A passive sonar data filtering method based on the sliding window method

Abstract

The present invention discloses a passive sonar data filtering method based on a sliding window method, which is applied to a passive sonar multi-static detection and positioning system. The single-base passive detection data obtained is analyzed and processed in real time to form single-base target azimuth data with high target azimuth accuracy, few outliers, and clear batch relationships of measurement values, thereby improving the accuracy of multi-static passive positioning and the real-time positioning performance of the detection system.

Description

Technical Field

[0001] The present invention belongs to the technical field of passive sonar data filtering, and is mainly used for filtering the data of underwater target detection by a passive line array sonar. Background Art

[0002] Passive sonar relies on the noise generated by target engines, motors, propellers, etc. as the sound source for detection, which is beneficial to maintaining its own concealment, and the target position can be estimated by the method of multi-base detection and cross-location. However, due to the influence of complex underwater acoustic channels and marine environments, passive detection using target radiated noise as the sound source often faces high false alarm situations, there are many outliers in the measured azimuth, the target azimuth angle resolution is low, and passive detection only focuses on obtaining the target azimuth. Especially in the case of multiple targets, the batch relationship of the measured values at different times cannot be determined, and it is necessary to analyze and judge the results of a single detection array to perform cross-location. Summary of the Invention

[0003] In view of the disadvantages of high false alarm rate, many outliers, low angle resolution and the inability to determine the batch relationship of measured values in passive sonar detection data, and the high time complexity and poor real-time performance of traditional filtering methods, the inventor has carried out research and improvement. The present invention aims to study the single-array detection data of passive line array sonars, and explore how to interpolate, associate, interpolate (predict), and filter the data with low target azimuth accuracy, many outliers, and unclear batch relationship of measured values, so as to effectively improve the resolution of single-array detection targets. The algorithms involved in the present invention include: K-nearest neighbor data association algorithm, curve fitting algorithm based on a sliding window, etc.

[0004] The present invention creatively proposes a passive sonar data filtering method based on the sliding window method, which is applied to a passive sonar multi-base detection and positioning system, and performs real-time analysis and processing on the obtained single-array passive detection data to form single-array target azimuth data with high target azimuth accuracy, few outliers, and clear batch relationship of measured values, thereby improving the accuracy of multi-base passive positioning and the real-time positioning performance of the detection system. Specifically, the present invention is implemented as follows:

[0005] Step S1: Initialize the algorithm parameters, set the sliding window length l, set the angle association test threshold e, and set the correct association test times N0; where 2 ≤ N0 ≤ l;

[0006] Step S2: Use the K-nearest neighbor algorithm to associate the azimuth angles of adjacent batches of measured values less than the association test threshold e to form an associated azimuth angle sequence g(i) with the number of associations greater than or equal to N0, where i is the number of successfully associated targets;

[0007] Step S3: Calculate the average value g(i) of each azimuth angle sequence;

[0008] Step S4: Move the sliding window backward by l lengths;

[0009] Step S5: Repeat Step S2 to obtain the azimuth angle sequence within the next sliding window.

[0010] Step S6: Repeat Step S3 to obtain the average value of the azimuth angle sequence within the next sliding window.

[0011] Step S7: Use the K-nearest neighbor association algorithm to associate adjacent azimuth angle sequences.

[0012] Step S8: Repeat Steps S4, S5, S6, and S7 until the average values of 5 groups of azimuth angle sequences are calculated.

[0013] Step S9: Use the curve fitting algorithm to fit the associated azimuth angle sequences to obtain the target azimuth change curve.

[0014] Step S10: Take the difference of the data according to the measurement batch time interval and perform a Kalman filter once to obtain the single-array target azimuth data for cross-location calculation.

[0015] Step S11: Repeat Steps S4, S5, S6, S7, S9, and S10 until the filtering is completed.

[0016] Further, in Step S1, the sliding window length l is set based on simultaneously processing the detection data of adjacent l batches; the angle association test threshold e is set based on the threshold of e degrees for associating angles using the k-nearest neighbor method; the correct association test count N0 is set based on selecting the angle sequence with the number of successful associations not less than N0 within the sliding window length l as the input for the next calculation.

[0017] Working principle and beneficial effects of the present invention: The method involved in the present invention is mainly applied to a multi-target passive sonar positioning and tracking system in water. The underwater targets include underwater vehicles, surface ships and other vehicles that can navigate underwater / surface and emit radiated noise. A passive sonar data filtering method based on the sliding window method is proposed. The correlation relationship of the target azimuth angle between batches of single-array measurements is obtained through the K-nearest neighbor algorithm. The outliers in the detection data are removed through thresholds and weights, and the target azimuth angle change curve is obtained through the sliding window curve fitting method, so as to obtain a target azimuth with higher angular resolution, effectively improving the target accuracy of multi-base cross-positioning, while reducing the spatial and time complexity of cross-positioning calculations, improving the calculation speed of subsequent target track association and fusion, and the real-time processing ability of the system, thus providing a technical basis for the real-time autonomous tracking and positioning of underwater unmanned passive sonar systems. This method successfully processes the target azimuth angle data of a single linear array of passive sonar through algorithms such as association, sliding window, curve fitting, interpolation and filtering, obtains a more accurate underwater target azimuth estimation, effectively removes outliers, reduces the time complexity of subsequent multi-base cross-positioning calculations, and improves the real-time positioning ability of passive sonar for multiple targets; forms single-array target azimuth data with high target azimuth accuracy, few outliers and clear batch relationships of measurement values, thereby improving the accuracy of multi-base passive positioning and the real-time positioning performance of the detection system. Description of the Drawings

[0018] Figure 1 It is a flow chart of a passive sonar data filtering method based on the sliding window method;

[0019] Figure 2 It is a schematic diagram of the window method in the embodiment of the present invention;

[0020] Wherein: the dot · represents the lateral data of a single array of passive sonar, a row l represents a batch of measurement data, and the solid line box represents the sliding window. Detailed Embodiments

[0021] To make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the specific embodiments and with reference to the accompanying drawings. It should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present invention. In addition, in the following description, the descriptions of well-known structures and technologies are omitted to avoid unnecessarily confusing the concepts of the present invention.

[0022] The specific implementation process of the present invention is as follows:

[0023] Step 1: Initialization of algorithm parameters. The sliding window length \(l\) represents processing adjacent \(l\) batches of detection data simultaneously, usually set to 5; the angular association test threshold \(e\) represents the threshold for associating angles using the \(k\)-nearest neighbor method as \(e\) degrees, usually set to 3; the correct association test count \(N_0\) represents selecting an angle sequence with no less than \(N_0\) successful associations within the sliding window length \(l\) as the input for the next calculation, usually set to 3.

[0024] Step 2: Use the \(K\)-nearest neighbor algorithm to associate the azimuth angles of adjacent batches of measurements that are less than the association test threshold \(e\) to obtain the following azimuth angle sequence:

[0025] \(g(i)=\{a\) i (1) \(a\) i (2) … \(a\) i (j)}, \(i = 1, 2, \ldots, n\); \(3\leq j\leq5\), where \(i\) is the target number with successful association, \(n\) is the number of successful associations, and \(j\) is the number of successful associations of the current target azimuth angle within the sliding window.

[0026] Step 3: Calculate the average value of each azimuth angle sequence

[0027] Step 4: Move the sliding window backward by \(l\) lengths;

[0028] Step 5: Repeat Step 2 to obtain the azimuth angle sequence within the next sliding window, \(g_2(i)=\{a\) i (1) \(a\) i (2) … \(a\) i (j)}, \(i = 1, 2, \ldots, n\); \(3\leq j\leq5\);

[0029] Step 6: Repeat Step 3 to obtain the average value of the azimuth angle sequence within the next sliding window,

[0030] Step 7: Use the \(K\)-nearest neighbor association algorithm to associate adjacent azimuth angle sequences:

[0031] Step 8: Repeat Steps 4, 5, 6, and 7 until the average values of 5 groups of azimuth angle sequences are calculated and associated to obtain the target azimuth angle sequence with successful association within the current sliding window;

[0032] Step 9: Use the 5-point 3rd-order curve fitting algorithm to fit the associated azimuth angle sequence to obtain the azimuth angle change curves of each target within the current sliding window,

[0033] Step 10: Interpolate the data according to the measurement batch time interval. According to the parameters set in Step 1, the number of interpolations for the first time is \(5\times l\), and \(l\) are inserted each time after the sliding window. As attached Figure 2As shown, before the first interpolation, the average values of 5 groups of azimuth series need to be calculated in step S8, so the number of interpolations is 5×l; after interpolation, in order to improve the data accuracy, a simple Kalman filter can be used to process the interpolated data, so as to obtain single-array target azimuth data with higher accuracy for cross-location calculation.

[0034] Step 11: Repeat steps 4, 5, 6, 7, 9, and 10 until the filtering is completed. After the 5th sliding window is performed, each sliding window completes an association detection, averaging, association with the data after historical interpolation filtering, interpolation, and filtering.

[0035] It should be understood that the above specific embodiments of the present invention are only used for exemplary illustration or explanation of the principle of the present invention, and do not constitute a limitation to the present invention. Therefore, any modifications, equivalent replacements, improvements, etc. made without departing from the spirit and scope of the present invention shall be included within the protection scope of the present invention. In addition, the appended claims of the present invention are intended to cover all changes and modification examples falling within the scope and boundaries of the appended claims, or equivalent forms of such scope and boundaries.

Claims

1. A passive sonar data filtering method based on the sliding window method, characterized in that, It includes the following steps: Step S1: Initialize the algorithm parameters, set the sliding window length \(l\), set the angular association test threshold \(e\), and set the correct association test times \(N_0\); where \(2\leq N_0\leq l\). Step S2: Use the K-nearest neighbor algorithm to associate the azimuth angles of adjacent batch measurements less than the association test threshold \(e\) to form an associated azimuth angle sequence \(g(i)\) with the number of associations greater than or equal to \(N_0\), where \(i\) is the number of successfully associated targets. Step S3: Calculate the average value of each azimuth angle sequence Step S4: Move the sliding window backward by \(l\) lengths. Step S5: Repeat Step S2 to obtain the azimuth angle sequence within the next sliding window. Step S6: Repeat Step S3 to obtain the average value of the azimuth angle sequence within the next sliding window. Step S7: Use the K-nearest neighbor association algorithm to associate two adjacent azimuth angle sequences. Step S8: Repeat Steps S4, S5, S6, and S7 until the average values of 5 groups of azimuth angle sequences are calculated. Step S9: Use the curve fitting algorithm to fit the associated azimuth angle sequence to obtain the target azimuth change curve. Step S10: Differentiate the data according to the measurement batch time interval and perform a Kalman filter once to obtain the single-base array target azimuth data for cross-location calculation. Step S11: Repeat Steps S4, S5, S6, S7, S9, and S10 until the filtering is completed.

2. The passive sonar data filtering method according to claim 1, wherein In Step S1, the sliding window length \(l\) is set based on simultaneously processing the detection data of adjacent \(l\) batches; the angular association test threshold \(e\) is set based on the threshold of \(e\) degrees for associating angles using the k-nearest neighbor method; the correct association test times \(N_0\) are set based on selecting an angle sequence with the number of successful associations not less than \(N_0\) within the sliding window length \(l\) as the input for the next calculation.

3. The passive sonar data filtering method according to claim 1, characterized in that In Step S2, the azimuth angle sequence is as follows: g(i) = {a i (1)a i (2)...a i (j)}, where i = 1, 2,..., n; 3 ≤ j ≤ 5, where i is the target number with successful association, n is the number of successful associations, and j is the number of successful associations of the current target azimuth angle within the sliding window.

4. The passive sonar data filtering method according to claim 1, wherein In Step S10, it includes: differentiating the data according to the measurement batch time interval. According to the parameters set in Step S1, the number of interpolations for the first time is \(5\times l\), and \(l\) are inserted each time after the sliding window; use a simple Kalman filter to process the interpolated data to obtain higher-precision single-base array target azimuth data for cross-location calculation.

Images