A normalization method for active sonar imaging data

Abstract

The present invention relates to the field of underwater acoustic information processing technology, and provides a method for normalizing active sonar imaging data, comprising: S1, constructing an azimuth-distance matrix: obtaining azimuth-distance data through a sonar system and storing it in a polar coordinate system; S2, smoothing the data: after constructing the azimuth-distance matrix, smoothing each row of data to reduce the impact of noise; S3, calculating expected echo intensity: after completing the smoothing process, calculating the expected echo intensity of each detection point using a data-driven method; normalizing each row of data according to the corresponding expected echo intensity; S4, coordinate system conversion: after the normalization process, converting the processed azimuth-distance matrix data from a polar coordinate system to a rectangular coordinate system. The present invention can reduce the interference of background reverberation and noise on the detection of small targets at long distances.

Description

Technical Field

[0001] The present invention relates to the technical field of underwater acoustic information processing, and in particular to a normalization method for active sonar imaging data. Background Art

[0002] Ensuring the security of ports and critical waters is crucial in coastal defense systems. In recent years, with the continuous evolution of underwater threats, detection tasks in shallow waters have become increasingly challenging. Typical threat targets include underwater intrusion devices such as small submersibles and divers. These targets are highly concealed and maneuverable, posing a serious threat to ports, ships, and critical underwater facilities. Active sonar is the primary means of detecting these underwater targets. Utilizing the principles of sound wave propagation, sonar systems transmit sound waves and receive echoes to determine the target's location and characteristics. However, in practical applications, sonar detection often faces complex background noise and reverberation issues. Reverberation refers to the reflection and scattering of active sonar signals by various scatterers in the ocean environment. These scattered signals are superimposed on the echoes, creating strong background noise and significantly interfering with target echo detection. In shallow waters, strong seabed scattering is the primary source of reverberation. This is especially true in areas with irregular seafloor formations or dense sediments. Seabed reverberation can be highly similar to target echo signals, further complicating target detection. Furthermore, echo intensity typically varies significantly with the target's distance from the sonar system. Typically, the closer a target is to a sonar system, the stronger its echo intensity. However, reverberation intensity also exhibits a similar decreasing trend with distance, making it difficult to distinguish between target and reverberation. Directly using raw echo intensity to generate sonar images fails to accurately reflect the effects of reverberation, making it difficult for subsequent target detection algorithms to set appropriate detection thresholds. Therefore, effectively normalizing the effect of reverberation's irregular distribution with distance and angle on imaging quality and detection results is a pressing issue in the sonar detection field. Summary of the Invention

[0003] The present invention provides a normalization method for active sonar imaging data, which can reduce the interference of background reverberation and noise on the detection of small targets at long distances.

[0004] A method for normalizing active sonar imaging data according to the present invention comprises the following steps:

[0005] S1. Construction of bearing-distance matrix: Obtain bearing-distance data through the sonar system and store it in the polar coordinate system;

[0006] S2, data smoothing: After constructing the azimuth distance matrix, each row of data is smoothed to reduce the impact of noise;

[0007] S3. Calculate expected echo intensity: After smoothing, calculate the expected echo intensity of each detection point using a data-driven method; normalize each row of data according to the corresponding expected echo intensity;

[0008] S4. Coordinate system conversion: After normalization, the processed azimuth distance matrix data is converted from the polar coordinate system to the rectangular coordinate system.

[0009] Preferably, in S1, the rows of the azimuth distance matrix represent different distance dimensions, and the columns represent different azimuth angles. Each element (i, j) represents the echo intensity at the i-th distance and the j-th angle. The echo intensity is obtained by beamforming and matched filtering. The azimuth distance matrix is ​​the basis for normalization processing.

[0010] Preferably, in S2, the smoothing method includes a weighted average method of K adjacent rows, and the weights are a rectangular window, a Hann window or a Hamming window.

[0011] Preferably, in S2, specifically:

[0012] Smoothing is performed by distance to reduce the noise of the distance sonar data. For each row of the azimuth distance matrix M, the adjacent K rows are selected and weighted averaged. The matrix after row smoothing is recorded as M'. The calculation process is expressed as:

[0013]

[0014] in[·] i,: represents the i-th row of the matrix, [·] j,: represents the jth row of the matrix, a j Represents a weighted value.

[0015] As a preference, in S3, if only the distance factor is considered, specifically:

[0016] A) Calculate the expected echo intensity at the same distance; calculate the average value of each row of M' to obtain the expected echo intensity at the distance The calculation process is expressed as:

[0017]

[0018] Where Mean means calculating the average value of the elements in the vector;

[0019] B) Normalize the data using the expected echo intensity at each distance, and the normalized matrix is It is calculated as:

[0020]

[0021] Where δ is a small positive number to avoid the denominator being 0 during normalization;

[0022] If distance and orientation factors are taken into consideration, the specific factors are:

[0023] a) Divide the matrix M' into N blocks by column, each block has L columns, and calculate the expected echo intensity at the same distance in each block; calculate the average value of each row of M' by block to obtain the expected echo intensity in the angle area under the distance The calculation process is expressed as:

[0024]

[0025] in[·] i,j-(L-1) / 2:j+(L-1) / 2 Represents the elements from j-(L-1) / 2 to j+(L-1) / 2 columns in the i-th row of the matrix;

[0026] b) Normalize the data using the expected echo intensity at each distance, and the normalized matrix is It is calculated as:

[0027]

[0028] Preferably, when dividing the sonar imaging matrix into non-overlapping blocks, if discontinuity occurs at the block edges, the following method is used for calculation:

[0029] 1) Construct the circulant matrix M':

[0030] M”=[[M'] :,Q-S+1:Q ,M',[M'] :,1:S ]

[0031] in[·] :,Q-S+1:Q represents the elements of all rows from the Q-S+1th column to the Qth column of the matrix, where S is a positive integer not less than 1 / 2 of the filter length; [·] :,1:S Represents the elements of all rows from the 1st column to the Sth column of the matrix;

[0032] 2) Use an FIR filter of length 2S to filter each row of M”;

[0033] 3) Take the middle Q columns of the M' matrix and record them as matrix N, that is:

[0034] N=[M] :,S+1:S+Q

[0035] [·] :,S+1:S+Q Represents the elements of all rows from the S+1th column to the S+Qth column of the matrix. This step is used to remove the edge effect when filtering the matrix row by row;

[0036] 4) Use matrix N to normalize matrix M', that is:

[0037]

[0038] where ⊙ denotes element-wise division of matrices.

[0039] The beneficial effects of the present invention are as follows:

[0040] This invention uses normalization to compare echo data from different distances and angles against the same baseline, thereby reducing the irregular distribution of reverberation in the image and improving target detectability, especially in complex shallow-water environments. This invention can simplify parameter settings in subsequent target detection tasks, reduce the complexity of manual debugging, and improve the overall detection performance of the sonar system.

[0041] The present invention further refines the normalization process based on the differences in reverberation across different angular regions, using block normalization, edge filtering, and other techniques to eliminate discontinuities at the edges of the blocks and improve the accuracy of the normalization process. Furthermore, the present invention proposes smoothing continuous sonar data frames in the time dimension to further improve the accuracy of estimating the expected echo intensity, thereby enhancing the stability of the entire system in complex environments. Through the above-mentioned method, the present invention significantly improves the quality of active sonar imaging data and reduces the interference of reverberation on detection results, making it particularly suitable for long-range small target detection tasks in complex shallow-water environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 4 is a flow chart of a method for normalizing active sonar imaging data in an embodiment. DETAILED DESCRIPTION

[0043] In order to further understand the content of the present invention, the present invention is described in detail with reference to the accompanying drawings and embodiments. It should be understood that the embodiments are merely for explaining the present invention and are not intended to limit the present invention.

[0044] Example

[0045] like Figure 1 As shown, this embodiment provides a normalization method for active sonar imaging data, which includes the following steps:

[0046] S1. Construction of bearing-distance matrix: Obtain bearing-distance data through the sonar system and store it in the polar coordinate system;

[0047] In S1, the rows of the azimuth distance matrix represent different distance dimensions, and the columns represent different azimuth angles. Each element (i, j) represents the echo intensity at the i-th distance and the j-th angle. The echo intensity is obtained by beamforming and matched filtering. The azimuth distance matrix is ​​the basis for normalization processing.

[0048] This embodiment uses a sonar system with a circular receiving array. After the transmitting element emits a linear frequency modulation signal, beamforming is used to collect reflected signals from different angles and perform correlation matching. For each set receiving angle, a vector of length P is formed based on the echo delay. The above process is repeated for Q angles as shown in 102, forming a P×Q azimuth-range sonar imaging matrix as shown in 103. If the above operation is repeated continuously, a series of azimuth-range sonar imaging matrices at different times are formed, i.e., a frame of data. The azimuth-range matrix at a specific time is denoted as M.

[0049] S2. Data smoothing: After constructing the azimuth-distance matrix, each row of data (i.e., echo intensities at different distances under the same azimuth) is smoothed to reduce the impact of noise.

[0050] In S2, the smoothing method involves weighted averaging of K adjacent rows, with weights applied using rectangular, Hann, or Hamming windows. This smoothing process effectively reduces high-frequency noise in the distance dimension, resulting in smoother background echo data. Due to the slowly varying nature of the reverberant background, smoothing is not limited to the distance dimension; it can also be performed in the time dimension, further eliminating temporal fluctuations.

[0051] In S2, specifically:

[0052] Smoothing is performed by distance to reduce the noise of the distance sonar data. For each row of the azimuth distance matrix M, the adjacent K rows are selected and weighted averaged. The weights can be constructed using common window functions (such as rectangular window, Hann window, Hamming window). The matrix after row smoothing is denoted as M', and its calculation process is expressed as:

[0053]

[0054] in[·] i,: represents the i-th row of the matrix, [·] j,: represents the jth row of the matrix, a j Represents a weighted value.

[0055] S3. Calculate expected echo intensity: After smoothing, calculate the expected echo intensity of each detection point using a data-driven method; normalize each row of data (i.e., echo intensity at different distances in the same direction) according to the corresponding expected echo intensity;

[0056] The expected echo intensity is primarily related to the distance from the detection point and can therefore be estimated by calculating the average echo intensity at each distance. However, the reverberation background may be unevenly distributed across angles, so when calculating the expected echo intensity, a more refined calculation can be performed based on the characteristics of the angular sector. The purpose of this step is to estimate the background echo intensity at different distances and angles through statistical data, thereby providing a baseline value for normalization.

[0057] The goal of normalization is to reduce the unevenness caused by echo enhancement at close range or attenuation at long range, thereby balancing the background reverberation in the sonar image and highlighting any abnormal echo intensities associated with the target. To avoid zero denominators during normalization, a small positive number, \delta, is introduced into the calculation to ensure the stability and reliability of the normalization process.

[0058] In S3, if only the distance factor is considered, it is as follows:

[0059] A) Calculate the expected echo intensity at the same distance; since each row of the matrix M' represents the echo reflection intensity at the same distance and different directions, the average value of each row of M' is calculated to obtain the expected echo intensity at the distance The calculation process is expressed as:

[0060]

[0061] Where Mean means calculating the average value of the elements in the vector;

[0062] B) Normalize the data using the expected echo intensity at each distance, and the normalized matrix is It is calculated as:

[0063]

[0064] Where δ is a small positive number to avoid the denominator being 0 during normalization;

[0065] Considering the characteristics of the sonar deployment environment, the reverberation may have very large differences in different angle areas. In this case, normalization needs to consider not only the distance factor, but also the azimuth factor. If distance and azimuth factors are considered, the specific results are:

[0066] a) Divide the matrix M' into N blocks by column, with each block having L columns (assuming L is an odd number without loss of generality), and calculate the expected echo intensity at the same distance in each block; since each row of the matrix M' represents the echo reflection intensity at different azimuths at the same distance, and each column represents different echo angles, the expected echo intensity within the angle region at the same distance is obtained by calculating the average value of each row of M' by block. The calculation process is expressed as:

[0067]

[0068] in[·] i,j-(L-1) / 2:j+(L-1) / 2 Represents the elements from j-(L-1) / 2 to j+(L-1) / 2 columns in the i-th row of the matrix;

[0069] b) Normalize the data using the expected echo intensity at each distance, and the normalized matrix is It is calculated as:

[0070]

[0071] When dividing the sonar imaging matrix into non-overlapping blocks, if discontinuity occurs at the block edges, the following calculation method is used:

[0072] 1) Construct the circulant matrix M':

[0073] M”=[[M'] :,Q-S+1:Q, M',[M'] :,1:S ]

[0074] in[·] :,Q-S+1:Q represents the elements of all rows from the Q-S+1th column to the Qth column of the matrix, where S is a positive integer not less than 1 / 2 of the filter length; [·] :,1:S Represents the elements of all rows from the 1st column to the Sth column of the matrix;

[0075] 2) Use an FIR filter of length 2S to filter each row of M". The value of 2S should not be much larger than the number of pixels occupied by the possible target in terms of angle. The FIR filter can use a rectangular window or a Hann window.

[0076] 3) Take the middle Q columns of the M' matrix and record them as matrix N, that is:

[0077] N=[M] :,S+1:S+Q

[0078] [·] :,S+1:S+Q Represents the elements of all rows from the S+1th column to the S+Qth column of the matrix. This step is used to remove the edge effect when filtering the matrix row by row;

[0079] 4) Use matrix N to normalize matrix M', that is:

[0080]

[0081] where ⊙ denotes element-wise division of matrices.

[0082] Considering the continuous acquisition of data frames, the time frame obtained at time t is M t,The estimation of the expected echo strength can also be further enhanced by time smoothing.

[0083] S4. Coordinate system conversion: After normalization, the processed azimuth distance matrix data is converted from polar coordinates to rectangular coordinates. The rectangular coordinate system is more convenient for subsequent data analysis and visualization, especially when fusion analysis with other types of sensor data is required.

[0084] This embodiment uses normalization to compare echo data at different distances and angles against the same baseline, thereby reducing the irregular distribution of reverberation in the image and improving target detectability, especially in complex shallow water environments. This embodiment can simplify parameter settings in subsequent target detection tasks, reduce the complexity of manual debugging, and improve the overall detection performance of the sonar system.

[0085] The above is a schematic description of the present invention and its embodiments, which is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. Therefore, if a person skilled in the art is inspired by this and, without departing from the purpose of the present invention, designs a structure and embodiment similar to this technical solution without inventiveness, they shall fall within the scope of protection of the present invention.

Claims

1. A normalization method for active sonar imaging data, characterized by: The following steps are involved: S1. Construction of bearing-distance matrix: Obtain bearing-distance data through the sonar system and store it in the polar coordinate system; S2, data smoothing: After constructing the azimuth distance matrix, each row of data is smoothed to reduce the impact of noise; S3. Calculate expected echo intensity: After smoothing, calculate the expected echo intensity of each detection point using a data-driven method; normalize each row of data according to the corresponding expected echo intensity; If distance and orientation factors are taken into consideration, the specific factors are: a) Divide the azimuth distance matrix M′ into N blocks by column, with L columns in each block. Calculate the expected echo intensity at the same distance in each block; calculate the average value of each row of M′ by block to obtain the expected echo intensity in the angle region under the distance. The calculation process is expressed as: in[·] i,j-(L-1) / 2:j+(L-1) / 2 Represents the elements from j-(L-1) / 2 to j+(L-1) / 2 columns in the i-th row of the matrix; b) Normalize the data using the expected echo intensity at each distance, and the normalized matrix is It is calculated as: S4, coordinate system conversion: after normalization, the processed azimuth distance matrix data is converted from the polar coordinate system to the rectangular coordinate system; When dividing the sonar imaging matrix into non-overlapping blocks, if discontinuity occurs at the block edges, the following calculation method is used: 1) Construct the circulant matrix M': M”=[[M′] :,Q-S+1:Q ,M′,[M′] :,1:S ] in[·] :,Q-S+1:Q represents the elements of all rows from the Q-S+1th column to the Qth column of the matrix, where S is a positive integer not less than 1 / 2 of the filter length; [·] :,1:S Represents the elements of all rows from the 1st column to the Sth column of the matrix; 2) Use an FIR filter of length 2S to filter each row of M”; 3) Take the middle Q columns of the M' matrix and record them as matrix N, that is: N=[M] :,S+1:S+Q [·] :,S+1:S+Q Represents the elements of all rows from the S+1th column to the S+Qth column of the matrix. This step is used to remove the edge effect when filtering the matrix row by row; 4) Use matrix N to normalize matrix M': where ⊙ denotes element-wise division of matrices.

2. The method for normalizing active sonar imaging data according to claim 1, wherein: In S1, the rows of the azimuth distance matrix represent different distance dimensions, and the columns represent different azimuth angles. Each element (i, j) represents the echo intensity at the i-th distance and the j-th angle. The echo intensity is obtained by beamforming and matched filtering. The azimuth distance matrix is ​​the basis for normalization processing.

3. The method for normalizing active sonar imaging data according to claim 2, wherein: In S2, the smoothing method includes the weighted average method of K adjacent rows, and the weights are rectangular window, Hann window or Hamming window.

4. The method for normalizing active sonar imaging data according to claim 3, wherein: In S2, specifically: Smoothing is performed by distance to reduce the noise of the distance sonar data. For each row of the azimuth distance matrix M, the adjacent K rows are selected and weighted averaged. The matrix after row smoothing is recorded as M'. The calculation process is expressed as follows: in[·] i,: represents the i-th row of the matrix, [·] j,: represents the jth row of the matrix, a j Represents a weighted value.

Images