A method for screening echo signals in sea clutter regions

Abstract

This invention discloses a method for screening echo signals in sea clutter regions, comprising the following steps: Step 1, beam illumination area data extraction; Step 2, smoothing filtering and clutter-to-noise ratio estimation; Step 3, multi-domain feature estimation; Step 4, feature label vector generation; Step 5, comprehensive screening and region filling expansion. The screening method disclosed in this invention addresses the problem of screening clutter region signals from large amounts of sea clutter data under various parameter conditions. When the amplitude fluctuations between range gates are large and various local strong values ​​may be present, it can retain strong clutter signals such as sea spikes as much as possible for further in-depth research on sea clutter characteristics while eliminating non-clutter signals.

Description

Technical Field

[0001] This invention belongs to the field of sea clutter data preprocessing, and specifically relates to a method for rapid screening of echo signals in sea clutter regions. Background Technology

[0002] Before conducting a comprehensive and in-depth analysis of sea clutter characteristics, a large amount of real and valid sea clutter data under various parameter conditions is typically required, relying on the long-term accumulation of multiple data sources. The sudden increase in data volume naturally brings about a need for rapid and effective preprocessing. However, long-term sea clutter observations cannot always be conducted under completely ideal testing environments. After obtaining the actual observation data, many problems need to be solved or corrected, such as radar beam pointing and pattern correction, radar external calibration and radar system constant calculation, valid data identification, and outlier data removal. Achieving standardized, batch-automatic extraction of valid regions from large amounts of sea clutter data and signal screening within these regions is a crucial aspect of clutter data preprocessing. Currently, there is limited literature on rapid preprocessing of large amounts of sea clutter data. How to quickly and effectively complete echo signal screening and extract pure clutter regions is an important foundation for subsequent sea clutter characteristic analysis and can also provide a large number of reliable learning samples for big data-based sea clutter characteristic analysis.

[0003] The most direct and simple rapid screening method can generally distinguish strong interference or anomalous signals from clutter signals by different amplitude intensity. However, for high-resolution sea clutter data, the amplitude fluctuations between different sample units are large, especially under complex marine environmental parameters. For example, under high sea state conditions, clutter regions may exhibit strong sea spike echoes. It is difficult to completely distinguish sea clutter signals from other interference signals solely from the amplitude domain. Setting the screening threshold too low will cause too many false alarms, while setting it too high will reject strong clutter signals such as sea spikes. However, to improve the screening accuracy, introducing too many complex detection algorithms into the screening of a large amount of data in the preprocessing stage will greatly increase the computational load. Therefore, if the screening results of a sea clutter region can be obtained quickly and simultaneously by designing certain characteristic calculation criteria and combined operations while batch calculating commonly used clutter characteristic data such as amplitude, time correlation, and Doppler spectrum, then pure clutter regions can be extracted for further in-depth analysis of sea clutter characteristics in the later stages. Summary of the Invention

[0004] The technical problem to be solved by this invention is to address the need for reduced processing complexity and time cost in the preprocessing of large amounts of sea clutter data, and to provide a method for rapidly screening echo signals within sea clutter regions to provide pure clutter regions.

[0005] The present invention adopts the following technical solution:

[0006] An improved method for screening echo signals in sea clutter regions includes the following steps:

[0007] Step 1, data extraction of the beam illumination area:

[0008] First, the range gates corresponding to the near 3dB and far 3dB points of the beam in the range dimension are calculated based on the actual illumination angle of the radar beam center and the elevation beam width. Then, the starting and ending sampling range gate numbers corresponding to the beam elevation illumination area are converted by combining the difference between the sampled range gate and the actual range gate. At the same time, the range dimension interception range is obtained by avoiding the radar blind zone. The radar sea clutter data is read and the echo signal of this segment is intercepted.

[0009] Step 2, Smoothing Filtering and Spurious-to-Noise Ratio Estimation:

[0010] The median amplitude of the clutter data at each range gate in the sea clutter region captured in step 1 is taken. The median amplitude curve is then subjected to local weighted linear regression smoothing filter. The difference between the curves before and after smoothing is used to obtain the amplitude deviation curve that varies with the range gate, which is used as the amplitude characteristic quantity. At the same time, the median amplitude curve is used to compare the echo intensity in the noise region, estimate the clutter-to-noise ratio, and determine whether it meets the set minimum clutter-to-noise ratio.

[0011] Step 3, Multi-domain feature estimation:

[0012] Extract data that meets the set noise-to-noise ratio requirements, perform time correlation calculation and average Doppler spectrum estimation for each range gate to obtain the average correlation time, average spectral frequency shift, and spectral intensity for each range gate. Then, perform smoothing filtering and subtraction operations in step 2 on the curves of correlation time and spectral intensity as a function of the range gate to obtain the corresponding deviation curves as time correlation features and spectral intensity features. Subtract the frequency shift curve as a function of the range gate from the median of the curve to obtain the deviation curve as the frequency shift feature.

[0013] The average correlation time is the mean of the correlation times obtained from multiple short-time sample sequences for each range gate. The average spectral shift is the frequency shift value obtained by estimating the average Doppler spectrum of the sample sequence for each range gate using the Welch method and the centroid method. The spectral intensity is the relative spectral intensity at the peak of the average Doppler spectrum curve for each range gate.

[0014] Step 4, Feature label vector generation:

[0015] The amplitude, time-correlation, frequency-shift, and spectral intensity features of each distance gate obtained in steps 2 and 3 are processed as follows: The difference between the high and low quantile values ​​of the features is statistically analyzed. The screening threshold for each feature is set to 3 to 6 times the corresponding difference. Points greater than the threshold are marked as 1, and the rest as 0, resulting in initial label vectors for each feature. Then, short line segments with a length of 5 units are used as structuring elements, and morphological dilation is performed on each initial label vector to obtain the corrected amplitude label vector A. i Time-related label vector C i Frequency shift label vector F i and the spectral intensity label vector P i i is the distance gate number;

[0016] Step 5, Comprehensive screening and region filling expansion:

[0017] The feature label vectors obtained in step 4 are processed by combining and merging to obtain the merged label vector M. i M i =A i ·C i ||A i ·F i ||C i ·P i Then, the merged tag vector is subjected to hollow filling and boundary expansion processing to finally obtain the comprehensive screening result of the clutter region, which is the tag vector M. i The following formula is used for hollow filling and boundary expansion:

[0018]

[0019] B1 and B2 are binary structure elements, using short line segments. The symbol represents expansion, and ⊙ represents erosion. The size of structural element B1 is determined by M. i The number of gates with the longest continuous distance marked as 1 is set, and the size of struct B2 is half that of B1;

[0020] The range gate marked as 0 represents the pure clutter region, and the range gate marked as 1 represents the interference or other abnormal signal region. This completes the screening and labeling of pure clutter signals and other signals within the sea clutter region.

[0021] Furthermore, robust local regression processing is added to the locally weighted linear regression smoothing filter in step 2, where robust weight w is added to each data sample point within the smoothing interval. k The calculation is as follows:

[0022]

[0023] In the above formula, sk M represents the residual error resulting from the regression smoothing filter operation on the k-th data sample point. x M is the median of the absolute values ​​of the residual errors at each point within the smoothing interval. x =median(|s|).

[0024] The beneficial effects of this invention are:

[0025] The screening method disclosed in this invention addresses the problem of screening clutter region signals in large amounts of sea clutter data under various parameter conditions. When the amplitude fluctuations between range gates are large and various local strong values ​​may be mixed in, it can achieve the following improvement effects: On the one hand, it can retain strong clutter signals such as sea spikes as much as possible while eliminating non-clutter signals for in-depth research on sea clutter characteristics in the later stage; on the other hand, the multi-domain feature quantity adopted is designed based on the batch calculation of basic characteristic data in clutter preprocessing. Compared with the complex processing algorithms using other types of feature values, it does not increase the amount of additional calculation, which is very suitable for screening and judging large amounts of sea clutter data in the preprocessing stage. It can quickly obtain the screening result of a sea clutter region, provide a large amount of pure clutter region data, and effectively support the in-depth analysis of sea clutter characteristics in the later stage. Attached Figure Description

[0026] Figure 1 This is a schematic flowchart of the screening method of the present invention;

[0027] Figure 2(a) is the echo amplitude diagram of data sample X01 in the sea clutter region;

[0028] Figure 2(b) is the echo amplitude diagram of data sample X02 in the sea clutter region;

[0029] Figure 3(a) is a schematic diagram of the amplitude characteristics of data sample X01 in the sea clutter region;

[0030] Figure 3(b) is a schematic diagram of the amplitude characteristics of data sample X02 in the sea clutter region;

[0031] Figure 4(a) is a schematic diagram comparing the median amplitude curve of sea clutter for data sample X01 and the average intensity of noise signal under the same radar transmission parameters.

[0032] Figure 4(b) is a schematic diagram showing the median amplitude curve of sea clutter for data sample X02 and the average intensity of noise signal under the same radar transmission parameters.

[0033] Figure 5(a) is a schematic diagram of the time-related features, frequency shift features, and spectral intensity features obtained from the various distance gates of the data sample X01 in the sea clutter region.

[0034] Figure 5(b) is a schematic diagram of the time-correlation characteristics, frequency shift characteristics, and spectral intensity characteristics obtained from each distance gate of the data sample X02 in the sea clutter region;

[0035] Figure 6(a) is a schematic diagram of the feature label vectors corresponding to data sample X01 in the sea clutter region;

[0036] Figure 6(b) is a schematic diagram of the feature label vectors corresponding to data sample X02 in the sea clutter region;

[0037] Figure 7(a) is a schematic diagram of the signal comprehensive screening results corresponding to data sample X01 in the sea clutter region;

[0038] Figure 7(b) is a schematic diagram of the signal comprehensive screening results corresponding to data sample X02 in the sea clutter region. Detailed Implementation

[0039] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0040] Example 1, such as Figure 1 As shown, this embodiment discloses a method for screening echo signals in sea clutter regions. Taking high-resolution radar sea clutter data in a certain frequency band as an example, data sample X01 and data sample X02 are taken respectively, including the following steps:

[0041] Step 1, data extraction of the beam illumination area:

[0042] First, the range gates corresponding to the near 3dB and far 3dB points of the beam in the range dimension are calculated based on the actual illumination angle of the radar beam center and the elevation beam width. Then, the starting and ending sampling range gate numbers corresponding to the beam elevation illumination area are converted by combining the difference between the sampled range gate and the actual range gate. At the same time, the range dimension interception range is obtained by considering avoiding the radar blind zone. The radar sea clutter data is read and the echo signal of this segment is intercepted.

[0043] The absolute value of the echo data from the intercepted sea clutter region is taken to obtain the amplitude diagram as shown below. Figure 2(a) , 2(b) As shown, 2(a) and 2(b) correspond to data sample X01 and data sample X02, respectively.

[0044] Step 2, Smoothing Filtering and Spurious-to-Noise Ratio Estimation:

[0045] The median amplitude of the clutter data at each range gate in the sea clutter region captured in step 1 is taken. The median amplitude curve is then subjected to local weighted linear regression smoothing filter. The difference between the curves before and after smoothing is used to obtain the amplitude deviation curve that varies with the range gate, which is used as the amplitude characteristic quantity. At the same time, the median amplitude curve is used to compare the echo intensity in the noise region, estimate the clutter-to-noise ratio, and determine whether it meets the set minimum clutter-to-noise ratio.

[0046] In locally weighted linear regression smoothing filtering, robust local regression processing is added to reduce the distortion of the smoothing curve caused by deviation points. The robust weight w for each data sample point within the smoothing interval is... k The calculation is as follows:

[0047]

[0048] In the above formula, s k M represents the residual error resulting from the regression smoothing filter operation on the k-th data sample point. x M is the median of the absolute values ​​of the residual errors at each point within the smoothing interval. x =median(|s|).

[0049] Figure 3(a) and 3(b) The diagrams show the amplitude characteristics, i.e., the clutter amplitude deviation curves, corresponding to data samples X01 and X02, respectively. Figure 4(a) and 4(b) The clutter amplitude median curves and the average noise signal intensity under the same radar transmission parameters are shown for data samples X01 and X02, respectively. Assume a minimum required clutter-to-noise ratio of 10 dB, which is met in both regions.

[0050] Step 3, Multi-domain feature estimation:

[0051] Extract data that meets the set noise-to-noise ratio requirements, perform time correlation calculation and average Doppler spectrum estimation for each range gate to obtain the average correlation time, average spectral frequency shift, and spectral intensity for each range gate. Then, perform smoothing filtering and subtraction operations in step 2 on the curves of correlation time and spectral intensity as a function of the range gate to obtain the corresponding deviation curves as time correlation features and spectral intensity features. Subtract the frequency shift curve as a function of the range gate from the median of the curve to obtain the deviation curve as the frequency shift feature.

[0052] The average correlation time is the mean of the correlation times obtained from multiple short-time sample sequences for each range gate. The average spectral shift is the frequency shift value obtained by estimating the average Doppler spectrum of the sample sequence for each range gate using the Welch method and the centroid method. The spectral intensity is the relative spectral intensity at the peak of the average Doppler spectrum curve for each range gate.

[0053] Figure 5(a) and5(b) The time-related features, frequency shift features, and spectral intensity features are obtained for each distance gate corresponding to data samples X01 and X02, respectively.

[0054] Step 4, Feature label vector generation:

[0055] The amplitude, time-correlation, frequency-shift, and spectral intensity features of each distance gate obtained in steps 2 and 3 are processed as follows: The difference between the high and low quantile values ​​of the features is statistically analyzed. The screening threshold for each feature is set to 3 to 6 times the corresponding difference. Points greater than the threshold are marked as 1, and the rest as 0, resulting in initial label vectors for each feature. Then, short line segments with a length of 5 units are used as structuring elements, and morphological dilation is performed on each initial label vector to obtain the corrected amplitude label vector A. i Time-related label vector C i Frequency shift label vector F i and the spectral intensity label vector P i i is the distance gate number;

[0056] In this embodiment, the high quantile and low quantile values ​​are selected as 0.7 and 0.2, respectively. The multiples of the screening thresholds for each feature are set to 4, 6, 6, and 5, respectively. A short line segment of length 3 is selected as the structuring element for morphological dilation. Finally, the feature label vectors corresponding to data samples X01 and X02 are obtained as follows: Figure 6(a) and 6(b) As shown.

[0057] Step 5, Comprehensive screening and region filling expansion:

[0058] The feature label vectors obtained in step 4 are processed by combining and merging to obtain the merged label vector M. i M i =A i ·C i ||A i ·F i ||C i ·P i Then, the merged tag vector is subjected to hollow filling and boundary expansion processing to finally obtain the comprehensive screening result of the clutter region, which is the tag vector M′. i The following formula is used for hollow filling and boundary expansion:

[0059]

[0060] B1 and B2 are binary structure elements, using short line segments. The circle represents the expansion operation, and the circle represents the erosion operation, meaning that B1 is first applied to M. iThe process involves expansion, followed by erosion using B1, and finally expansion using B2. The size of the structuring element B1 is determined by M. i The number of gates with the longest continuous distance marked as 1 is set, and the size of struct B2 is half that of B1;

[0061] The range gate marked as 0 represents the pure clutter region, while the range gate marked as 1 represents the interference or other abnormal signal region. This completes the screening and labeling of pure clutter signals and other signals within the sea clutter region. Depending on the requirements, the data blocks corresponding to the range gates marked as 0 can be extracted separately, or different data blocks within the same group can be merged to obtain the pure clutter region.

[0062] Figure 7(a) , 7(b) The final screening results are for data samples X01 and X02. In Figure 7(a), the area marked with 1 corresponds to the non-clutter area, while the entire segment of Figure 7(b) is a pure clutter area.

[0063] By performing the above processing on a large amount of sea clutter data, regions marked as 0 can be extracted in batches, and regions marked as 1 can be removed to obtain pure clutter regions, thereby providing a large number of data samples for further in-depth research on clutter characteristics.

Claims

1. A method for screening echo signals in sea clutter regions, characterized in that, Includes the following steps: Step 1, data extraction of the beam illumination area: First, the range gates corresponding to the near 3dB and far 3dB points of the beam in the range dimension are calculated based on the actual illumination angle of the radar beam center and the elevation beam width. Then, the starting and ending sampling range gate numbers corresponding to the beam elevation illumination area are converted by combining the difference between the sampled range gate and the actual range gate. At the same time, the range dimension interception range is obtained by avoiding the radar blind zone. The radar sea clutter data is read and the echo signal of this segment is intercepted. Step 2, Smoothing Filtering and Spurious-to-Noise Ratio Estimation: The median amplitude of the clutter data at each range gate in the sea clutter region captured in step 1 is taken. The median amplitude curve is then subjected to local weighted linear regression smoothing filter. The difference between the curves before and after smoothing is used to obtain the amplitude deviation curve that varies with the range gate, which is used as the amplitude characteristic quantity. At the same time, the median amplitude curve is used to compare the echo intensity in the noise region, estimate the clutter-to-noise ratio, and determine whether it meets the set minimum clutter-to-noise ratio. Step 3, Multi-domain feature estimation: Extract data that meets the set noise-to-noise ratio requirements, perform time correlation calculation and average Doppler spectrum estimation for each range gate to obtain the average correlation time, average spectral frequency shift, and spectral intensity for each range gate. Then, perform smoothing filtering and subtraction operations in step 2 on the curves of correlation time and spectral intensity as a function of the range gate to obtain the corresponding deviation curves as time correlation features and spectral intensity features. Subtract the frequency shift curve as a function of the range gate from the median of the curve to obtain the deviation curve as the frequency shift feature. The average correlation time is the mean of the correlation times obtained from multiple short-time sample sequences for each range gate. The average spectral shift is the frequency shift value obtained by estimating the average Doppler spectrum of the sample sequence for each range gate using the Welch method and the centroid method. The spectral intensity is the relative spectral intensity at the peak of the average Doppler spectrum curve for each range gate. Step 4, Feature label vector generation: The amplitude, time-correlation, frequency-shift, and spectral intensity features of each distance gate obtained in steps 2 and 3 are processed as follows: The difference between the high and low quantile values ​​of the features is statistically analyzed. The screening threshold for each feature is set to 3 to 6 times the corresponding difference. Points greater than the threshold are marked as 1, and the rest as 0, resulting in initial label vectors for each feature. Then, short line segments with a length of 5 units are used as structuring elements, and morphological dilation is performed on each initial label vector to obtain the corrected amplitude label vector A. i Time-related label vector C i Frequency shift label vector F i and the spectral intensity label vector P i i is the distance gate number; Step 5, Comprehensive screening and region filling expansion: The feature label vectors obtained in step 4 are processed by combining and merging to obtain the merged label vector M. i M i =A i ·C i ||A i ·F i ||C i ·P i Then, the merged tag vector is subjected to hollow filling and boundary expansion processing to finally obtain the comprehensive screening result of the clutter region, which is the tag vector M′. i The following formula is used for hollow filling and boundary expansion: B1 and B2 are binary structure elements, using short line segments. The symbol represents expansion, and ⊙ represents erosion. The size of structural element B1 is determined by M. i The number of gates with the longest continuous distance marked as 1 is set, and the size of struct B2 is half that of B1; The range gate marked as 0 represents the pure clutter region, and the range gate marked as 1 represents the interference or other abnormal signal region. Thus, the screening and labeling of pure clutter signals and other signals within the sea clutter region is completed.

2. The method for screening echo signals in the sea clutter region according to claim 1, characterized in that: In step 2, robust local regression processing is added to the locally weighted linear regression smoothing filter, and the robust weight w for each data sample point within the smoothing interval is increased. k The calculation is as follows: In the above formula, s k M represents the residual error resulting from the regression smoothing filter operation on the k-th data sample point. x M is the median of the absolute values ​​of the residual errors at each point within the smoothing interval. x =median(|s|).

Images