Passive distributed joint detection method based on spatial domain windowing de-convolution

Abstract

The present application relates to the technical field of sonar signal processing, and in particular to a passive distributed joint detection method of spatial windowing and unwrapping, which is a sonar passive distributed joint detection method of one-dimensional azimuth space energy spectrum, and is used for passive target positioning, and the present application does not need prior information such as target sound source level and detection threshold, and can effectively eliminate false peak points of cross positioning.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of sonar signal processing, and particularly relates to a passive distributed joint detection method of spatial domain windowing and unwrapping. BACKGROUND

[0002] Passive underwater detection has good application prospects due to its good concealment, but its shortcomings are also obvious, that is, it cannot estimate the distance of the target. Passive distributed joint detection aims to integrate multi-platform detection information to confirm the target and improve the detection probability of the target, and can directly locate the real position of the target. The traditional passive distributed fusion detection is to divide a scanning grid in a concerned area, map the original spatial energy spectrum of each node to the grid range, and detect and locate the target in space through the intersection of the passive spatial energy spectrum of multiple nodes. In the case of not increasing the spatial scanning information, the passive spatial energy spectrum only contains energy spectrum information of different directions, so the intersection point after the fusion of the spatial spectrum of each node will have the problem of false peaks, and the elimination of false peaks is a problem that needs in-depth study.

[0003] Document [1, A direct position determination approach for underwater acoustic sensor networks, Lu Wang, 2020] proposes to perform near-field scanning in the two-dimensional plane of x and y, and to detect and locate the passive target through the strength relationship of the intersection point after the fusion of the spatial spectrum of multiple platforms. However, the spatial processing is performed in the two-dimensional plane, which results in a geometric growth of the operation amount of spatial spectrum estimation, and at the same time, when the spatial energy spectrum of two-dimensional scanning is fused, the data transmission amount of each platform is sharply increased compared with the traditional one-dimensional scanning, which increases the bandwidth burden of the data transmission link.

[0004] Document [2, Superdirective beamforming applied to SWellEx96 horizontal array data for source localization, T C Yang, 2019] proposes to perform cross positioning after unwrapping and post-processing of the spatial spectrum of each platform, which improves the spatial resolution of the small aperture array and weakens the influence of spatial sidelobe leakage. However, when the target is in the near-field range, the focusing and adaptation of the steering vector will cause beam bifurcation, and the output signal-to-noise ratio will be reduced. At the same time, when the sidelobe leakage of the strong interference in space is much larger than the beam output of the weak target, simply increasing the number of iterations cannot unwrap the weak target beam output.

[0005] Document [3, Multiple-array passive acoustic source locakization in shallowwater, D Tollefsen, ] proposes to use the matched field method to cross locate the passive target, which needs the environmental information such as sound velocity and water depth, and is sensitive to model adaptation, which is not conducive to engineering practice, and also needs to scan and calculate in the two-dimensional plane of depth and distance, and the operation amount is large. SUMMARY

[0006] In view of the deficiencies in the prior art, the purpose of the present application is to provide a sonar passive distributed joint detection method using only one-dimensional azimuth space energy spectrum, which does not need prior information such as target sound source level and detection threshold, can effectively eliminate the false peak points of cross location, and can locate the passive target.

[0007] To achieve the above purpose, the present application provides the following technical scheme: a spatial domain windowing deconvolution passive distributed joint detection method, characterized in that the steps are as follows:

[0008] Step 1: Constructing a scanning vector of a windowed beamformer

[0009]

[0010] x l and y l are the x-axis vector and y-axis vector of the array l respectively, indicates vector dot product, wind l is the spatial domain windowing amplitude weighting function,

[0011] After normalizing the scanning steering vector,

[0012] Obtaining the spatial spectrum estimation of the windowed beamformer

[0013]

[0014] Wherein, |||2 represents the two-norm of the vector;

[0015] Step 2: Using the R-L deconvolution algorithm in the form of discrete integral, iteratively processing the output of step 1 to give the energy distribution estimation of the target

[0016]

[0017] Wherein, is the beam pattern under the condition that each main lobe is directed after energy correction, indicates the spatial spectrum estimation of the windowed beamformer. Set the initial value of ​After i+1 iterations, the final spatial energy distribution is obtained

[0018] Step three: the spatial energy spectrum of each frequency point in the processing frequency band obtained in step two is non-coherently superimposed in the frequency domain to obtain the broadband spatial energy spectrum of the method Wherein, f d And f h The lower limit frequency and the upper limit of the broadband non-coherent frequency domain beam forming are respectively, the minimum maximum point of the spatial spectrum is used as the estimation of the background noise, and the spatial spectrum background average after the unwrapping processing,

[0019] PB l (θ)=PB l (θ)+min(peak(PB l (θ))

[0020] Wherein, peak( ) represents all maximum values of the spatial spectrum, and min( ) represents the minimum value of the sequence;

[0021] Step four: for the spatial spectrum PB l (θ) obtained after unwrapping of each platform, energy compensation of two-dimensional grid scanning points is completed based on the spherical expansion assumption, and the scanning point coordinates are assumed to be The distance and the direction of the array l reference origin are respectively The two-dimensional scanning beam angle is calculated The difference value of each platform scanning beam, the beam with the smallest distance is directly used for energy compensation by spherical expansion, and the sound source level estimation of the scanning point is obtained, and the beam angle θ j From Obtain

[0022]

[0023] After obtaining the two-dimensional scanning spatial energy spectrum distribution of each platform, the sound source level estimation of each platform at the two-dimensional space scanning point is averaged to obtain the fused passive distributed detection spatial spectrum

[0024]

[0025] Wherein, L is the number of platforms;

[0026] Step five: the standard deviation of the two-dimensional scanning spatial energy spectrum of each platform is calculated to measure the authenticity of the point

[0027]

[0028] The single-frame processing result of the method is obtained by correcting the step four result by using the above formula result

[0029]

[0030] Normalization processing is to eliminate the influence of dimension;

[0031] Step six: based on the difference of fluctuation of signal and noise between snapshots, the nonlinear filter is used to fuse the multi-frame processing results to improve the fusion performance of the distributed joint detection,

[0032]

[0033] wherein, represents the processing result of the first batch of the method, k represents the order of the nonlinear filter,

[0034] The maximum value search is performed on the intersection point of the output PB non (x i , y i ), and through a reasonable detection threshold, the positioning result estimation of the spatial distribution passive target can be obtained, and the coordinate estimation value of the target is output.

[0035] Compared with the prior art, the beneficial effects of the present application are: first, the spatial energy spectrum of each narrow sub-band in the processing frequency band is obtained by using a spatial windowed beamformer, so as to alleviate the problems of spatial spectrum widening and output signal-to-noise ratio reduction caused by the near-field propagation model. In order to improve the spatial resolution and reduce the influence of interference sidelobe leakage, the spatial spectrum of each narrow sub-band is deconvolved, and a point spread function is constructed by using the windowed beamformer. Since the equal-angle spatial scanning mode is adopted, the point spread function does not satisfy the shift-invariance, and the integral form R-L deconvolution iterative algorithm is adopted. Then, the non-coherent superposition of the outputs of each narrow sub-band is obtained, and the wideband spatial energy spectrum estimation is obtained. The two-dimensional scanning grid points are divided in the concerned region, the propagation loss is compensated by using the spherical spreading model or the sound field calculation result, and the sound source level of each scanning grid point is inversely calculated based on the one-dimensional wideband energy spectrum. The fusion two-dimensional spatial energy spectrum is obtained by averaging the inverse calculation results of each platform. In order to eliminate the false peak problem of cross positioning, the standard deviation of each scanning point is calculated by using the inverse calculation sound source level of each platform, the fusion sound source level is weighted and outputted, so as to weaken the output of the false target position and improve the output of the real target position. The simulation experiment and sea trial results verify the effectiveness of the algorithm. BRIEF DESCRIPTION OF DRAWINGS

[0036] Figure 1 The processing flowchart of the present application;

[0037] Figure 2 The overall situation diagram of the platform and the target under the simulation condition;

[0038] Figure 3 (a) Spatial energy spectrum of platform 1 after spatial windowed deconvolution and two-dimensional grid energy spectrum;

[0039] Figure 3 (b) Spatial energy spectrum and two-dimensional grid energy spectrum of platform 2 after spatial window deconvolution

[0040] Figure 3 (c) Spatial energy spectrum and two-dimensional grid energy spectrum of platform 3 after spatial window deconvolution

[0041] Figure 4 (a) Spatial energy spectrum and two-dimensional grid energy spectrum of platform 1 of conventional beamformer

[0042] Figure 4 (b) Spatial energy spectrum and two-dimensional grid energy spectrum of platform 2 of conventional beamformer

[0043] Figure 4 (c) Spatial energy spectrum and two-dimensional grid energy spectrum of platform 3 of conventional beamformer

[0044] Figure 5 Spatial spectrum after multiplication of output results of all platforms of spatial window deconvolution

[0045] Figure 6 Spatial spectrum after averaging of output results of all platforms of spatial window deconvolution

[0046] Figure 7 (a) Spatial spectrum after summation of output results of each platform of conventional beamformer

[0047] Figure 7 (b) Spatial spectrum output by algorithm in document [2]

[0048] Figure 8 Spatial spectrum processed by method of the application in single frame

[0049] Figure 9 Spatial spectrum processed by method of the application in multiple frames

[0050] Figure 10 (a) Sea trial route map

[0051] Figure 10 (b) Sea trial situation map

[0052] Figure 11 (a) Fusion spatial spectrum of conventional beamformer under high signal-to-noise ratio in sea trial

[0053] Figure 11 (b) Fusion spatial spectrum of document [2] under high signal-to-noise ratio in sea trial

[0054] Figure 11 (c) Fusion spatial spectrum of method of the application under high signal-to-noise ratio in sea trial

[0055] Figure 12 (a) Back-propagation sound pressure level of north array

[0056] Figure 12 (b) Nanjian counterfactual sound pressure level;

[0057] Figure 13 (a) Under low SNR in sea trial, conventional beamformer fusion spatial spectrum;

[0058] Figure 13 (b) Under low SNR in sea trial, literature [2] fusion spatial spectrum;

[0059] Figure 13 (c) Under low SNR in sea trial, the method of the present application fusion spatial spectrum. DETAILED DESCRIPTION

[0060] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0061] Consider the near-field passive propagation model. Suppose that there is a passive target in the near-field of the array, and the distances from the target to each array element l are Suppose that the array element l is the reference element, and the angle between the signal incident direction and the negative half of the x-axis is θ. The signal model of the kth frequency-domain snapshot received by the array l can be expressed as

[0062] X l (f, k) = a l (θ, r, f) s(f, k) + N l (f, k) (1)

[0063] wherein a l (θ, r, f) is the steering vector of the array l, s(f, k) is the frequency-domain representation of the signal, and N l (f, k) is the frequency-domain noise of the array l.

[0064]

[0065] wherein is the distance of the target relative to the reference origin of the array l, and c is the sound speed. is the distance of the target relative to the m # channel of the array l, wherein is the coordinate of the m # channel of the array l, is the coordinate of the reference origin of the array l.

[0066] Based on the element frequency-domain data of K frequency-domain snapshots, the cross-spectral density matrix of the array is obtained

[0067]

[0068] The relevant theory and specific content of step one are as follows:

[0069] In order to alleviate the problem of spatial spectrum beam broadening and output signal-to-noise ratio reduction caused by near-field propagation scanning mismatch, a spatial amplitude weighting window beamforming process is adopted. Then, taking the spatial windowed beam pattern as a point spread function, a deconvolution algorithm in the form of normalized integral is used to deconvolve the spatial energy spectrum of the windowed beamformer output.

[0070] The scanning steering vector of the spatial amplitude weighting window beamformer of the array l can be expressed as x l and y l are the x-axis vector and y-axis vector of the array l, respectively. denotes vector dot product, wind l is a spatial window amplitude weighting function. In order to estimate the input energy level of the target, the scanning steering vector is normalized to obtain the spatial spectrum estimation of the windowed beamformer

[0071]

[0072] wherein || ||2 denotes the two-norm of a vector.

[0073] The relevant theory and specific content of step two are as follows:

[0074] Since the spatial scanning angle is equiangular, the spatial beam response, i.e., the point spread function, does not satisfy the shift-invariance at this time, and an integral form deconvolution processing method is adopted to obtain the spatial energy spectrum estimation. Under the condition of not considering noise, the output of the beamformer can be expressed in the form of discrete integral

[0075]

[0076] wherein is the beam pattern under the condition that each main lobe is directed after energy correction, and it can be seen that the beam pattern after energy normalization satisfies beam pattern

[0077] denotes the real energy distribution of the target in space, and is as follows

[0078]

[0079] wherein |A j | 2 is the signal power, and δ( ) is a delta function, i.e., a unit impulse function.

[0080] The energy distribution estimation of the target is given by the R-L iterative deconvolution algorithm in the form of discrete integral

[0081]

[0082] where, is the spatial spectrum estimation of the windowed beamformer, for The initial value of can be assigned as After i+1 iterations, the final spatial energy distribution is obtained It can be seen that, compared with the iterative deconvolution processing in the form of convolution, the iterative deconvolution processing in the form of integral does not have the problem of boundary blurring, so it is not necessary to extend the data of P wCBF (f, θ). Since the deconvolution processing in the form of integral is adopted, the algorithm in the present application is also applicable to spatial spectrum deconvolution processing under non-shift-invariant conditions such as arbitrary array and near-field propagation.

[0083] The windowed deconvolution post-processing avoids the problem of reduction of output signal-to-noise ratio caused by the influence of various mismatches on the traditional high-resolution algorithm, sharpens the main lobe of the beamformer, reduces the side lobe level, and avoids the influence of strong interference on weak target detection.

[0084] The relevant theory and specific content of step three are as follows:

[0085] The windowed deconvolution post-processing is performed on the spatial energy spectrum of each frequency point in the processing frequency band, and then non-coherent superposition processing is performed in the frequency domain to obtain the wideband spatial energy spectrum of the method in the present application

[0086]

[0087] where, f d and f h are the lower limit frequency and the upper limit of the wideband non-coherent frequency domain beamforming, respectively.

[0088] In order to avoid the influence of the zero point of deconvolution in some directions on the subsequent processing, the minimum maximum value point of the spatial spectrum is used as the estimation of the background noise, and the spatial spectrum after deconvolution processing is averaged.

[0089] PB l (θ) = PB l (θ) + min(peak(PB l (θ)) (9)

[0090] Where, peak() represents all maximum values of the spatial spectrum, and min() represents the minimum value of the sequence.

[0091] The relevant theory and specific content of step four are as follows:

[0092] The spatial spectrum PB obtained after the spatial spectrum of each platform is unfolded l (θ), and the energy compensation of the two-dimensional grid scanning point is completed based on the spherical expansion assumption. Assuming that the scanning point coordinates are The distance and azimuth of the array l reference origin are The corresponding energy value can be obtained by spatial spectrum back calculation, and when the scanning grid point is in different quadrants, the angle estimation is corrected.

[0093]

[0094] Calculate the two-dimensional scanning beam angle The difference between the scanning beams of each platform is obtained, and the beam with the smallest distance is directly used for energy compensation based on spherical expansion, and the sound source level estimation of the scanning point is obtained. The beam angle θ is obtained by Obtain

[0095]

[0096] After obtaining the two-dimensional scanning spatial energy spectrum distribution of each platform, the sound source level estimation of each platform at the two-dimensional scanning point is averaged to obtain the fused passive distributed detection spatial spectrum.

[0097]

[0098] Wherein, L is the number of platforms.

[0099] The related theory and specific content of step five are as follows:

[0100] The above average operation cannot eliminate the false peak generated by cross positioning, and at the same time, the problem of greater output energy with greater distance from the platform cannot be solved, and the real position of the passive target cannot be estimated. Therefore, the variance of the sound source level estimation of each scanning point by multiple platforms is calculated to measure the fluctuation characteristics of the sound source level of the point. When there is a real target at a point, the sound source level estimation of the point by multiple platforms should be similar, and the corresponding sound source level variance is small. When a point does not contain a target or is a false peak point of cross positioning, the sound source level estimation of the point by multiple platforms is different, and the corresponding sound source level variance is large. Therefore, the standard deviation of the two-dimensional scanning spatial energy spectrum of each platform can be used to measure the authenticity of the point.

[0101]

[0102] The output result of the method of the present application is obtained by correcting formula (14) using the results of the above formula as follows

[0103]

[0104] The normalization processing is to eliminate the influence of dimension.

[0105] The relevant theory and specific content of step six are as follows:

[0106] Based on the difference in fluctuation of signals and noises between snapshots, if it is assumed that the amplitude fluctuation of signals between snapshots is small and the amplitude fluctuation of noises between snapshots is large, the non-linear filter can emphasize the energy output with small amplitude fluctuation and suppress the noise energy output with large amplitude fluctuation, thereby improving the output signal-to-noise ratio of the real target position. For the target source level of multi-platform spatial spectrum backstepping, at the real intersection point, the power output of the slowly varying target beam has small amplitude fluctuation, and the power output of the noise beam has large amplitude fluctuation. Thus, the non-linear filtering can be used to improve the fusion performance of distributed joint detection between multiple snapshots.

[0107]

[0108] wherein, represents the processing result of the first batch of the method, and k represents the order of the non-linear filter.

[0109] The output PS non (x i , y i ) intersection point is subjected to maximum value search, and the positioning result estimation of the spatially distributed passive target can be obtained by using a reasonable detection threshold, and the coordinate estimation value of the target is output.

[0110] Embodiment of the present application

[0111] Simulation experiment: three platforms are used in the simulation, and the three platforms are 70-meter uniform linear arrays with an element spacing of 3.75 meters (corresponding to 200 Hz) half wavelength. The first element of platform 1 is the reference element, which is located at the coordinate origin of the absolute coordinate system, and the array is arranged horizontally along the x-axis. The first element of platform 2 has a coordinate of (20000, 0) meters, and is also arranged horizontally along the x-axis. The first element of platform 3 has a coordinate of (10000, 100000) meters, and is arranged vertically along the y-axis. The distance between target one and the absolute coordinate origin is 10000 meters, the angle between target one and the positive half axis of the y-axis is 20°, and the sound source level is 145 dB. The distance between target two and the absolute coordinate origin is 15000 meters, the angle between target two and the positive half axis of the y-axis is 60°, and the sound source level is 120 dB. The distance between target three and the absolute coordinate origin is 20000 meters, the angle between target three and the positive half axis of the y-axis is 40°, and the sound source level is 130 dB. The overall situation of each platform in the whole scene is shown in Figure 2 .

[0112] The signal from the target to each array element obeys the energy attenuation law of spherical wave expansion. The ambient noise level is 45dB. The frequency domain snapshot number is 5. The 3dB beam width of the base station in the simulation condition is 1.46°-1.62° in the positive transverse direction of the base station, and the corresponding main lobe covers the spatial distance of 127m-141m at 5000. Thus, the scanning beam angle of the three platforms covers 0°-360°, and the scanning interval is 1°, so as to ensure the spatial azimuth resolution, wherein 0° is the angle with the negative half of the x-axis. The scanning grid x-axis range is [0, 20000]m, the y-axis range is [0, 15000]m, and the scanning interval of the x-axis and the y-axis is 50m, so as to avoid the adjacent scanning grid points falling into the same main lobe. In order to avoid the energy fluctuation existing in the single frequency processing and the problem of target side lobe energy leakage, a wideband processing is adopted, and the processing frequency band is [180-200]Hz, and the frequency resolution is 1Hz. The spatial domain window adopts the Hamming window, and the deconvolution iteration number is 20.

[0113] Firstly, the passive spatial energy spectrum of each platform and the spatial energy spectrum mapped to the two-dimensional scanning grid after the compensation of spherical propagation loss are obtained as shown in Figure 3 (a), (b) and (c). As a comparison, the true position of the target is given by a black circle in the grid spatial spectrum. As a comparison, the spatial spectrum of the three platform conventional beamformers and the spatial spectrum mapped to the two-dimensional scanning grid are shown in Figure 4 (a), (b) and (c). It can be seen that in the spatial spectrum mapped to the two-dimensional scanning grid, each target radiates an energy line from the platform reference origin, and the passive target is located by using the energy ray. The side lobe leakage of the strong target of the conventional beamformer is serious, and multiple side lobe caused pseudo energy lines appear in the two-dimensional grid energy spectrum, so that the detector false alarm is easily generated. At the same time, affected by the wide main lobe, the spatial widening of the energy line radiated by the target from the platform is more and more serious, so that the spatial resolution after the fusion of each node will be seriously affected. On the other hand, due to the ambiguity of the left and right flanks of the linear array, there are symmetrical spectrum peaks of each target in the scanning azimuth of 360°. Compared with the results of the conventional beamforming processing, the deconvolution effectively weakens the side lobe leakage of the target, sharpens the main lobe of the target energy spectrum, improves the spatial resolution of the base station, suppresses the appearance of the side lobe pseudo peak, improves the detection ability of the weak target under the influence of strong interference. A finer passive target radiation energy ray is obtained in the two-dimensional scanning space, and the spatial resolution of the two-dimensional distributed fusion detection is improved.

[0114] Figure 5 The spatial energy spectrum fused by the multiplication criterion after the deconvolution post-processing is given. It can be seen that simply multiplying the energy spectra of the three platforms can obtain the intersection result of the passive targets, but cannot solve the problem of the intersection positioning pseudo peak, and multiple intersection pseudo peaks are generated in the two-dimensional energy spectrum, which are difficult to eliminate.Figure 6 The spatial energy spectrum after deconvolution and post-processing is given by additive fusion. It can be seen that, as with the result of multiplying the energy spectrum, although the energy lines of each platform will have intersection points, the problem of false peaks in intersection positioning cannot be solved, and the strength relationship of each intersection point is less obvious, making it more difficult to determine the true position of the target. As a comparison, the spatial spectrum results after fusion of conventional beamformer and algorithm proposed in reference [2] are given as follows. Figure 7 As shown in (a)(b).

[0115] Compare Figure 6 and Figure 7 It can be seen that the energy rays of conventional beamformers are severely broadened, and false energy rays caused by sidelobe leakage are scattered around the target, with false crossover peaks increasing exponentially, making it impossible to accurately determine the target's true location. Traditional deconvolution algorithms, due to near-field focusing mismatch, although sharpening the main lobe and corresponding energy rays and improving spatial resolution, still suffer from significant effects from false energy rays caused by sidelobe leakage, resulting in numerous crossover points and making it impossible to correctly identify and locate targets with low input signal-to-noise ratios.

[0116] Finally, the result of single-frame spatial spectrum fusion using the method of this invention is given as follows: Figure 8 As shown, the method of this invention effectively eliminates redundant energy rays, avoids the false peak problem of passive cross-positioning, and can estimate the target's position by utilizing the intensity relationship of the bright spots. Simultaneously, since the intensity of the intersecting bright spots differs between the port and starboard sides, the method of this invention also solves the problem of port and starboard ambiguity in linear arrays.

[0117] Using the processing results of 5 frames, nonlinear filtering was performed to obtain the multi-frame processing results of the method of this invention, as shown below. Figure 9 As shown, it can be seen that, based on the characteristic that the amplitude fluctuation of the signal is small between snapshots, while the amplitude fluctuation of the noise is large between snapshots, the nonlinear filter emphasizes the energy output with smaller amplitude fluctuations and suppresses the noise energy output with larger amplitude fluctuations. This further suppresses spurious peak output, improves the output signal-to-noise ratio of the true target position, and is conducive to the accurate positioning of the true target.

[0118] Sea Trial Data: The algorithm performance was verified using publicly available Swellex96 test data. Detailed test information can be found at the publicly available website. For data processing of the northern and southern linear arrays, since the spacing between channels 14 and 15 in the northern array is much larger than the spacing between other elements, and the spacing between channels 13 and 14 in the southern array is also much larger than the spacing between other elements, elements 1-14 of the northern array are considered as Northern Array 1, and elements 15-27 are considered as Northern Array 2. Similarly, elements 1-13 of the southern array are considered as Southern Array 1, and elements 14-28 are considered as Southern Array 2. Coherent processing was used for the channels within each subarray. The test configuration is as follows: Figure 10The signal sampling rate is 3276.8 Hz, and 16384-point time-domain samples with 50% overlap are used for short-time Fourier transform to obtain the array element frequency-domain data, and the array element cross-spectral density matrix is obtained by using 5 batches of frequency-domain snapshots (corresponding to an integration time of 12.5 s).

[0119] a) High signal-to-noise ratio condition

[0120] First, consider the high signal-to-noise ratio scenario. The high source level signal emitted by the sound source ship contains a signal of 148 Hz, and the processing frequency band is selected as 140 Hz-160 Hz with a frequency interval of 0.2 Hz, and the processing frequency point contains the frequency point of 148 Hz. The space is scanned at an interval of 1° from 0° to 360°, the spatial window is a Hamming window, and the number of iterations of deconvolution is 50. The two-dimensional grid is divided into regions, the x-axis range is 0-2000 meters, the y-axis range is -3500-0 meters, and the grid interval is 50 meters. After obtaining the spatial spectrum estimation, the voltage level is converted into the input sound pressure level by using the sensitivity. Since the north array and the south array are arranged at approximately the same depth (200 meters deep and 198 meters deep), and the target moves along the isobath, the propagation loss of the target can be simplified, and considering the limited interface of the shallow sea environment, the sound source level of the two-dimensional grid point is obtained after compensating the propagation loss by (15*lgr) dB.

[0121] The fusion results of the conventional beamforming, the algorithm in document [2], and the method of the application at T0 are shown in (a), (b), and (c) respectively. Figure 11 As can be seen, the energy ray of the algorithm in document [2] has a lower intensity at the intersection point than the fusion intensity at the extension line, and cannot correctly detect the target by peak value discrimination. The method of the application effectively sharpens the beam main lobe, reduces the influence range of the target energy ray, weakens the energy ray leakage of the non-target point, and obtains an isolated energy output at the target position. The target can be detected by setting a suitable peak value detection threshold, and the absolute coordinates can be estimated.

[0122] The sound source level of the 148 Hz emission signal is estimated by using the spatial energy spectrum estimation of the north array and the south array, and the results of cross-location at each time are compensated for propagation loss by using a spherical expansion model, i.e., (15*lgr) dB, and the estimation history of the 1000 s north array and south array reverse sound source level is as shown in (a) and (b). Figure 12 (a) and (b). The source level of the sound source design is 158 dB, which is given together in the figure for comparison. As can be seen, due to the un-compensated propagation loss caused by seawater absorption, etc., the source level estimation result of the target fluctuates around the actual emission source level, and the estimation result basically conforms to the distribution of the designed emission source level. At the same time, the energy fluctuation characteristics of the single-frequency signal caused by sound field propagation are also present.

[0123] b) Low signal-to-noise ratio condition

[0124] Next, consider the scenario of reduced signal-to-noise ratio. The low source level signal emitted by the sound source ship contains a signal of 154 Hz, the processing frequency band is selected as 152-155 Hz, the frequency interval is unchanged, the processing frequency point contains the frequency point of 154 Hz, and other processing parameters are set as in the foregoing high signal-to-noise ratio condition.

[0125] At time T0, the fusion results of the conventional beamforming fusion, the algorithm in document [2] and the method of the application are as shown in (a), (b) and (c) respectively. Figure 13 As can be seen, the method of the application reduces the influence range of the energy rays, effectively suppresses the redundant energy rays and the false intersection points, avoids the detection and screening of false targets, and obtains isolated energy output at the target position, and has good practicability.

[0126] According to the implementation examples, it can be seen that the method of the application can solve the position information of the target in real time without the prior information of the original sound source level of the target, the detection threshold, the characteristic parameters and the like, has good false peak suppression and anti-space domain interference capabilities, and has good application prospects.

[0127] Although the embodiments of the application have been shown and described, it can be understood by those of ordinary skill in the art that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and spirits of the application, and the scope of the application is defined by the appended claims and their equivalents.

Claims

1. A method for spatially windowed unwrapping of passive distributed joint detection, characterized in that: The steps are as follows: Step one: constructing the scanning vector of the windowed beamformer and are arrays of of axis vectors and axis vectors, denotes vector dot product, is a spatial domain windowed magnitude weighting function, Normalizing scan steering vectors After, Obtaining the broadband spatial energy spectrum estimation of the windowed beamformer wherein denotes the two-norm of a vector; Step two: using the R-L unwrapping algorithm in the form of discrete integral to iteratively process the output of step one to give the energy distribution estimation of the target wherein, is the energy-corrected beam pattern for each main lobe pointing condition, denotes the wideband spatial energy spectrum estimate of the windowed beamformer, which is initialized as and after i+1 iterations the final spatial energy distribution is obtained; Step three: the spatial energy spectrum of each frequency point in the processing frequency band obtained in step two is non-coherently superimposed in the frequency domain to obtain a wideband spatial energy spectrum wherein, and are the lower limit frequency and the upper limit frequency of the wideband non-coherent frequency domain beamforming respectively, the minimum maximum point of the wideband spatial energy spectrum is used as the estimation of the background noise, and the wideband spatial energy spectrum after the de-convolution processing is background averaged, wherein denotes finding all maxima of the wideband spatial energy spectrum, denotes finding the sequence minima; Step four: the wideband spatial energy spectrum obtained after unwrapping each platform , the energy compensation of the two-dimensional grid scanning point is completed based on the spherical expansion assumption, assuming that the scanning point coordinates are , the distance and orientation of the array from the reference origin are , , the difference between the two-dimensional scanning beam angle and the scanning beam of each platform is calculated, and the beam with the smallest distance is directly energy compensated by spherical expansion to serve as the sound source level estimation of the scanning point. The beam angle is obtained from ​ After obtaining the two-dimensional scanning spatial energy spectrum distribution of each platform, the sound source level estimation of each platform at the two-dimensional scanning point is averaged to obtain the fused passive distributed detection broadband spatial energy spectrum wherein, Np is the number of platforms; Step five: calculating the standard deviation of the two-dimensional scanning spatial energy spectrum of each platform to measure the authenticity of the point Using the results of the above formula to correct the results of step four to obtain the single-frame processing result The normalization processing is to eliminate the influence of dimension; Step six: based on the difference in fluctuation between signals and noise between snapshots, using nonlinear filtering to fuse the multi-frame processing results to output, so as to improve the fusion performance of the distributed joint detection, wherein represents the first processing result of the batch, represents the order of the non-linear filter, On the output The maximum value search is made at the cross point, and the positioning result estimation of the spatial distribution passive target is obtained by a reasonable detection threshold, and the coordinate estimation value of the target is output.

Images