Distributed MIMO sonar single-frame and multi-frame integrated reverberation suppression and target enhancement method

Abstract

In order to solve the problem that the detection performance of an active sonar is seriously reduced due to channel fluctuation, target intensity fluctuation and strong reverberation backgrounds, the invention provides a single-frame and multi-frame integrated reverberation suppression and target enhancement method for a distributed MIMO sonar, which comprises the following steps: firstly, carrying out space-time processing on all receiving ends of the distributed MIMO sonar to obtain transmitting-receiving channel output; carrying out incoherent delay summation on all transmitting-receiving channel outputs to obtain single-frame detection output; then, single-frame detection output serves as input of multi-frame processing, and a multi-frame Gaussian mixture model is used for restraining reverberation background and enhancing moving target bright spots; and finally, processing the output of the multi-frame Gaussian mixture model by using the multi-frame high-order nonlinear cumulant. Compared with a traditional method, the method has the advantages that under the conditions of channel fluctuation, target intensity fluctuation and strong reverberation background, the strong reverberation background can be effectively inhibited, and more stable and higher-quality target bright spot detection output is obtained.

Description

Technical Field

[0001] This invention belongs to the field of electronic information, and specifically relates to a method for integrated reverberation suppression and target enhancement in single-frame and multi-frame distributed MIMO sonar. Background Technology

[0002] Monostatic active sonar is a common sonar detection system that requires only one platform to carry both transmitting and receiving equipment, reducing system complexity and cost, and facilitating deployment and maintenance (Urick, and Robert J. Principles of underwater sound / -3rd ed. McGraw-Hill Book Company, 1983.). However, traditional monostatic active sonar faces drawbacks in complex and variable fluctuating channel environments, such as difficulty in suppressing reverberation background fluctuations and unstable target intensity, making it difficult to effectively detect targets.

[0003] Unlike monostatic sonar detection methods, distributed MIMO sonar detection methods can utilize a large-spacing array to detect targets simultaneously from multiple perspectives, thereby obtaining spatial diversity gain and effectively suppressing channel fluctuations (Jiang Jingning. Research on Distributed MIMO Sonar Target Detection and Imaging Methods [D]. Zhejiang: Zhejiang University, 2020.). Therefore, compared to traditional monostatic sonar, distributed MIMO sonar is more likely to obtain robust single-frame detection output. However, in real-world environments, there are simultaneous channel fluctuations, target intensity fluctuations, and strong reverberation backgrounds. Existing distributed MIMO sonar detection methods cannot effectively suppress the background or enhance the target in these conditions, resulting in a severe degradation in detection performance and failing to meet application requirements.

[0004] Therefore, there is an urgent need to propose a new method to solve the problems that cause the detection performance to degrade and make it difficult to meet the needs of use. Summary of the Invention

[0005] To address the severe performance degradation of active sonar due to channel fluctuations, target intensity fluctuations, and strong reverberation backgrounds, this invention proposes a single-frame and multi-frame integrated reverberation suppression and target enhancement method for distributed MIMO sonar. First, space-time processing is performed on all receivers of the distributed MIMO sonar to obtain the transmit-receive channel output. Then, incoherent delay summation is applied to all transmit-receive channel outputs to obtain a single-frame detection output, overcoming channel fluctuations and target intensity fluctuations and obtaining a stable single-frame target highlight. Next, the single-frame detection output is used as input for multi-frame processing, employing a multi-frame Gaussian mixture model to suppress the reverberation background and enhance the moving target highlight. Finally, multi-frame high-order nonlinear cumulants are used to process the multi-frame Gaussian mixture model output, further suppressing the reverberation background and enhancing the continuous moving target highlight, ultimately achieving a stable and high-quality target highlight detection output despite channel fluctuations, target intensity fluctuations, and strong reverberation backgrounds.

[0006] The technical solution of this invention is as follows:

[0007] A method for integrated reverberation suppression and target enhancement in single-frame and multi-frame distributed MIMO sonar is characterized by the following specific steps:

[0008] Step 1: Utilize the spatial diversity characteristics of the distributed channel and the intra-node waveform coherence and inter-node channel orthogonality characteristics of the target signal to perform space-time processing on all receiving nodes of the distributed MIMO sonar to obtain the detection output on the transmit-receive channel, and perform incoherent delay summation on the transmit-receive channel outputs between all nodes to obtain a single frame detection output.

[0009] Step 2: Utilize the statistical distribution consistency features of multi-frame reverberation and the outlier features of the target bright spots to model the multi-frame amplitude of each pixel in the single-frame detection output obtained in Step 1 as a Gaussian mixture model. Then, use the multi-frame Gaussian mixture model to suppress the reverberation background and retain the target bright spots with outlier features to obtain the multi-frame Gaussian mixture model output.

[0010] Step 3: Utilize the motion continuity features of the target highlights in multiple frames, and use multi-frame high-order nonlinear accumulation to process the output of the multi-frame Gaussian mixture model obtained in Step 2, and finally output the multi-frame high-order nonlinear accumulation quantity.

[0011] Furthermore, step 1 specifically includes:

[0012] Step 1.1: Arrangement Each launch node Distributed MIMO sonar at each receiving node forms a... There are 1 transmit / receive channel, with each transmitting node using a single transmit transducer and each receiving node using... A uniform linear array is used to calculate the echo received by the receiving node;

[0013] Step 1.2: [The text appears to be incomplete and contains several grammatical errors. A more accurate translation would require The echoes on each receiving array element are matched and filtered with the transmitted signal;

[0014] Step 1.3: Divide the detection area into a uniform grid structure, then calculate the angle of each grid point relative to the receiving node, and then calculate the beamforming vector corresponding to each grid point. Finally, perform conventional beamforming on the matched filtering results obtained in Step 1.2.

[0015] Step 1.4: Calculate the target detection results of a single transceiver channel of the distributed MIMO sonar based on the time delay and beam output value corresponding to each grid point;

[0016] Step 1.5: For The channel detection results are incoherently superimposed to obtain the distributed MIMO sonar detection results.

[0017] Furthermore, in step 1.1, the receiving transducer... The echo received by each array element is:

[0018]

[0019] In the formula,

[0020] The scattering intensity of the target; To transmit signals from the transducer; The propagation delay from the launching node to the target; The propagation delay from the target to the receiving node; This refers to a specific moment when the echo is received. for Time of the first The first receiving transducer The noise received by each array element.

[0021] Furthermore, in step 1.2, the matched filtering result is in the form of a temporal convolution:

[0022]

[0023] In the formula,

[0024] For the first The echo of each receiving array element; Let be the impulse response function of the matched filter corresponding to the transmitted signal; superscript Indicates taking the conjugate; The length of a single transmitted signal.

[0025] Furthermore, in step 1.3, the beam output result is as follows:

[0026]

[0027] In the formula,

[0028] Beam output for a single transmit / receive channel; The weighting vector used for beamforming is expressed as:

[0029]

[0030] in, The imaginary unit; The center frequency of the echo signal; To represent the time delay of the i-th array element relative to the first array element; The element spacing of a uniform linear array, The angle between the echo and the array normal direction. This is the speed of sound in water.

[0031] Furthermore, in step 1.5, the distributed MIMO sonar detection results are as follows:

[0032]

[0033] In the formula,

[0034] The target detection results for a single transmit / receive channel in step 1.4 are given, where, For the first The grid point relative to the first The first launch node and the first The latency of each receiving node is expressed as:

[0035]

[0036] in, It is the first The location of each grid point; For the first The location of each transmission node; It is the first The location of each receiving node; c is the speed of sound in water.

[0037] Furthermore, step 2 specifically includes:

[0038] Step 2.1: Establish a multi-frame Gaussian mixture model and initialize parameters such as mean, variance, and weights;

[0039] Step 2.2: After initialization, update the mean, variance, weight and other parameters of each Gaussian component, and sort each Gaussian component in descending order of its weight value;

[0040] Step 2.3: Match the multi-frame amplitude of each pixel with the existing multi-frame Gaussian mixture model, and compare the amplitude of each pixel in the current frame with the mean of each Gaussian component in the multi-frame Gaussian mixture model: if the judgment condition is met, the pixel is determined to be the target bright spot and is retained; otherwise, the pixel is determined to be the reverberant background and the process returns to step 2.2 to update the parameters of the multi-frame Gaussian mixture model.

[0041] Step 2.4: Combine the target highlights retained in Step 2.3 to form the final multi-frame Gaussian mixture model output.

[0042] Furthermore, the multi-frame Gaussian mixture model established in step 2.1 is as follows:

[0043]

[0044] In the formula,

[0045] This indicates the position of each pixel in the sonar image sequence output from a single frame. This indicates its corresponding amplitude; It represents the number of Gaussian components, with a range of values ​​of 1. ; It is the first The weights of each Gaussian component; Let be the probability density function of a Gaussian distribution, expressed as:

[0046]

[0047] in, and These represent the first and second frames in the multi-frame Gaussian mixture model, respectively. The mean and variance of each Gaussian component.

[0048] The parameter initialization in step 2.1 includes:

[0049] initial mean With initial variance :

[0050]

[0051]

[0052] The intensity of the pixels in the first frame image is used as the mean of a certain Gaussian component in the multi-frame Gaussian mixture model, and a large weight is assigned to it. Meanwhile, the mean of other Gaussian components is set to 0, and a large variance and a small weight are assigned.

[0053] The basis for updating the parameters in step 2.2 is:

[0054]

[0055]

[0056]

[0057] In the formula,

[0058] The weight learning rate, and ; The learning rate is the sum of the mean and variance, and ; , and They represent the first The weights, mean, and variance of each Gaussian component.

[0059] Furthermore, the determination condition in step 2.3 is: the difference between the amplitude of a pixel in the current frame and the mean of each Gaussian component exceeds a set threshold, i.e. ,in, Take 2.5.

[0060] Furthermore, in step 3, the multi-frame high-order nonlinear cumulative quantity is:

[0061]

[0062] In the formula,

[0063] The number of frames for multi-frame high-order nonlinear cumulative quantities; For the first The pixel value of the frame; The order of the multi-frame high-order nonlinear cumulant.

[0064] The beneficial effects of this invention are as follows:

[0065] This invention proposes an integrated reverberation suppression and target enhancement method for single-frame and multi-frame distributed MIMO sonar. In single-frame processing, space-time processing is performed on all receiving nodes of the distributed MIMO sonar to obtain the detection output on the transmit-receive channel. Incoherent delay summation is then applied to the transmit-receive channel detection outputs between all nodes to initially suppress channel fluctuations, target intensity fluctuations, and reverberant background, enhancing the target bright spots in single-frame detection. Multiple consecutive single-frame detection outputs are then used as input for multi-frame processing. In multi-frame processing, a multi-frame Gaussian mixture model is used to suppress the reverberant background while retaining target bright spots with outlier characteristics, enhancing the target bright spots in multi-frame detection. Finally, the output of the multi-frame Gaussian mixture model is processed using multi-frame high-order nonlinear cumulants to further suppress channel fluctuations, target intensity fluctuations, and reverberant background, significantly enhancing the multi-frame continuous moving bright spots of the target. Ultimately, stable and high-quality target bright spot detection outputs are obtained even with channel fluctuations, target intensity fluctuations, and strong reverberant backgrounds.

[0066] The basic principles and implementation schemes of this invention have been verified by computer numerical simulation. The results show that, compared with traditional methods, under the conditions of channel fluctuations, target intensity fluctuations, and strong reverberation background, the proposed method can effectively suppress strong reverberation background and obtain more stable and higher quality target bright spot detection output. Attached Figure Description

[0067] The above or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0068] Figure 1 The main steps of the method proposed in this invention are as follows:

[0069] Figure 2 The process for processing distributed MIMO sonar echoes in step 1;

[0070] Figure 3 This is a schematic diagram of the grid division of the detection area;

[0071] Figure 4 This refers to the reverberation suppression process using a multi-frame Gaussian mixture model in step 2.

[0072] Figure 5 A schematic diagram of a multi-frame Gaussian mixture model for a reverberant background;

[0073] Figure 6 For a single Gaussian model and a Gaussian mixture model, there are probability density functions.

[0074] Figure 7 This is a reverberation suppression process using multiple frames of high-order nonlinear cumulative quantities in step 3;

[0075] Figure 8The diagrams show the node location distribution of the traditional method and the proposed method, where (a) is the node location distribution diagram of the traditional method and (b) is the node location distribution diagram of the proposed method.

[0076] Figure 9 The results are shown below, with (a) being the result of the traditional method and the proposed method after processing the background reverberation and the target enhancement, respectively. (b) is the result of the traditional method after processing the 10th frame, (c) is the result of the traditional method after processing the 10th frame, (d) is the result of the proposed method after processing the 20th frame, (e) is the result of the traditional method after processing the 30th frame, and (f) is the result of the proposed method after processing the 30th frame.

[0077] Figure 10 The results are obtained by accumulating 50 consecutive frames of data after applying the traditional method to suppress reverberation and enhance the target, respectively, and the proposed method. (a) shows the result of accumulating 50 consecutive frames of data after applying the traditional method to suppress reverberation and enhance the target, and (b) shows the result of accumulating 50 consecutive frames of data after applying the proposed method to suppress reverberation and enhance the target.

[0078] Figure 11 The integral sidelobe ratio deviations after reverberation suppression and target enhancement using the traditional method and the proposed method are respectively, where (a) is the integral sidelobe ratio deviation of target 1, (b) is the integral sidelobe ratio deviation of target 2, and (c) is the integral sidelobe ratio deviation of target 3. Detailed Implementation

[0079] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention 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, and therefore should not be construed as a limitation of the invention. The invention will be further described below with reference to the accompanying drawings.

[0080] See Figure 1 The main contents of this invention are as follows:

[0081] 1. Starting from the characteristics of the environment and the target, we explore the feature differences between them. Based on this, we design single-frame and multi-frame processing to jointly suppress channel fluctuations, target intensity fluctuations, and strong reverberation background, and enhance the target highlights. In single-frame processing, we utilize the spatial diversity characteristics of the distributed channel and the intra-node waveform coherence and inter-node channel orthogonality characteristics of the target signal. We adopt intra-node spatiotemporal processing and inter-node incoherent delay summation processing under the distributed MIMO sonar architecture to effectively suppress channel fluctuations, target intensity fluctuations, and reverberation background, and enhance the single-frame target highlights. In multi-frame processing, we utilize the statistical distribution consistency characteristics of multi-frame reverberation and the outlier and motion continuity characteristics of the target highlights. We adopt a multi-frame Gaussian mixture model and multi-frame high-order nonlinear cumulants to suppress strong reverberation background and target fluctuations, and enhance the multi-frame target highlights.

[0082] 2. Based on the single-frame and multi-frame processing methods adopted, the following process was designed: First, single-frame processing was used to perform space-time processing on all receiving nodes of the distributed MIMO sonar to obtain the detection output on the transmit-receive channel. Incoherent delay summation was then applied to the transmit-receive channel detection outputs between all nodes to initially suppress channel fluctuations, target intensity fluctuations, and reverberation background, thereby enhancing the target highlights in single-frame detection. Next, multiple consecutive single-frame detection outputs were used as input for multi-frame processing. A multi-frame Gaussian mixture model was used to suppress the reverberation background and retain target highlights with outlier characteristics, enhancing the target highlights in multi-frame detection. Then, the output of the multi-frame Gaussian mixture model was processed using multi-frame high-order nonlinear cumulants to further suppress channel fluctuations, target intensity fluctuations, and reverberation background, significantly enhancing the target's continuous motion highlights across multiple frames. Finally, stable and high-quality target highlight detection outputs were obtained amidst channel fluctuations, target intensity fluctuations, and strong reverberation background.

[0083] 3. Computer numerical simulations were used to present the results of traditional monostatic sonar reverberation suppression and target enhancement methods and the proposed distributed MIMO sonar single-frame and multi-frame integrated reverberation suppression and target enhancement methods. The results show that under the conditions of channel fluctuations, target intensity fluctuations, and strong reverberation background, compared with the traditional monostatic sonar reverberation suppression and target enhancement method (hereinafter referred to as the "traditional method" for simplicity), the proposed distributed MIMO sonar single-frame and multi-frame integrated reverberation suppression and target enhancement method (hereinafter referred to as the "proposed method") can obtain more stable and higher quality target bright spot detection output.

[0084] like Figure 1 As shown, the technical solution adopted by the present invention to solve the existing problems can be divided into the following three steps:

[0085] Step 1: Utilizing the spatial diversity characteristics of the distributed channel and the intra-node waveform coherence and inter-node channel orthogonality characteristics of the target signal, space-time processing is performed on all receiving nodes of the distributed MIMO sonar to obtain the probe output on the transmit-receive channel. Then, incoherent delay summation is performed on the transmit-receive channel outputs between all nodes to obtain a single-frame probe output. The main process is as follows: Figure 2 As shown, it specifically includes:

[0086] Step 1.1: Assuming distributed MIMO sonar is used Each launch node Each receiving node forms a There are 1 transmit / receive channel, with each transmitting node using a single transmit transducer and each receiving node using... A uniform linear array, assuming the signal of the transmitting transducer is... Then the receiving transducer The echo received by each array element is:

[0087] (1)

[0088] In the formula,

[0089] The scattering intensity of the target; The propagation delay from the launching node to the target; The propagation delay from the target to the receiving node; This refers to a specific moment when the echo is received. for Time of the first The first receiving transducer Noise received by each array element;

[0090] Step 1.2: [The text appears to be incomplete and contains several grammatical errors. A more accurate translation would require The echoes from each receiving array element are subjected to matched filtering with the transmitted signal. The result of the matched filtering is in the form of a time-domain convolution.

[0091] (2)

[0092] In the formula,

[0093] For the first The echo of each receiving array element; Let be the impulse response function of the matched filter corresponding to the transmitted signal; superscript Indicates taking the conjugate; The length of a single transmitted signal;

[0094] Step 1.3: Divide the detection area into, as follows Figure 3The uniform grid structure shown is used to calculate the angle of each grid point relative to the receiving node after grid division, and then calculate the beamforming vector corresponding to each grid point. Then, the matched filtering result obtained by equation (2) is subjected to conventional beamforming.

[0095] The weighting vector used in beamforming is:

[0096] (3)

[0097] In the formula,

[0098] The imaginary unit; The center frequency of the echo signal; This represents the time delay of the i-th array element relative to the first array element, and the time delay of the first array element relative to itself is 0; The element spacing of a uniform linear array; The angle between the echo and the array normal direction; The speed of sound in water; since the path difference between two adjacent array elements is According to the relationship that the path difference divided by the speed of sound equals the time delay, that is... , can be further expanded into equation (3);

[0099] The beam output results are as follows:

[0100] (4)

[0101] In the formula,

[0102] Beam output for a single transmit / receive channel;

[0103] Step 1.4: Calculate the target detection results of a single transceiver channel of the distributed MIMO sonar based on the time delay and beam output value corresponding to each grid point;

[0104] The time delay corresponding to each grid point is:

[0105] (5)

[0106] In the formula,

[0107] For the first The grid point relative to the first The first launch node and the first The latency of each receiving node; It is the first The location of each grid point; For the first The location of each transmission node; It is the first The location of each receiving node; The speed of sound in water;

[0108] Target detection results of a single transmit / receive channel :

[0109] (6)

[0110] Step 1.5: For By incoherently superimposing the detection results from each channel, the distributed MIMO sonar detection results can be obtained:

[0111] (7)

[0112] The output result is an incoherent superposition of multiple output results from multiple transceiver nodes after space-time processing. At this time, multiple target bright spots will be superimposed at the target location, which can further improve the intensity of the target bright spots and the robustness of detection. At the same time, it initially suppresses channel fluctuations, target intensity fluctuations and reverberation background, and obtains multiple consecutive single-frame detection outputs as input for step 2.

[0113] Step 1 above mainly involves utilizing the spatial diversity characteristics of distributed channels and the intra-node waveform coherence and inter-node channel orthogonality characteristics of the target signal to perform single-frame processing of the echoes from all receiving nodes of a distributed MIMO sonar. In this step: firstly, The received echo of each receiving node and The transmitted signals from each transmitting node are subjected to matched filtering. Then, based on the grid division of the detection area, the scan vector corresponding to each grid point is calculated. Beamforming is then performed on the output of the matched filter according to the scan vector of each grid point to obtain the beam output. The time delay of each grid point relative to the transmitting and receiving nodes is then taken on the beam output. Finally, the above steps are repeated to obtain the beam output. The values ​​of the grid points are incoherently superimposed to obtain the single-frame detection output of the distributed MIMO sonar, and multiple consecutive single-frame detection outputs are used as input for multi-frame processing.

[0114] Step 2: Utilizing the statistical distribution consistency characteristics of multi-frame reverberation and the outlier characteristics of target bright spots, a multi-frame Gaussian mixture model is used to suppress the reverberant background and retain target bright spots with outlier characteristics. The main process is as follows: Figure 4 As shown, it specifically includes:

[0115] Step 2.1: Establish a multi-frame Gaussian mixture model and initialize parameters such as mean, variance, and weights;

[0116] like Figure 5 As shown, the position of each pixel in the sonar image sequence output by a single frame can be obtained using... This indicates that its corresponding amplitude can be expressed as follows: This means that the magnitude of each pixel can be represented by a weighted sum of multiple Gaussian components, i.e., a multi-frame Gaussian mixture model:

[0117] (8)

[0118] In the formula,

[0119] It is the number of Gaussian components, usually The range of values ​​is ; It is the first The weights of each Gaussian component; Let be the probability density function of a Gaussian distribution, such as Figure 6 As shown, the Gaussian probability density function at this time is:

[0120] (9)

[0121] in, and These represent the first and second frames in the multi-frame Gaussian mixture model, respectively. The mean and variance of each Gaussian component;

[0122] Based on the obtained single-frame detection outputs, the initial mean of the multi-frame Gaussian mixture model is calculated at each pixel. With initial variance :

[0123] (10)

[0124] (11)

[0125] To speed up the modeling process, the intensity of the pixels in the first frame of the image can usually be used as the mean of a certain Gaussian component in the multi-frame Gaussian mixture model and given a large weight. At the same time, the mean of other Gaussian components is set to 0, and a large variance and a small weight are set.

[0126] Step 2.2: After initialization, update the mean, variance, weight, and other parameters of each Gaussian component;

[0127] At this point, the intensity of each pixel in the current image is sequentially matched with the existing multi-frame Gaussian mixture model. If the intensity of a pixel matches any of the Gaussian components, then the parameter of that Gaussian component is updated using that value.

[0128] (12)

[0129] (13)

[0130] (14)

[0131] In the formula,

[0132] The weight learning rate, and ; The learning rate is the sum of the mean and variance, and ; , and They represent the first The weights, mean, and variance of each Gaussian component;

[0133] Step 2.3: Separate the foreground and background in the detection output;

[0134] Specifically, the multi-frame amplitude values ​​of each pixel are matched with existing multi-frame Gaussian mixture models to determine:

[0135] When satisfied hour( If 2.5 is generally taken, then the pixel is considered to be the background of the image, and the process returns to step 2.2, using equations (12)-(14) to update the parameters of the multi-frame Gaussian mixture model;

[0136] Conversely, when If the pixel is considered a target bright spot, it will be retained.

[0137] Step 2.4: The target highlights retained in Step 2.3 are used to form the final multi-frame Gaussian mixture model output, which is then used as the input for Step 3.

[0138] Step 2 above mainly concerns the method of using the statistical distribution consistency characteristics of multi-frame reverberation and the outlier characteristics of target bright spots to suppress reverberant background and retain target bright spots with outlier characteristics using a multi-frame Gaussian mixture model, thereby enhancing the target bright spots in multi-frame detection. In this step: First, the multi-frame amplitude of each pixel in the multiple single-frame detection outputs is modeled as a Gaussian mixture model, and the mean, variance, and weight of each Gaussian component in the multi-frame Gaussian mixture model are initialized and updated; then, during the update process, each Gaussian component is sorted from largest to smallest according to its weight value, and the amplitude of each pixel in the current frame is compared with the mean of each Gaussian component in the multi-frame Gaussian mixture model. If the difference between the amplitude of a pixel in the current frame and the mean of each Gaussian component exceeds a set threshold, the pixel is determined to be a target bright spot and is retained; otherwise, the pixel is determined to be reverberant background, and the process returns to step 2.2 to update the parameters of the multi-frame Gaussian mixture model again; finally, the reverberation suppression result is output.

[0139] Step 3: Utilizing the motion continuity characteristics of target highlights across multiple frames, the output of the multi-frame Gaussian mixture model is processed using multi-frame high-order nonlinear accumulation. The main process is as follows: Figure 7 As shown, it specifically includes:

[0140] The output of the multi-frame Gaussian mixture model is used as the input to the multi-frame higher-order nonlinear accumulation processing, and the output of the multi-frame Gaussian mixture model is incoherently accumulated using the multi-frame higher-order nonlinear accumulation.

[0141] Due to the continuous motion of the target, when a bright spot passes through a pixel in the image, the intensity of the pixel and its vicinity fluctuates significantly. However, clutter and reverberant background are relatively stable in intensity, with less noticeable fluctuations. Therefore, using multi-frame high-order nonlinear cumulants can effectively describe the differences in the inter-frame fluctuation characteristics of the target bright spot and reverberation, further separating the reverberant background from the target bright spot. It can be represented as:

[0142] (15)

[0143] In the formula,

[0144] The number of frames for multi-frame high-order nonlinear cumulative quantities; For the first The pixel value of the frame; The order of the multi-frame high-order nonlinear cumulant;

[0145] Finally, the output of the multi-frame high-order nonlinear cumulative quantity is used as the final output for reverberation suppression and target enhancement.

[0146] Step 3 above mainly involves a method that utilizes the motion continuity characteristics of multi-frame target highlights, employs multi-frame high-order nonlinear cumulative quantity processing to further suppress channel fluctuations, target intensity fluctuations, and reverberation background, and significantly enhances the multi-frame continuous motion highlights of the target. This method further suppresses channel fluctuations, target intensity fluctuations, and reverberation background, significantly enhances the multi-frame continuous motion highlights of the target, and ultimately obtains stable and high-quality target highlight detection output.

[0147] Implementation Examples

[0148] The method proposed in this invention can be used for robust detection of frogmen, unmanned underwater vehicles (UUVs), surface ships, submarines, marine life, and other moving targets in strongly reverberant environments. For simplicity, this embodiment uses a typical underwater slow-moving small target detection process as an example to verify the effectiveness of the proposed method. The embodiment utilizes computer simulation for numerical simulation to verify the performance of the proposed method.

[0149] Setting the sonar detection mode and transmission signal parameters: Assuming the transmission signal is a sound wave with a propagation speed of 1500 m / s underwater, the detection period is set to 1 s when simulating multi-frame target echoes. Three targets are set in this simulation experiment: the positions of targets one, two, and three are (-200m, 400m), (50m, 450m), and (250m, 400m), respectively. The target motion states are shown in Table 1.

[0150] Table 1 Target Motion Status

[0151]

[0152] The node location distributions of the traditional method and the proposed method are as follows: Figure 8 As shown, the settings for the traditional method and the proposed method are as follows:

[0153] Traditional method: Target detection is performed using a single transceiver node with coordinates (0m, 0m). The transmitting node uses a single transmitting transducer, and the receiving node uses a 64-element uniform linear array. The transmitting node transmits an LFM signal with a center frequency of 50kHz, a bandwidth of 5kHz, and a pulse width of 5ms.

[0154] The proposed method involves two co-located transmit and receive nodes with a large spacing between them. Node 1 is located at (0m, 0m), and Node 2 at (200m, 0m). The transmitting node still uses a single transmitting transducer, and the receiving node still uses a 64-element uniform linear array. To ensure the orthogonality of the transmitted signals, the two nodes transmit radio frequency divided orthogonal LFM signals. Node 1 transmits an LFM signal with a center frequency of 50kHz, a bandwidth of 5kHz, and a pulse width of 5ms; Node 2 transmits an LFM signal with a center frequency of 60kHz, a bandwidth of 5kHz, and a pulse width of 5ms.

[0155] The reverberation suppression parameters are set using both traditional and proposed methods as follows: The parameters for the multi-frame Gaussian mixture model are as follows: Number of Gaussian components. initial weight values initial value of variance Weight learning rate The minimum weight ratio of the background model is 0.7; the parameters of the multi-frame high-order nonlinear cumulant are as follows: 4 cumulative frames, 3rd order.

[0156] For cases where the integral sidelobe ratio deviation takes negative or zero values ​​in subsequent simulations, these are uniformly referred to as "missed detections," and the value of the integral sidelobe ratio deviation is uniformly set to 0dB.

[0157] The results at frames 10, 20, and 30 after reverberation suppression and target enhancement are as follows: Figure 9As shown: Under the traditional method, the bright spot of target 2 disappears, and the intensity of the bright spot of target 3 is very weak; while under the proposed method, all three bright spots exist, and their intensity and area are significantly greater than the results of the traditional method.

[0158] Reverberation suppression was performed using both traditional and proposed methods. The bright spot distribution of the three targets after target enhancement and multi-frame accumulation is shown below. Figure 10 As shown, compared with the results of traditional methods, the intensity and area of ​​the three target bright spots in the proposed method are significantly increased, and there is basically no "missed detection" of target 2 bright spot.

[0159] The integral sidelobe ratio deviation of the three targets after reverberation suppression and target enhancement is as follows: Figure 11 As shown: For objective 1, the integral sidelobe ratio deviation of the proposed method is stable at around 50dB, with almost no missed detections, while the integral sidelobe ratio deviation of the traditional method fluctuates significantly across multiple frames, and shows a sharp decline in some frames; For objective 2, the integral sidelobe ratio deviation of the proposed method remains at around 30dB, at least 15dB higher than that of the traditional method, and the traditional method shows "missed detections" in many frames; For objective 3, the integral sidelobe ratio deviation of the proposed method is stable at around 50dB, at least 5dB higher than that of the traditional method, while the integral sidelobe ratio deviation of the traditional method fluctuates significantly across multiple frames, and shows a sharp decline in the integral sidelobe ratio deviation in some frames.

Claims

1. A method for integrated reverberation suppression and target enhancement in single-frame and multi-frame distributed MIMO sonar, characterized in that, The specific steps include: Step 1: Utilize the spatial diversity characteristics of the distributed channel and the intra-node waveform coherence and inter-node channel orthogonality characteristics of the target signal to perform space-time processing on all receiving nodes of the distributed MIMO sonar to obtain the detection output on the transmit-receive channel, and perform incoherent delay summation on the transmit-receive channel outputs between all nodes to obtain a single frame detection output. Step 2: Utilize the statistical distribution consistency features of multi-frame reverberation and the outlier features of the target bright spots to model the multi-frame amplitude of each pixel in the single-frame detection output obtained in Step 1 as a Gaussian mixture model. Then, use the multi-frame Gaussian mixture model to suppress the reverberation background and retain the target bright spots with outlier features to obtain the multi-frame Gaussian mixture model output. Step 3: Utilize the motion continuity features of the target highlights in multiple frames, and use multi-frame high-order nonlinear accumulation to process the output of the multi-frame Gaussian mixture model obtained in Step 2, and finally output the multi-frame high-order nonlinear accumulation quantity.

2. The distributed MIMO sonar single-frame and multi-frame integrated reverberation suppression and target enhancement method as described in claim 1, characterized in that, Step 1 specifically includes: Step 1.1: Arrangement Each launch node Distributed MIMO sonar at each receiving node forms a... There are 1 transmit / receive channel, with each transmitting node using a single transmit transducer and each receiving node using... A uniform linear array is used to calculate the echo received by the receiving node; Step 1.2: [The text appears to be incomplete and contains several grammatical errors. A more accurate translation would require the full context.] The echoes on each receiving array element are matched and filtered with the transmitted signal; Step 1.3: Divide the detection area into a uniform grid structure, then calculate the angle of each grid point relative to the receiving node, and then calculate the beamforming vector corresponding to each grid point. Finally, perform conventional beamforming on the matched filtering results obtained in Step 1.

2. Step 1.4: Calculate the target detection results of a single transceiver channel of the distributed MIMO sonar based on the time delay and beam output value corresponding to each grid point; Step 1.5: For The channel detection results are incoherently superimposed to obtain the distributed MIMO sonar detection results.

3. The distributed MIMO sonar single-frame and multi-frame integrated reverberation suppression and target enhancement method as described in claim 2, characterized in that, In step 1.1, the receiving transducer... The echo received by each array element is: In the formula, The scattering intensity of the target; To transmit signals from the transducer; The propagation delay from the launching node to the target; The propagation delay from the target to the receiving node; This refers to a specific moment when the echo is received. for Time of the first The first receiving transducer The noise received by each array element.

4. The distributed MIMO sonar single-frame and multi-frame integrated reverberation suppression and target enhancement method as described in claim 2, characterized in that, In step 1.2, the matched filtering result is in the form of a temporal convolution: In the formula, For the first The echo of each receiving array element; Let be the impulse response function of the matched filter corresponding to the transmitted signal; superscript Indicates taking the conjugate; The length of a single transmitted signal.

5. The distributed MIMO sonar single-frame and multi-frame integrated reverberation suppression and target enhancement method as described in claim 2, characterized in that, In step 1.3, the beam output result is as follows: In the formula, Beam output for a single transmit / receive channel; The weighting vector used for beamforming is expressed as: in, The imaginary unit; The center frequency of the echo signal; To represent the time delay of the i-th array element relative to the first array element; The element spacing of a uniform linear array, The angle between the echo and the array normal direction. This is the speed of sound in water.

6. The distributed MIMO sonar single-frame and multi-frame integrated reverberation suppression and target enhancement method as described in claim 2, characterized in that, In step 1.5, the distributed MIMO sonar detection results are as follows: In the formula, The target detection results for a single transmit / receive channel in step 1.4 are given, where, For the first The grid point relative to the first The first launch node and the first The latency of each receiving node is expressed as: in, It is the first The location of each grid point; For the first The location of each transmission node; It is the first The location of each receiving node; c is the speed of sound in water.

7. The distributed MIMO sonar single-frame and multi-frame integrated reverberation suppression and target enhancement method as described in claim 1, characterized in that, Step 2 specifically includes: Step 2.1: Establish a multi-frame Gaussian mixture model and initialize parameters such as mean, variance, and weights; Step 2.2: After initialization, update the mean, variance, weight and other parameters of each Gaussian component, and sort each Gaussian component in descending order of its weight value; Step 2.3: Match the multi-frame amplitude of each pixel with the existing multi-frame Gaussian mixture model, and compare the amplitude of each pixel in the current frame with the mean of each Gaussian component in the multi-frame Gaussian mixture model: if the judgment condition is met, the pixel is determined to be the target bright spot and is retained; otherwise, the pixel is determined to be the reverberant background and the process returns to step 2.2 to update the parameters of the multi-frame Gaussian mixture model. Step 2.4: Combine the target highlights retained in Step 2.3 to form the final multi-frame Gaussian mixture model output.

8. The distributed MIMO sonar single-frame and multi-frame integrated reverberation suppression and target enhancement method as described in claim 7, characterized in that, The multi-frame Gaussian mixture model established in step 2.1 is as follows: In the formula, This indicates the position of each pixel in the sonar image sequence output from a single frame. This indicates its corresponding amplitude; It represents the number of Gaussian components, with a range of values ​​of 1. ; It is the first The weights of each Gaussian component; Let be the probability density function of a Gaussian distribution, expressed as: in, and These represent the first and second frames in the multi-frame Gaussian mixture model, respectively. The mean and variance of each Gaussian component. The parameter initialization in step 2.1 includes: initial mean With initial variance : The intensity of the pixels in the first frame image is used as the mean of a certain Gaussian component in the multi-frame Gaussian mixture model, and a large weight is assigned to it. Meanwhile, the mean of other Gaussian components is set to 0, and a large variance and a small weight are assigned. The basis for updating the parameters in step 2.2 is: In the formula, The weight learning rate, and ; The learning rate is the sum of the mean and variance, and ; , and They represent the first The weights, mean, and variance of each Gaussian component.

9. The distributed MIMO sonar single-frame and multi-frame integrated reverberation suppression and target enhancement method as described in claim 7, characterized in that, The determination condition in step 2.3 is: the difference between the magnitude of a pixel and the mean of each Gaussian component in the current frame exceeds a set threshold, i.e. ,in, Take 2.

5.

10. The distributed MIMO sonar single-frame and multi-frame integrated reverberation suppression and target enhancement method as described in claim 1, characterized in that, In step 3, the multi-frame high-order nonlinear cumulative amount is: In the formula, The number of frames for multi-frame high-order nonlinear cumulative quantities; For the first The pixel value of the frame; The order of the multi-frame high-order nonlinear cumulant.

Images