A long-range high-precision marine target automatic recognition and determination method and system

Abstract

The present invention specifically discloses a method and system for automatic identification and determination of long-range high-precision marine targets, the method comprising the following steps: S1, training an acoustic model network and pre-storing it in a signal processing subcard; S2, the signal processing subcard receives the interference light signal of the optical fiber hydrophone array cable; S3, the analog-to-digital conversion module performs photoelectric conversion and analog-to-digital conversion on the received interference light signal and sends it to the corresponding modulation and demodulation module for phase modulation and demodulation to restore the sound signal of the marine target, and then uses the corresponding AI recognition module to perform voice extraction and target matching on the signal; S4, the host computer performs data scheduling and runs the pre-stored display and control software. The present invention uses the networked optical fiber hydrophone array cable to collect the sound signal of the marine target, and uses multiple signal processing subcards to perform parallel calculations on multiple data channels, effectively improving the target detection area and data processing efficiency, so that the target can be matched and identified accurately and quickly.

Description

Technical Field

[0001] The present invention relates to the technical field of marine target recognition, and in particular to a method and system for automatically recognizing and determining a remote and high-precision marine target. Background Art

[0002] Traditional marine ship identification requires drones to collect target images and make target judgments through the identification system. This is limited by the drone's detection range and the degree of target camouflage. At the same time, the amount of target image data collected by drones is large and not conducive to transmission, and the real-time performance is poor. Drones also have the disadvantage of being easily detected and destroyed. Moreover, marine underwater target identification generally requires sonar devices. Traditional sonar devices have low sensitivity, short detection distance and difficulty in networking.

[0003] In view of this, studying a long-range high-precision ocean target automatic recognition and determination method and system is a technical problem that people in this technical field need to solve urgently. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a method for automatic identification and determination of long-range high-precision marine targets, the method comprising the following steps:

[0005] S1. Train an acoustic model network based on the specific sound frequency of the ocean target sample, and pre-store the trained acoustic model network in the signal processing subcard;

[0006] S2, using an optical transmitter to send modulated light to a fiber optic hydrophone array cable, and sending the interference light signal containing the sound signal in the detection area to a plurality of signal processing subcards that are arranged in parallel and can process the interference light signal in parallel through the fiber optic hydrophone array cable;

[0007] S3. The analog-to-digital conversion module in each signal processing subcard performs analog-to-digital conversion on the received interference light signal to convert the optical signal of the interference light signal into an electrical signal, and then sends the electrical signal corresponding to the interference light signal to the corresponding modulation and demodulation module for phase modulation and demodulation, thereby restoring the sound signal of the ocean target, and then uses the corresponding AI recognition module to perform sound recognition and target matching on the restored ocean target sound signal, and finally sends the matching result to the host computer;

[0008] S4. The host computer receives the matching result sent by the signal processing subcard, and then displays the matching result and plays back the information through the display control software pre-stored in the host computer.

[0009] Preferably, the acoustic model network is a CD-DNN-HMM architecture acoustic model network.

[0010] Preferably, the specific process of phase adjustment and demodulation performed by the modulation and demodulation module in step S3 is: the modulation and demodulation module performs mixing and multiplication, low-pass filtering, differential cross multiplication, subtraction and integration operations on the electrical signal corresponding to the interfering optical signal in sequence to realize phase modulation and demodulation of the signal, and then restore the sound signal of the ocean target.

[0011] Preferably, the specific implementation method of the AI ​​recognition module in step S3 performing sound recognition and target matching on the restored ocean target sound signal includes:

[0012] S31, performing speech framing on the received sound signal by using a moving window function, and then performing MFCC acoustic feature extraction on the sound signal after the speech framing, so as to convert the interference light signal corresponding to each frame of the sound signal into a multi-dimensional vector;

[0013] S32, combining the multidimensional vectors corresponding to all the frame sound signals in step S31 to form an observation sequence matrix, and inputting the observation sequence matrix into a pre-stored acoustic model network;

[0014] S33. Use the Viterbi algorithm of dynamic programming pruning to find the global optimal path in the acoustic network model to complete the matching of ocean targets, thereby realizing automatic recognition and judgment of ocean targets.

[0015] The present invention also provides a long-range high-precision marine target automatic identification and determination system, which identifies and determines marine targets based on the long-range high-precision marine target automatic identification and determination method described above, and includes a main control unit and a fiber optic hydrophone array cable network, wherein:

[0016] The fiber optic hydrophone array cable network includes a plurality of fiber optic hydrophone array cables arranged in parallel and used for collecting ocean target sound signals to form interference light waveforms;

[0017] The main control unit includes a host computer, an optical transmitter and several signal processing sub-cards arranged in parallel. The host computer is used to schedule the data of the signal processing sub-card and display the operation status of the signal processing sub-card. The optical transmitter is used to send modulated light to the fiber optic hydrophone. The fiber optic hydrophone array cable network is used to transmit the interference light signal containing the sound signal to the signal processing sub-card. Each signal processing sub-card includes an analog-to-digital conversion module, a modulation and demodulation module and an AI recognition module connected in sequence. The analog-to-digital conversion module is used to perform photoelectric conversion on the received interference light signal to convert the light signal into an electrical signal. The modulation and demodulation module is used to perform phase modulation and demodulation on the electrical signal corresponding to the interference light signal to restore the sound signal of the ocean target. The AI ​​recognition module is used to perform sound recognition and target matching on the restored ocean target sound signal.

[0018] Preferably, the exterior of each of the fiber optic hydrophone array cables is armored, and each of the fiber optic hydrophone array cables includes a plurality of fiber optic hydrophones isolated by PU tubes and connected in series, and adjacent fiber optic hydrophones are connected by optical fibers.

[0019] Preferably, each of the signal processing subcards is provided with a SERDES universal interface for performing data exchange with a host computer.

[0020] Preferably, the optical transmitter and the fiber optic hydrophone array cable network are connected by an optical cable to achieve data transmission.

[0021] Preferably, the fiber optic hydrophone array cable network and the signal processing subcard are connected by optical cables to achieve data transmission.

[0022] Preferably, the host computer adopts Feiteng CPU and Kylin operating system.

[0023] Compared with the prior art, the advantages of the present invention are mainly reflected in the following aspects:

[0024] 1. The present invention utilizes a networked fiber optic hydrophone array cable to collect sound signals of marine targets, which greatly improves the detection area and detection accuracy. At the same time, multiple signal processing subcards arranged in parallel perform parallel calculations on multiple data channels, effectively improving data processing efficiency, thereby being able to accurately and quickly identify targets, and having the characteristics of low power consumption and high accuracy.

[0025] 2. In the present invention, by setting a SERDES universal interface on the signal processing subcard, the real-time performance of data transmission between the host computer and the signal processing subcard can be effectively improved.

[0026] 3. The present invention uses an AI recognition module for recognition, which has higher recognition accuracy, lower power consumption and lower data delay.

[0027] 4. The main control unit and the fiber optic hydrophone array cable in the present invention can both be domestically produced products, which can solve the problems of foreign technology blockade and equipment embargo. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is a flow chart of a method for automatic identification and determination of long-range high-precision marine targets in the present invention.

[0029] Figure 2 This is a structural block diagram of a long-range high-precision marine target automatic recognition and determination system in the present invention.

[0030] Figure 3 It is a modulation and demodulation flow chart of the modulation and demodulation module in the present invention.

[0031] In the figure: 1. Host computer, 2. Fiber optic hydrophone array cable, 3. Signal processing daughter card, 31. Analog-to-digital conversion module, 32. Modulation and demodulation module, 33. AI recognition module, 4. Optical transmitter, 5. SERDES general interface. DETAILED DESCRIPTION

[0032] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings.

[0033] like Figure 1-Figure 3 As shown, a method for automatic identification and determination of long-range high-precision marine targets comprises the following steps:

[0034] S1, training an acoustic model network based on the specific sound frequency of the ocean target sample, and pre-storing the trained acoustic model network in the host computer 1, wherein the acoustic model network first constructs an acoustic model based on the existing CD-DNN-HMM (context-dependent-deep neural network-hidden Markov model) architecture, and then inputs the sound data of the ocean environment and optimizes the branch paths of each network in the training acoustic model to obtain the trained acoustic model network;

[0035] S2, using the optical transmitter 4 to send the modulated light to the optical fiber hydrophone array cable 2, and sending the interference light signal containing the sound signal in the detection area to multiple signal processing subcards 3 which are arranged in parallel and can process the interference light signal in parallel through the optical fiber hydrophone array cable 2;

[0036] S3, the analog-to-digital conversion module 31 in each signal processing subcard 3 performs analog-to-digital conversion on the received interference light signal to convert the optical signal of the interference light signal into an electrical signal, and then sends the electrical signal corresponding to the interference light signal to the corresponding modulation and demodulation module 32 for phase modulation and demodulation, thereby restoring the sound signal of the ocean target, and then uses the corresponding AI recognition module 33 to perform sound recognition and target matching on the restored ocean target sound signal, and finally sends the matching result to the host computer 1;

[0037] S4 , the host computer 1 receives the matching result sent by the signal processing subcard 3 , and then displays the matching result and plays back the information through the display control software pre-stored in the host computer 1 .

[0038] In this embodiment, the specific training process of the acoustic model network belongs to the prior art and will not be repeated here. Since each ocean target has its own specific sound frequency, and the sound is transmitted over a long distance in water with low loss, the sound can cause the inner and outer walls of the fiber optic hydrophone to deform, thereby causing the optical fiber wound on the fiber optic hydrophone to produce an optical path difference, and the optical path difference causes interference between the two coherent light beams to form interference light. Even subtle sound signals can be reflected in the changes in the interference light. Therefore, the sound detection and identification of ocean targets through the fiber optic hydrophone has the characteristics of high sensitivity and long detection distance. First, an acoustic model network is trained according to the specific sound frequency of the ocean target and pre-stored in the host computer 1; then, the optical fiber hydrophone array cable composed of optical fiber hydrophones is used to network the detection area to generate an interference light signal containing a sound signal; then, the optical signal of the interference light signal is converted into an electrical signal by the analog-to-digital conversion module 31 and sent to the modulation and demodulation module 32 for phase modulation and demodulation to restore the sound signal of the ocean target, and then the AI ​​recognition module 33 performs a silent removal VAD operation and sends it to the host computer 1; finally, the host computer 1 performs target matching on the received electrical signal based on the pre-stored acoustic model network, thereby realizing automatic recognition and judgment of the ocean target. The method uses the networked optical fiber hydrophone array cable 2 to collect the sound signal of the ocean target, which greatly improves the detection area and detection accuracy. At the same time, through the parallel calculation of multiple signal processing sub-cards 3 arranged in parallel for multi-channel data, the data processing efficiency is effectively improved, so that the target can be accurately and quickly identified, and it has the characteristics of low power consumption and high accuracy.

[0039] In this embodiment, the acoustic model network is a CD-DNN-HMM architecture acoustic model network.

[0040] like Figure 3 As shown, the specific process of the phase adjustment and demodulation performed by the modulation and demodulation module 33 is: the modulation and demodulation module 33 performs mixing and multiplication, low-pass filtering, differential cross multiplication, subtraction and integration operations on the electrical signal corresponding to the interference light signal in sequence to realize phase modulation and demodulation of the signal, and then restore the sound signal of the ocean target.

[0041] In this embodiment, since the modulated light sent by the optical transmitter 4 adds phase carrier modulation, the interference optical signal I PGC The formula is:

[0042]

[0043] In formula (1), η represents the insertion loss of the interferometer that generates the interference light signal, k represents the visibility of the interference fringes, t represents time, A and B represent the DC term and AC term of the interference light signal, respectively, C represents the phase modulation amplitude (in this embodiment, C is set to 2.63 rad), ω0 represents the modulation frequency, Φ(t) represents the phase signal of the interferometer, and the interference light signal I is obtained by performing Bessel expansion on formula (1). PGC Converts to:

[0044]

[0045] In formula (2), J0, J 2k , J 2k+1 Both represent Bessel constants (which are constants);

[0046] When the electrical signal corresponding to the interference light signal enters the modulation and demodulation module 32, it is mixed and multiplied (that is, the interference light signal is multiplied by the carrier signal cos(ω0t) and the carrier double frequency signal cos(2ω0t) respectively), and then low-pass filtered to obtain:

[0047]

[0048] In formula (3), J1 and J2 both represent Bessel constants (which are constants);

[0049] Then differentiate equation (3) and get:

[0050]

[0051]

[0052] Then, by cross-multiplying equation (3) and equation (4) (i.e., calculating D1(t)×L2(t)-D2(t)×L1(t)), we can obtain:

[0053]

[0054] Finally, the cross-multiplication result in equation (5) is integrated to obtain:

[0055] B 2 J1(C)J2(C)Φ(t) (6)

[0056] From formula (6), we can see that B, J1(C) and J2(C) in formula (6) are all constants, and the function Φ(t) that changes with time can be obtained.

[0057] The specific implementation method of the AI ​​recognition module 33 in step S3 performing sound recognition and target matching on the restored ocean target sound signal includes:

[0058] S31, performing speech framing on the received sound signal by using a moving window function, and then performing MFCC acoustic feature extraction on the sound signal after the speech framing, so as to convert the interference light signal corresponding to each frame of the sound signal into a multi-dimensional vector;

[0059] S32, combining the multidimensional vectors corresponding to all the frame sound signals in step S31 to form an observation sequence matrix, and inputting the observation sequence matrix into a pre-stored acoustic model network;

[0060] S33. Use the Viterbi algorithm of dynamic programming pruning to find the global optimal path in the acoustic network model to complete the matching of ocean targets, thereby realizing automatic recognition and judgment of ocean targets.

[0061] like Figure 2 As shown, a long-range high-precision marine target automatic identification and determination system is provided. The system identifies and determines marine targets based on the long-range high-precision marine target automatic identification and determination method described above. The system includes a main control unit and a fiber optic hydrophone array cable network, wherein:

[0062] The fiber optic hydrophone array cable network includes a plurality of fiber optic hydrophone array cables 2 arranged in parallel and used to collect ocean target sound signals to form interference light signals;

[0063] The main control unit includes a host computer 1, an optical transmitter 4 and a plurality of signal processing sub-cards 3 arranged in parallel. The host computer 1 is used to schedule the data of the signal processing sub-card 3 and display the operation status of the signal processing sub-card 3. The optical transmitter 4 is used to send modulated light to the optical fiber hydrophone. The optical fiber hydrophone array cable network is used to transmit the interference optical signal containing the sound signal to the signal processing sub-card 3. Each signal processing sub-card 3 includes an analog-to-digital conversion module 31, a modulation and demodulation module 32 and an AI recognition module 33 connected in sequence. The analog-to-digital conversion module 31 is used to perform photoelectric conversion on the received interference optical signal to convert the optical signal into an electrical signal. The modulation and demodulation module 32 is used to perform phase modulation and demodulation on the electrical signal corresponding to the interference optical signal to restore the sound signal of the ocean target. The AI ​​recognition module 33 is used to perform sound recognition and target matching on the restored ocean target sound signal.

[0064] In this embodiment, the host computer 1 adopts Feiteng CPU and Kylin operating system. By using domestic products for the main control unit and the optical fiber hydrophone array cable 2, the problem of foreign technology blockade and equipment embargo can be solved.

[0065] Among them, the outside of each of the fiber optic hydrophone array cables 2 is armored, and each of the fiber optic hydrophone array cables 2 includes a number of fiber optic hydrophones isolated and connected in series by PU tubes (a pneumatic pressure hose, POLYURETHANE TUBING), and adjacent fiber optic hydrophones are connected by optical fibers. In this embodiment, armor is used on the outside of each of the fiber optic hydrophone array cables 2, and a number of fiber optic hydrophones isolated and connected in series by PU tubes inside are connected by optical fibers, and silicone oil is used to enhance the isolation and sealing performance, so that a number of fiber optic hydrophone array cables 2 can be distributed and networked, which is conducive to expanding the detection area.

[0066] Each of the signal processing subcards 3 is provided with a reserved SERDES (short for SERializer serializer / DESerializer deserializer) general interface 5 for data interaction with the host computer 1. In this embodiment, the SERDES general interface 5 is provided to facilitate high-speed data transmission between the signal processing subcard 3 and the host computer 1, and to facilitate flexible networking between the signal processing subcards 3 and between the signal processing subcards 3 and the host computer 1.

[0067] Among them, optical cables are used to connect the optical transmitter 4 and the optical fiber hydrophone array cable network, and the optical fiber hydrophone array cable network and the signal processing subcard 3 to achieve data transmission. In this embodiment, optical cables are provided to achieve data transmission between the optical transmitter 4 and the optical fiber hydrophone array cable network, and between the optical fiber hydrophone array cable network and the signal processing subcard 4, which not only ensures the reliability of data transmission, but also ensures the timeliness of data transmission.

[0068] The above is a detailed introduction to a long-range high-precision marine target automatic recognition and determination method and system provided by the present invention. This article uses specific examples to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the core idea of ​​the present invention. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of the present invention, the present invention can also be improved and modified, and these improvements and modifications also fall within the scope of protection of the claims of the present invention.

Claims

1. A method for automatic identification and determination of long-range high-precision marine targets, characterized in that: The method comprises the following steps: S1. Train an acoustic model network based on the specific sound frequency of the ocean target sample, and pre-store the trained acoustic model network in the signal processing subcard; S2, using an optical transmitter to send modulated light to a fiber optic hydrophone array cable, and sending the interference light signal containing the sound signal in the detection area to a plurality of signal processing subcards that are arranged in parallel and can process the interference light signal in parallel through the fiber optic hydrophone array cable; S3. The analog-to-digital conversion module in each signal processing subcard performs analog-to-digital conversion on the received interference light signal to convert the optical signal of the interference light signal into an electrical signal, and then sends the electrical signal corresponding to the interference light signal to the corresponding modulation and demodulation module for phase modulation and demodulation, thereby restoring the sound signal of the ocean target, and then uses the corresponding AI recognition module to perform sound recognition and target matching on the restored ocean target sound signal, and finally sends the matching result to the host computer; S4, the host computer receives the matching result sent by the signal processing subcard, and then displays the matching result and plays back the information through the display control software pre-stored in the host computer; The acoustic model network in step S1 is a CD-DNN-HMM architecture acoustic model network; The specific implementation method of the AI ​​recognition module in step S3 performing sound recognition and target matching on the restored ocean target sound signal includes: S31, performing speech framing on the received sound signal by a moving window function, and then performing MFCC acoustic feature extraction on the sound signal after the speech framing, so as to convert each frame of the sound signal into a multidimensional vector; S32, combining the multidimensional vectors corresponding to all the frame sound signals in step S31 to form an observation sequence matrix, and inputting the observation sequence matrix into a pre-stored acoustic model network; S33. Use the Viterbi algorithm of dynamic programming pruning to find the global optimal path in the acoustic network model to complete the matching of ocean targets, thereby realizing automatic recognition and judgment of ocean targets.

2. The long-range high-precision marine target automatic recognition and determination method according to claim 1, characterized in that: The specific process of the phase adjustment and demodulation performed by the modulation and demodulation module in step S3 is: the modulation and demodulation module performs mixing and multiplication, low-pass filtering, differential cross multiplication, subtraction and integration operations on the electrical signal corresponding to the interference light signal in sequence to realize phase modulation and demodulation of the signal, and then restore the sound signal of the ocean target.

3. A long-range high-precision marine target automatic recognition and determination system, characterized in that: The system automatically identifies and determines marine targets based on the long-range high-precision marine target automatic identification and determination method according to any one of claims 1 to 2, and comprises a main control unit and a fiber optic hydrophone array cable network, wherein: The fiber optic hydrophone array cable network includes a plurality of fiber optic hydrophone array cables arranged in parallel and used for collecting ocean target sound signals to form interference light signals; The main control unit includes a host computer, an optical transmitter and several signal processing sub-cards arranged in parallel. The host computer is used to schedule the data of the signal processing sub-card and display the operation status of the signal processing sub-card. The optical transmitter is used to send modulated light to the fiber optic hydrophone. The fiber optic hydrophone array cable network is used to transmit the interference light signal containing the sound signal to the signal processing sub-card. Each signal processing sub-card includes an analog-to-digital conversion module, a modulation and demodulation module and an AI recognition module connected in sequence. The analog-to-digital conversion module is used to perform photoelectric conversion on the received interference light signal to convert the light signal into an electrical signal. The modulation and demodulation module is used to perform phase modulation and demodulation on the electrical signal corresponding to the interference light signal to restore the sound signal of the ocean target. The AI ​​recognition module is used to perform sound recognition and target matching on the restored ocean target sound signal.

4. The long-range high-precision marine target automatic recognition and determination system as claimed in claim 3, characterized in that: The outside of each of the fiber optic hydrophone array cables is armored, and each of the fiber optic hydrophone array cables includes a plurality of fiber optic hydrophones isolated by PU tubes and connected in series, and adjacent fiber optic hydrophones are connected by optical fibers.

5. The long-range high-precision marine target automatic recognition and determination system according to claim 4, characterized in that: Each of the signal processing subcards is provided with a SERDES universal interface for performing data exchange with a host computer.

6. The long-range high-precision marine target automatic recognition and determination system according to claim 4, characterized in that: The optical transmitter and the optical fiber hydrophone array cable network are connected by optical cables to achieve signal transmission.

7. The long-range high-precision marine target automatic recognition and determination system according to claim 6, characterized in that: The fiber optic hydrophone array cable network and the signal processing subcard are connected by optical cables to achieve signal transmission.

8. The long-range high-precision marine target automatic recognition and determination system according to claim 7, characterized in that: The host computer uses Feiteng CPU and Kylin operating system.

Images