A Fusion Detection Method Based on Weighted Fourier Integral Method

Abstract

The invention provides a detection method based on weighted Fourier integral method, which adopts double triangular window, double cosine window and double Kaiser window multi-group windowing mode to shift left and right and select small value for beam diagram of weighted Fourier integral method. Since beam diagrams of three weighting modes have differences in sidelobe level, the invention can reduce the sidelobe level of beam diagram by selecting small value for blending, reduce the background level of target estimation result, improve the detection performance of weak target, accurately estimate the position of weak target, and overcome the problem that weak target is difficult to detect under high background caused by high sidelobe level of beam diagram when single window function is weighted; the invention obtains narrower main lobe of beam by using the method of shifting left and right of beam diagram, overcomes the problem that main lobe of beam diagram is wide under single window function, and improves the display effect of target in azimuth estimation result.

Description

Technical Field

[0001] The present invention belongs to the technical field of array signal processing, and in particular relates to a fusion detection method based on a weighted Fourier integral method. Background Art

[0002] Conventional beamforming technology plays an important role in signal processing of passive sonar platforms. However, due to the limited size of passive sonar platforms, the physical size of detection sonars is often constrained, resulting in insufficient angular resolution when estimating the angle of arrival of the sonar, and limited noise suppression capabilities.

[0003] In order to improve the azimuth estimation performance of passive sonar, some scholars use the idea of ​​Fourier integral method to perform Toeplitz averaging on the data sampling covariance matrix, that is, to perform arithmetic averaging on all elements parallel to the main diagonal direction, and to reorganize the averaged results into new data in order, which is called virtual array output. The scanning vector is constructed with the number of array elements of the virtual array, and the virtual array output is weighted by triangular window to obtain the conventional beamforming result, but the resolution of this method is often poor; the virtual array output is uniformly weighted to obtain the Fourier integral method azimuth estimation result, which has the effect of high resolution and noise suppression, but this high resolution result is often accompanied by the problem of high side lobes.

[0004] The existing technologies all use a single weighted window function. Although different weighting methods can accurately estimate the target position, the height difference of the side lobes causes insufficient detection performance for weak targets, and the position of weak targets cannot be accurately estimated. At the same time, under the weighting of a single window function, the main lobe is wider and the display effect is poor. Summary of the invention

[0005] In order to overcome the problem of insufficient performance in weak target detection under a single window function weighting, the present invention proposes a fusion detection method based on the weighted Fourier integral method. The present invention designs a series of weighted window functions, utilizes the differences in the sidelobe positions of the results of different weighted window functions, and selects the minimum sidelobe value among all the results to achieve high resolution and noise suppression while maintaining low sidelobes, thereby bringing better weak target detection effects; at the same time, the beam pattern is shifted left and right to improve the main lobe of the beam and further improve the display effect.

[0006] To achieve the above purpose, the technical solution provided by the present invention is:

[0007] A fusion detection method based on weighted Fourier integral method comprises the following steps:

[0008] Step 1: Perform Fourier transform on the base array time domain signal to frequency domain broadband signal, and divide it into I sub-bands in the frequency domain; construct the sampling covariance matrix of the sub-band data, perform Toeplitz averaging on the sampling covariance matrix, and obtain the virtual array output; use different weighted Fourier integral methods to estimate the target direction;

[0009] Step 2: The virtual array output obtained in step 1 is processed by multi-weighted Fourier integral method to obtain a beam pattern;

[0010] Step 3: Determine the beam pattern obtained in step 2, adjust the scanning vector, and shift the beam pattern, and the offset angle position of the beam pattern after the shift corresponds to the position where the beam pattern power is maximum when it is not deflected;

[0011] Step 4: Introduce the amplitude control coefficient, adjust the value of the offset angle after the offset to the maximum power of the beam pattern when it is not deflected, and obtain the scan vector required to control the offset of the beam pattern by adjusting the scan vector pointed by the beam and the coefficient of the control amplitude;

[0012] Step 5: Introduce the left and right offset factors and the left and right offset magnification factors of the scanning vector to obtain the beam pattern after beam offset and amplitude adjustment;

[0013] Step 6: Using the scanning vector of the beam pattern offset obtained in step 5, the sub-band data is processed to obtain the target detection results of left and right offsets under different Fourier integral methods;

[0014] Step 7: From the n (n ≥ 2) target detection results without offset and left and right offset under the Fourier integral method with different windowing modes, select the minimum value at each direction, which is the optimized target detection result.

[0015] Furthermore, the step 1 includes the following sub-steps:

[0016] Step 1.1: The sampling covariance matrix is:

[0017]

[0018] in,[·] H represents the conjugate transpose, R k,l represents the covariance estimate of the data of the kth array element and the data of the lth array element, k, l = 1, 2, ..., M, f i is the subband frequency, B is the subband bandwidth, and each subband satisfies the narrowband condition f i ≥10B; S(f i ) is the frequency domain expression of each sub-band signal, which is an M×L dimensional matrix, where M is the number of array elements and L is the number of data frequency points of each sub-band;

[0019] Step 1.2: The virtual array output is:

[0020]

[0021] Where q represents the number of virtual array elements, p = 1-M, 2-M, ..., -1, 0, 1, ..., M-1, C(q)

[0022] represents the virtual array element output, [·] T Represents transpose:

[0023] C(q)=[C(1-M),C(2-M),…C(M-1)] T (2)

[0024] Step 1.3: Introduce subband frequency f i , construct the scan vector a(f i ,θ):

[0025] a(f i ,θ)=exp[-j2πf i (1-M,2-M,…,M-1) T dsinθ / c0] (1)

[0026] Where, d is the array element spacing, c0 is the speed of sound in water;

[0027] Step 1.4: Use the weighted Fourier integral method to obtain the target estimation result:

[0028] p(f i ,θ)=[a(f i ,θ)⊙w] H C (2)

[0029] Where w represents the value of the weighted window function, p = [p(f1,θ),p(f2,θ),…,p(f L ,θ)].

[0030] Furthermore, in step 2, the beam pattern is:

[0031] B1(f i ,θ)=[a(f i ,θ)⊙w] H CX

[0032] Where B1(f i ,θ) is the frequency f i The beam pattern at the time of sub-band frequency is f i , CX is the virtual array output.

[0033] Furthermore, in step 3, the left and right offset angles of the beam pattern are defined as BW and BW. - and BW +, and |BW - |=|BW + |,|·| represent absolute values; B1(f i ,θ) corresponds to the angle θ ML ,θ ML Meet | BW - |+|BW + |=2θ ML .

[0034] Furthermore, in step 4, the scanning vector for controlling the left and right offset of the beam pattern is constructed as

[0035]

[0036] Among them, the angular interval between the two offset angles is recorded as BW, and the left and right offset angles are less than the half-power beam width; M is the number of array elements, and c0 is the speed of sound in water.

[0037] Furthermore, in step 5, the scanning vector is offset left and right by a factor S ± for:

[0038]

[0039] Among them, S + Indicates the beam pattern left deviation factor, S - Indicates the beam pattern right deviation factor;

[0040] Left and right offset magnification A ± for:

[0041] A ± =1 / B2(f i ,BW ± ) (5)

[0042] Among them, A + Indicates the multiple by which the beam pattern needs to be amplified when it is offset to the left. A - Indicates the multiple by which the beam pattern needs to be magnified to shift right;

[0043] Then the beam pattern after beam offset and amplitude adjustment can be expressed as:

[0044]

[0045] in, Indicates the left offset beam pattern, beam pattern Represents a right-shifted beam pattern.

[0046] Furthermore, in step 6, the left and right offset magnification A is combined ± and left and right offset factor S ± , the result of the left and right deviation of the azimuth estimation is expressed as:

[0047]

[0048] Take P ± (f i ), and delete all values ​​less than 0, which is the result of the azimuth estimation after left and right offsets.

[0049] Furthermore, in step 7, the known weighted window function expression and the left and right offset magnification A are used. ± and left and right offset factor S ± Substituting into equation (7), we can obtain the left and right offset target estimation results of different weighted windows, the number of which is n. The minimum value of these n results at each direction is the final fusion result.

[0050] Furthermore, the known weighted window functions are weighted window functions of double triangular window, double cosine window and double Kaiser window, and the expressions are respectively:

[0051]

[0052]

[0053]

[0054] Among them, w Tri , w Cos , w Kaiser They represent the weighted values ​​of the double triangular window, double cosine window and double Kaiser window respectively, β represents the parameter controlling the sidelobe attenuation of the Kaiser window, is the zero-order Bessel function of the first kind.

[0055] Beneficial Effects

[0056] The advantage of the present invention is that the present invention selects the minimum amplitude at the same azimuth from the beam diagram of the weighted Fourier integral method under different weighted window functions and the beam diagram after the main lobe pointing is shifted left and right by a certain angle (within 1 / 2 half-power beam width), so as to obtain the optimal beam diagram.

[0057] This method not only improves the resolution and noise suppression capabilities, but also optimizes the beam amplitude under different weighting values ​​to keep the sidelobes at a low level; in addition, the left and right offset of the beam pattern further reduces the width of the beam main lobe, thereby obtaining better angular resolution performance.

[0058] The present invention focuses on using a double triangular window, a double cosine window and a double Kaiser window multi-group windowing method to shift the beam pattern of the weighted Fourier integral method left and right and select a small value for optimization. Since the beam patterns of the three weighting methods have differences in sidelobe levels, the present invention can reduce the sidelobe level of the beam pattern by selecting a small value for optimization, reduce the background level of the target estimation result, improve the detection performance of weak targets, accurately estimate the position of weak targets, and overcome the problem that the weak target is difficult to detect under a high background caused by a high sidelobe level of the beam pattern when a single window function is weighted; the present invention uses a method of shifting the beam pattern left and right to obtain a narrower beam main lobe, overcomes the problem that the main lobe of the beam pattern is wide under a single window function, and improves the display effect of the target in the azimuth estimation result.

[0059] The basic principle and implementation scheme of the present invention have been verified by computer numerical simulation. Fig.15 and Fig.19 The comparison results show that compared with the conventional beamforming method, the method proposed in the present invention makes full use of the characteristics of background reduction brought by frequency selection and main lobe narrowing brought by beam left and right deviation, so that Fig.19 The result has lower background and narrower target track, which has better detection effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 Overall implementation flow chart

[0061] Figure 2 Step 1) Implement the flow chart

[0062] Figure 3 Specific implementation method Step 2) Implementation flow chart

[0063] Figure 4 Diagram of three weighted window functions: double triangular window, double cosine window and double Kaiser window

[0064] Figure 5 Beam patterns of three weighting methods: double triangular window, double cosine window and double Kaiser window

[0065] Figure 6 Double triangular window weighted Fourier integral method beam pattern and left and right offset 1 / 2 half power point beam width beam pattern

[0066] Figure 7 Double cosine window weighted Fourier integral method beam pattern and left and right offset 1 / 2 half power point beam width beam pattern

[0067] Figure 8 Double Kaiser window weighted Fourier integral method beam pattern and left and right offset 1 / 2 half power point beam width beam pattern

[0068] Fig. 9Specific implementation method Step 3) Implementation flow chart

[0069] Fig.10 Comparison of the azimuth estimation results of the double triangular window, double cosine window and double Kaiser window weighted sum optimization method for a single target with an internal signal-to-noise ratio of -20dB

[0070] Fig.11 Comparison of the azimuth estimation results of the double triangular window, double cosine window and double Kaiser window weighted sum optimization methods for a single target with an internal signal-to-noise ratio of -30dB

[0071] Fig.12 Comparison of the azimuth estimation results of the double triangular window, double cosine window and double Kaiser window weighted and fusion methods for weak targets under interference

[0072] Fig.13 Comparison of the azimuth estimation results of the weighted and optimized methods of dual-target dual-triangular window, dual-cosine window and dual-Kaiser window

[0073] Fig.14 Comparison of the azimuth estimation results of multi-target double triangular window, double cosine window and double Kaiser window weighted and fusion methods

[0074] Fig.15 Target time and position history diagram of conventional beamforming method

[0075] Fig.16 Double Kaiser window weighted target time-direction history diagram

[0076] Fig.17 Double triangle window weighted target time and direction history diagram

[0077] Fig.18 Double cosine window weighted target time-direction history diagram

[0078] Fig.19 Optimization method target time direction process diagram DETAILED DESCRIPTION

[0079] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0080] The main contents of the present invention are:

[0081] (1) In the processing frequency band of passive sonar broadband energy detection [f l ,f h ] range, f l is the low frequency of the frequency band, f h For high frequency

[0082] Frequency, using the narrowband division principle i ≥10B(f iis the sub-band frequency, B is the sub-band bandwidth), the processing frequency band [f l ,f h ] is divided into multiple sub-bands, and the scanning vector a(θ) and sampling covariance matrix R corresponding to each sub-band are constructed, where a(θ) is an M×1-dimensional column vector, M represents the number of array elements, and θ is the scanning angle; R is an M×M-dimensional square matrix; the sampling covariance matrix is ​​Toeplitz averaged to obtain the virtual array output C, which is a (2M-1)×1-dimensional column vector. The virtual array output C of each sub-band is processed by the weighted Fourier integral method of double triangular window, double cosine window and double Kaiser window respectively to obtain the azimuth estimation results of the three.

[0083] (2) Using the processing frequency band range and each sub-band frequency in 1, a noise-free array signal is designed according to the sub-band frequency, and the weighted Fourier integral method of double triangular window, double cosine window and double Caesar window is used for processing to obtain the three beam patterns respectively, and the main lobe of the beam pattern is shifted left and right by a certain angle, and the output value at the target azimuth in the shifted beam pattern is set to 1; the scanning vector after left and right shift can be obtained, and then the vector is used to process the sub-band data in 1 to obtain the left and right shifted data processing results, and the minimum value corresponding to each azimuth in the 9 results of the three conventional processing results and the left and right deflection processing results is selected, that is, the small selection and optimal processing is performed. A high-resolution and low-background processing result can be obtained.

[0084] (3) Computer numerical simulation is used to give the results of the three methods alone and the target detection results of the method of the present invention, thereby proving that the method proposed in the present invention can achieve underwater target detection performance that is better than that of a single method under the conditions of strong surface interference and strong background noise.

[0085] The technical solution adopted by the present invention to solve the existing problems can be divided into the following three steps:

[0086] 1) Perform Fourier transform on the time domain broadband signal to the frequency domain, divide it into I sub-bands in the frequency domain, obtain the sampling covariance matrix of the received data of each sub-band, perform Toeplitz averaging on the sampling covariance matrix, and use the weighted Fourier integral method of double triangular window, double cosine window and double Kaiser window to estimate the target direction. The target detection results of the three weighted methods have differences in sidelobe height.

[0087] 2) Use the subband frequency in 1 to construct the scanning vector corresponding to each subband, use the subband frequency to construct a noise-free narrowband array receiving signal, and use the weighted Fourier integral method of double triangular window, double cosine window and double Kaiser window to process the receiving signal to obtain the beam patterns of three methods. Determine the half-power beam width of the beam patterns of the three methods respectively, adjust the scanning vector, add (subtract) the offset angle to the scanning angle in the scanning vector, and offset the beam pattern to the left (right). At this time, the offset angle position of the offset beam pattern corresponds to the position with the maximum power of the beam pattern when it is not deflected; by multiplying by an amplitude control coefficient, the value of the offset angle position after the offset is adjusted to the maximum power. By adjusting the scanning vector pointing to the beam and adjusting the coefficient of the control amplitude, we can obtain the scanning vector required to control the offset of the beam pattern.

[0088] 3) Using the offset vector to process the sub-band data in 1) can obtain the target detection results of the three methods with left and right offsets. From the 9 target detection results of the three weighted methods without offset and with left and right offsets, the minimum value at each direction is selected, which is the target detection result after fusion.

[0089] The overall implementation process is as follows Figure 1 shown.

[0090] Each step of the present invention is described in detail below:

[0091] The specific contents involved in step 1) are as follows:

[0092] Perform Fourier transform on the time domain broadband signal received by the linear array to obtain the frequency domain broadband signal. In the frequency domain, the broadband signal is divided into I sub-bands, each of which satisfies the narrowband condition (f i ≥10B)(f i is the subband frequency, B is the subband bandwidth), the frequency domain expression of each subband signal is S(f i ), is an M×L dimensional matrix, where M is the number of array elements and L is the number of data frequency points of each subband, that is, the data length. Then the covariance matrix can be obtained by formula (11).

[0093]

[0094] in,[·] H represents the conjugate transpose, R k,l Represents the covariance estimate of the data of the kth array element and the data of the lth array element, k,l=1,2,…,M.

[0095] The virtual array output C is obtained by performing Toeplitz averaging on R obtained from formula (11).

[0096]

[0097] Wherein, q represents the number of virtual array elements, and p = 1-M, 2-M,…,-1, 0, 1,…, M-1.

[0098] C(q)=[C(1-M),C(2-M),…C(M-1)] T (13)

[0099] represents the virtual array element output, [·] T Indicates transpose.

[0100] Then, using the subband frequency f i Construct the scan vector a(f i ,θ).

[0101] a(f i ,θ)=exp[-j2πf i (1-M,2-M,…,M-1) T dsinθ / c0] (14)

[0102] Where d is the array element spacing and c0 is the speed of sound in water.

[0103] Then the weighted Fourier integral method can be used to obtain the target estimation result.

[0104] p(f i ,θ)=[a(f i ,θ)⊙w] H C (15)

[0105] Where w represents the value of the weighted window function, p = [p(f1,θ),p(f2,θ),…,p(f L ,θ)], represents the target estimation result.

[0106] Implementation process such as Figure 2 shown.

[0107] The specific contents involved in step 2) are as follows:

[0108] Using the subband frequency f in step 1) i , construct a noise-free narrowband array signal X i , X i For an M×L dimensional matrix, construct X i The sampling covariance matrix RX and the virtual array output CX of step 1) are transformed into X using equations (1) to (2) i After processing, beam patterns under three window functions are obtained respectively.

[0109] B1(f i ,θ)=[a(f i ,θ)⊙w] H ·CX (16)

[0110] B1(f i ,θ) is the frequency f i The beam pattern at the time of - , B.W. + , and |BW - |=|BW + |, |·| represent absolute values. The angle interval between the two offset angles is recorded as BW, and the left and right offset angles are less than the half-power beam width. B1(f i ,θ) corresponds to the angle θ ML ,θ ML Meet | BW - |+|BW + |=2θ ML .

[0111] The scanning vector that controls the left and right offset of the beam pattern can be constructed as

[0112]

[0113] Dividing the offset scan vector by the unoffset scan vector can obtain a factor that controls the vector offset, denoted by S.

[0114]

[0115] Among them, S + Indicates the beam pattern left deviation factor, S - Indicates the beam pattern right shift factor.

[0116] Then the shifted beam pattern can be obtained by equation (19).

[0117]

[0118] in, is the frequency f i The beam pattern is shifted left and right when

[0119] Then order Medium θ ML The power at the position is related to B(f i ,θ) ML The power at the position is equal, both are 1, let B(f i ,θ) ML The power at the location is divided by Medium θ ML The power at the position can be used to obtain the multiple by which the beam pattern needs to be amplified, denoted by A.

[0120] A ± =1 / B2(f i ,BW± ) (20)

[0121] Among them, A + Indicates the multiple by which the beam pattern needs to be amplified when it is offset to the left. A - Indicates the multiple that needs to be amplified to shift the beam pattern to the right.

[0122] Then the beam pattern after beam shift and amplitude adjustment can be expressed as formula (6).

[0123]

[0124] in, Indicates the left offset beam pattern, beam pattern Represents a right-shifted beam pattern.

[0125] Then, using equations (16) to (6), the original beam pattern and the beam pattern after left and right shifting of the weighted Fourier integral method of the double triangular window, double cosine window and double Kaiser window are obtained respectively. The weighted window functions of the double triangular window, double cosine window and double Kaiser window are expressed as equations (8) to (10).

[0126]

[0127]

[0128]

[0129] Among them, w Tri , w Cos , w Kaiser They represent the weighted values ​​of the double triangular window, double cosine window and double Kaiser window respectively, β represents the parameter controlling the sidelobe attenuation of the Kaiser window, is the zero-order Bessel function of the first kind.

[0130] Step 2) Implement the process as follows Figure 3 shown.

[0131] The weighted values ​​of the Fourier integral method of double triangular window, double cosine window and double Kaiser window weight are as follows Figure 4 As shown, the comparison of the three original beam patterns is as follows Figure 5 As shown, the beam patterns of the three methods with a left and right offset of 1 / 2 half-power beam width are respectively as follows Figure 6-Figure 8 shown.

[0132] The specific contents involved in step 3) are as follows:

[0133] The magnification factor A of the left and right offset calculated in step 2) ± and left and right offset factor S ±It is used to estimate the target direction of the actual data. The result of the left and right deviation of the direction estimation can be expressed as

[0134] P ± (f i )=[A ± ·w⊙S ± ⊙a(f i )] H ·C (25)

[0135] Take P ± (f i ), and delete all values ​​less than 0, which is the result of the azimuth estimation after left and right offsets.

[0136] The window function of equation (8)-equation (10) and the magnification factor A of the left and right offset obtained in step 2 are ± and left and right offset factor S ± Substituting into equation (7), we can obtain the left and right offset target estimation results of double triangular window, double cosine window and double Kaiser window respectively.

[0137] At this point, we will obtain the original estimation results and left and right offset estimation results of the double triangular window, double cosine window and double Kaiser window, a total of 9 target estimation results, and then take the minimum value of these 9 results at each direction as the final fusion result.

[0138] The implementation process of step 3) is as follows Fig. 9 shown.

[0139] Taking a 128-element uniform linear array as an example, an example implementation of the present invention is given. The implementation example uses computer simulation to perform numerical simulation to verify the effect of the method proposed by the present invention.

[0140] Basic simulation parameters:

[0141] Assume that the real speed of sound waves propagating underwater is 1500 m / s. The receiving array is a 64-element uniform linear array, the base array design frequency is 100 Hz, the array element spacing is half a wavelength, the sampling frequency is 6000 Hz, the sampling time is 20 s, the beam scanning interval is 0.1°, and the frequency band of the simulated broadband signal used is 50 Hz-100 Hz, which is divided into ten sub-bands with a bandwidth of 5 Hz.

[0142] Simulation signal parameters:

[0143] Simulation 1): The simulation signal used is a 50Hz-100Hz broadband signal, the target azimuth is located at 0°, the signal strength is 0dB, and the in-band noise strength is 20dB and 30dB. The double triangle window, double cosine window and double Kaiser window weighted Fourier integral method are used to estimate the target azimuth.

[0144] Simulation 2) Simulate weak target detection under strong interference. The simulation signal used is a 50Hz-100Hz broadband target, the target angle is at 4°, the signal strength is 0dB, the interference used is a 50Hz-100Hz broadband interference, the interference angle is at 0°, the interference strength is 10dB, and the in-band noise strength is 30dB. The double triangular window, double cosine window and double Kaiser window weighted Fourier integral method are used to estimate the target direction.

[0145] Simulation 3) Simulate the detection of dual targets with close angles. The simulation signals used are two 50Hz-100Hz broadband targets. The target angles are located at 0° and 2.5°. The signal strength is 0dB and the in-band noise strength is 30dB. The double triangular window, double cosine window and double Kaiser window weighted Fourier integral method are used to estimate the target direction.

[0146] Simulation 4) Simulate multi-target detection, the simulation signals used are multiple 50Hz-100Hz broadband targets, the target angles are located at -26°, -5° and 0°, the signal strengths are 3dB, 0dB and 0dB, the interference angles are located at -78°, -20° and 20°, the interference strengths are 20dB, 12dB and 15dB, and the in-band noise strength is 35dB. The double triangular window, double cosine window and double Kaiser window weighted Fourier integral method are used for target azimuth estimation.

[0147] according to Figure 1 Process and Figure 4 The beam pattern of the method of the present invention is obtained by simulating the weighted method of Figure 5-Figure 8 It can be seen that the present invention can obtain a beam pattern with a narrower main lobe by using the beam pattern left and right offset technology, thereby improving the target detection effect.

[0148] according to Figure 1 The process and Figure 4 The broadband signal in simulation 1) is processed by using a weighted method, and the results of double triangular window, double cosine window, double Kaiser window and optimization are plotted on Figure 10-11 In. By Fig.10 It can be seen that the proposed blending method accurately estimates the angle information of the target, and the blending method adopted by the present invention has the effect of the narrowest main lobe and the lowest background; Fig.11 Simulation results show that the proposed method maintains good target detection performance under low signal-to-noise ratio and has good robustness.

[0149] according to Figure 1 The process and Figure 4 The broadband signal in simulation 2) is processed by weighting method, and the double triangle window, double cosine window and double Kaiser window as well as the optimized results are plotted on Fig.12 In, by Fig.12 It can be seen that the method used in the present invention not only retains the weak target detection capability under the weighting of the three window functions, but also further reduces its main lobe and improves the display effect. The simulation example shows that the method proposed in the present invention can effectively detect weak targets through fusion optimization processing.

[0150] according to Figure 1 The process and Figure 4 The broadband signal in simulation 3) is processed by weighting method, and the double triangle window, double cosine window and double Kaiser window as well as the optimized results are plotted on Fig.13 In, by Fig.13 It can be seen that the method proposed in the present invention not only retains the target resolution ability of the weighted Fourier integral method, but also has the low noise background brought by the optimization processing and the narrow main lobe characteristics brought by the left and right offset. The simulation example shows that the method proposed in the present invention can effectively detect dual targets with close orientations through the optimization processing.

[0151] according to Figure 1 The process and Figure 4 The broadband signal in simulation 4) is processed by weighting method, and the double triangle window, double cosine window and double Kaiser window as well as the optimized results are plotted on Fig.14 In, by Fig.14 It can be seen that under low signal-to-noise ratio, this method retains the multi-target resolution performance and noise suppression characteristics of the Fourier integral method, while the left and right offsets and the small optimal selection make the main lobe narrower and the background lower. The simulation example shows that the method proposed in the present invention can effectively detect multiple targets through optimal processing.

[0152] In order to further compare the method proposed in the present invention, a signal of 40 blocks is processed, each block has a time length of 5s, and the signal form is the broadband signal in simulation 4). Conventional beamforming technology and the three window function weighting methods and the fusion optimization method mentioned in the present invention are used for processing, and the results are displayed in the form of a time azimuth history diagram for comparison. The results are as follows: Figure 15-19 As shown. Fig.15 The time-azimuth history diagram of conventional beamforming shows that the conventional method can no longer observe the target's azimuth due to the low signal-to-noise ratio of the target; Figure 16-18 The time-direction history diagrams under the three weighting methods are compared Fig.15 The background is reduced, and targets at -26°, -5°, and 0° can be observed, but the trajectory observation at -5° and 0° is difficult due to the background noise; Fig.19 The optimal method of Zhongfu can significantly reduce the noise background, make the target estimation clear at -5° and 0°, and make the target track narrower and the display effect better due to the left and right offset.

[0153] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should be included in the protection scope of the present invention.

Claims

1. A fusion detection method based on weighted Fourier integral method, characterized in that: The following steps are involved: Step 1: Perform Fourier transform on the base array time domain signal to frequency domain broadband signal, and divide it into I sub-bands in the frequency domain; construct the sampling covariance matrix of the sub-band data, perform Toeplitz averaging on the sampling covariance matrix, and obtain the virtual array output; use different weighted Fourier integral methods to estimate the target direction; Step 2: The virtual array output obtained in step 1 is processed by multi-weighted Fourier integral method to obtain a beam pattern; Step 3: Determine the beam pattern obtained in step 2, adjust the scanning vector, and shift the beam pattern, and the offset angle position of the beam pattern after the shift corresponds to the position where the beam pattern power is maximum when it is not deflected; Step 4: Introduce the amplitude control coefficient, adjust the value of the offset angle after the offset to the maximum power of the beam pattern when it is not deflected, and obtain the scan vector required to control the offset of the beam pattern by adjusting the scan vector pointed by the beam and the coefficient of the control amplitude; Step 5: Introduce the left and right offset factors and the left and right offset magnification factors of the scanning vector to obtain the beam pattern after beam offset and amplitude adjustment; Step 6: Using the scanning vector of the beam pattern offset obtained in step 5, the sub-band data is processed to obtain the target detection results of left and right offsets under different Fourier integral methods; Step 7: From the n (n ≥ 2) target detection results without offset and left and right offset under the Fourier integral method with different windowing modes, select the minimum value at each direction, which is the optimized target detection result.

2. The optimal blending detection method based on the weighted Fourier integral method according to claim 1, characterized in that: The step 1 includes the following sub-steps: Step 1.1: The sampling covariance matrix is: in,[·] H represents the conjugate transpose, R k,l represents the covariance estimate of the data of the kth array element and the data of the lth array element, k, l = 1, 2, ..., M, f i is the subband frequency, B is the subband bandwidth, and each subband satisfies the narrowband condition f i ≥10B; S(f i ) is the frequency domain expression of each sub-band signal, which is an M×L dimensional matrix, where M is the number of array elements and L is the number of data frequency points of each sub-band; Step 1.2: The virtual array output is: Where q represents the number of virtual array elements, p = 1-M, 2-M, ..., -1, 0, 1, ..., M-1, C(q) represents the virtual array element output, [·] T Represents transpose: Step 1.3: Introduce subband frequency f i , construct the scan vector a(f i ,θ): a(f i ,θ)=exp[-j2πf i (1-M,2-M,…,M-1) T dsinθ / c0] (3) Where, d is the array element spacing, c0 is the speed of sound in water; Step 1.4: Use the weighted Fourier integral method to obtain the target estimation result: p(f i ,θ)=[a(f i ,θ)⊙w] H C (4) Where w represents the value of the weighted window function, p = [p(f1,θ),p(f2,θ),…,p(f L ,θ)].

3. The optimal fusion detection method based on weighted Fourier integral method according to claim 2, characterized in that: In step 2, the beam pattern is: B1(f i ,θ)=[a(f i ,θ)⊙w] H ·CX Where B1(f i ,θ) is the frequency f i The beam pattern at the time of sub-band frequency is f i , CX is the virtual array output.

4. The optimal fusion detection method based on weighted Fourier integral method according to claim 3, characterized in that: In step 3, the left and right offset angles of the beam pattern are defined as BW and - and BW + , and |BW - |=|BW + |, |·| represent absolute values; B1(f i ,θ) corresponds to the angle θ ML ,θ ML Meet | BW - |+|BW + |=2θ ML .

5. The optimal fusion detection method based on weighted Fourier integral method according to claim 4, characterized in that: In step 4, the scanning vector for controlling the left and right offset of the beam pattern is constructed as follows: Among them, the angular interval between the two offset angles is recorded as BW, and the left and right offset angles are less than the half-power beam width; M is the number of array elements, and c0 is the speed of sound in water.

6. The optimal fusion detection method based on weighted Fourier integral method according to claim 5, characterized in that: In step 5, the scanning vector is offset left and right by a factor S ± for: Among them, S + Indicates the beam pattern left deviation factor, S - Indicates the beam pattern right deviation factor; Left and right offset magnification A ± for: A ± =1 / B2(f i ,BW ± ) (2) Among them, A + Indicates the multiple by which the beam pattern needs to be amplified when it is offset to the left. A - Indicates the multiple by which the beam pattern needs to be magnified to shift right; Then the beam pattern after beam offset and amplitude adjustment can be expressed as: in, Indicates the left offset beam pattern, beam pattern Represents a right-shifted beam pattern.

7. The optimal fusion detection method based on weighted Fourier integral method according to claim 6, characterized in that: In step 6, the left and right offset magnification A is combined ± and left and right offset factor S ± , the result of the left and right deviation of the azimuth estimation is expressed as: P ± (f i ,θ)=[A ± w⊙S ± ⊙a(f i ,i)] H C (4) Take P ± (f i ,θ), and delete all values ​​less than 0, which is the azimuth estimation result after left and right offset.

8. The optimal fusion detection method based on weighted Fourier integral method according to claim 7, characterized in that: In step 7, the known weighted window function expression and the left and right offset magnification A are used. ± and left and right offset factor S ± Substituting into equation (4), we can obtain the left and right offset target estimation results of different weighted windows, the number of which is n. The minimum value of these n results at each direction is the final fusion result.

9. The optimal fusion detection method based on weighted Fourier integral method according to claim 8, characterized in that: The known weighted window functions are weighted window functions of double triangular window, double cosine window and double Kaiser window, and the expressions are respectively: Among them, w Tri , w Cos , w Kaiser They represent the weighted values ​​of the double triangular window, double cosine window and double Kaiser window respectively, β represents the parameter controlling the sidelobe attenuation of the Kaiser window, is the zero-order Bessel function of the first kind.

Images