A method for underwater target detection in transmitted acoustic field based on principal component dimension optimization

Abstract

The present application belongs to the technical field of underwater acoustic target detection. The present application provides a method for underwater target detection in a transmitted sound field based on principal component dimensional optimization. The disclosed embodiment is based on the abnormal characteristics of the transmitted sound field caused by the target, optimizes the selected principal components using a small amount of labeled data, extracts the principal component dimensions with high correlation with the target to train the filter weight vector, and obtains the optimized filter weight vector, which enhances the interference suppression and target feature extraction capabilities and improves the robustness of the target detection process. The filter is then used to process the signal to be tested to achieve real-time detection of underwater targets. The trained filter is used to process the actual data, which greatly improves the signal processing speed and improves the real-time and robustness of target detection.

Description

Technical Field

[0001] The disclosed embodiments relate to the technical field of underwater acoustic target detection, and in particular to a method for underwater target detection in a transmitted sound field based on principal component dimensional optimization. Background Art

[0002] In transmitted sound detection, when the target is located near the transceiver connection between the sound source and the receiver, an abnormal signal of the transmitted sound field (hereinafter referred to as the target signal) will be stimulated. However, the masking effect of strong background interference (sound waves received when there is no target) in the time, frequency and space domains makes it difficult to directly extract the target signal, and background interference suppression and weak target feature extraction processing are required. The traditional method calculates the principal component based on the maximum variance direction of the received data, and uses the projection measurement anomaly caused by the target signal on certain principal component dimensions to achieve target detection. However, for the selection of the principal component dimension where the target feature is located, the traditional method often relies on engineering experience, and it is impossible to determine that the selected dimension is insensitive to interference and sensitive to the target feature, and lacks theoretical guidance. At the same time, the principal component analysis method requires feature decomposition processing of the input data at each moment, and the performance of the method depends on a large amount of data to be measured. Therefore, there are problems such as high computational cost in engineering practice.

[0003] Therefore, it is necessary to improve one or more problems existing in the above-mentioned related technical solutions.

[0004] It should be noted that this section is intended to provide background or context for the technical solutions of the present disclosure stated in the claims. The description herein is not admitted to be prior art by virtue of being included in this section. Summary of the invention

[0005] In order to avoid the shortcomings of the prior art, the present invention provides a method for underwater target detection in a transmitted sound field based on principal component dimensional optimization, so as to solve the problems in the prior art that it is impossible to determine whether the selected dimension is insensitive to interference while being sensitive to target characteristics, and there is a lack of theoretical guidance.

[0006] According to an embodiment of the present disclosure, a method for underwater target detection in a transmission sound field based on principal component dimensional optimization is provided, the method comprising:

[0007] Preprocess the training signal to obtain a signal matrix and a label matrix;

[0008] The filter weights are optimized and trained using the signal matrix and the label matrix to obtain an optimized filter weight vector;

[0009] The optimized filter weight vector is used to process the test signal to obtain the target detection result.

[0010] Furthermore, the step of preprocessing the training signal to obtain a signal matrix and a label matrix includes:

[0011] Performing pulse compression processing on the training signal;

[0012] The training signal after pulse compression processing is sorted and divided according to the signal period to obtain the signal matrix X and the label matrix Y ;

[0013] The signal matrix X Contains multi-cycle received data with or without target, label matrix Y Representing signal matrix X The label of each row of data is 1 if there is a target and 0 if there is no target.

[0014] Furthermore, the step of training the filter weights using the signal matrix and the label matrix to obtain an optimized filter weight vector includes:

[0015] Signal Matrix X and the label matrix Y Perform standardization to obtain the independent variable signal matrix X 0 and dependent variable label matrix Y 0;

[0016] Use the dependent variable principal component to correct the direction of the independent variable principal component to obtain the first unit vector weight coefficient and the second unit vector weight coefficient ;

[0017] According to the independent variable signal matrix X 0. Dependent variable label matrix Y 0. First unit vector weight coefficient and the second unit vector weight coefficient , to obtain the first principal component; wherein the first principal component includes the first signal principal component and the first label principal component ;

[0018] The independent variable signal matrix X 0 and independent variable label matrix Y 0 are projected onto the principal component direction to obtain the first projection length and the first projection length ;

[0019] The first signal principal component From the first projection length To obtain the first signal residual matrix X 1. Label the first principal component From the first projection length Peel off to get the first label residual matrix Y 1;

[0020] According to the first signal residual matrix X 1 and the first label residual matrix Y 1, to obtain the regression relationship between the independent variable and the dependent variable;

[0021] Based on the regression relationship between the independent variable and the dependent variable, the first signal residual matrix X 1 and the first label residual matrix Y 1 Iterate and solve to the Nth principal component to obtain the training result;

[0022] The training results are cross-checked, the number of principal components corresponding to the minimum error of the cross-check results is selected, and the optimized filter weight vector is obtained according to the number of principal components.

[0023] Furthermore, the independent variable signal matrix X 0 and dependent variable label matrix Y The expressions for 0 are:

[0024] (1)

[0025] in, X is the signal matrix, Y is the label matrix, n is the number of rows in the signal matrix and the label matrix, X i，j The signal matrix i Line j Column elements, X j For the j Column signal matrix, Y i For the i The row label matrix, var () is the standard deviation.

[0026] Furthermore, the principal component of the dependent variable is used to correct the direction of the principal component of the independent variable, and then formula (2) is obtained:

[0027] (2)

[0028] Solving formula (2) by the Lagrange multiplier method, we can obtain , , we can get the first unit vector weight coefficient and the second unit vector weight coefficient , It means taking the second norm.

[0029] Furthermore, the first projection length and the first projection length The expressions are:

[0030] (3)

[0031] The first signal residual matrix X 1 and the first label residual matrix Y The expressions of 1 are:

[0032] (4)

[0033] in, is the first signal principal component, is the first label principal component.

[0034] Furthermore, the regression relationship between the independent variable and the dependent variable is:

[0035] (5)

[0036] in, is the label matrix Y 0 in the first signal principal component t 1 is the projection length.

[0037] Furthermore, the step of obtaining an optimized filter weight vector according to the number of principal components includes:

[0038] Generate through iterative solution a principal components, the data matrix is ​​represented as:

[0039] (6)

[0040] in, is the optimized filter weight vector, For the i Unit vector weight coefficient, The label matrix is i The projection length onto the principal components of the signal, .

[0041] Further, the step of processing the test signal using the optimized filter weight vector to obtain the target detection result includes:

[0042] Processing the signal to be tested into a signal residual matrix;

[0043] Based on the signal residual matrix and the optimized filter weight vector, the target detection result is obtained; wherein, the expression of the target detection result is:

[0044] (7)

[0045] In the formula, X c is the signal residual matrix, Y c Target detection result.

[0046] The technical solution provided by the embodiments of the present disclosure may have the following beneficial effects:

[0047] In the embodiments of the present disclosure, the above-mentioned method for underwater target detection in a transmitted sound field based on principal component dimension optimization is used. On the one hand, based on the abnormal characteristics of the transmitted sound field caused by the target, a small amount of labeled data is used to optimize the selected principal components, and the principal component dimensions with high correlation with the target are extracted to train the filter weight vector to obtain the optimized filter weight vector, thereby enhancing the interference suppression and target feature extraction capabilities and improving the robustness of the target detection process. The filter is then used to process the signal to be tested to achieve real-time detection of underwater targets. On the other hand, the trained filter is used to process the actual data, and the signal processing speed is greatly improved, thereby improving the real-time and robustness of target detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The accompanying drawings herein are incorporated into the specification and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the specification are used to explain the principles of the present disclosure. Obviously, the accompanying drawings described below are only some embodiments of the present disclosure, and for ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without creative work.

[0049] Figure 1 A step diagram showing a method for underwater target detection in a transmission sound field based on principal component dimension optimization in an exemplary embodiment of the present disclosure;

[0050] Figure 2 The measured sound velocity profile in the exemplary embodiment of the present disclosure is shown;

[0051] Figure 3 A schematic diagram of transmitted sound detection in an exemplary embodiment of the present disclosure is shown;

[0052] Figure 4 A detection schematic diagram and a target motion trajectory in an exemplary embodiment of the present disclosure are shown;

[0053] Figure 5 A specific flow chart of the underwater target detection method of the transmission sound field based on the optimization of the principal component dimension in the exemplary embodiment of the present disclosure is shown;

[0054] Figure 6 The received signal matrix in the exemplary embodiment of the present disclosure is shown;

[0055] Figure 7 The cross-check result in the exemplary embodiment of the present disclosure is shown;

[0056] Figure 8 The detection output curve in the exemplary embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0057] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that the disclosure will be more comprehensive and complete and to fully convey the concepts of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0058] In addition, the accompanying drawings are only schematic illustrations of the embodiments of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated descriptions will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities.

[0059] This example embodiment provides a method for underwater target detection in a transmitted sound field based on principal component dimension optimization. Figure 1 As shown in , the underwater target detection method of the transmitted sound field based on principal component dimension optimization may include: step S101 to step S103.

[0060] Step S101: preprocessing the training signal to obtain a signal matrix and a label matrix;

[0061] Step S102: Optimizing and training the filter weights using the signal matrix and the label matrix to obtain an optimized filter weight vector;

[0062] Step S103: Process the signal to be tested using the optimized filter weight vector to obtain a target detection result.

[0063] Through the above-mentioned underwater target detection method of the transmitted sound field based on the optimization of the principal component dimension, on the one hand, based on the abnormal characteristics of the transmitted sound field caused by the target, a small amount of labeled data is used to optimize the selected principal components, and the principal component dimensions with high correlation with the target are extracted to train the filter weight vector to obtain the optimized filter weight vector, which enhances the interference suppression and target feature extraction capabilities and improves the robustness of the target detection process. The filter is then used to process the signal to be tested to achieve real-time detection of underwater targets. On the other hand, the trained filter is used to process the actual data, and the signal processing speed is greatly improved, which improves the real-time and robustness of target detection.

[0064] Next, we will refer to Figures 1 to 8 Each step of the above-mentioned method for underwater target detection in a transmitted sound field based on principal component dimensional optimization in this example implementation is described in more detail.

[0065] In step S101, the transmitting end is composed of a single sound source, the receiving end is composed of a single hydrophone, and the detection area is near the two-dimensional vertical plane between the sound source and the hydrophone. The received signal is pulse compressed and sorted into a matrix form according to the signal period T. The number of rows and columns of the matrix are the number of detection periods and the number of sampling points in one period, respectively. The matrix is ​​classified into a signal matrix according to the category X and the label matrix Y , signal matrix X Contains multi-cycle receive data with and without target, label matrix Y Representing signal matrix X The label of each row of data is 1 if there is a target and 0 if there is no target.

[0066] In step S102, a signal matrix containing labels is used X and the label matrix Y The filter weights are trained as follows:

[0067] Step 1: Signal Matrix X and the label matrix Y Perform standardization (remove the mean by column and divide by the variance) to obtain the independent variable signal matrix and the dependent variable label matrix .

[0068] (1)

[0069] Step 2: The principal component is a linear combination of the column vectors of the data matrix, and there is a first unit vector weight coefficient Let the first signal principal component ; Similarly, there is a second unit vector weight coefficient Let the first label principal component The essence of using label data to optimize the selection of principal components is to use the principal component of the dependent variable to correct the direction of the principal component of the independent variable, and solve the weight coefficient of the first unit vector according to formula (2): and the second unit vector weight coefficient .

[0070] (2)

[0071] Solve the above equation by Lagrange multiplier method: , , so the two unit vectors are the eigenvectors corresponding to the maximum eigenvalues ​​of the two combined matrices.

[0072] Step 3: Obtain the first signal principal component and the first label principal component Then, the independent variable signal matrix and the dependent variable label matrix Project them respectively into the direction of the principal component and obtain the first projection length and the first projection length ,have:

[0073] (3)

[0074] After the principal component is stripped from the data matrix, the first signal residual matrix X 1 and the first label residual matrix Y 1 are:

[0075] (4)

[0076] Since the dependent variable can be expressed by the principal components of the independent variables to be solved, we have:

[0077] (5)

[0078] Formula (5) shows that the regression relationship between the independent variable and the dependent variable can be represented by the projection of the dependent variable on the principal component of the independent variable.

[0079] Step 4: Use the first signal residual matrix X 1 and the first label residual matrix Y 1 Continue to solve the second to Nth principal components according to the second and third steps (N is the total number of principal components). Perform a cross-check on the training results and select the number of principal components corresponding to the minimum error of the test results. a Among them, cross-checking is an existing technology and will not be described in detail here.

[0080] Step 5: Generate the solution after the above iterations a principal components, the data matrix can be expressed as:

[0081] (6)

[0082] in is the optimized filter weight vector.

[0083] In step S103, the signal to be tested is converted into a residual matrix of the signal to be tested according to step S101. X c, the optimized filter weight vector obtained by training according to formula (7) The residual matrix of the test signal X c is processed to obtain the target detection output Y c.

[0084] (7)

[0085] Based on the abnormal characteristics of the transmitted sound field caused by the target, a small amount of labeled data is used to optimize the principal component feature subspace, and a feature subspace that is sensitive to the target characteristics and insensitive to background interference is selected. The feature projection measurement value in the subspace is used as the detection quantity, which improves the target detection performance. Compared with the feature extraction method based on subspace decomposition, this application uses a trained filter to process the actual data, and the signal processing speed is greatly improved. This method can improve the real-time and robustness of target detection.

[0086] In a specific embodiment, the background information is as follows: the test water depth is 35m, and the measured sound velocity profile is as follows Figure 2 As shown in Figure 1, the depth of 0m-20m is an isothermal layer, and the depth of 20m-35m is a strong negative gradient. Using a single high-frequency sound source and a single hydrophone to arrange the detection scene with separate transmitters and receivers, the results are as follows: Figure 3 As shown in the figure, the receiving and transmitting distance is 1.1 km. The sound source depth is 10 m, and the receiving depth is 20 m. The moving target is an AUV with an outer diameter of 324 mm, a navigation depth of 40 m, and a speed of 4 knots. The AUV is sailing in a straight line at a constant depth and at a constant speed, following the predetermined track (such as Figure 4 The solid line shows the Figure 4 dotted line) 1 time.

[0087] The implementation process of the underwater target detection method based on principal component dimension optimization in the transmitted sound field is as follows: Figure 5 As shown, the specific process is as follows:

[0088] Signal preprocessing:

[0089] The high-frequency sound source periodically radiates a linear frequency modulation pulse signal with a center frequency of 45kHz, a bandwidth of 4kHz, a pulse width of 200ms, and a period of 0.5s. The receiver arranged at a distance performs pulse compression processing on the data and organizes it into a matrix according to the signal period T. The number of rows and columns of the matrix are the number of detection cycles and the number of sampling points in one cycle, respectively. According to the data category, the matrix is ​​divided into a signal matrix X and a label matrix Y, where X contains a total of 432s of target data, and Y contains 1 and 0 for target or no target, respectively.

[0090] The filter weights are trained using the first 190 seconds of labeled data, where 165s to 190s contain the target. The specific processing process is as follows:

[0091] Step 1: Signal Matrix X and the label matrix Y Perform standardization (remove the mean by column and divide by the variance) to obtain the independent variable signal matrix and the dependent variable label matrix .

[0092] (1)

[0093] Step 2: The principal component is a linear combination of the column vectors of the data matrix, and there is a first unit vector weight coefficient Let the first signal principal component ; Similarly, there is a second unit vector weight coefficient Let the first label principal component The essence of using label data to optimize the selection of principal components is to use the principal component of the dependent variable to correct the direction of the principal component of the independent variable, and solve the weight coefficient of the first unit vector according to formula (2): and the second unit vector weight coefficient .

[0094] (2)

[0095] Solve the above equation by Lagrange multiplier method: , , so the two unit vectors are the eigenvectors corresponding to the maximum eigenvalues ​​of the two combined matrices.

[0096] Step 3: Obtain the first signal principal component and the first label principal component Then, the independent variable signal matrix and the dependent variable label matrix Project them respectively into the direction of the principal component and obtain the first projection length and the first projection length ,have:

[0097] (3)

[0098] After the principal component is stripped from the data matrix, the first signal residual matrix X 1 and the first label residual matrix Y 1 are:

[0099] (4)

[0100] Since the dependent variable can be expressed by the principal components of the independent variables to be solved, we have:

[0101] (5)

[0102] Formula (5) shows that the regression relationship between the independent variable and the dependent variable can be represented by the projection of the dependent variable on the principal component of the independent variable.

[0103] Step 4: Use the first signal residual matrix X 1 and the first label residual matrix Y1 Continue to solve the second to Nth principal components according to the second and third steps (N is the total number of principal components). The classification results of the training data in the first two dimensions are as follows Figure 6 As shown in , the targetless data (class 0) and the target data hardly overlap on the feature plane formed by dimension 1 and dimension 2, indicating that the selected dimensional space can achieve a good distinction between target features and background interference. The training results were subjected to a 10-fold cross-validation (the training set was evenly divided into 10 parts, one of which was taken as the test set each time without repetition, and the other 9 were taken as the training set. The minimum mean square error of the method on the test set when different numbers of principal components were taken was calculated and averaged). The results of the error changing with the number of principal components are shown in Figure 6 As shown, the number of principal components corresponding to the minimum error is selected a =12.

[0104] Step 5: After the above iterative solution and verification process, the a =12 principal components, the data matrix can be expressed as:

[0105] (6)

[0106] in, is the optimized filter weight vector, For the i Unit vector weight coefficient, The label matrix is i The projection length onto the principal components of the signal, .

[0107] Preprocess the test signal to generate the matrix to be processed X c, according to formula (7), the optimized filter weight vector after training is used Process the data to be tested and obtain the target detection output Y c.

[0108] (7)

[0109] Repeat the above process for all 432s of received signals to obtain the corresponding detection output values ​​in each cycle, and arrange them in time to form a target detection curve as shown in the figure Figure 8 As shown by the solid line, the actual label at the corresponding moment is Figure 8 The dotted line shows that the data of the first 190s is the same as the training data, so the results of this section are more stable; 190s~432s are untrained test data. It can be seen that the detection curve of the processed results has large fluctuations in 165s~255s, and the detection output is close to 1, indicating that the target crossing event is successfully detected. During the period without the target, the detection output is stable and remains near the background value of 0. The above detection results are consistent with the results recorded by the submersible, indicating that this method can effectively detect the target.

[0110] This example shows that the method based on the transmission sound detection of a single sound source and a single hydrophone can achieve rapid real-time detection of the AUV small target crossing process in a shallow water environment. The detection gain is high when the target appears, and the detection background is stable when there is no target.

[0111] Compared with traditional methods, this method has the following advantages:

[0112] This method optimizes the feature selection dimension using a small amount of labeled data, enhances the target detection performance, and has better robustness.

[0113] This method uses trained filters to process the test data. Compared with the traditional feature decomposition method, it significantly improves the data processing speed and can achieve real-time processing.

[0114] Through the above-mentioned underwater target detection method of the transmitted sound field based on the optimization of the principal component dimension, on the one hand, based on the abnormal characteristics of the transmitted sound field caused by the target, a small amount of labeled data is used to optimize the selected principal components, and the principal component dimensions with high correlation with the target are extracted to train the filter weight vector to obtain the optimized filter weight vector, which enhances the interference suppression and target feature extraction capabilities and improves the robustness of the target detection process. The filter is then used to process the signal to be tested to achieve real-time detection of underwater targets. On the other hand, the trained filter is used to process the actual data, and the signal processing speed is greatly improved, which improves the real-time and robustness of target detection.

[0115] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the embodiments of the present disclosure, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0116] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example" or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification.

[0117] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any modification, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present disclosure are indicated by the appended claims.

Claims

1. A method for underwater target detection in a transmission acoustic field based on principal component dimension optimization, characterized in that: The method includes: Preprocess the training signal to obtain a signal matrix and a label matrix; The filter weights are optimized and trained using the signal matrix and the label matrix to obtain an optimized filter weight vector; The optimized filter weight vector is used to process the test signal to obtain the target detection result; wherein, The step of preprocessing the training signal to obtain a signal matrix and a label matrix includes: Performing pulse compression processing on the training signal; The training signal after pulse compression processing is sorted and divided according to the signal period to obtain the signal matrix X and the label matrix Y ; Among them, the signal matrix X Contains multi-cycle received data with and without a target, label matrix Y Representing signal matrix X The label of each row of data is 1 if there is a target and 0 if there is no target; The step of training the filter weights using the signal matrix and the label matrix to obtain an optimized filter weight vector includes: Signal Matrix X and the label matrix Y Perform standardization to obtain the independent variable signal matrix X 0 and dependent variable label matrix Y 0; Use the dependent variable principal component to correct the direction of the independent variable principal component to obtain the first unit vector weight coefficient and the second unit vector weight coefficient ; According to the independent variable signal matrix X 0. Dependent variable label matrix Y 0. First unit vector weight coefficient and the second unit vector weight coefficient , to obtain the first principal component; wherein the first principal component includes the first signal principal component and the first label principal component ; The independent variable signal matrix X 0 and independent variable label matrix Y 0 are projected onto the principal component direction to obtain the first projection length and the first projection length ; The first signal principal component From the first projection length To obtain the first signal residual matrix X 1. Label the first principal component From the first projection length Peel off to get the first label residual matrix Y 1; According to the first signal residual matrix X 1 and the first label residual matrix Y 1, to obtain the regression relationship between the independent variable and the dependent variable; Based on the regression relationship between the independent variable and the dependent variable, the first signal residual matrix X 1 and the first label residual matrix Y 1 Iterate and solve to the Nth principal component to obtain the training result; The training results are cross-checked, the number of principal components corresponding to the minimum error of the cross-check results is selected, and the optimized filter weight vector is obtained according to the number of principal components.

2. The underwater target detection method in the transmission sound field based on principal component dimension optimization according to claim 1 is characterized in that: Independent variable signal matrix X 0 and dependent variable label matrix Y The expressions for 0 are: （1） in, X is the signal matrix, Y is the label matrix, n is the number of rows in the signal matrix and the label matrix, Xi,j The signal matrix i Line j Column elements, Xj For the j Column signal matrix, Yi For the i The row label matrix, var () is the standard deviation.

3. The underwater target detection method based on principal component dimension optimization in the transmission sound field according to claim 2 is characterized in that: Using the principal component of the dependent variable to correct the direction of the principal component of the independent variable, we have formula (2): （2） Solving formula (2) by the Lagrange multiplier method, we can obtain , , we can get the first unit vector weight coefficient and the second unit vector weight coefficient , It means taking the second norm.

4. The underwater target detection method based on principal component dimension optimization in the transmission sound field according to claim 3 is characterized in that: First projection length and the first projection length The expressions are: （3） The first signal residual matrix X 1 and the first label residual matrix Y The expressions of 1 are: （4） in, is the first signal principal component, is the first label principal component.

5. The underwater target detection method in the transmission sound field based on principal component dimension optimization according to claim 4 is characterized in that: The regression relationship between the independent variable and the dependent variable is: （5） in, is the label matrix Y 0 in the first signal principal component t 1 is the projection length.

6. The underwater target detection method based on principal component dimension optimization in a transmitted sound field according to claim 5, characterized in that: The step of obtaining an optimized filter weight vector according to the number of principal components includes: Generate through iterative solution a principal components, the data matrix is ​​represented as: （6） in, is the optimized filter weight vector, For the i Unit vector weight coefficient, The label matrix is i The projection length onto the principal components of the signal, .

7. The underwater target detection method based on principal component dimension optimization in a transmitted sound field according to claim 6 is characterized in that: The step of processing the test signal using the optimized filter weight vector to obtain the target detection result includes: Processing the signal to be tested into a signal residual matrix; Based on the signal residual matrix and the optimized filter weight vector, the target detection result is obtained; wherein, the expression of the target detection result is: （7） In the formula, X c is the residual matrix of the signal to be tested, Y c is the target detection result.

Images