Radar signal pre-sorting method suitable for multi-parameter overlapping
Based on the DBSCAN algorithm, the clustering method of gradually expanding from the smallest neighborhood and absorbing new points is solved, and the traditional DBSCAN algorithm has poor sorting effect in density uneven signal processing is achieved, and multi-parameter overlapping is achieved Effective separation of signals.
Patent Information
- Application Number
- CN202411573271.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-06
- Publication Date
- 2025-02-14
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The traditional DBSCAN algorithm has poor sorting effect when processing radar signals with uneven density, and is sensitive to clustering parameters, so the artificial setting is not reliable.
A clustering method is adopted to gradually expand the neighborhood to absorb new points from the smallest neighborhood, and secondary sorting is performed through signal sorting software, clustering is performed using Euclidean distance and core distance, and the neighborhood is updated until it cannot be found. Until the new point is added.
The correct separation of multiple radar signals is achieved in a signal environment with multi-parameter overlapping, which improves the reliability and effectiveness of signal sorting.
Smart Images
Figure CN119439109A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of electromagnetic environment signal monitoring, and in particular to a radar signal pre-sorting method. Background Art
[0002] In modern electronic warfare, due to the emergence of a large number of new radar systems with complex signal characteristics, the battlefield electromagnetic signal environment has become increasingly dense and complex. The electronic reconnaissance system provides intelligence support for battlefield situation awareness and auxiliary interference decision-making by intercepting dense information flows in the electromagnetic space.
[0003] Radar signal sorting is the initial link of the electronic reconnaissance system. The signal sorting problem is essentially the process of separating the signals of different radar radiation sources detected by the reconnaissance receiver. It is an important prerequisite for the successful implementation of electronic warfare and has a great impact on the entire electronic warfare system.
[0004] Signal sorting is divided into pre-sorting and main sorting; traditional pre-sorting generally uses multi-parameter correlation, and the correlation uses a fixed tolerance, which cannot adapt to the signal environment where multiple parameters overlap. Therefore, clustering analysis technology in data mining is introduced, which can divide data objects into multiple classes according to the similarities between data objects.
[0005] The more classic one is the DBSCAN algorithm. The traditional DBSCAN algorithm is sensitive to clustering parameters, and the clustering parameters are artificially set based on prior knowledge, so the reliability is not high. At the same time, the DBSCAN algorithm is suitable for processing data with relatively uniform density. When the density of different clusters varies greatly, the DBSCAN sorting effect is not good. Summary of the invention
[0006] The technical problem to be solved by the present invention is: based on the above technical background, on the basis of the traditional DBSCAN algorithm, combined with the characteristics of the parameters of the radar signal itself and the characteristics of the clustered distribution of the pulses of a single radar signal, a clustering method is found that starts from the minimum neighborhood and gradually expands the neighborhood to absorb new points, so as to solve the problems that the traditional DBSCAN algorithm is sensitive to clustering parameters, and the clustering parameters are artificially set according to prior knowledge, and the reliability is not high; at the same time, the DBSCAN algorithm is suitable for processing data with relatively uniform density. When the density of different clusters varies greatly, the DBSCAN sorting effect is not good, and the present invention is suitable for engineering applications in electronic reconnaissance equipment.
[0007] The technical solution to be adopted by the present invention to solve its technical problem is: a radar signal pre-sorting method suitable for multi-parameter overlap, including signal sorting software, the signal sorting software includes a primary traditional sorting method and a secondary radar signal pre-sorting method, characterized in that when the signal sorting software fails to be sorted by the primary traditional sorting method, the radar signal pre-sorting method is used for secondary sorting; when the radar signal pre-sorting method is used for secondary sorting, clustering is created from the minimum core; the maximum neighborhood of the core circle density point is counted as a new neighborhood, and the existing clustering is searched, each cluster point is used as the core, the new neighborhood is used as the radius, and the newly added point is found, and the neighborhood is updated, and the existing clustering is searched, and the above process is repeated. If no newly added point is found, the clustering ends; the clustering ends, that is, the clustering result is evaluated, if it is valid, the separation is successful, and then the next valid core is selected, and the remaining pulses start the next round of clustering; if it is invalid, the next valid core is also selected, and the next round of clustering starts, until the valid core is not found, the entire clustering, that is, the radar signal pre-sorting method process ends.
[0008] When the present invention works, the specific clustering method is as follows:
[0009] Step 1: Calculate the Euclidean distance and generate a distance matrix;
[0010] Step 2: Count the core distance of each pulse and sort them in ascending order as the initial set of free core sequences;
[0011] Step 3: Create clusters starting from the minimum core, the initial neighborhood is the minimum core distance, and the initial cluster set includes the four core circle density points of the minimum core;
[0012] Step 4: Find the largest neighborhood of the core circle density point as the new neighborhood;
[0013] Step 5: Use the new neighborhood as the radius and each point in the cluster set as the core to search for new points;
[0014] Step 6: If there are new points, the new points are included in the cluster set, the largest neighborhood of the new points is found, and then go to step 5;
[0015] Step 7: If there are no new points, the clustering ends;
[0016] Step 8: Evaluate the clustering results. If the clustering is invalid, find the next one from the free core set as the minimum core and go to step 3. If the clustering is valid, reorganize the free core set, take its minimum core and go to step 3.
[0017] Step 9: If the free core set is empty, the entire clustering ends.
[0018] Through the above processing method, when the signal sorting software fails in traditional sorting, if the input signal channel contains multiple radar signals with multiple overlapping parameters, the correct separation of multiple radar signals can be achieved and the signal sorting is successful.
[0019] The clustering principle of the present invention is defined as follows:
[0020] Input pulse data set {Vi}, i = 1, 2, ..., N, where Vi is the three-dimensional space composed of carrier frequency (RF), angle of arrival (DOA), and pulse width (PW) in the pulse description word (PDW), and N is the total number of pulse samples.
[0021] 1) Distance: The Euclidean distance between two pulses vi and vj in three-dimensional space, calculated as:
[0022]
[0023] Wherein: i, j = 1, 2, ..., N; rf represents the RF feature; doa represents the DOA feature; pw represents the PW feature; and parameter normalization is considered in actual calculation.
[0024] 2) Core distance: The distance between pulse vi and the fourth point closest to it is defined as the core distance of pulse vi.
[0025] 3) Minimum core: The pulse with the smallest core distance in the input pulse set.
[0026] 4) Neighborhood: distance radius.
[0027] 5) Core circle density points: the set of all points within the core distance.
[0028] 6) Newly added points: points that are detected to belong to the same category as the cluster set but have not yet been added to the cluster set.
[0029] 7) Free core set: a set of pulse points that have not been clustered; after each successful clustering, the points in the cluster set are deducted from the free core set. BRIEF DESCRIPTION OF THE DRAWINGS
[0030] The present invention will be further described below by means of the accompanying drawings, tables and examples;
[0031] Figure 1 It is a software flow chart of the radar signal pre-sorting method with multiple parameters overlapping.
[0032] Table 1 is the parameter table of the four-part parameter overlapping radar.
[0033] Table 2 is a table of the input PDW sequence and its pulse core distances.
[0034] Table 3 is the Euclidean distance matrix.
[0035] Table 4 is a schematic diagram of the clustering process.
[0036] Table 5 is the PRI evaluation table of the clustering results.
[0037] Table 6 is a schematic diagram of the clustering process. Table 2 corresponds to radar 3.
[0038] Table 7 is a schematic diagram of the clustering process. Table 3 corresponds to radar 2.
[0039] Table 8 is a schematic diagram of the clustering process. Table 4 corresponds to radar 1. DETAILED DESCRIPTION
[0040] like Figure 1 As shown in Tables 1-8, the technical problem to be solved by the present invention is: based on the traditional DBSCAN algorithm, combined with the characteristics of the parameters of the radar signal itself and the characteristics of the clustered distribution of the pulses of a single radar signal, a clustering method is found that starts from the minimum neighborhood and gradually expands the neighborhood to absorb new points, so as to solve the problem that the traditional DBSCAN algorithm is sensitive to clustering parameters, and the clustering parameters are artificially set according to prior knowledge, and the reliability is not high. At the same time, the DBSCAN algorithm is suitable for processing data with relatively uniform density. When the density of different clusters varies greatly, the DBSCAN sorting effect is not good, and the method is suitable for engineering applications in electronic reconnaissance equipment.
[0041] The technical solution to be adopted by the present invention to solve the technical problem is: a radar signal pre-sorting method suitable for multi-parameter overlap, including signal sorting software, the signal sorting software includes a primary traditional sorting method and a secondary radar signal pre-sorting method, when the signal sorting software fails to be sorted by the primary traditional sorting method, the radar signal pre-sorting method is used for secondary sorting; when the radar signal pre-sorting method is used for secondary sorting, clusters are created starting from the minimum core; the maximum neighborhood of the core circle density point is counted as a new neighborhood, and the existing clusters are searched, each cluster point is used as the core and the new neighborhood is used as the radius, and new points are found, and the neighborhood is updated, and the existing clusters are searched, and the above process is repeated. If no new points are found, the clustering ends;
[0042] At the end of clustering, the clustering results are evaluated. If they are valid, the separation is successful, and then the next valid core is selected, and the remaining pulses start the next round of clustering;
[0043] If it is invalid, the next valid core is selected and the next round of clustering is started until no valid core is found. Then the whole clustering process, i.e. the radar signal pre-sorting method, ends.
[0044] When the present invention is working, in order to realize the application of the above method in practical problems, 100 PDWs, i.e., the number of pulses that the signal sorting software routinely enters the sorting process, are simulated according to the problem scenario, including four radar signals. In the three dimensions of azimuth, frequency, and pulse width, the four radars are mutually overlapped. In the same time, two radars have more pulses and two have less pulses. The specific radar signal parameters are as shown in Table 1:
[0045]
[0046] According to the existing detection system parameter measurement resolution: carrier frequency 0.5MHz, azimuth 0.01°, pulse width 25ns, 100 PDW sequences are generated as shown in Table 2, PDW sequence number, frequency, arrival time, pulse width, azimuth, and the radar number of each pulse is recorded at the same time for the subsequent verification algorithm, which are 0, 1, 2, and 3 respectively. Table 2:
[0047]
[0048]
[0049]
[0050] After traditional sorting, no signal was extracted from this channel, so this method was used for sorting. The steps are as follows:
[0051] Step 1: Calculate the Euclidean distance and generate a distance matrix;
[0052] Use the maximum absolute value normalization method to perform normalization processing, and then calculate the distance between each point of the input pulse according to the above Euclidean distance formula to generate a distance matrix. The distance matrix of this example is shown in Table 3:
[0053] PDW serial number 0 1 2 3 4 5 6～97 98 99 0 0 0.619069 0.993533 1.002398 0.915975 0.233071 ... 0.97233 0.0868 1 0.619069 0 0.639565 0.649411 0.503526 0.448228 ... 0.567665 0.591597 2 0.993533 0.639565 0 1.09499 0.962207 0.761681 ... 0.968392 0.920647 3 1.002398 0.649411 1.09499 0 0.152778 0.930102 ... 0.133406 0.995119 4 0.915975 0.503526 0.962207 0.152778 0 0.820184 ... 0.097941 0.903787 5 0.233071 0.448228 0.761681 0.930102 0.820184 0 ... 0.871623 0.171286 6～97 ... ... ... ... ... ... ... ... ... 98 0.97233 0.567665 0.968392 0.133406 0.097941 0.871623 ... 0 0.954821 99 0.0868 0.591597 0.920647 0.995119 0.903787 0.171286 ... 0.954821 0
[0054] Step 2: Count the core distance of each pulse and sort them in ascending order as the initial set of free core sequences;
[0055] According to the distance matrix, sort the distances from each pulse to other pulses in ascending order, and find the core distance of each pulse, that is, the distance to the fourth pulse closest to the reference pulse; the core distance in this example is shown in the "core distance" column of Table 2;
[0056] Sort the core distances in ascending order to generate the initial set of free core sequences:
[0057] {85,16,55,17,98,60,49,97,43,10,99,71,23,86,89,25,44,22,41,30,80,47,0,48,67,96,34,61,7,18,57,13,8,69,36,9,88,93,56,21,64,87,62,40,73,52,74,51,19,8 4,79,92,11,90,35,65,78,68,38,66,63,12,29,28,37,94,75,83,31,32,4,91,33,6,42,3,54,81,24,77,5,15,82,59,53,70,14,45,26,72,2,39,46,95,20,58,76,50,1,27}
[0058] The set element is the original PDW number corresponding to the sorted core distance;
[0059] Step 3: Create clusters starting from the minimum core, the initial neighborhood is the minimum core distance, and the initial cluster set includes the four core circle density points of the minimum core;
[0060] According to the above definition, in this case, the minimum core is 85, the initial neighborhood is 0.03262, and the initial cluster set is {85; 43; 49; 97}, as shown in Table 2;
[0061] Step 4: Find the largest neighborhood of the core circle density point as the new neighborhood;
[0062] According to the above definition, the maximum neighborhood of the core circle density point {85; 43; 49; 97} is 43, which corresponds to a core distance of 0.048925;
[0063] Step 5: Use the new neighborhood as the radius and each point in the cluster set as the core to search for new points;
[0064] In this example, the first clustering starts from the initial cluster set and is cycled five times. The status of each search for new points is shown in Table 4:
[0065]
[0066] Step 6: If there are new points, the new points are included in the cluster set, the largest neighborhood of the new points is found, and then go to step 5;
[0067] After the new cluster is established in this example, the first updated neighborhood is 0.048925, with 7 new points; the next updated neighborhood is 0.072236, with 64, 28, and 10 new points.
[0068] Step 7: If there are no new points, the clustering ends;
[0069] In this example, after five cycles, the new neighborhood is 0.115134, and there are no more new points, so clustering ends;
[0070] Step 8: Evaluate the clustering results. If the clustering is invalid, find the next one from the free core set as the minimum core and go to step 3. If the clustering is valid, reorganize the free core set, take its minimum core and go to step 3.
[0071] The clustering result is evaluated by: sorting the cluster set in ascending order by sequence number, obtaining the corresponding PDW, and making a first-level histogram of the arrival time difference. If more than 80% of the PRI distribution rules are within two times, the clustering is valid; otherwise, the clustering is invalid.
[0072] The reorganization of the free core set is to remove all the points that are valid for clustering, and the remaining points form a new free core set;
[0073] If the clustering is valid and the number of pulses is greater than or equal to the minimum number of sorting pulses, the separation is successful, and the cluster set pulses are transferred to the main sorting. When the processing is completed, the cluster set pulses are deducted and the remaining pulses are counted. If there are remaining pulses, go to step 3 for the next clustering. If there are no remaining pulses, the entire clustering is completed; if the clustering is valid and the number of pulses is less than the minimum number of sorting pulses, the cluster set is transferred to the temporary storage channel for accumulation;
[0074] The first clustering result of this example is 42 pulses, see Table 4; sort them in ascending order, get the corresponding PDW, and make a first-level histogram of their arrival time differences. The result is shown in Table 5:
[0075]
[0076]
[0077] The PRI range is 615 to 798, within two times, clustering is effective, the number of pulses is 42, and the minimum number of sorting pulses is exceeded, which is generally configured as 10 to 30, that is, the main sorting is carried out, and the radar separation is successful;
[0078] Step 9: If the free core set is empty, the entire clustering ends;
[0079] The successfully separated PDWs are removed, and the remaining pulses are clustered, and the cluster sets are obtained in turn:
[0080] Table 6 is a schematic diagram of the clustering process. Table 2 corresponds to radar 3.
[0081]
[0082]
[0083] Table 7 is a schematic diagram of the clustering process. Table 3 corresponds to radar 2.
[0084] order Core distance Core circle density 1 0.098131 77；20；39；2； 2 0.150391 77；20；39；2；95；58；
[0085] Table 8 is a schematic diagram of the clustering process. Table 4 corresponds to radar 1.
[0086] order Core distance Core circle density 1 0.31239 76；27；50；1；
[0087] At this point, clustering is complete;
[0088] Through the above processing method, the four radars were separated successfully.
[0089] The present invention is not limited to the above-mentioned embodiments. On the basis of the technical solution disclosed in the present invention, technicians in this field can make some substitutions and deformations to some technical features therein according to the disclosed technical content without creative labor, and these substitutions and deformations are all within the protection scope of the present invention.
Claims
1. A radar signal pre-sorting method suitable for multi-parameter overlap, comprising signal sorting software, wherein the signal sorting software comprises a primary traditional sorting method and a secondary radar signal pre-sorting method, characterized in that: When the signal sorting software fails in the first traditional sorting method, the radar signal pre-sorting method is used for secondary sorting; when the radar signal pre-sorting method is used for secondary sorting, clusters are created starting from the smallest core; the maximum neighborhood of the core circle density point is counted as a new neighborhood, and the existing clusters are searched, with each cluster point as the core and the new neighborhood as the radius, new points are found, and the neighborhood is updated, and the existing clusters are searched, and the above process is repeated. If no new points are found, the clustering ends; At the end of clustering, the clustering results are evaluated. If they are valid, the separation is successful, and then the next valid core is selected, and the remaining pulses start the next round of clustering; If it is invalid, the next valid core is selected and the next round of clustering is started until no valid core is found. Then the whole clustering process, i.e. the radar signal pre-sorting method, ends.
2. A radar signal pre-sorting method suitable for multi-parameter overlap according to claim 1, characterized in that: The clustering method comprises: Step 1: Calculate the Euclidean distance and generate a distance matrix; Step 2: Count the core distance of each pulse and sort them in ascending order as the initial set of free core sequences; Step 3: Create clusters starting from the minimum core, the initial neighborhood is the minimum core distance, and the initial cluster set includes the four core circle density points of the minimum core; Step 4: Find the largest neighborhood of the core circle density point as the new neighborhood; Step 5: Use the new neighborhood as the radius and each point in the cluster set as the core to search for new points; Step 6: If there are new points, the new points are included in the cluster set, the largest neighborhood of the new points is found, and then go to step 5; Step 7: If there are no new points, the clustering ends; Step 8: Evaluate the clustering results. If the clustering is invalid, find the next one from the free core set as the minimum core and go to step 3. If the clustering is valid, reorganize the free core set, take its minimum core and go to step 3. Step 9: If the free core set is empty, the entire clustering ends; Through the above steps, when the signal sorting software fails after a traditional sorting method, if the input signal channel contains multiple radar signals with multiple overlapping parameters, the correct separation of multiple radar signals can be achieved, and the signal sorting is successful.
3. The radar signal pre-sorting method applicable to multi-parameter overlap according to claim 1, characterized in that: The relevant definitions of the clustering principle are as follows: Input pulse data set {Vi}, i = 1, 2, ..., N, where Vi is the three-dimensional space consisting of carrier frequency (RF), angle of arrival (DOA), and pulse width (PW) in the pulse description word (PDW), and N is the total number of pulse samples; The distance is defined as the Euclidean distance between two pulses vi and vj in three-dimensional space, and the calculation formula is: Where: i, j = 1, 2, ..., N; rf represents the RF feature; doa represents the DOA feature; pw represents the PW feature; Consider the normalization of parameters during actual calculations; The core distance is defined as follows: the distance between the pulse vi and the fourth point closest to it is defined as the core distance of the pulse vi; The definition of the minimum core is: the pulse with the smallest core distance in the input pulse set; The neighborhood is defined as: distance radius; The core circle density point is defined as: the set of all points within the core distance; The definition of the newly added point is: a point that is detected to belong to the same class as the cluster set but has not yet been added to the cluster set; The free core set is a set of pulse points that have not been clustered. When each clustering is successful, the points in the clustering set are deducted from the free core set.