the underwater electromagnetic signals; An electromagnetic monitoring and intelligent processing subsystem: comprising: An electromagnetic monitoring unit: in communication with the electromagnetic detection subsystem through the wired network, acquiring electromagnetic data samples of the underwater electromagnetic signals and storing them in an electromagnetic database; An electromagnetic monitoring unit: in communication with the electromagnetic detection subsystem through the wired network, acquiring electromagnetic data samples of the underwater electromagnetic signals and storing them in an electromagnetic database; The electromagnetic data preprocessing unit is configured to denoise electromagnetic data samples of underwater electromagnetic signals acquired by the electromagnetic monitoring unit, and further decompose the electromagnetic data samples into time-domain signal sequences, frequency-domain power spectrum signal sequences and time-frequency power spectrum maps; The electromagnetic data intelligent processing unit comprises: a time-domain signal deep learning network configured to take the time-domain signal sequences as input and identify whether there is an underwater target; a frequency-domain power spectrum signal deep learning network configured to take the frequency-domain power spectrum signal sequences as input and identify whether there is an underwater target; a time-frequency power spectrum signal deep learning network configured to take the time-frequency power spectrum maps as input and identify whether there is an underwater target; The comprehensive decision unit is connected with output ends of the time-domain signal deep learning network, the frequency-domain power spectrum signal deep learning network and the time-frequency power spectrum signal deep learning network, and is configured to generate a final underwater target identification result according to an underwater target identification result of the time-domain signal deep learning network, an underwater target identification result of the frequency-domain power spectrum signal deep learning network and an underwater target identification result of the time-frequency power spectrum signal deep learning network according to a comprehensive decision rule.

8. The intelligent electromagnetic underwater target detection system of claim 7, wherein, Each electromagnetic detection node comprises an electric field sensor, a magnetic field sensor, an electric field pre-signal processing circuit, a magnetic field pre-signal processing circuit, an acquisition circuit and a main controller; the acquisition circuit is a multi-channel synchronous sampling analog-digital conversion chip; the main controller comprises an FPGA and an MCU; the FPGA is connected with the acquisition circuit to control multi-channel synchronous sampling; the electric field sensor is connected with the acquisition circuit through the electric field pre-signal processing circuit; and the magnetic field sensor is connected with the acquisition circuit through the magnetic field pre-signal processing circuit.

9. The underwater target intelligent electromagnetic detection system according to claim 8, wherein: the electric field pre-signal processing circuit comprises sequentially connected input impedance matching units, chopping modulation units, audio transformer isolation amplification units, high input impedance pre-amplification units, low-noise integrated operational amplifier amplification units, program-controlled gain units, chopping demodulation units, low-pass filter units and output impedance matching units; the input impedance matching units are connected with the electric field sensor, and the output impedance matching units are connected with the acquisition circuit; the magnetic field pre-signal processing circuit comprises input impedance matching units, low-noise integrated operational amplifier amplification units and output impedance matching units; the input impedance matching units are connected with the magnetic field sensor, and the output impedance matching units are connected with the acquisition circuit.

Images