Target line spectrum extraction method and system for improving the azimuth estimation capability of vector hydrophones

Abstract

This invention provides a method and system for extracting target line spectra to improve the azimuth estimation capability of vector hydrophones. The method includes the following steps: S1: acquiring vector hydrophone data and performing normalization processing; the vector hydrophone data includes scalar channel data; S2: obtaining high-order temporal intermittent features based on the vector hydrophone data and extracting the target line spectrum; S3: calculating the spectral intensity of the target line spectrum to obtain the target azimuth. This invention utilizes the effective spatial texture information of the signal's time-frequency image to obtain high-order temporal intermittent features. After extracting the target line spectrum, interfering line spectra are removed or their spectral intensity is reduced. Then, a weighted histogram statistical method is used to highlight the role of the strong target line spectrum, reduce the influence of interfering line spectra, and improve the azimuth estimation capability.

Description

Technical Field

[0001] This invention relates to the field of sonar, and more specifically, to a method and system for extracting target line spectra to improve the azimuth estimation capability of vector hydrophones. Background Technology

[0002] Single-vector hydrophone detection is widely used in small-scale detection platforms due to its high concealment, low power, small scale, and flexible application. Single-vector hydrophone azimuth estimation refers to the process of directly estimating the spatial azimuth information of a target by processing the received sound pressure and vibration velocity information. The acoustic energy flow method based on complex sound intensity analysis is a typical example. Applying the frequency domain implementation of the acoustic energy flow method to histogram statistics can obtain more comprehensive target azimuth information. The weighted histogram statistical method is suitable for detecting broadband signals containing line spectra and can highlight the role of strong line spectra. By weighting the estimated azimuth of all frequency points according to spectral intensity and statistically allocating them to the corresponding azimuth information intervals, the azimuth estimation curve at a certain moment is obtained. The peak of the curve corresponds to the estimated azimuth value of a target. This method is also called the weighted histogram statistical method. If the spectral intensity of the interfering line spectrum is strong, the weighted histogram statistical method may easily generate non-target curve peaks.

[0003] To address the aforementioned problems, common methods include using filtering to extract the signal within the target bandwidth and reduce line spectrum interference outside the bandwidth, or using integration methods to remove non-stationary, incoherent line spectrum signals from the data. These two methods, through special processing, aim to eliminate interfering line spectra. However, these methods are ineffective for unknown target signal frequency bands and stationary line spectra in the environment. Currently, single-vector hydrophones are susceptible to azimuth estimation errors or false alarms due to interference line spectra. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the purpose of this invention is to provide a method and system for extracting target line spectra that improves the azimuth estimation capability of vector hydrophones.

[0005] A target line spectrum extraction method for improving the azimuth estimation capability of a vector hydrophone, provided by the present invention, includes:

[0006] Step S1: Acquire vector hydrophone data and perform normalization processing;

[0007] The vector hydrophone data includes scalar channel data;

[0008] Step S2: Based on vector hydrophone data, obtain high-order temporal intermittent features and extract the target line spectrum;

[0009] Step S3: Calculate the spectral intensity of the target line spectrum to obtain the target's orientation.

[0010] Preferably, step S1 includes the following sub-steps:

[0011] Step S1.1: Obtain one frame of data from the vector hydrophone, and form a one-dimensional array after Fourier transform. Each number in the array represents the amplitude of a frequency point. After Fourier transforming multiple frames of data arranged in time sequence, accumulate them to form a two-dimensional matrix representing time and frequency.

[0012] Step S1.2: Normalize the two-dimensional matrix.

[0013] Preferably, step S2 includes the following sub-steps:

[0014] Step S2.1: Combine the one-dimensional array formed by the Fourier transform of the current data with the n-row array corresponding to the previous t seconds to form an n+1-row sub-two-dimensional matrix, and solve for its time interval characteristics:

[0015]

[0016] Among them, P t Represents the pixel value, and E{·} is the expected value of the data;

[0017] Step S2.2: Process the temporal intermittency of the sub-two-dimensional matrix with n+1 rows using higher-order central distances to obtain higher-order temporal intermittency features:

[0018]

[0019] Preferably, if it is possible to extract a segment of data from the current environmental sonar image database that contains only background and no target line spectrum using expert experience, then a high-order temporal intermittent feature extraction method based on background learning should be selected.

[0020] If it is impossible to extract a segment of data from the current environmental sonar image database that contains no threatening targets but only data containing both interference and targets, then the option is to use integral accumulation of high-order temporal intermittent features without prior knowledge.

[0021] Preferably, the high-order temporal intermittent feature extraction process based on background learning includes:

[0022] In the database, select the mean U of background data without targets within d seconds, and m sets of data within d seconds:

[0023]

[0024] Select n sets of data within a time step t seconds before the current time, and calculate the difference between each set and U before processing:

[0025]

[0026] Where k represents the order.

[0027] The preferred high-order temporal interstitial feature extraction process without prior knowledge includes:

[0028] Select the historical mean U of the data set nm within the time step s seconds ago:

[0029]

[0030] Where n represents all the data currently stored in the device;

[0031] Selecting t seconds as the time span, the data from v groups within the t seconds prior to the current time are subtracted from U and then processed.

[0032]

[0033] Preferably, step S3 includes calculating the spectral intensity of the signal line spectrum after removing interference line spectra using higher-order time intervals, and using a statistical weighted histogram method that uses acoustic energy flow weighted by spectral intensity, with the peak of the fitted curve being the target azimuth.

[0034] A target line spectrum extraction system for improving the azimuth estimation capability of a vector hydrophone, provided by the present invention, includes:

[0035] Module M1: Acquires vector hydrophone data and performs normalization processing;

[0036] The vector hydrophone data includes scalar channel data;

[0037] Module M2: Based on vector hydrophone data, obtain high-order temporal intermittent features and extract target line spectra;

[0038] Module M3: Calculates the spectral intensity of the target line spectrum to obtain the target's orientation.

[0039] Preferably, module M1 includes the following sub-modules:

[0040] Module M1.1: Obtains one frame of data from the vector hydrophone, and forms a one-dimensional array after Fourier transform, where each number in the array represents the amplitude of a frequency point; it accumulates multiple frames of data arranged in time sequence after Fourier transform to form a two-dimensional matrix representing time and frequency.

[0041] Module M1.2: Normalizes two-dimensional matrices.

[0042] Preferably, module M2 includes the following sub-modules:

[0043] Module M2.1: Combines the one-dimensional array formed by Fourier transforming the current data with the n-row array corresponding to the previous t seconds, forming an n+1-row sub-two-dimensional matrix, and solves for the time intermittency of this matrix.

[0044]

[0045] Among them, P t Represents the pixel value, and E{·} is the expected value of the data;

[0046] Module M2.2: Processes the temporal intermittency of the (n+1)-row sub-two-dimensional matrix using higher-order central distances to obtain higher-order temporal intermittency characteristics:

[0047]

[0048] Preferably, if it is possible to extract a segment of data from the current environmental sonar image database that contains only background and no target line spectrum using expert experience, then a high-order temporal intermittent feature extraction method based on background learning should be selected.

[0049] If it is impossible to extract a segment of data from the current environmental sonar image database that contains no threatening targets but only data containing both interference and targets, then the option is to use integral accumulation of high-order temporal intermittent features without prior knowledge.

[0050] Preferably, the high-order temporal intermittent feature extraction process based on background learning includes:

[0051] In the database, select the mean U of background data without targets within d seconds, and m sets of data within d seconds:

[0052]

[0053] Select n sets of data within a time step t seconds before the current time, and calculate the difference between each set and U before processing:

[0054]

[0055] Where k represents the order.

[0056] The preferred high-order temporal interstitial feature extraction process without prior knowledge includes:

[0057] Select the historical mean U of the data set nm within the time step s seconds ago:

[0058]

[0059] Where n represents all the data currently stored in the device;

[0060] Selecting t seconds as the time span, the data from v groups within the t seconds prior to the current time are subtracted from U and then processed.

[0061]

[0062] Preferably, module M3 includes calculating the spectral intensity of the signal line spectrum after removing interference line spectra using high-order time intervals, and using a statistical weighted histogram method that uses acoustic energy flow weighted by spectral intensity, with the peak of the fitted curve being the target azimuth.

[0063] Compared with the prior art, the present invention has the following beneficial effects:

[0064] 1. This invention utilizes the effective spatial texture information of the time-frequency image of the signal to obtain high-order temporal intermittent features. After extracting the target line spectrum, it removes or reduces the spectral intensity of the interfering line spectrum. Then, it uses the weighted histogram statistical method to highlight the role of the target strong line spectrum, reduce the influence of the interfering line spectrum, and improve the azimuth estimation capability.

[0065] 2. The algorithm used in this invention is simple in structure, requires no prior setting of any parameters, is flexible and fast, and can remove interference line spectra without prior knowledge of the interference signal frequency and the target signal frequency. It can still be extracted when the target line spectrum produces a Doppler frequency shift.

[0066] 3. This invention can remove stationary interference signal lines in the environment that cannot be removed during integral processing, and can also effectively suppress interference lines, improve azimuth estimation capability, and reduce false alarm rate.

[0067] Other beneficial effects of the present invention will be explained in detail through the introduction of specific technical features and technical solutions in specific embodiments. Those skilled in the art should be able to understand the beneficial technical effects brought about by these technical features and technical solutions through the introduction of these technical features and technical solutions. Attached Figure Description

[0068] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0069] Figure 1 This is a flowchart of the method of the present invention.

[0070] Figure 2 This is a flowchart of the processing flow of the present invention.

[0071] Figure 3 This is a flowchart of the HOT-Lac feature extraction method based on background learning according to the present invention.

[0072] Figure 4 This is a time-frequency diagram containing the target line spectrum and interference line spectrum, without HOT-Lac feature line spectrum extraction in this invention.

[0073] Figure 5 This is a schematic diagram showing the results of the HOT-Lac feature line spectrum extraction method based on background learning according to the present invention.

[0074] Figure 6 This is a flowchart of the HOT-Lac feature extraction method for integral accumulation without prior knowledge, as described in this invention.

[0075] Figure 7 This is a diagram showing the azimuth estimation method using weighted histogram statistics before and after the extraction of the target line spectrum in this invention. Detailed Implementation

[0076] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0077] The problem addressed by this invention is that single-vector hydrophones are affected by interference line spectra, leading to incorrect azimuth estimation or false alarms. The basic idea is to utilize the effective spatial texture information of the signal's time-frequency image to obtain high-order temporal intermittency (HOT-Lac) features, extract the target line spectrum, eliminate interference line spectra or reduce the spectral intensity of interference line spectra, and then use a weighted histogram statistical method to highlight the role of the strong target line spectrum and reduce the influence of interference line spectra, thereby improving the azimuth estimation capability.

[0078] Interlacity is a specialized term in fractal geometry that provides a measure of how a pattern fills space, reflecting the degree of deviation between actual pixel values ​​and expected values, and is defined as follows:

[0079]

[0080] P t Pixel value.

[0081] As can be seen from the formula, the more uneven the pixel distribution of the image, the larger the value of L.

[0082] Time-frequency images utilize joint time and frequency features to characterize the signals acquired by these vector hydrophones. Target and interference line spectra exhibit differences in temporal or frequency characteristics; that is, their intensity or frequency changes differ over time. Since the non-stationary behavior of the target often causes greater variations in the target line spectrum, higher-order central moments can be used to improve the ability to distinguish between them. Using higher-order central moments better characterizes the heavy-tailed properties of interference noise, increases the inter-class distance between the target and interference line spectra, and achieves effective differentiation between them. Different orders can extract different inter-class distances for the target and interference line spectra, enabling applications in different scenarios. Therefore, utilizing higher-order temporal gap (HOT-Lac) features can better remove relatively stationary interference line spectra.

[0083] Reference Figure 1 As shown, a target line spectrum extraction method for improving the azimuth estimation capability of a vector hydrophone includes:

[0084] Step 1: Obtain one frame of data from the vector hydrophone, perform a Fourier transform to form a one-dimensional array. Each number in the array represents the amplitude at a frequency point. Multiple frames of data arranged in time sequence are accumulated after Fourier transforms to form a two-dimensional matrix representing time and frequency. This matrix is ​​the time-frequency diagram of the signal, with row numbers representing frequency on the X-axis, column numbers representing time on the Y-axis, and the elements in the array representing the amplitude at each frequency.

[0085] Step 2: Normalize the matrix representing time and location after accumulation.

[0086] Step 3: Combine the one-dimensional array formed by the Fourier transform of the current data with the n-row array corresponding to the previous t seconds to form an n+1-row sub-two-dimensional matrix. t seconds represents the time step size of the sliding motion on the image. t cannot be too large to prevent the loss of characteristic features of the target and interference spectra, while ensuring a sufficiently large t to obtain rich target spectra information. Its temporal interval characteristic is calculated as follows:

[0087]

[0088] Among them, P t Represents the pixel value, and E{·} is the expected value of the data;

[0089] Step 4: Process the temporal gap characteristics of the (n+1)-row sub-two-dimensional matrix using the higher-order central distance to obtain the higher-order temporal gap characteristic (HOT-Lac). As shown in the following equation:

[0090]

[0091] L kThe larger the value, the stronger the probability that the target spectral line is present in the corresponding azimuth. k The lower the value, the stronger the probability that the corresponding direction contains interfering spectral lines.

[0092] Step 5: Calculate the spectral intensity of the signal line spectrum after removing interference line spectra using higher-order time intervals. Use the acoustic energy flow to perform a statistical weighted histogram method based on spectral intensity weighting. The peak of the fitted curve is the target azimuth.

[0093] This invention utilizes the effective spatial texture information of the time-frequency image of the signal to obtain high-order temporal intermittent features. After extracting the target line spectrum, it removes or reduces the spectral intensity of the interfering line spectrum. Then, it uses the weighted histogram statistical method to highlight the role of the target strong line spectrum, reduce the influence of the interfering line spectrum, and improve the azimuth estimation capability.

[0094] The above are basic embodiments of the present invention. The technical solution of the present invention will be further described below through a preferred embodiment.

[0095] Example 1

[0096] Reference Figure 1 As shown, the proposed method for improving the azimuth estimation capability of vector hydrophones through target line spectrum extraction includes the following steps:

[0097] Step 1: Perform a Fourier transform on the data from the scalar channel of the vector hydrophone.

[0098] The method in this embodiment uses the spectral intensity of the vector hydrophone's scalar channel as the weighting coefficient of the weighted histogram, while the extraction of high-order time-interval (HOT-Lac) feature lines only requires processing the vector hydrophone's scalar channel signal.

[0099] Step 2: Refer to Figure 2 As shown, based on the criterion of whether expert experience can be used to extract data from the current environment's sonar database that is targetless and contains interference line spectra, the choice is to use either the background learning-based HOT-Lac feature extraction method or the integral accumulation HOT-Lac feature extraction method without prior knowledge.

[0100] 1. If we can use expert experience to extract a segment of data from the current environmental sonar image database that contains only background and no target line spectrum, we should choose to use the HOT-Lac feature extraction method based on background learning.

[0101] Reference Figure 3 As shown, the HOT-Lac feature extraction method based on background learning consists of the following steps:

[0102] 1. By manually selecting the mean U of background data without targets within d seconds in the database, and m sets of data within d seconds, as shown in the following formula:

[0103]

[0104] 2. Select n sets of data within a time step t seconds before the current time, and calculate the difference between each set and U, as shown in the following formula:

[0105]

[0106] Where k represents the order.

[0107] Reference Figure 4 and Figure 5 As shown, the algorithm is validated using a set of data. This data contains interference line spectra, and the target sound source emits three line spectra with different intensities. Comparing the images before and after processing, it can be seen that the HOT-Lac feature extraction method based on background learning can effectively remove interference line spectra and retain the target line spectra.

[0108] 2. If it is not possible to extract a segment of the current environmental sonar image database that contains no threatening targets and only data that includes both interference and targets, then choose to use integral cumulative HOT-Lac feature extraction without prior knowledge.

[0109] Reference Figure 6 As shown, the workflow of the HOT-Lac feature extraction method without prior knowledge is as follows:

[0110] 1) The integral accumulation HOT-Lac feature extraction method without prior knowledge differs from the background learning-based HOT-Lac feature extraction method. Its computational cost is higher than that of the background learning-based HOT-Lac feature extraction method. The integral accumulation HOT-Lac feature extraction method without prior knowledge selects the historical mean U of nm groups of data within a time step of s seconds ago (corresponding to m groups), where n is all the data currently stored in the device, as shown in the following formula:

[0111]

[0112] 2) Select t seconds as the time span. For the v sets of data within t seconds before the current time, calculate the difference between each set and U, and then process the data as follows:

[0113]

[0114] The differences between the background learning-based HOT-Lac feature extraction method and the integral accumulation-based HOT-Lac feature extraction method without prior knowledge are as follows:

[0115] a. Compared to the background learning-based HOT-Lac feature extraction method, the integral accumulation HOT-Lac feature extraction method without prior knowledge requires a period of data accumulation and processing, which may lead to loss of the target line spectrum or failure to completely remove interfering line spectra.

[0116] b. The integral accumulation HOT-Lac feature extraction method without prior knowledge requires a large amount of data. If the amount of data is insufficient, m and v in equations (6) and (7) need to be appropriately selected according to the actual situation.

[0117] c. The integral accumulation HOT-Lac feature extraction method without prior knowledge can be used regardless of whether the target line spectrum is present. The background learning-based HOT-Lac feature extraction method is mainly used when background data without the target line spectrum can be extracted from the current environmental sonar signal.

[0118] Step 3: Azimuth estimation using the vector hydrophone weighted histogram method after line spectrum extraction.

[0119] Compare the weighted histogram-based azimuth estimation before and after line spectrum extraction, such as... Figure 7 As shown, the peaks were weakened, while the peaks at the target's azimuth were highlighted, thus achieving the goal of extracting the target's spectral density. It can be seen that the target's azimuth was highlighted, and the peaks were reduced.

[0120] The algorithm used in this invention is simple in structure, requires no prior parameter setting, is flexible and fast, and can remove interference line spectra without prior knowledge of the interference signal frequency and the target signal frequency. It can still be extracted when the target line spectrum produces a Doppler frequency shift. It can remove stationary interference signal line spectra in the environment that cannot be removed in the integral processing, and can effectively suppress interference line spectra, improve the azimuth estimation capability, and reduce the false alarm rate.

[0121] This invention also provides a target line spectrum extraction system for improving the azimuth estimation capability of a vector hydrophone. The target line spectrum extraction system for improving the azimuth estimation capability of a vector hydrophone can be implemented by executing the process steps of the target line spectrum extraction method for improving the azimuth estimation capability of a vector hydrophone. That is, those skilled in the art can understand the target line spectrum extraction method for improving the azimuth estimation capability of a vector hydrophone as a preferred embodiment of the target line spectrum extraction system for improving the azimuth estimation capability of a vector hydrophone.

[0122] Specifically, a target line spectrum extraction system for improving the azimuth estimation capability of vector hydrophones includes:

[0123] Module M1: Acquires vector hydrophone data and performs normalization processing;

[0124] The vector hydrophone data includes scalar channel data;

[0125] Module M2: Based on vector hydrophone data, obtain high-order temporal intermittent features and extract target line spectra;

[0126] Module M3: Calculates the spectral intensity of the target line spectrum to obtain the target's orientation.

[0127] The module M1 includes the following sub-modules:

[0128] Module M1.1: Obtains one frame of data from the vector hydrophone, and forms a one-dimensional array after Fourier transform, where each number in the array represents the amplitude of a frequency point; it accumulates multiple frames of data arranged in time sequence after Fourier transform to form a two-dimensional matrix representing time and frequency.

[0129] Module M1.2: Normalizes two-dimensional matrices.

[0130] Module M2 includes the following sub-modules:

[0131] Module M2.1: Combines the one-dimensional array formed by Fourier transforming the current data with the n-row array corresponding to the previous t seconds, forming an n+1-row sub-two-dimensional matrix, and solves for the time intermittency of this matrix.

[0132]

[0133] Among them, P t Represents pixel value;

[0134] Module M2.2: Processes the temporal intermittency of the (n+1)-row sub-two-dimensional matrix using higher-order central distances to obtain higher-order temporal intermittency characteristics:

[0135]

[0136] If it is possible to extract a segment of data from the current environmental sonar image database that contains only background and no target line spectrum using expert experience, then a high-order temporal intermittent feature extraction method based on background learning should be selected.

[0137] If it is impossible to extract a segment of data from the current environmental sonar image database that contains no threatening targets but only data containing both interference and targets, then the option is to use integral accumulation of high-order temporal intermittent features without prior knowledge.

[0138] The process of extracting higher-order temporal intermittent features based on background learning includes:

[0139] In the database, select the mean U of background data without targets within d seconds, and m sets of data within d seconds:

[0140]

[0141] Select n sets of data within a time step t seconds before the current time, and calculate the difference between each set and U before processing:

[0142]

[0143] The high-order temporal interstitial feature extraction process without prior knowledge through integral accumulation includes:

[0144] Select the historical mean U of the data set nm within the time step s seconds ago:

[0145]

[0146] Where n represents all the data currently stored in the device;

[0147] Selecting t seconds as the time span, the data from v groups within the t seconds prior to the current time are subtracted from U and then processed.

[0148]

[0149] The module M3 includes calculating the spectral intensity of the signal line spectrum after removing interference line spectra using high-order time intervals, and using a statistical weighted histogram method that uses acoustic energy flow weighted by spectral intensity. The peak of the fitted curve is the target azimuth.

[0150] Those skilled in the art will understand that, besides implementing the system and its various devices, modules, and units provided by this invention in the form of purely computer-readable program code, the same functions can be achieved entirely through logical programming of the method steps, making the system and its various devices, modules, and units of this invention function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, the system and its various devices, modules, and units provided by this invention can be considered as a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered as structures within the hardware component; alternatively, the devices, modules, and units for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0151] In the description of this application, it should be understood that the terms "upper", "lower", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0152] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A method for extracting target line spectra to improve the azimuth estimation capability of vector hydrophones, characterized in that, include: Step S1: Acquire vector hydrophone data and perform normalization processing; The vector hydrophone data includes scalar channel data; Step S2: Based on the vector hydrophone data, obtain high-order temporal intermittent features and extract the target line spectrum; Step S3: Calculate the spectral intensity of the target line spectrum to obtain the target's bearing; If it is possible to extract a segment of data from the current environmental sonar image database that contains only background and no target line spectrum using expert experience, then a high-order temporal intermittent feature extraction method based on background learning should be selected. If it is impossible to extract a segment of data from the current environmental sonar image database that contains no threatening targets but only both interference and targets, then we choose to use integral accumulation of high-order temporal intermittent features without prior knowledge for extraction. The process of extracting higher-order temporal intermittent features based on background learning includes: In the database, select the mean U of background data without targets within d seconds, and m sets of data within d seconds: Select n sets of data within a time step t seconds before the current time, and calculate the difference between each set and U before processing: Where k represents the order; The high-order temporal interstitial feature extraction process without prior knowledge includes: Select the historical mean U of the data set nm within the time step s seconds ago: Where n represents all the data currently stored in the device; Selecting t seconds as the time span, the data from v groups within the t seconds prior to the current time are subtracted from U and then processed. Step S3 includes calculating the spectral intensity of the signal line spectrum after removing interference line spectra using higher-order time intervals, and using the acoustic energy flow to perform a statistical weighted histogram method based on spectral intensity weighting. The peak of the fitted curve is the target azimuth.

2. The target line spectrum extraction method for improving the azimuth estimation capability of vector hydrophones according to claim 1, characterized in that, Step S1 includes the following sub-steps: Step S1.1: Obtain one frame of data from the vector hydrophone, and form a one-dimensional array after Fourier transform. Each number in the array represents the amplitude of a frequency point. After Fourier transforming multiple frames of data arranged in time sequence, accumulate them to form a two-dimensional matrix representing time and frequency. Step S1.2: Normalize the two-dimensional matrix.

3. The target line spectrum extraction method for improving the azimuth estimation capability of vector hydrophones according to claim 2, characterized in that, Step S2 includes the following sub-steps: Step S2.1: Combine the one-dimensional array formed by the Fourier transform of the current data with the n-row array corresponding to the previous t seconds to form an n+1-row sub-two-dimensional matrix, and solve for its time interval characteristics: Among them, P t Represents the pixel value, and E{·} is the expected value of the data; Step S2.2: Process the temporal intermittency of the sub-two-dimensional matrix with n+1 rows using higher-order central distances to obtain higher-order temporal intermittency features:

4. A target line spectrum extraction system for improving the azimuth estimation capability of vector hydrophones, characterized in that, include: Module M1: Acquires vector hydrophone data and performs normalization processing; The vector hydrophone data includes scalar channel data; Module M2: Based on vector hydrophone data, obtain high-order temporal intermittent features and extract target line spectra; Module M3: Calculates the spectral intensity of the target line spectrum to obtain the target's bearing; If it is possible to extract a segment of data from the current environmental sonar image database that contains only background and no target line spectrum using expert experience, then a high-order temporal intermittent feature extraction method based on background learning should be selected. If it is impossible to extract a segment of data from the current environmental sonar image database that contains no threatening targets but only both interference and targets, then we choose to use integral accumulation of high-order temporal intermittent features without prior knowledge for extraction. The process of extracting higher-order temporal intermittent features based on background learning includes: In the database, select the mean U of background data without targets within d seconds, and m sets of data within d seconds: Select n sets of data within a time step t seconds before the current time, and calculate the difference between each set and U before processing: Where k represents the order; The high-order temporal interstitial feature extraction process without prior knowledge includes: Select the historical mean U of the data set nm within the time step s seconds ago: Where n represents all the data currently stored in the device; Selecting t seconds as the time span, the data from v groups within the t seconds prior to the current time are subtracted from U and then processed. Step S3 includes calculating the spectral intensity of the signal line spectrum after removing interference line spectra using higher-order time intervals, and using the acoustic energy flow to perform a statistical weighted histogram method based on spectral intensity weighting. The peak of the fitted curve is the target azimuth.

5. The target line spectrum extraction system for improving the azimuth estimation capability of vector hydrophones according to claim 4, characterized in that, The module M1 includes the following sub-modules: Module M1.1: Obtains one frame of data from the vector hydrophone, and forms a one-dimensional array after Fourier transform, where each number in the array represents the amplitude of a frequency point; it accumulates multiple frames of data arranged in time sequence after Fourier transform to form a two-dimensional matrix representing time and frequency. Module M1.2: Normalizes two-dimensional matrices.

6. The target line spectrum extraction system for improving the azimuth estimation capability of vector hydrophones according to claim 5, characterized in that, The module M2 Includes the following sub-modules: Module M2.1: Combines the one-dimensional array formed by Fourier transforming the current data with the n-row array corresponding to the previous t seconds, forming an n+1-row sub-two-dimensional matrix, and solves for the time intermittency of this matrix. Among them, P t Represents the pixel value, and E{·} is the expected value of the data; Module M2.2: Processes the temporal intermittency of the (n+1)-row sub-two-dimensional matrix using higher-order central distances to obtain higher-order temporal intermittency characteristics:

Images