Small-size moving obstacle detection method based on entropy under underwater dense interference source condition
The JCESM method efficiently detects small-volume moving targets by using joint conditional entropy similarity measurement to suppress underwater interference, reducing computational complexity and improving detection efficiency in near-port waters.
Patent Information
- Application Number
- CN202411336181.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-24
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2044-09-24
AI Technical Summary
Existing methods for underwater obstacle detection in near-port waters struggle to efficiently distinguish small-volume moving targets from dense underwater interference sources, particularly due to high computational complexity and sensitivity to parameter adjustments, and require multi-frame data association.
A method utilizing joint conditional entropy similarity measurement (JCESM) for single-frame detection, combining low-rank sparse matrix decomposition and alternating direction multi-multiplier optimization to suppress reverberation interference and detect moving targets.
The method significantly reduces computational complexity and improves detection efficiency, achieving rapid suppression of single-frame reverberation and dynamic background interference, with calculation times under 14 seconds compared to over 1 minute for traditional methods.
Smart Images

Figure CN119064907A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the fields of hydroacoustic engineering, underwater obstacle avoidance detection, and relates to a method for detecting small-volume moving obstacles based on entropy under conditions of dense underwater interference sources. In particular, it relates to a low-rank sparse matrix decomposition algorithm based on joint conditional entropy similarity measurement, which can quickly detect small-volume moving obstacles under conditions of shielding from dense underwater interference sources, and is suitable for near-port waters scenes based on active sonar arrays. Background Art
[0002] Faced with the challenge of underwater traffic collision avoidance under dense obstacle conditions, it is crucial to distinguish fixed obstacles from mobile obstacles in the initial stage of underwater target detection while ensuring reduced computational complexity and high reliability. However, in the near-port waters application scenario relying on active sonar arrays, it may be extremely difficult to extract the path of a small-volume moving target from the fixed interference gap. Analyzing the active sonar echo signal in shallow water, its background interference mainly comes from the reverberation of the sea surface and seabed, and the scattered echoes between underwater obstacles; among them, the reverberation interference of multiple frames shows a low-rank characteristic in the time-azimuth history matrix, while the obstacle has a very limited space occupied by the beam scanning, and its echo has a significant sparse characteristic. Based on the difference between reverberation interference and targets, the low-rank matrix decomposition (LMD) can effectively suppress the reverberation interference and extract the target echo to a certain extent. See "Chu Wenhua, Wang Weiguo, Zeng Xueying. A non-convex model for sparse low-rank matrix decomposition [J]. Journal of Ocean University of China (Natural Science Edition), 2018, 48(S2): 223-227." However, this method has poor suppression effect on dense interference sources in complex environments and is sensitive to parameters. Its dereverberation effect and target detection performance are heavily dependent on manual parameter adjustment. In addition, the LMD method requires multi-frame data association calculation, which has extremely high computational complexity and takes a long time. Therefore, it is necessary to design a more efficient post-processing algorithm based on the non-optimal suppression results of a single-frame LMD to perform fast small target detection under the background of dense underwater interference sources. Summary of the invention
[0003] Technical issues to be solved
[0004] In order to avoid the shortcomings of the prior art, the present invention proposes an entropy-based small-volume moving obstacle detection method under the condition of dense underwater interference sources. In order to detect small-volume moving targets with interference from fixed obstacles in a strong reverberation background, a target detection method based on joint conditional entropy similarity measure (JCESM) is proposed. Different from the LMD algorithm of multi-frame association, the present invention only uses the time-azimuth history matrix of the first frame echo as a reference to detect moving targets on subsequent single-frame echoes.
[0005] Technical Solution
[0006] A method for detecting small-volume moving obstacles based on entropy under conditions of dense underwater interference sources, characterized by the following steps:
[0007] Step 1: Perform beamforming on the active echo signal to obtain the time-position history of the i-th frame echo
[0008] Step 2: Use the LMD algorithm to perform low-rank sparse decomposition on the beam data of each frame of the time-azimuth history matrix to obtain a low-rank matrix L i and the sparse matrix S i ;
[0009] B i =L i +S i ;
[0010] Step 3: Use the alternating direction multi-multiplier method to optimize the low-rank sparse decomposition of the matrix;
[0011] Through steps 2 to 3, a low-rank matrix L1 and a sparse matrix S1 whose time-direction history of the first frame reference signal is B1 are obtained;
[0012] The time-direction history of the i-th frame reference signal is obtained as B i The low-rank matrix L i and sparse matrices
[0013] Step 4: S i and S1 are vectorized first, and vec(S i ) and vec(S1) joint conditional entropy
[0014] For L i and S i First perform vectorization and calculate the matrix L i The sum of the sparse matrix S i The joint conditional entropy
[0015] H(vec(S i )|vec(S1)) and H(vec(S i )|vec(L i )) is converted back to a matrix of size n1*n2, defined as H(X i |X1), H(X i |L i );
[0016] Step 5: Calculate H(S i |S1) and H(S i |L i ) H , as well as L1 and the matrix L i The Euclidean distance of L ;
[0017] Step 6: Calculate Ε H and The Pearson correlation coefficient of the target bright spot under the condition of dense interference sources is obtained:
[0018] The time-position history of the i-th frame echo The expression is a matrix of size n1×n2, where n1 represents the dimension in the time direction and n2 represents the dimension in the azimuth direction. The expression is as follows:
[0019]
[0020] in: Corresponding to the orientation matrix B i The value at the n1th moment and the n2th angle.
[0021] The decomposition process of step 2 is robust principal component analysis (RPCA), and its constraints follow:
[0022]
[0023] sB i =L i +S i
[0024] Among them, ||L i || * Calculate the matrix L i The sum of the singular values is the nuclear norm; ||S i ||1 is used to calculate the matrix S i The sum of the absolute values of all elements is the L1 norm; λ represents the balance parameter between the low-rank matrix and the sparse matrix; since the matrix L iWith low-rank characteristics, the low-rank matrix L is i Relax the representation and use the L1 norm to sparse matrix S i Perform a relaxation representation.
[0025] The optimization solution of step 3 using the alternating direction multi-multiplier method for low-rank sparse decomposition of the matrix is:
[0026] Definition||L i || * and ||S i ||1's matrix variable is f(L i ) and g(S i ), and the constraints are expressed as:
[0027]
[0028] sB i =L i +S i
[0029] The augmented Lagrangian multi-multiplier expression of the above formula is expressed as:
[0030]
[0031] Among them, Y i is the dual variable, ρ is the penalty factor, and k is the number of iterations of the optimization calculation;
[0032] In the alternating variable solution process, the matrix corresponding to the kth iteration is as follows:
[0033]
[0034] Y i,k =Y i,k-1 +L i,k +S i,k -B i,k
[0035] in, Formula L i,k Japanese Style i,k The solution is obtained by singular value soft threshold operator and soft threshold operator.
[0036] The penalty factor ρ is set as a variable, and the expression is as follows:
[0037]
[0038] Se i,k =||S i.k -S i,k-1 || F / ||S i.k ||F
[0039] Le i,k =||L i.k -L i,k-1 || F / ||L i.k || F
[0040] Among them, τ1, τ2 and μ are small constants, Le i,k and Se i,k Respectively represent the iterative residuals of the k-th low-rank matrix and sparse matrix and the k-1-th low-rank matrix and sparse matrix; the smaller the residual, the closer the matrix decomposition result is to the convergence threshold; ε, Se i,k ,Le i,k <ε.
[0041] The value range of τ1, τ2 and μ is usually between 0 and 10.
[0042] The calculation sparse matrix S i and the joint conditional entropy of the sparse matrix S1 First, S i S1 is first vectorized to obtain a column vector vec(S i ) and vec(S1):
[0043]
[0044] Calculate vec(S i ) and vec(S1) are:
[0045]
[0046] Among them, s i ,s 1 Corresponding to vec(S i ) and elements of vec(S1).
[0047] The calculation matrix L i and the sparse matrix S i The joint conditional entropy First, L i and S i First, vectorization is performed to obtain a column vector vec(L) with n1×n2 dimensions and 1 dimension. i ) and vec(S1):
[0048]
[0049] Joint Conditional Entropy The calculation expression is as follows:
[0050]
[0051] Among them, l i is the vector vec(L i ), vec(L i ) represents the matrix L i The converted columns are n1×n2 dimensional and the rows are 1-dimensional column vectors.
[0052] The Euclidean distance expression of step 5 is as follows:
[0053]
[0054] The step 1 performs beam processing on the active echo signal using a conventional beamforming algorithm, CBF, to obtain a time-direction history.
[0055] Beneficial Effects
[0056] The present invention proposes an entropy-based small-volume moving obstacle detection method under the condition of dense underwater interference sources. The method first uses the LMD algorithm to decompose the time-direction history matrix to obtain a low-rank matrix and a sparse matrix; on this basis, a joint conditional entropy matrix and a low-rank similarity matrix are established respectively; finally, based on the Pearson correlation coefficient of the low-rank similarity matrix and the joint conditional entropy similarity matrix, the position of the small target in the matrix is located to achieve rapid detection and tracking of the moving small target.
[0057] Compared with traditional methods, the proposed method significantly improves the computational efficiency and generalization of the model. It aims to solve the problem of detecting small mobile obstacles under the cover of dense underwater interference near the port. The present invention combines variational mode decomposition (VMD) and joint conditional entropy matching to optimize the low-rank matrix decomposition (LMD) method, which overcomes the problem of high cost caused by the need for LMD to calculate multiple frames continuously, and achieves rapid suppression of single-frame reverberation and dynamic background, and the target detection effect is better than the conventional LMD algorithm. The average calculation time of the present invention is within 12 to 14 seconds, which is more than 5 times the calculation efficiency of the continuous multi-frame LMF algorithm of more than 1 minute.
[0058] Compared with the prior art, the present invention has the following beneficial effects:
[0059] a) Compared with the traditional LMD method, the present invention does not require multi-frame correlation calculation, which greatly reduces the calculation complexity and saves calculation time.
[0060] b) Compared with the traditional LMD method, the present invention adopts an adaptive post-processing method based on joint conditional entropy and Euclidean distance similarity measurement, which reduces the sensitivity of the algorithm to parameter selection and improves the robustness of the method.
[0061] c) The present invention overcomes the problem of suppressing reverberation while removing the influence of dense interference sources at fixed positions, thereby effectively realizing the azimuth detection of small targets in motion. BRIEF DESCRIPTION OF THE DRAWINGS
[0062] Figure 1 Schematic diagram of the target detection method based on joint conditional entropy similarity measurement;
[0063] Figure 2 It is the time-azimuth history diagram of the reference signal of the first frame and the echo of the 233rd frame;
[0064] (a) Azimuth history diagram of the reference signal in the first frame;
[0065] (b) The azimuth history diagram of the reference signal at frame 233;
[0066] Figure 3 It represents the low-rank matrix and sparse matrix obtained by low-rank sparse decomposition of the time orientation history matrix of the 233rd frame;
[0067] (a) low-rank matrix; (b) sparse matrix;
[0068] Figure 4 Denotes the joint conditional entropy matrix H(S i |S1) and H(S i |L i );
[0069] (a)H(S i |S1); (b)H(S i |L i );
[0070] Figure 5 Represents the similarity measure matrix E H and
[0071] (a) H ; (b)
[0072] Figure 6 represents the Pearson correlation coefficient matrix, and the bright spots in the figure are the locations of the target bright spot echoes under the condition of dense interference sources;
[0073] (a) Single-frame JCESM algorithm detects the target, and the red circle frames the target bright spot echo;
[0074] (b) Conventional LMD interference suppression result of 4-frame correlation. DETAILED DESCRIPTION
[0075] The present invention will now be further described with reference to the embodiments and the accompanying drawings:
[0076] A target detection method based on joint conditional entropy similarity measure (JCESM), characterized by: firstly, using LMD algorithm to decompose the time-direction history matrix to obtain a low-rank matrix and a sparse matrix; then, respectively establishing a joint conditional entropy matrix of the sparse matrix of the reference signal and the sparse matrix of the echo signal, a low-rank similarity matrix of the low-rank matrix of the reference signal and the low-rank matrix of the echo signal, and a joint conditional entropy matrix of the low-rank matrix of the echo signal and the sparse matrix; then, using Euclidean distance to measure the similarity of the two joint conditional entropy matrices to obtain a joint conditional entropy similarity matrix; finally, calculating the correlation coefficient of the low-rank similarity matrix and the joint conditional entropy similarity matrix to find the corresponding position of the mobile target of the fixed interference source that is abnormal from the entropy approximation. It can be summarized as the following steps:
[0077] Step 1: Perform beamforming processing on the active echo signal. For the echo signal of the i-th active sonar pulse, define its beamforming processing result as a matrix of size n1×n2 Where n1 represents the dimension in the time direction (which can be converted into distance by multiplying by the speed of sound), and n2 represents the dimension in the azimuth direction. The expression is as follows:
[0078]
[0079] Here, B i Indicates the time-direction history of the i-th frame echo. Similarly, the time-direction history of the first frame data as the reference signal is defined as B1; Corresponding to the orientation matrix B i The value at the n1th moment and the n2th angle.
[0080] Step 2: Perform low-rank sparse decomposition on the beamforming data of each frame of echo. Assuming that there is a moving target in the active sonar echo, the azimuth-range spatial spectrum of the target exists in the matrix B i medium and sparse, and B i It also contains relatively stable reverberation fluctuations. i Perform a low-rank matrix L i and the sparse matrix S i :
[0081] B i =L i +S i (2)
[0082] The decomposition process of the above formula is also called robust principal component analysis (RPCA), and its constraints follow:
[0083]
[0084] Among them, ||L i || * Calculate the matrix L i The sum of the singular values, also known as the nuclear norm, ||S i ||1 is used to calculate the matrix S i The sum of the absolute values of all elements is also called the L1 norm. λ is represented as a balance parameter between a low-rank matrix and a sparse matrix. i With low-rank characteristics, the low-rank matrix L is i Relax the representation and use the L1 norm to sparse matrix S i Perform a relaxation representation.
[0085] Step 3: Use the alternating direction multi-multiplier method to optimize the low-rank sparse decomposition of the matrix. Define ||L i || * and ||S i ||1's matrix variable is f(L i ) and g(S i ), and reformulate formula (3) as:
[0086]
[0087] The augmented Lagrangian multi-multiplier expression of formula (4) can be expressed as:
[0088]
[0089] Among them, Y i is the dual variable, ρ is the penalty factor, and k is the number of iterations of the optimization calculation. In the process of solving the alternating variable, the matrix corresponding to the kth iteration is as follows:
[0090]
[0091]
[0092] Y i,k =Y i,k-1 +L i,k +S i,k -B i,k (8)
[0093] in, Equations (6) and (7) can be solved by the singular value soft threshold operator and the soft threshold operator. In order to converge to the optimal solution quickly, the penalty factor ρ k Usually set to a variable, the expression is as follows:
[0094]
[0095] Among them, τ1, τ2 and μ are small constants. According to experience, their values usually range from 0 to 10. i,k and Se i,k Represent the iterative residuals of the k-th low-rank matrix and sparse matrix and the k-1-th low-rank matrix and sparse matrix respectively. The smaller the residual, the closer the matrix decomposition result is to the convergence threshold. ε,Se i,k ,Le i,k <ε
[0096] Step 4: Define the time-direction history of the reference signal of the first frame as B1. After steps 2 and 3, the low-rank matrix decomposed is L1 and the sparse matrix is S1. Correspondingly, the time-direction history of the reference signal of the i-th frame is B i After steps 2 and 3, the low-rank matrix L i and sparse matrices
[0097] First, S i S1 is first vectorized to obtain a column vector vec(S i ) and vec(S1):
[0098]
[0099] Next, calculate vec(S i ) and vec(S1) are:
[0100]
[0101] Among them, s i ,s 1 Corresponding to vec(S i ) and elements of vec(S1).
[0102] Similarly, calculate the matrix L i The sum of the sparse matrix S i The joint conditional entropy of is as follows:
[0103]
[0104] Joint Conditional Entropy The calculation expression is as follows:
[0105]
[0106] Among them, l i is the vector vec(L i ), vec(Li ) represents the matrix L i The converted columns are n1×n2 dimensional and the rows are 1-dimensional column vectors.
[0107] Then, H(vec(S i )|vec(S1)) and H(vec(S i )|vec(L i )) is converted back to a matrix of size n1*n2, defined as H(X i |X1), H(X i |L i ).
[0108] Step 5: Calculate H(S i |S1) and H(S i |L i ) H , as well as L1 and the matrix L i The Euclidean distance of L , the expression is as follows:
[0109]
[0110] Step 6: Calculate E H and The Pearson correlation coefficient of the target bright spot under the condition of dense interference sources is obtained:
[0111]
[0112] The present invention aims to solve the problem of detecting small mobile obstacles under the cover of dense underwater interference objects near the port. The present invention combines variational mode decomposition (VMD) and joint conditional entropy matching to optimize the low-rank matrix decomposition (LMD) method, thereby overcoming the problem that LMD requires multiple frames of continuous calculation, resulting in excessive costs, and achieves rapid suppression of single-frame reverberation and dynamic background, and the target detection effect is better than the conventional LMD algorithm. The average calculation time of the present invention is within 12 to 14 seconds, which is more than 5 times the calculation efficiency of the continuous multi-frame LMF algorithm of more than 1 minute.
[0113] Table 1 Comparison of the average computation time of single-frame JCESM and multi-frame LMD Unit: seconds
[0114] method Frame number 1 Frame rate 4 Frame rate 6 Frame rate 8 Frame rate: 10 JCESM 12.69 \ \ \ \ LMD \ 65.44 106.85 167.53 197.2
[0115] Table 1 shows the comparison results between the present invention and the conventional LMD algorithm. To achieve the reverberation suppression effect, LMD needs to associate at least 4 frames of matrices. The calculation time is much higher than that of the single-frame JCESM algorithm. Moreover, as the number of associated frames increases, the calculation cost continues to increase. Figure 6By comparison, it is obvious that the JCESM algorithm used in this patent can obtain a complete target movement trajectory, while the conventional LMD algorithm is difficult to find the target bright spot and requires further processing to detect the target, and the calculation time is much higher than that of the present invention.
[0116] The present invention has achieved obvious implementation effects in typical embodiments. Figure 6 The continuous target trajectory tail can be clearly detected in the red box of (a), while the LMD method cannot obtain obvious target echo highlights. Compared with the prior art (LMD associated with 4 frames usually takes more than 1 minute to calculate), the JCESM algorithm proposed in the present invention only takes 5.4 seconds to calculate a single frame, and its advantage is that it can quickly locate targets shielded by dense interference sources.
Claims
1. A method for detecting small-volume moving obstacles based on entropy under conditions of dense underwater interference sources, characterized in that Here are the steps: Step 1: Perform beamforming on the active echo signal to obtain the time-position history of the i-th frame echo Step 2: Use the LMD algorithm to perform low-rank sparse decomposition on the beam data of each frame of the time-azimuth history matrix to obtain a low-rank matrix L i and the sparse matrix S i ; B i =L i +S i ; Step 3: Use the alternating direction multi-multiplier method to optimize the low-rank sparse decomposition of the matrix; Through steps 2 to 3, a low-rank matrix L1 and a sparse matrix S1 whose time-direction history of the first frame reference signal is B1 are obtained; The time-direction history of the i-th frame reference signal is obtained as B i The low-rank matrix L i and sparse matrices Step 4: S i and S1 are vectorized first, and vec(S i ) and vec(S1) joint conditional entropy For L i and S i First perform vectorization and calculate the matrix L i The sum of the sparse matrix S i The joint conditional entropy H(vec(S i )|vec(S1)) and H(vec(S i )|vec(L i )) is converted back to a matrix of size n1*n2, defined as H(X i |X1), H(X i |L i ); Step 5: Calculate H(S i |S1) and H(S i |L i ) H , as well as L1 and the matrix L i The Euclidean distance of L ; Step 6: Calculate Ε H and The Pearson correlation coefficient of the target bright spot under the condition of dense interference sources is obtained:
2. According to claim 1, the entropy-based small-volume moving obstacle detection method under the condition of dense underwater interference sources is characterized by: The time-position history of the i-th frame echo The expression is a matrix of size n1×n2, where n1 represents the dimension in the time direction and n2 represents the dimension in the azimuth direction. The expression is as follows: in: Corresponding to the orientation matrix B i The value at the n1th moment and the n2th angle.
3. According to claim 1, the entropy-based small-volume moving obstacle detection method under the condition of dense underwater interference sources is characterized by: The decomposition process of step 2 is robust principal component analysis (RPCA), and its constraints follow: Among them, ||L i || * Calculate the matrix L i The sum of the singular values is the nuclear norm; ||S i ||1 is used to calculate the matrix S i The sum of the absolute values of all elements is the L1 norm; λ represents the balance parameter between the low-rank matrix and the sparse matrix; since the matrix L i With low-rank characteristics, the low-rank matrix L is i Relax the representation and use the L1 norm to sparse matrix S i Perform a relaxation representation.
4. According to claim 1, the entropy-based small-volume moving obstacle detection method under the condition of dense underwater interference sources is characterized by: The optimization solution of step 3 using the alternating direction multi-multiplier method for low-rank sparse decomposition of the matrix is: Definition||L i || * and ||S i ||1's matrix variable is f(L i ) and g(S i ), and the constraints are expressed as: The augmented Lagrangian multi-multiplier expression of the above formula is expressed as: Among them, Y i is the dual variable, ρ is the penalty factor, and k is the number of iterations of the optimization calculation; In the alternating variable solution process, the matrix corresponding to the kth iteration is as follows: in, Formula L i,k Japanese Style i,k The solution is obtained by singular value soft threshold operator and soft threshold operator.
5. According to claim 4, the entropy-based small-volume moving obstacle detection method under the condition of dense underwater interference sources is characterized by: The penalty factor ρ is set as a variable, and the expression is as follows: Yes i,k =||S i.k -S i,k-1 || F / ||S i.k || F THE i,k =||L i.k -L i,k-1 || F / ||L i.k || F Among them, τ1, τ2 and μ are small constants, Le i,k and Se i,k Respectively represent the iterative residuals of the k-th low-rank matrix and sparse matrix and the k-1-th low-rank matrix and sparse matrix; the smaller the residual, the closer the matrix decomposition result is to the convergence threshold; ε, Se i,k ,Le i,k <ε.
6. According to claim 5, the entropy-based small-volume moving obstacle detection method under the condition of dense underwater interference sources is characterized by: The value range of τ1, τ2 and μ is usually between 0 and 10.
7. According to claim 1, the entropy-based small-volume moving obstacle detection method under the condition of dense underwater interference sources is characterized by: The calculation sparse matrix S i and the joint conditional entropy of the sparse matrix S1 First, S i S1 is first vectorized to obtain a column vector vec(S i ) and vec(S1): Calculate vec(S i ) and vec(S1) are: Among them, s i ,s 1 Corresponding to vec(S i ) and elements of vec(S1).
8. According to claim 1, the entropy-based small-volume moving obstacle detection method under the condition of dense underwater interference sources is characterized by: The calculation matrix L i and the sparse matrix S i The joint conditional entropy First, L i and S i First, vectorization is performed to obtain a column vector vec(L) with n1×n2 dimensions and 1 dimension. i ) and vec(S1): Joint Conditional Entropy The calculation expression is as follows: Among them, l i is the vector vec(L i ), vec(L i ) represents the matrix L i The converted columns are n1×n2 dimensional and the rows are 1-dimensional column vectors.
9. According to claim 1, the entropy-based small-volume moving obstacle detection method under the condition of dense underwater interference sources is characterized by: The Euclidean distance expression of step 5 is as follows:
10. The entropy-based small-volume moving obstacle detection method under the condition of dense underwater interference sources according to claim 1, characterized in that: The step 1 performs beam processing on the active echo signal using a conventional beamforming algorithm, CBF, to obtain a time-direction history.
Citation Information
Patent Citations
Phi-OTDR underwater acoustic signal processing method and device based on adaptive VMD
CN114077854A
Low-rank matrix sparsity compensation method for improving reverberation suppression robustness
CN114818773A
Complex electromagnetic interference signal decomposition method with high time-frequency aggregation degree
CN115994300A
Sonar image small target detection method based on matrix decomposition
CN116403100A
Multi-frame joint detection method, system and terminal for moving target under deep sea reverberation background
CN118151157A
Cited By
Underwater active sonar reverberation suppression method and device
CN122260291A
A method and device for reverberation suppression of an active sonar underwater
CN122260291B
