A shipboard radar signal sorting method

Abstract

The application discloses a ship-borne radar signal sorting method, comprising the following steps: S01, time domain merging is performed on the detected PDW pulse in a single channel, and different PDW pulses are classified according to spatial position distribution by using unsupervised clustering analysis in a machine learning method according to the azimuth and the elevation of the detected PDW pulse; S02, unsupervised clustering analysis is performed on the results of multiple single-channel classification according to the azimuth and the elevation, time domain merging is performed on different PDW pulses of the same azimuth, and PDW pulses of different frequency ranges are classified by using unsupervised clustering analysis; S03, time domain analysis is performed on the PDW pulses classified according to the frequency ranges; and S04, the EDW result is calculated and generated according to the pulse-in-frequency analysis of the pulse-in modulation characteristics, the PDW pulse first arrival time, the signal power and the signal arrival angle.

Description

Technical Field

[0001] This invention relates to the field of reconnaissance signal processing. More specifically, it relates to a method for sorting shipborne radar signals. Background Technology

[0002] Radar source signal sorting technology is crucial for radar reconnaissance signal processing and a fundamental prerequisite for such processing. Only by correctly identifying and separating the pulses from different targets within a complex, high-density, and highly mixed electromagnetic signal can subsequent reconnaissance signal processing be completed. Therefore, the effectiveness of source signal sorting directly determines the outcome of reconnaissance signal processing and is key to electronic battlefield situational awareness. With the increasing types and numbers of radars, the electromagnetic environment at sea is becoming increasingly complex, posing growing challenges to electronic reconnaissance.

[0003] The increasing complexity of the electromagnetic environment at sea is mainly reflected in the rising pulse current density and the growing complexity of signal modulation. Under special conditions, pulse density can reach millions of data streams per second. Conventional single-frequency signals and fixed-cycle signals are becoming less common, while the proportion of complex modulation types such as linear frequency modulation, phase coding, and frequency agility is increasing.

[0004] Faced with increasingly complex changes in the sea surface electromagnetic environment, traditional signal sorting methods based on histogram statistics are insufficient to meet practical needs. Therefore, a method combining unsupervised clustering analysis from machine learning with PDW key parameter analysis can effectively and promptly solve the problem of sorting signals from complex radiation sources. Summary of the Invention

[0005] The purpose of this invention is to provide a method for sorting shipborne radar signals.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A method for sorting shipborne radar signals, characterized by comprising the following steps:

[0008] S01: For the detected PDW pulses, time-domain combining is performed in a single channel, and based on the azimuth and elevation angles of the detected PDW pulses, unsupervised clustering analysis in machine learning methods is used to classify different PDW pulses according to their spatial distribution.

[0009] S02: For the results of multiple single-channel classifications, unsupervised clustering analysis is performed based on azimuth and elevation angles. Different PDW pulses at the same azimuth angle are merged in the time domain, and unsupervised clustering analysis is used to classify PDW pulses in different frequency ranges.

[0010] S03: Perform time-domain analysis on the PDW pulses classified by frequency range. If the PRI type of the PDW pulse cannot be correctly identified and the pulse width PW has similar results, repeat step S02. If the PRI type of the PDW pulse can be correctly identified and the pulse width PW has similar results, proceed to step S04.

[0011] S04: Based on the pulse frequency analysis, the pulse modulation characteristics are analyzed, and the PDW pulse first arrival time, signal power, and signal arrival angle are calculated to generate the EDW result.

[0012] Preferably, in step S01, the PDW pulse time domain is merged into

[0013]

[0014] The arrival times and pulse widths of two adjacent pulses are respectively , , , The time-domain merging threshold is , This represents the pulse end time of the former.

[0015] Preferably, if the two pulses are merged, the end time of the former pulse is...

[0016] .

[0017] Preferably, the step of using unsupervised clustering analysis to perform clustering analysis on the azimuth and elevation angles of the detected signals in step S01 further includes: determining the clustering range of the clustering analysis based on the angle measurement accuracy. The data to be analyzed is initially prepared, and then analyzed using an improved version of the K-means method.

[0018]

[0019]

[0020] , It is clustering The mean of the samples contained in the sample represents the mean of the j-th cluster center; This represents the number of samples in the j-th cluster. Indicates sample The Euclidean distance from the j-th cluster center Error range of cluster analysis .

[0021] Preferably, the analysis of the pulse repetition period (PRI) characteristic parameters further includes: identifying the PRI pattern of the PDW pulse and calculating the arrival time sequence (TOA) of the PDW pulse sequence.

[0022]

[0023] Where b represents the intercept, k represents the slope, which is also the center value of the PRI sequence, and M represents the sequence length.

[0024] Preferably, the slope k is

[0025] .

[0026] Preferably, identifying the unevenness type of the PRI sequence involves subtracting the PRI sequence from the slope k and summing the results to calculate the average jitter of the PRI sequence. :

[0027] .

[0028] Preferably, the analysis of the PDW pulse repetition period PRI characteristic parameter further includes: determining and identifying the fixed PRI of the pulse; after the PRI sequence value of the pulse sequence is processed by cluster analysis, the fixed repetition frequency PRI, after harmonic removal, the result of cluster analysis is only one value.

[0029] Preferably, the analysis of the pulse repetition period (PRI) characteristic parameters further includes: identifying and judging slip PRI: performing a difference operation on the PRI sequence of length M-1 to obtain a difference sequence of length M-2. ,right Symbolization ,

[0030] .

[0031] Preferably, the determination and identification of slippage PRI analysis further includes: weighting the symbolic results to obtain the difference change value. This allows for the identification of the slip type PRI.

[0032] .

[0033] The beneficial effects of this invention are as follows:

[0034] This invention utilizes unsupervised clustering analysis in machine learning to analyze parameters such as angle and frequency, and combines it with parameters such as repetition frequency (PRI) and pulse width (PW) for joint analysis. This allows for the accurate sorting and identification of complex modulation types of radiation source signals, such as linear frequency modulation, frequency agility, and staggered repetition frequency, in complex electromagnetic environments. Attached Figure Description

[0035] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0036] Figure 1 A flowchart of the present invention is shown.

[0037] Figure 2 A schematic diagram of time-domain merging according to the present invention is shown.

[0038] Figure 3 The flowchart of the unsupervised clustering analysis of the present invention is shown. Detailed Implementation

[0039] To more clearly illustrate the present invention, the following description, in conjunction with preferred embodiments and accompanying drawings, further explains the invention. Similar components in the drawings are indicated by the same reference numerals. Those skilled in the art should understand that the specific description below is illustrative rather than restrictive and should not be construed as limiting the scope of protection of the present invention.

[0040] A method for sorting shipborne radar signals, characterized by comprising the following steps:

[0041] S01: For the detected PDW pulses, time-domain combining is performed in a single channel, and based on the azimuth and elevation angles of the detected PDW pulses, unsupervised clustering analysis in machine learning methods is used to classify different PDW pulses according to their spatial distribution.

[0042] S02: For the results of multiple single-channel classifications, unsupervised clustering analysis is performed based on azimuth and elevation angles. Different PDW pulses at the same azimuth angle are merged in the time domain, and unsupervised clustering analysis is used to classify PDW pulses in different frequency ranges.

[0043] S03: Perform time-domain analysis on the PDW pulses classified by frequency range. If the PRI type of the PDW pulse cannot be correctly identified and the pulse width PW has similar results, repeat step S02. If the PRI type of the PDW pulse can be correctly identified and the pulse width PW has similar results, proceed to step S04.

[0044] S04: Based on the pulse frequency analysis, the pulse modulation characteristics are analyzed, and the PDW pulse first arrival time, signal power, and signal arrival angle are calculated to generate the EDW result.

[0045] Preferably, in step S01, the PDW pulse time domain is merged into

[0046]

[0047] The arrival times and pulse widths of two adjacent pulses are respectively , , , The time-domain merging threshold is , This represents the pulse end time of the former.

[0048] Preferably, if the two pulses are merged, the end time of the former pulse is...

[0049] .

[0050] Preferably, the step of using unsupervised clustering analysis to perform clustering analysis on the azimuth and elevation angles of the detected signals in step S01 further includes: determining the clustering range of the clustering analysis based on the angle measurement accuracy. The data to be analyzed is initially prepared, and then analyzed using an improved version of the K-means method.

[0051]

[0052]

[0053] , It is clustering The mean of the samples contained in the sample represents the mean of the j-th cluster center; This represents the number of samples in the j-th cluster. Indicates sample The Euclidean distance from the j-th cluster center Error range of cluster analysis .

[0054] Preferably, the analysis of the pulse repetition period (PRI) characteristic parameters further includes: identifying the PRI pattern of the PDW pulse and calculating the arrival time sequence (TOA) of the PDW pulse sequence.

[0055]

[0056] Where b represents the intercept, k represents the slope, which is also the center value of the PRI sequence, and M represents the sequence length.

[0057] Preferably, the slope k is

[0058] .

[0059] Preferably, identifying the unevenness type of the PRI sequence involves subtracting the PRI sequence from the slope k and summing the results to calculate the average jitter of the PRI sequence. :

[0060] .

[0061] Preferably, the analysis of the PDW pulse repetition period PRI characteristic parameter further includes: determining and identifying the fixed PRI of the pulse; after the PRI sequence value of the pulse sequence is processed by cluster analysis, the fixed repetition frequency PRI, after harmonic removal, the result of cluster analysis is only one value.

[0062] Preferably, the analysis of the pulse repetition period (PRI) characteristic parameters further includes: identifying and judging slip PRI: performing a difference operation on the PRI sequence of length M-1 to obtain a difference sequence of length M-2. ,right Symbolization ,

[0063] .

[0064] Preferably, the determination and identification of slippage PRI analysis further includes: weighting the symbolic results to obtain the difference change value. This allows for the identification of the slip type PRI.

[0065] .

[0066] A specific example:

[0067] A method for sorting shipborne radar signals:

[0068] 1. Step (1) Single-channel space-time processing: Based on the PDW signal detected by the detection end, firstly perform time-domain merging and eliminate narrow pulses; then, based on the measured azimuth angle... Pitch angle Unsupervised clustering analysis, a machine learning method, is used to classify different radar pulses according to their spatial distribution.

[0069] Step (2) Multi-channel spatial and temporal processing: The classification results calculated in Step 1 are then processed again based on the azimuth angle. Pitch angle Cluster analysis was performed, and different pulses at the same azimuth angle were merged in the time domain.

[0070] Step (3) Frequency Domain Analysis and Processing: Based on the results of the second step, frequency analysis is performed. Unsupervised clustering analysis is used to classify the data in different frequency ranges;

[0071] Step (4) Based on the frequency analysis, perform time domain analysis and judgment, analyze characteristic parameters such as pulse repetition period PRI, and determine whether it is necessary to return to step 3 for re-analysis based on the analysis results, so as to achieve the correct analysis of frequency agile signals.

[0072] If step 4 fails to correctly identify the PRI type and the pulse width PW has a similar result, repeat step 3.

[0073] Step (5) Generate EDW results: Analyze the pulse modulation characteristics based on the pulse frequency, pulse first arrival time, signal power, signal arrival angle and other parameters, and calculate and generate EDW results.

[0074] Preferably, other parameters include angle, frequency, and pulse repetition period (PRI).

[0075] 2. The pulse time domain merging described in steps (1) and (2) is performed; unsupervised clustering analysis is used to perform clustering analysis on the azimuth and elevation angles of the detected signal.

[0076] 2.1(a) Merging of different pulses in the same orientation: Merging of time-domain pulses is required both within a single channel and between multiple channels to solve the problem of pulse cutting for large bandwidth signals in channelization detection. The specific process is shown in the figure.

[0077]

[0078] The arrival times and pulse widths of two adjacent pulses are respectively , , , Temporal merging threshold , This indicates the end time of the former pulse; if so, the two pulses are merged.

[0079]

[0080] 2.1(b) Determine the clustering range for cluster analysis based on the angular measurement accuracy. Then, the data that needs to be analyzed is preliminarily organized.

[0081] The processed data was then analyzed using an improved version of the K-means method:

[0082]

[0083]

[0084] , It is clustering The mean of the samples contained in the sample represents the mean of the j-th cluster center; This represents the number of samples contained in the j-th cluster. Indicates sample The Euclidean distance from the j-th cluster center The error range of cluster analysis: if the Euclidean distance is less than the error range, the sample data will be assigned to that class.

[0085] 3. The variation law of the pulse repetition period PRI described in step (4) specifically includes the following steps:

[0086] 3.1(a) Determining the pattern of pulse PRI: The arrival time series TOA of the pulse sequence is linearly fitted using the least squares method, i.e.

[0087]

[0088] Where b represents the intercept, k represents the slope, which is also the center value of the PRI sequence, and M represents the sequence length. The expression for the slope k is:

[0089]

[0090] The average jitter of the PRI sequences can be obtained by subtracting and summing the differences between all PRI sequences and k. It can identify PRI sequences with uneven patterns.

[0091]

[0092] 3.1(b) Identifying and determining the fixed PRI of the pulse: After the PRI sequence values ​​of the pulse sequence are processed by cluster analysis, the PRI of the fixed repetition frequency, after harmonic removal, will have only one value in the cluster analysis result, namely the fixed repetition frequency PRI.

[0093] 3.1(c) Identifying the slip transition PRI: Perform a difference operation on the PRI sequence of length M-1 to obtain a difference sequence of length M-2. ,right Symbolization ,

[0094]

[0095] Furthermore, the difference change value is obtained by weighting the symbolization results. It can identify the slip type PRI.

[0096]

[0097] 3.1(d) Through rigorous PRI parameter analysis and pulse width analysis, determine whether to return to step 3 to re-perform frequency analysis, thereby identifying the frequency cutoff radiation source signal.

[0098] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. For those skilled in the art, other variations or modifications can be made based on the above description. It is impossible to exhaustively list all the implementation methods here. All obvious variations or modifications derived from the technical solutions of the present invention are still within the protection scope of the present invention.

Claims

1. A method for sorting shipborne radar signals, characterized in that, Includes the following steps: S01: For the detected PDW pulses, time-domain combining is performed in a single channel, and based on the azimuth and elevation angles of the detected PDW pulses, unsupervised clustering analysis in machine learning methods is used to classify different PDW pulses according to their spatial distribution. S02: For the results of multiple single-channel classifications, unsupervised clustering analysis is performed based on azimuth and elevation angles. Different PDW pulses at the same azimuth angle are merged in the time domain, and unsupervised clustering analysis is used to classify PDW pulses in different frequency ranges. S03: Perform time-domain analysis on the PDW pulses classified by frequency range. If the PRI type of the PDW pulse cannot be correctly identified and the pulse width PW has similar results, repeat step S02. If the PRI type of the PDW pulse can be correctly identified and the pulse width PW has similar results, proceed to step S04. S04: Based on the pulse frequency analysis, the pulse modulation characteristics are analyzed, and the PDW pulse first arrival time, signal power, and signal arrival angle are calculated to generate the EDW result.

2. The method as described in claim 1, characterized in that, In step S01, the PDW pulse time domain is merged into The arrival times and pulse widths of two adjacent pulses are respectively , , , The time-domain merging threshold is , This represents the pulse end time of the former.

3. The method as described in claim 2, characterized in that, If the two pulses are merged, the end time of the former pulse is... 。 4. The method as described in claim 1, characterized in that, The step of using unsupervised clustering analysis to perform clustering analysis on the azimuth and elevation angles of the PDW pulse signal in step S01 further includes: determining the clustering range of the clustering analysis based on the angle measurement accuracy. The data to be analyzed is initially prepared, and then analyzed using an improved version of the K-means method. , It is clustering The mean of the samples contained in the sample represents the mean of the j-th cluster center; This represents the number of samples in the j-th cluster. Indicates sample The Euclidean distance from the j-th cluster center Indicates the error range of the cluster analysis. .

5. The method as described in claim 1, characterized in that, Identifying and determining the PRI type of the PDW pulse further includes: identifying the PRI pattern of the PDW pulse and calculating the arrival time sequence (TOA) of the PDW pulse sequence. Where b represents the intercept, k represents the slope, which is also the center value of the PRI sequence, and M represents the sequence length.

6. The method as described in claim 5, characterized in that, The slope k is 。 7. The method as described in claim 6, characterized in that, Identifying the jitter type of PRI sequences involves subtracting the PRI sequence from the slope k and summing the results to calculate the average jitter of the PRI sequence. : 。 8. The method as described in claim 5, characterized in that, Identifying the PRI type of the PDW pulse further includes: determining the fixed PRI of the pulse; after the PRI sequence value of the pulse sequence is processed by cluster analysis, the PRI of the fixed repetition frequency is processed by harmonic removal, and the result of cluster analysis is only one value.

9. The method as described in claim 8, characterized in that, Identifying the PDW pulse PRI type further includes: identifying the slip PRI: performing a difference operation on the PRI sequence of length M-1 to obtain a difference sequence of length M-2. , right Symbolization .

10. The method as described in claim 9, characterized in that, The judgment and identification of slippage PRI analysis further includes: weighting the symbolic results to obtain the difference change value. This allows for the identification of the slip type PRI. 。

Images