Target detection method and system based on underwater acoustic signal

Abstract

The invention discloses a target detection method and system based on an underwater acoustic signal, and the method comprises the steps: dividing the underwater acoustic signal into a high-frequency-band signal and a low-frequency-band signal, carrying out the line spectrum detection of the high-frequency-band signal and the low-frequency-band signal, and extracting the line spectrum feature information; performing azimuth angle estimation on the low-frequency-band signal by adopting a weighted histogram method based on energy search to obtain a vector array multi-target direction finding result; performing azimuth angle estimation on the high-frequency band signal by adopting an adaptive beam forming algorithm to obtain a scalar array direction finding result; filtering a direction finding result, which is close to the direction finding result of the scalar array, in the vector array multi-target direction finding result to obtain a target direction finding result; and performing feature extraction on the low-frequency-band signal corresponding to the target azimuth direction finding result so as to perform identification, classification and judgment on the target corresponding to the low-frequency-band signal. The interference signals can be effectively filtered out, the azimuth information of the target is reserved while the interference signals are removed, the false alarm rate is reduced, and the detection and analysis capability of the weak target is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of underwater monitoring, and in particular to a target detection method and system based on underwater acoustic signals. BACKGROUND

[0002] The underwater unmanned platform has the advantages of simple deployment and controllable cost, and can work in a passive manner for a long time to obtain information of underwater targets within the warning range. At present, a large amount of research has been conducted on underwater unmanned platforms at home and abroad. According to the target detection device, most underwater unmanned platforms have poor self-adaptive ability to complex underwater environment, low information dimension, poor suppression effect on strong interference, and a large amount of interference causing high false alarm, which further leads to insufficient detection ability of weak targets. SUMMARY

[0003] In order to solve the problems in the prior art, the present application adopts the following technical solutions:

[0004] In a first aspect, the present application provides a target detection method based on underwater acoustic signals, which is applied to an underwater unmanned detection platform, and the target detection method comprises the following steps:

[0005] The underwater acoustic signal is divided into a high-frequency signal and a low-frequency signal, and line spectrum detection is performed on the high-frequency signal and the low-frequency signal respectively, and the line spectrum feature information of the low-frequency signal and the line spectrum feature information of the high-frequency signal are extracted;

[0006] The low-frequency signal is subjected to differential pressure vector conversion to obtain vector low-frequency line spectrum feature information, and a weighted histogram method based on energy search is used to estimate the azimuth angle of the vector low-frequency line spectrum feature information to obtain a vector array multi-target direction finding result;

[0007] Based on the line spectrum feature information of the high-frequency signal, an adaptive beam forming algorithm is used to estimate the azimuth angle of the high-frequency signal to obtain a scalar array direction finding result;

[0008] The direction finding result in the vector array multi-target direction finding result that is within a preset angle range from the azimuth of the scalar array direction finding result is filtered out to obtain a target azimuth direction finding result;

[0009] Feature extraction is performed on the low-frequency signal corresponding to the target azimuth direction finding result to obtain the LOFAR spectrum and the DEMON spectrum of the low-frequency signal, and the target corresponding to the low-frequency signal is identified and classified based on the LOFAR spectrum and the DEMON spectrum.

[0010] In summary, the target detection method based on underwater acoustic signals provided in the application classifies underwater acoustic signals by frequency bands, uses a vector and scalar joint processing method to estimate the azimuth of low-frequency signals and high-frequency signals respectively, obtains the vector array multi-target direction finding result of the low-frequency signals and the scalar array direction finding result of the high-frequency signals, filters out the direction finding result similar to the azimuth of the scalar array direction finding result from the vector array multi-target direction finding result, and obtains the target azimuth direction finding result after filtering through a spatial filtering process. The target azimuth direction finding result obtained after filtering mainly corresponds to weak targets in the low-frequency band region. By using the characteristics of the scalar array and the vector array, interference signals are effectively filtered out, the azimuth information of the target is retained while the interference signals are removed, the false alarm rate is reduced, and the detection and analysis capability for weak targets is improved.

[0011] Further, the weighted histogram method based on energy search is used to estimate the azimuth of the vector low-frequency line spectrum feature information, including:

[0012] The statistical interval is determined based on the weighted histogram, and the window length and the sliding step distance of the sliding window are configured. The statistical value of each statistical interval is calculated by the sliding window method.

[0013] The statistical interval with the maximum statistical value is the statistical window where the sound source azimuth of the underwater acoustic signal is located, and the statistical window where the sound source azimuth is located is taken as the vector array multi-target direction finding result.

[0014] Further, the statistical value of each statistical interval is calculated by the sliding window method, including:

[0015] The starting point of the statistical interval is taken as the data starting point of the sliding window, and the energy weighted mean value is calculated according to the window length of the sliding window to obtain the calculation result.

[0016] After the calculation is completed, the sliding window is moved by the length of the sliding step distance, the window length of the sliding window is kept unchanged, and the calculation is performed again. The calculation is repeated in the statistical interval until all data in the statistical interval are calculated.

[0017] The maximum value in all calculation results is selected as the statistical value of the statistical interval.

[0018] Further, before the frequency band division of the underwater acoustic signal, the adaptive line spectrum enhancement technology based on time reversal convolution interference suppression is used to enhance the feature of the underwater acoustic signal.

[0019] Further, the signal-to-noise ratio of the underwater acoustic signal is improved by the following time reversal convolution interference suppression expression, and the time reversal convolution interference suppression expression is:

[0020] y(t)=x(-t)*x(t)·E(t)·I(t);

[0021] In the formula, "*" represents the convolution operation, x(t) represents the underwater acoustic signal mixed with noise, y(t) represents the output of time-domain deconvolution interference suppression, E(t) represents the time-domain amplitude equalization window, and we have:

[0022]

[0023] A represents the amplitude of the underwater acoustic signal, and T represents the segmented processing time;

[0024] I(t) represents the interference suppression gate, and we have:

[0025]

[0026] Furthermore, the stability and uniqueness of the target azimuth direction finding results are judged. If the judgment is successful, the low-frequency band signal corresponding to the target azimuth direction finding results is extracted. If the judgment is unsuccessful, the process is repeated to re-estimate the azimuth angle of the low-frequency band signal and the high-frequency band signal.

[0027] Furthermore, an adaptive dual-threshold technique is used to perform line spectrum detection on the feature-enhanced underwater acoustic signal. The adaptive dual-threshold technique includes mean constant false alarm rate detection and ordered constant false alarm rate detection.

[0028] Furthermore, the LOFAR spectrum of the low-frequency signal is extracted using short-time Fourier transform.

[0029] Furthermore, broadband noise envelope modulation analysis technology is used to measure the fundamental frequency and harmonics of the low-frequency signal in order to extract the DEMON spectrum of the low-frequency signal.

[0030] Secondly, this application also provides a target detection system based on underwater acoustic signals, which is applied to an underwater unmanned exploration platform and uses the aforementioned target detection method based on underwater acoustic signals. Attached Figure Description

[0031] Figure 1 This is a flowchart illustrating the steps of a target detection method provided in one embodiment of this application;

[0032] Figure 2 This is a simulation experiment of the signal radiation power spectrum received by an underwater unmanned platform in one embodiment of this application;

[0033] Figure 3 The radiated noise power spectrum of a water target in a simulation experiment provided in one embodiment of this application;

[0034] Figure 4 The radiated noise power spectrum of a water surface target in a simulation experiment provided in one embodiment of this application;

[0035] Figure 5The LOFAR spectrum of the original underwater acoustic signal in a simulation experiment provided in one embodiment of this application;

[0036] Figure 6 LOFAR spectrum of underwater acoustic signal after processing by conventional ALE method in a simulation experiment provided in one embodiment of this application;

[0037] Figure 7 LOFAR spectrum of underwater acoustic signal after processing by the target detection method provided in this application in a simulation experiment provided in one embodiment of this application;

[0038] Figure 8 A schematic diagram of the detection results of underwater acoustic signal line spectrum detection using the CA-CFAR method in a simulation experiment provided in one embodiment of this application;

[0039] Figure 9 A schematic diagram of the detection results of underwater acoustic signal line spectrum detection by OS-CFAR method in a simulation experiment provided in one embodiment of this application;

[0040] Figure 10 This is a schematic diagram of the detection results of underwater acoustic signal line spectrum detection by combining the CA-CFAR method and the OS-CFAR method in a simulation experiment provided in one embodiment of this application;

[0041] Figure 11 A schematic diagram of the direction finding results of low-frequency signals using the weighted histogram method based on energy search in a simulation experiment provided in one embodiment of this application;

[0042] Figure 12 This is a schematic diagram illustrating the error of the weighted histogram method based on energy search for azimuth direction finding of signals of different frequencies in a simulation experiment provided in one embodiment of this application.

[0043] Figure 13 A schematic diagram of the direction finding results of the original underwater acoustic signal by using the weighted histogram method based on energy search in a simulation experiment provided in one embodiment of this application;

[0044] Figure 14 The spectrum of the original underwater acoustic signal in a simulation experiment provided in one embodiment of this application;

[0045] Figure 15 A spectrum of a low-frequency signal obtained by the target detection method provided in this application in a simulation experiment according to an embodiment of this application;

[0046] Figure 16 A schematic diagram of the direction finding results of underwater acoustic signals using the target detection method provided in this application in a simulation experiment provided in one embodiment of this application;

[0047] Figure 17 The LOFAR spectrum of a low-frequency signal extracted by the target detection method provided in this application in a simulation experiment provided in one embodiment of this application;

[0048] Figure 18 The DEMON spectrum of a low-frequency signal extracted by the target detection method provided in this application in a simulation experiment provided in one embodiment of this application;

[0049] Figure 19 This application provides a simulation experiment in which the first set of lake test data is processed before and after orientation spectrum diagrams.

[0050] Figure 20 The image shows the azimuth spectrum of the second set of lake test data before and after processing in a simulation experiment provided in one embodiment of this application. Detailed Implementation

[0051] The present application will be described in detail below with reference to the specific embodiments shown in the accompanying drawings. However, these embodiments do not limit the present application. Any structural, methodological, or functional modifications made by those skilled in the art based on these embodiments are included within the protection scope of the present application.

[0052] To address the shortcomings of existing technologies, firstly, embodiments of this application provide a target detection method based on underwater acoustic signals. This target detection method is applied to underwater unmanned detection platforms, such as... Figure 1 As shown, target detection methods include:

[0053] Step S11: Divide the underwater acoustic signal into high-frequency signal and low-frequency signal, perform line spectrum detection on the high-frequency signal and low-frequency signal respectively, and extract the line spectrum feature information of the low-frequency signal and the line spectrum feature information of the high-frequency signal.

[0054] Step S12: Perform pressure difference vector conversion on the low-frequency band signal to obtain vector low-frequency line spectrum feature information. Use the weighted histogram method based on energy search to estimate the azimuth of the vector low-frequency line spectrum feature information to obtain the multi-target direction finding results of the vector array.

[0055] Step S13: Based on the line spectrum feature information of the high-frequency signal, an adaptive beamforming algorithm is used to estimate the azimuth angle of the high-frequency signal to obtain the scalar array direction finding results.

[0056] Step S14: Filter out the direction finding results in the multi-target direction finding results of the vector array that differ from the direction finding results of the scalar array within a preset angle range, and obtain the target azimuth direction finding results.

[0057] Step S15: Extract features from the low-frequency signal corresponding to the target orientation measurement result, obtain the LOFAR spectrum and DEMON spectrum of the low-frequency signal, and identify and classify the target corresponding to the low-frequency signal based on the LOFAR spectrum and DEMON spectrum.

[0058] Specifically, based on the energy distribution patterns of radiated noise in depth signals—different types of targets (such as surface vessels and weak underwater targets like submarines) often have different dominant energy frequency bands—underwater acoustic signals are divided into high-frequency and low-frequency signals through frequency band division. These high-frequency and low-frequency signals are then processed separately, avoiding mutual interference between high and low frequency signals, which is common in traditional single-band signal processing. Frequency band division is typically based on the typical frequency distribution of the target's radiated noise; for example, low-frequency signals may cover tens to hundreds of hertz, while high-frequency signals may cover thousands of hertz or higher.

[0059] Line spectrum detection is performed on both high-frequency and low-frequency signals. This detection initially reveals the characteristic frequencies of potential targets and forms a set of line spectrum feature information, including the distribution pattern, energy intensity, and its time-varying trend. High-frequency line spectrum features reflect the high-frequency radiation characteristics of the corresponding high-frequency signal (e.g., the high-frequency line spectrum of a jamming ship's mechanical vibration), providing a data foundation for subsequent location estimation of high-frequency signals. Low-frequency line spectrum features reflect the radiation characteristics of the corresponding low-frequency signal (e.g., the low-frequency line spectrum of a submarine at low speed), providing a data foundation for subsequent location estimation of low-frequency signals.

[0060] For low-frequency signals, a pressure vector conversion is first performed to obtain vector low-frequency line spectrum feature information. Then, an energy-search-based weighted histogram method is used to estimate the azimuth of this feature information, thus obtaining the multi-target direction finding results for the vector array. Low-frequency signals often contain characteristics of weak targets, but due to the complexity of the marine environment, such as multipath propagation and environmental noise, direct azimuth estimation is often insufficient in accuracy. Pressure vector conversion converts the underwater acoustic signal received by the hydrophone into a vector signal containing sound pressure and particle velocity information. The vector signal includes sound pressure amplitude and the direction of sound wave propagation. Pressure vector conversion provides richer sound field information, thereby enhancing the accuracy of sound source azimuth determination.

[0061] For high-frequency signals, an adaptive beamforming algorithm is employed to estimate the azimuth angle based on the line spectrum characteristics of the high-frequency signals, thereby obtaining scalar array direction finding results. High-frequency signals are often associated with strong interfering targets (such as surface vessels), which tend to generate high-frequency noise components when moving at high speeds or operating mechanically. The adaptive beamforming algorithm enhances signals from specific directions while suppressing interference from other directions by dynamically adjusting the weighting coefficients of the sensor array. It automatically optimizes the beam pattern based on the real-time signal environment, accurately obtaining scalar array direction finding results for high-frequency signals.

[0062] Furthermore, scalar array direction finding exhibits high gain and directionality in the high-frequency region, enabling accurate detection of interfering targets; while vector array lateral direction finding provides richer azimuth information in the low-frequency region, aiding in the identification of weak targets. The multi-target direction finding results of the vector array are compared with those of the scalar array. Direction finding results from the vector array that differ from the scalar array results within a preset angle Δθ are filtered out, thus obtaining the target azimuth direction finding result. When two direction finding results are close in azimuth, it indicates the presence of an interfering target in that direction, and these results are filtered out. The preset angle Δθ can be set to within 10° (inclusive).

[0063] Through the aforementioned spatial filtering process, the target orientation results obtained after filtering mainly correspond to weak targets in the low-frequency region, such as submarines. By utilizing the characteristics of scalar and vector arrays, interference signals and weak targets are effectively separated, improving the specificity of detection. This ensures that while removing interference, the orientation information of the true target is preserved, reducing the false alarm rate and providing clean data input for subsequent feature extraction and target recognition.

[0064] The process involves acquiring low-frequency signals corresponding to the target's azimuth direction finding results, extracting features from these signals to obtain LOFAR and DEMON spectra, and then classifying and identifying the target based on these spectra. The LOFAR spectrum displays the changes in frequency components over time, revealing the target's line spectrum stability, modulation characteristics, and frequency distribution patterns. For example, a submarine's line spectrum might appear as a continuous and slowly changing frequency line, while noise interference might exhibit random fluctuations. The DEMON spectrum extracts the target's fundamental frequency and harmonics by demodulating the signal envelope. It reflects the target's rotating mechanical characteristics, thus helping to distinguish different types of targets. After acquiring the LOFAR and DEMON spectra, the targets corresponding to the low-frequency signals are identified and classified using these spectra.

[0065] Based on the above description, the target detection method based on underwater acoustic signals provided in this application classifies the underwater acoustic signals by frequency band and uses a scalar-vector joint processing method to estimate the azimuth angles of low-frequency and high-frequency signals respectively, obtaining vector array multi-target direction finding results for low-frequency signals and scalar array direction finding results for high-frequency signals. Direction finding results with azimuths close to the scalar array direction finding results in the vector array multi-target direction finding results are filtered out. Through a spatial filtering process, the target azimuth direction finding results obtained after filtering mainly correspond to weak targets in the low-frequency region. By utilizing the characteristics of scalar and vector arrays, interference signals are effectively filtered out. While eliminating interference signals, the azimuth information of the target is retained, reducing the false alarm rate and improving the detection and analysis capability of weak targets.

[0066] As an alternative implementation, before dividing the underwater acoustic signal into frequency bands, an adaptive line spectrum enhancement technique based on temporal inverse convolution interference suppression is used to enhance the underwater acoustic signal's features.

[0067] Specifically, underwater acoustic signals collected by unmanned underwater platforms in real marine environments typically have a low signal-to-noise ratio (SNR). Useful features of a target often manifest as stable and sharp line spectra, but under low SNR conditions, these line spectra are very weak and difficult to extract effectively using traditional signal processing techniques, leading to a high false negative rate in subsequent line spectrum detection steps. Before dividing the underwater acoustic signal into frequency bands, an adaptive line spectrum enhancement technique based on time-deconvolution interference suppression is used to preprocess and enhance the features of the original underwater acoustic signal, thereby increasing the line spectrum components. Time-deconvolution interference suppression, through convolution operations, effectively compresses the signal waveform and refocuses energy in the time domain, effectively improving the SNR of the underwater acoustic signal. Using time-deconvolution interference suppression as a preprocessing technique for the input signal of the adaptive line spectrum enhancement technique improves the SNR of the input signal beforehand, thus enhancing the effectiveness and reliability of line spectrum detection under low SNR conditions.

[0068] Furthermore, the signal-to-noise ratio of the underwater acoustic signal is improved by using the following expression for time-deconvolution interference suppression:

[0069] The expression for time-deconvolution interference suppression is:

[0070] y(t)=x(-t)*x(t)·E(t)·I(t);

[0071] In the formula, "*" represents the convolution operation, x(t) represents the underwater acoustic signal mixed with noise, y(t) represents the output of time-domain deconvolution interference suppression, E(t) represents the time-domain amplitude equalization window, and we have:

[0072]

[0073] A represents the amplitude of the underwater acoustic signal, and T represents the segmented processing time;

[0074] I(t) represents the interference suppression gate, and we have:

[0075]

[0076] As an optional implementation, an adaptive dual-threshold technique is used to perform line spectrum detection on the feature-enhanced underwater acoustic signal. The adaptive dual-threshold technique includes mean constant false alarm rate detection and ordered constant false alarm rate detection.

[0077] Specifically, after effectively enhancing the underwater acoustic signal's features, an adaptive dual-threshold technique is employed for line spectrum detection. This technique includes mean constant false alarm rate (CFAR) detection and ordered CFAR detection. First, mean CFAR detection detrends the underwater acoustic signal, suppressing overall fluctuations in background noise and reducing misjudgments caused by an unstable noise floor. Then, ordered CFAR detection accurately extracts the line spectrum from the detrended signal. Ordered CFAR detection adaptively adjusts the detection threshold by sorting the signal intensity of reference cells, effectively distinguishing target line spectra from interference line spectra even in scenarios with dense, strong interference. This allows for more comprehensive and reliable extraction of effective line spectrum features from the feature-enhanced underwater acoustic signal, improving detection performance and reliability in complex underwater environments.

[0078] As an optional implementation, after obtaining the line spectrum feature information of the low-frequency signal, the weighted histogram method based on energy search is used to perform azimuth direction finding on the low-frequency signal. This includes: defining statistical intervals based on the weighted histogram, configuring the window length and sliding step of the sliding window, calculating the statistical values ​​of each statistical interval using the sliding window method, wherein the statistical interval with the largest statistical value is the statistical window where the sound source azimuth of the underwater acoustic signal is located, and using the statistical window where the sound source azimuth is located as the multi-target direction finding result of the vector array.

[0079] Specifically, firstly, statistical intervals are defined based on weighted histograms, and the window length and sliding step size of the sliding window are configured using the sliding window method. The window length defines the size of the data range covered in each energy calculation, and the sliding step size determines the distance the window moves forward after each calculation. Statistical values ​​for each statistical interval are calculated using the sliding window method. By comparing the statistical values ​​of all statistical intervals, the interval with the largest statistical value is identified. According to acoustic principles, the energy radiated by a sound source propagates through space, and the direction in which the energy is most concentrated when it reaches the receiving array points to the true location of the sound source. Therefore, the statistical interval with the largest statistical value is determined as the region where the sound source of the underwater acoustic signal is most likely to exist. The statistical interval containing this sound source location is the result of multi-target direction finding for low-frequency signals using a vector array.

[0080] By using a weighted histogram method based on energy search to determine the azimuth of low-frequency signals, the energy concentration characteristic effectively suppresses random noise and interference, effectively extracting the target's azimuth information from complex acoustic environments, thus improving the accuracy and robustness of azimuth estimation.

[0081] Furthermore, as an optional implementation method, the calculation of statistical values ​​for each statistical interval using the sliding window method includes: taking the starting point of the statistical interval as the data starting point of the sliding window, calculating the energy-weighted mean according to the window length of the sliding window, and obtaining the calculation result; after the calculation is completed, the sliding window moves backward by the sliding step length, keeping the window length of the sliding window unchanged, and performing the calculation again, cyclically calculating within the statistical interval until all data within the statistical interval is statistically analyzed; the maximum value among all calculation results is selected as the statistical value of the statistical interval.

[0082] Specifically, the starting boundary of the current statistical interval is used as the data starting point of the sliding window. A segment of data is extracted according to the length of the sliding window for calculation, obtaining the first calculation result. Then, the data starting point moves backward by one sliding step length. After the sliding window moves, the window length remains unchanged, and the same calculation is performed again based on the new data segment to obtain the second calculation result. This process is repeated within the current statistical interval until the sliding window moves from the starting point of the statistical interval and scans to the ending point, ensuring that all data within the statistical interval is covered by the window and participates in the calculation. Through iterative calculation, a series of calculation results are obtained within a statistical interval. These results record the distribution of energy or related statistics at different local locations within the statistical interval.

[0083] The energy-weighted mean can be calculated using the following formula:

[0084]

[0085] In the formula, Ψ represents the azimuth estimation result, and R... i Let θ be the energy value of the i-th angle in the obtained statistical interval. i This corresponds to the angle value.

[0086] According to acoustic principles, the representative energy level of a statistical interval is determined by the local window where the energy is most concentrated. After completing the iterative calculation for the entire statistical interval, given multiple calculation results generated within that interval, the maximum value among all results is selected as the statistical value for that interval. Using a sliding window calculation effectively highlights the energy focal point within the statistical interval while suppressing the averaging effect from other lower-energy regions, ensuring that the statistical values ​​of each interval used for comparison clearly reflect the true possible location of the sound source.

[0087] As an optional implementation, after obtaining the target azimuth direction finding results, the stability and uniqueness of the target azimuth direction finding results are judged. If the judgment is successful, the low-frequency band signal corresponding to the target azimuth direction finding results is feature extracted. If the judgment is unsuccessful, the process is repeated to re-estimate the azimuth angle of the low-frequency band signal and the high-frequency band signal.

[0088] Specifically, in real and complex marine environments, the initial direction finding results obtained after the aforementioned series of processing steps may still be affected by various factors such as transient strong interference, multipath effects, non-stationary noise, or transient convergence of multiple targets, leading to occasional fluctuations or ambiguities in the results. Therefore, after obtaining the target azimuth direction finding results, the stability and uniqueness of the results should be assessed to improve the detection confidence of the method and reduce the false alarm rate.

[0089] Stability assessment primarily evaluates the consistency of target orientation results over time. A realistic underwater target's orientation should change continuously, or at least be relatively stable, over a short period. This is achieved by tracking and analyzing the orientation estimates of the same potential target within an observation window. If the target's orientation value exhibits continuous, irregular, and violent fluctuations, or is intermittent and unclear, then the orientation result is considered unstable and likely an artifact caused by environmental interference. Uniqueness assessment focuses on confirming the clarity of the orientation result in the spatial dimension. If the orientation result shows multiple adjacent peaks that are spatially indistinguishable, or if its orientation energy distribution is too diffuse to clearly identify it as a single target, it indicates that the current result may be confusing and unsuitable for subsequent decision-making.

[0090] As an optional implementation method, the stability and uniqueness criteria can be determined using the MN criterion. For example, if there are N seconds in the target azimuth direction finding results within an M-second time window where the azimuth result always varies within the range of Δθ, then the target is considered to meet the stability and uniqueness conditions. M and N can be flexibly adjusted based on the analysis results of the measured data; generally, M is less than 10 minutes and N is less than M / 2.

[0091] Optionally, the criteria for judging stability and uniqueness can be dynamically adjusted based on historical data, the overall acoustic characteristics of the current environment, and prior knowledge. Examples include the allowable range of azimuth fluctuations and the sharpness requirements of energy peaks. Through iterative optimization, these criteria can adapt to detection needs under different sea states, target characteristics, and signal-to-noise ratio levels.

[0092] When the target azimuth direction finding result passes the stability and uniqueness test, in-depth feature extraction is performed on the corresponding low-frequency signal to complete the final identification and classification. If the test fails, it indicates that the current direction finding result is of poor quality, and subsequent analysis is not based on this result. Instead, azimuth angle estimation is performed again for both the low-frequency and high-frequency signals. By judging the stability and uniqueness of the target azimuth direction finding result, the robustness and accuracy of the target detection method in autonomous detection in complex environments are improved.

[0093] As an optional implementation, after the target orientation measurement results pass stability and uniqueness checks, feature extraction is performed on the low-frequency signal corresponding to the target orientation measurement results. Short-Time Fourier Transform (SFT) is used to extract the LOFAR spectrum of the low-frequency signal. Specifically, when extracting features from the low-frequency signal corresponding to the target orientation measurement results, SFT is used to extract the LOFAR spectrum of the low-frequency signal. This transforms the low-frequency signal, which has already undergone orientation screening and is considered to originate from a potential real target, into intuitive features that can be used for final identification and classification. The LOFAR spectrum can simultaneously display the signal frequency components and their dynamic changes over time. By observing and analyzing the resulting LOFAR spectrum, analysts or automatic identification systems can obtain relevant information about the target's acoustic characteristics, thereby enabling target identification and classification.

[0094] As an optional implementation method, feature extraction is performed on the low-frequency signal corresponding to the target azimuth direction finding results. Broadband noise envelope modulation analysis technology is used to measure the fundamental frequency and harmonics of the low-frequency signal in order to extract the DEMON spectrum of the low-frequency signal.

[0095] Specifically, when extracting features from the low-frequency signals corresponding to the target's azimuth direction finding results, broadband noise envelope modulation analysis is used to extract the DEMON spectrum. The extracted DEMON spectrum has significant identification value. For example, underwater targets of different tonnages, types, and designs have different propeller blade numbers, rotational speeds, and mechanical structures. Therefore, the fundamental frequency, number of harmonics, and energy distribution relationships between harmonics in their DEMON spectra will exhibit unique characteristics. By analyzing the DEMON spectrum, the target's speed can be effectively estimated, certain structural parameters of its propeller can be inferred, and these features can ultimately be compared with a known target database to achieve refined target classification and identification, improving the accuracy and confidence of target identification.

[0096] To further illustrate the effectiveness of the target detection method based on underwater acoustic signals provided in this application, a simulation experiment is conducted below to verify the method. The simulation conditions are: sea state 3, sea depth 300 meters, seabed composed of mud and sand, and current velocity of 3 knots. In the simulation experiment, the underwater unmanned detection platform obtains the power spectrum of the radiated noise signal (i.e., underwater acoustic signal) received by the mooring as shown below. Figure 2 As shown, the radiated noise signal includes submarine radiated noise and surface crosstalk radiated noise. The power spectrum of the submarine radiated noise signal is shown below. Figure 3 As shown, the power spectrum of the water surface radiated noise signal is as follows: Figure 4 As shown, the radiated noise level of surface ships is much higher than that of submarine target signals, seriously interfering with submarine detection missions.

[0097] The line spectrum signal input to the simulation experiment is the simulated underwater target radiated noise line spectrum signal, with frequencies of 30Hz, 45Hz, 60Hz, 80Hz, and 90Hz, a sampling rate of 5kHz, and a signal-to-noise ratio of -15dB. The LOFAR spectrum of the original underwater acoustic signal is shown below. Figure 5 As shown, the LOFAR spectrum of the underwater acoustic signal after processing with traditional adaptive line spectrum enhancement technology is as follows: Figure 6 As shown, the LOFAR spectrum of the underwater acoustic signal after processing with adaptive line spectrum enhancement technology based on temporal deconvolution interference suppression is as follows: Figure 7 As shown, the adaptive line spectrum enhancement technology based on time-deconvolution interference suppression provided in this application produces a stronger and more significant gain in detecting underwater acoustic signals compared to traditional adaptive line spectrum enhancement technologies. Simulation experiments demonstrate that, against a background of Gaussian white noise, the adaptive line spectrum enhancement technology based on time-deconvolution interference suppression provided in this application can improve the signal-to-noise ratio and exhibits better line spectrum enhancement performance, which is beneficial for subsequent line spectrum detection, extraction, and analysis.

[0098] The underwater acoustic signal line spectrum is detected and tracked using an adaptive dual-threshold technique. Line spectrum detection of the underwater acoustic signal is performed solely through mean constant false alarm rate (CFAR) detection. Figure 8 As shown, the line spectrum detection of underwater acoustic signals using ordered constant false alarm rate detection alone is as follows: Figure 9 As shown, the line spectrum detection of underwater acoustic signals is performed by combining mean constant false alarm rate (CFAR) detection and ordered CFAR detection. Figure 10 As shown in the figure, the blue line represents the original power curve, the orange line represents the detection threshold, and the frequency points above the detection threshold are the detection line spectra. The figure demonstrates that the combined detection of mean constant false alarm rate (CFAR) and ordered CFAR can extract effective line spectrum information from the underwater acoustic signal.

[0099] Azimuth direction finding was performed on a signal with a signal-to-noise ratio of -10 dB and an incoming direction of 40° using a weighted histogram method based on energy search. The direction finding results are as follows: Figure 11 As shown, it can be observed that the method provided in this application embodiment can clearly distinguish the direction of arrival of the wave even at a low signal-to-noise ratio, and the interference from the direction of non-arrival is relatively weak. Using this method, angle measurement error simulations were performed on signals of different frequencies, and the root mean square error of the angle measurement as a function of frequency is shown below. Figure 12 As shown, it can be found that within the 10Hz to 100Hz operating frequency band, the maximum direction finding error is less than 0.15°, which meets the requirements for low-frequency angle measurement.

[0100] based on Figure 2 In the simulation experiment, the underwater unmanned platform received data and used an improved energy search-based cross-spectral histogram orientation method for azimuth estimation. The orientation results are as follows: Figure 13 As shown in the image, the energy proportion of the position estimation result for surface-interfering vessels is relatively large, severely affecting the judgment. The spectrum of the original underwater acoustic signal is shown below. Figure 14 As shown, based on this, the main line spectra from the bearings of surface interference vessels are filtered out from the original spectrum, such as... Figure 15 As shown, the remaining spectral lines are considered to originate from an underwater submarine target. Finally, azimuth estimation is performed using the method provided in this application embodiment, and the azimuth estimation result is as follows. Figure 16 As shown, the direction finding results are unaffected by surface vessels, and the effectiveness of the target detection method provided in this application's embodiments is verified through simulation experiments.

[0101] Based on the direction finding results at this time, the corresponding low-frequency signal is obtained, and feature extraction is performed on the low-frequency signal. The LOFAR spectrum of the low-frequency signal is as follows: Figure 17 As shown, the DEMON spectrum of the low-frequency signal is as follows: Figure 18 As shown. Through simulation and experimental data verification, the anti-interference target detection performance of the target detection method provided in this application embodiment was verified under lake test conditions. The results of the first set of lake test data and the second set of lake test data are shown below. Figure 19 and Figure 20 As shown, Figure 19 and Figure 20 The left column is the azimuth spectrum before processing, which includes the direction finding effects of surface interference vessels; Figure 19 and Figure 20 The right column shows the processed azimuth spectrum. The test results show that, using the target detection method provided in this application embodiment, the influence of interfering targets has been significantly filtered out, while information about weak targets has been preserved, effectively improving the feature extraction capability for weak targets.

[0102] Based on the above description, the target detection method based on underwater acoustic signals provided in this application classifies the underwater acoustic signals by frequency band and uses a scalar-vector joint processing method to estimate the azimuth angles of low-frequency and high-frequency signals respectively, obtaining vector array multi-target direction finding results for low-frequency signals and scalar array direction finding results for high-frequency signals. Direction finding results with azimuths close to the scalar array direction finding results in the vector array multi-target direction finding results are filtered out. Through a spatial filtering process, the target azimuth direction finding results obtained after filtering mainly correspond to weak targets in the low-frequency region. By utilizing the characteristics of scalar and vector arrays, interference signals are effectively filtered out. While eliminating interference signals, the azimuth information of the target is retained, reducing the false alarm rate and improving the detection and analysis capability of weak targets.

[0103] Secondly, this application also provides a target detection system based on underwater acoustic signals. This target detection system is applied to an underwater unmanned exploration platform. The target detection system uses the target detection method based on underwater acoustic signals described above. The target detection system provided by this application can effectively filter out interference signals, retain the target's orientation information while eliminating interference signals, reduce false alarm rate, and improve the detection and analysis capability of weak targets.

[0104] It is understood that the term "exemplary" as used herein means "as an example, illustration, or description." Any embodiment described as "exemplary" is not necessarily preferred or superior to other embodiments and / or does not exclude features in combination with other embodiments. It should be understood that certain features of this application described in the context of a single embodiment for clarity may also be provided in combination in a single embodiment. Conversely, various features of this application described in the context of a single embodiment for clarity may also be provided individually or in any suitable combination or as part of any other described embodiment of this application.

[0105] In the description of this application, unless otherwise stated, " / " means "or," for example, A / B can mean A or B. The "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. Furthermore, "at least one" means one or more, and "multiple" means two or more. The terms "first," "second," etc., do not limit the quantity or order of execution, and "first," "second," etc., do not necessarily imply differences.

[0106] The above-disclosed embodiments are merely preferred embodiments of this application, but are not intended to limit the scope of this application. Those skilled in the art will understand that any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and scope of this application and the appended claims are equivalent substitutions and still fall within the scope of this application.

Claims

1. A target detection method based on underwater acoustic signals, the target detection method being applied to an unmanned underwater exploration platform, characterized in that, The target detection method comprises: The underwater acoustic signal is divided into high-frequency signals and low-frequency signals, linear spectrum detection is performed on the high-frequency signals and the low-frequency signals respectively, linear spectrum feature information of the low-frequency signals and linear spectrum feature information of the high-frequency signals are extracted; The low-frequency signals are subjected to differential pressure vector conversion to obtain vector low-frequency linear spectrum feature information, and a weighted histogram method based on energy search is used to estimate the azimuth angle of the vector low-frequency linear spectrum feature information to obtain a vector array multi-target direction finding result; Based on the linear spectrum feature information of the high-frequency signals, an adaptive beam forming algorithm is used to estimate the azimuth angle of the high-frequency signals to obtain a scalar array direction finding result; The direction finding results in the vector array multi-target direction finding result that differ from the scalar array direction finding result by a preset angle range are filtered out to obtain a target azimuth direction finding result; Feature extraction is performed on the low-frequency signals corresponding to the target azimuth direction finding result, LOFAR spectrum and DEMON spectrum of the low-frequency signals are obtained, and based on the LOFAR spectrum and the DEMON spectrum, a target corresponding to the low-frequency signals is identified and classified.

2. The method of claim 1, wherein, The weighted histogram method based on energy search is used to estimate the azimuth angle of the vector low-frequency linear spectrum feature information, which comprises: Based on the weighted histogram, a statistical interval is determined, and the window length and sliding step distance of a sliding window are configured, and the statistical values of each statistical interval are calculated by the sliding window method, Wherein, the statistical interval with the maximum statistical value is the statistical window where the sound source azimuth of the underwater acoustic signal is located, and the statistical window where the sound source azimuth is located is taken as the vector array multi-target direction finding result.

3. The method of claim 2, wherein, The statistical values of each statistical interval are calculated by the sliding window method, which comprises: Taking the starting point of the statistical interval as the data starting point of the sliding window, the energy weighted mean value is calculated according to the window length of the sliding window to obtain the calculation result; After the calculation is completed, the sliding window moves backward by the length of the sliding step distance, the window length of the sliding window remains unchanged, and the calculation is performed again, and the calculation in the statistical interval is repeated until all the data in the statistical interval are calculated; The maximum value of all the calculation results is selected as the statistical value of the statistical interval.

4. The method of claim 1, wherein, Before the underwater acoustic signal is divided into frequency bands, an adaptive linear spectrum enhancement technology based on time reversal convolution interference suppression is used to enhance the feature of the underwater acoustic signal.

5. The method of claim 4, wherein, The signal-to-noise ratio of the underwater acoustic signal is improved by the following time reversal convolution interference suppression expression, The time reversal convolution interference suppression expression is: y(t) = x(-t) * x(t) E(t) I(t); In the formula, "*" represents convolution operation, x(t) represents underwater acoustic signal mixed with noise, y(t) represents the output of the time reversal convolution interference suppression; E(t) represents time domain amplitude equalization window, and has: A represents the amplitude of the underwater acoustic signal, T represents the segment processing time length; I(t) represents interference suppression gate, and has:

6. The method of claim 4, wherein, An adaptive double threshold technology is used to detect the linear spectrum of the feature enhanced underwater acoustic signal, and the adaptive double threshold technology comprises mean constant false alarm rate detection and ordered constant false alarm rate detection.

7. The underwater acoustic signal based target detection system according to claim 1, wherein, the target azimuth direction finding result is judged for stability and uniqueness, and if the judgment is passed, the low frequency band signal corresponding to the target azimuth direction finding result is extracted for features, and if the judgment is not passed, the low frequency band signal and the high frequency band signal are re-estimated for azimuth angle.

8. The underwater acoustic signal based target detection system of claim 1, wherein, the low frequency band signal corresponding to the target azimuth direction finding result is extracted for features, and the LOFAR spectrum of the low frequency band signal is extracted by using short-time Fourier transform.

9. The underwater acoustic signal based target detection system of claim 1, wherein, the low frequency band signal corresponding to the target azimuth direction finding result is extracted for features, and the DEMON spectrum of the low frequency band signal is extracted by using wideband noise envelope modulation analysis technology to measure the fundamental frequency and each harmonic of the low frequency band signal.

10. A target detection system based on underwater acoustic signals, the target detection system being applied to an unmanned underwater exploration platform, characterized in that, the target detection system applies the underwater acoustic signal based target detection method according to any one of claims 1 to 9.

Images