Signal sorting method and system based on improved mutual neighbor DBSCAN

Abstract

The invention discloses a mutual neighbor DBSCAN radar pulse sorting method and a mutual neighbor DBSCAN radar pulse sorting system, and mainly solves the problems of core point misjudgment and cross-cluster development of sorting clusters during radar pulse sorting in an existing sorting method. The implementation scheme is as follows: constructing a distance matrix of each pulse about a pulse description word by using K-nearest neighbor; performing mutual neighbor judgment on all the pulses, and taking the pulses meeting judgment conditions as core pulses; constructing a core pulse matrix, and carrying out sorting label initialization; performing cluster division on each core pulse and the density direct pulse thereof; constructing a temporary pulse vector for the density direct pulse, selecting a density reachable pulse according to a set scale factor, and performing cluster division on the density reachable pulse; and performing density direct pulse and density reachable pulse acquisition and cluster division on the core points which are not sorted yet until all core pulses are sorted. The method improves the accuracy of radar pulse sorting, and can be used for electronic reconnaissance and target number acquisition.

Description

Technical Field

[0001] The invention belongs to the technical field of radar signal processing, and in particular relates to a signal sorting method which can be used for electronic reconnaissance and obtaining the number of targets. Background Art

[0002] DBSCAN is a density-based spatial clustering algorithm proposed by Martin Ester, Hans-Peter Kriegel, Jörg Sander, and Xiaowei Xu in 1996. The core idea of ​​the algorithm is to identify cluster structures through the spatial density distribution of data points. It can effectively handle clusters of arbitrary shapes and automatically distinguish noise points. Its technical implementation depends on two key parameters: neighborhood radius And the minimum number of points MinPts, that is, first traverse each data point and calculate its The number of points in the neighborhood, if a point If the neighborhood contains at least MinPts points, it will be marked as a core point; then, starting from the core point, the cluster is recursively expanded through density reachability, and all points that are directly density-reachable and density-connected are included in the same cluster. At the same time, points that are not core points but are located at a core point are Points within the neighborhood are defined as boundary points, and points that do not belong to any cluster are defined as noise points. This definition process does not rely on the global density assumption, but instead determines the cluster boundaries by local density, thereby identifying multiple clusters with different densities. There are problems with the selection of MinPts, such as difficulty in selecting parameters and difficulty in adapting to density differences when the neighborhood radius is fixed.

[0003] To address the above-mentioned issues, the prior art proposes an improved RNN-DBSCA algorithm that reduces the number of parameters to one, which only depends on the parameter K. That is, when a point is marked as a K-nearest neighbor point by the other points more than K times, the point is defined as a core point. The K neighbor points based on the current core point and the K-nearest neighbor points of each neighbor point are defined as boundary points. Noise points are defined as data points that are neither boundary points nor core points. Although this improved RNN-DBSCAN algorithm reduces the number of parameters, it facilitates parameter selection and no longer relies on a fixed neighborhood radius. However, when performing signal sorting under compact data distribution, boundary points and noise points may be misjudged as core points, and when expanding the cluster, data points and noise points of other clusters may be expanded into the sorting cluster, affecting the sorting accuracy. Summary of the Invention

[0004] The present invention aims to address the above-mentioned deficiencies in the prior art and propose a signal sorting method and system based on an improved mutual nearest neighbor DBSCAN to improve the accuracy of signal sorting.

[0005] The present invention further improves the improved RNN-DBSCAN algorithm, that is, through K mutual nearest neighbors and proportional factors, it reduces the probability of core point misjudgment and cross-cluster expansion, and improves the sorting accuracy. Its implementation scheme includes:

[0006] 1. A signal sorting method based on an improved KMNN-DBSCAN, comprising:

[0007] 1) Use K nearest neighbors to construct the distance matrix of each pulse with respect to the pulse description word PDW;

[0008] 2) Obtain the K nearest neighbor pulses of each pulse through the distance matrix and perform mutual neighbor determination:

[0009] If the K nearest neighbors of each neighboring pulse all contain the pulse to be determined, the pulse to be determined is marked as a core point and step 3 is executed);

[0010] Otherwise, the pulse to be determined is temporarily marked as a noise pulse;

[0011] 3) Construct a core pulse matrix for all pulses that meet the core judgment and initialize the sorting labels;

[0012] 4) Select any unsorted core pulse from the core pulse matrix, select the K neighboring pulses of the core pulse as the density direct pulses of the current core pulse and perform cluster division;

[0013] 5) Construct a temporary pulse vector for the K density direct pulses of the core pulse, and according to the set scaling factor , select K of each density direct pulse The neighboring pulses are regarded as density-reachable pulses, and then they are clustered;

[0014] 6) Repeat 4) to 5) for the core points and boundary points that have not been sorted until all core pulses are sorted.

[0015] Furthermore, in 3) above, a core pulse matrix is ​​constructed for all pulses that meet the core determination, and sorting labels are initialized. The implementation thereof includes:

[0016] 3a) Extract all Z core pulses that meet the mutual neighbor judgment and the K neighbor pulses of each core pulse in the P-dimensional space from the M pulses, and merge these core pulses that meet the mutual neighbor judgment and the K neighbor pulses of each core pulse into one dimensional core pulse matrix, and the Z pulses in the matrix that meet the mutual neighbor judgment are used as the core points during sorting;

[0017] 3b) The sorting cluster labels of all core points are recorded as -1, that is, the pulse is not classified. The sorting cluster labels of the remaining pulses are recorded as 0, that is, these pulses are temporarily treated as noise points, and the initial number of sorting clusters is set to 0.

[0018] Furthermore, the clustering of density-reachable pulses in 5) is implemented by:

[0019] 5c) taking the first pulse in the temporary vector with a sorting cluster label of 0 or -1 as the pulse to be sorted;

[0020] 5d) In Find the pulse to be sorted in the first column of the dimensional matrix, and assume that the pulse to be sorted is located in The first column of the Y-th row in the dimensional matrix;

[0021] 5e) In Select the first row from the Yth row and the 2nd column to the Kth column in the matrix pulses, merge them into a temporary pulse vector and remove duplicate pulses;

[0022] 5f) Change the sorting cluster label corresponding to the pulse to be sorted to the current sorting cluster number i;

[0023] 5g) Repeat 5c) to 5f) until all pulses in the temporary pulse vector are over-sorted.

[0024] 2. A signal sorting system based on improved mutual nearest neighbor DBSCAN, comprising:

[0025] A distance calculation module is used to calculate the distance between each pulse and the rest of the pulses in the P-dimensional space;

[0026] A mutual neighbor determination module is used to determine whether a pulse satisfies the mutual neighbor determination;

[0027] A core pulse module is used to obtain core pulses that meet the mutual neighbor judgment among all pulses;

[0028] Density direct access module, used to obtain density direct access pulses of each core pulse;

[0029] Density reachable module, used to obtain density reachable pulses of each core pulse;

[0030] The sorting module is used to determine which sorting cluster each pulse belongs to.

[0031] Compared with the prior art, the present invention has the following advantages:

[0032] First, the present invention adopts the K mutual neighbor core determination method. When performing core pulse determination, the K mutual neighbor core determination is used instead of the existing reverse nearest neighbor core determination. Through the strict reciprocal relationship, the boundary pulse is avoided from being misjudged as the core pulse during core pulse selection.

[0033] Second, since the present invention introduces a scaling factor when expanding the sorting cluster, compared with the existing RNN-DBSCAN algorithm that uses K-nearest neighbors to expand the sorting cluster, the present invention adds a scaling factor when expanding the sorting cluster with K-nearest neighbors, and filters out noise or interference points of other clusters through the scaling factor to avoid cross-cluster development. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 This is a flow chart of the implementation of the signal sorting method based on the improved mutual nearest neighbor DBSCAN proposed in Example 1 of the present invention;

[0035] Figure 2 Schematic diagram of the mutual neighbor core points based on the improved mutual neighbor DBSCAN signal sorting in Example 1 of the present invention;

[0036] Figure 3 Schematic diagram of the scaling factor expansion cluster in Example 1 of the present invention;

[0037] Figure 4 This is a block diagram of a signal sorting system based on improved mutual nearest neighbor DBSCAN according to embodiment 2 of the present invention;

[0038] Figure 5 A comparison diagram of radar signal sorting simulation using the present invention and the existing RNN-DBSCAN and DBSCAN methods in dataset 1 using pulse descriptors corresponding to different numbers of radars when the K value is the same;

[0039] Figure 6 This is a simulation comparison diagram of sorting radar signals in the bandwidth PW-arrival angle DOA dimension using the pulse descriptors corresponding to different numbers of radars in data set 2 using the present invention and the existing RNN-DBSCAN and DBSCAN methods when the K value is the same. DETAILED DESCRIPTION

[0040] The present invention will be described in further detail below with reference to the accompanying drawings.

[0041] Example 1: A signal sorting method based on an improved mutual nearest neighbor DBSCAN.

[0042] This example's application scenario includes several radars and a receiver. The radars transmit radar signals, which the receiver receives and works with the signal processing unit to generate radar signal PDW parameters. The signal processing module then processes these signals using a modified mutual nearest neighbor (DBSCAN) signal sorting method, thereby sorting the radar waveforms emitted by each radar.

[0043] Reference Figure 1 The implementation steps of this example include the following:

[0044] Step 1: Use K nearest neighbors to construct a distance matrix for each pulse PDW.

[0045] 1.1) Assume that M pulses are sorted, and each pulse has P pulse descriptors PDW. Assume that the pulse descriptors include carrier frequency CF, bandwidth PW, and angle of arrival DOA. Each pulse is numbered.

[0046] 1.2) Build a Dimensional matrix and normalize it;

[0047] Find the maximum and minimum values ​​of each pulse description word PDW parameter in all pulses, and subtract the minimum value of the pulse description word PDW parameter in all pulses from the current pulse description word PDW parameter of the current pulse to be normalized to obtain the carrier frequency CF parameter after pulse X normalization :

[0048] ,

[0049] in, and They are respectively the minimum carrier frequency and the maximum carrier frequency among all pulse carrier frequency CF parameters.

[0050] Normalize the bandwidth PW parameter in the pulse description word PDW to obtain the normalized bandwidth PW parameter of pulse X :

[0051] ,

[0052] in, and They are the minimum bandwidth and maximum bandwidth of all pulse carrier frequency PW parameters respectively;

[0053] Normalize the arrival angle DOA parameter in the pulse description word PDW to obtain the normalized arrival angle DOA parameter of pulse X :

[0054] ,

[0055] in, and are the minimum arrival angle and the maximum arrival angle among all pulse carrier frequency DOA parameters respectively;

[0056] 1.3) Take each normalized pulse as a point in P-dimensional space, and the pulse description word PDW parameter of each normalized pulse as the coordinate of the pulse in P-dimensional space, and calculate the distance between pulse X and other pulses in P-dimensional space. , the formula is as follows:

[0057] ,

[0058] in, is the parameter carrier frequency CF in the pulse description word PDW after the normalization of pulse X, is the parameter carrier frequency CF in the normalized pulse description word PDW of pulse Y, is the angle of arrival (DOA) of the pulse description word (PDW) after normalization of pulse X. is the angle of arrival (DOA) parameter in the normalized pulse description word (PDW) of pulse Y, is the parameter bandwidth PW in the pulse description word PDW after the normalization of pulse X, is the parameter bandwidth PW in the normalized pulse description word PDW of pulse Y.

[0059] 1.4) Arrange the remaining M-1 pulses in order from nearest to farthest distance and construct a dimensional matrix, the first column of the matrix represents the sequence number of pulse X in M ​​pulses, and the subsequent M-1 columns are the sequence numbers of each pulse;

[0060] 1.5) For each pulse, construct the dimensional matrix, the final combination is dimensional matrix;

[0061] 1.6) Select according to parameter K Dimensional matrix Column, composition dimensional matrix, ignore dimensional matrix, reducing its dimensionality to dimensional distance matrix.

[0062] Step 2: Perform mutual neighbor determination on the K nearest neighbor pulses of each pulse in the P-dimensional space.

[0063] Reference Figure 2 , the implementation of this step includes the following:

[0064] 2.1) Utilization The M rows of the dimensional distance matrix represent a total of M pulses, and the sequence number of the first element in the Yth row among all the pulses is recorded as , and move the second to the The K pulses listed in the P-dimensional space from near to far are denoted as ;

[0065] 2.2) In The first column elements of the distance matrix are and find the elements in the second to K+1 columns of these row vectors :

[0066] If the second to K+1th columns of these row vectors all contain , then the pulse represented by the Yth row is a K-nearest neighbor point in the P-dimensional space;

[0067] Otherwise, the pulse to be determined is temporarily marked as a noise pulse;

[0068] 2.3) Count all Z pulses among the M pulses that meet the K mutual nearest neighbors.

[0069] Step 3: Construct a core pulse matrix that satisfies the mutual nearest neighbor judgment and initialize the labels.

[0070] 3.1) Extract all Z core pulses that meet the mutual neighbor judgment and the K neighbor pulses of each core pulse in the P-dimensional space from the M pulses, and merge these core pulses that meet the mutual neighbor judgment and the K neighbor pulses of each core pulse into one dimensional core pulse matrix, and the Z pulses in the matrix that meet the mutual neighbor judgment are used as the core points during sorting;

[0071] 3.2) The sorting cluster labels of all core points are recorded as -1, which means that the pulse is not classified, and the sorting cluster labels of the remaining pulses are recorded as 0, which means that these pulses are temporarily regarded as noise points;

[0072] 3.2) Set the initial number of sorting clusters to 0.

[0073] Step 4: Cluster the core pulse points and their dense direct pulses.

[0074] 4.1) Add 1 to the number of sorting clusters and let the sorting cluster at this time be numbered i;

[0075] 4.2) Randomly select a core pulse with a sorting cluster label of -1 from all the core pulses, that is, from all the elements in the first column of the core pulse matrix, and change the sorting cluster label of the current core pulse to be sorted to the number i of the current sorting cluster;

[0076] 4.3) Select the density direct pulses of the current pulse to be sorted, that is, the second element to the K+1th element in the row of the core pulse matrix where the current pulse to be sorted is located, and change the sorting cluster label corresponding to each density direct pulse to the current sorting cluster number i.

[0077] Step 5: Construct a temporary pulse vector for the K density direct pulses of the core pulse, and use the set scaling factor , select K of each density direct pulse The neighboring pulses are regarded as density-reachable pulses and then clustered.

[0078] Reference Figure 3 , the implementation of this step includes the following:

[0079] 5.1) In Find the current core pulse in the first column of the dimensional core pulse matrix;

[0080] 5.2) Assume that the current core pulse is located at The first column and the Yth row of the dimensional core pulse matrix are selected, and a total of K pulses from the 2nd column to the K+1th column of the Yth row are selected as temporary pulse vectors.

[0081] 5.3) The first pulse in the temporary vector with a sorting cluster label of 0 or -1 is used as the pulse to be sorted, and the sorting cluster label corresponding to the pulse to be sorted is changed to the current sorting cluster number i;

[0082] 5.4) In Find the pulse to be sorted in the first column of the dimensional matrix, and assume that the pulse to be sorted is located in The first column of the Y-th row in the dimensional matrix;

[0083] 5.5) According to the set scale factor ,exist Select the first row from the Yth row and the 2nd column to the Kth column in the matrix pulses as density-reachable pulses, merge these density-reachable pulses into the temporary pulse vector and remove repeated pulses to complete the update of the temporary pulse vector;

[0084] 5.6) Repeat 5.3) to 5.5) until all pulses in the temporary pulse vector are sorted.

[0085] Step 6: Summarize all pulses of all sorting clusters and output the sorting results.

[0086] Repeat steps 4 and 5 until all core pulses are sorted, and summarize all data points of all sorted clusters to output the sorting results.

[0087] The flowchart representation or method representation of the above embodiment can be understood as a module, segment or portion of code that includes one or a group of executable instructions configured to implement the steps of a specific logical function or process. The present invention is not limited to the disclosed preferred embodiments, and its implementation may not be carried out in the order shown or discussed.

[0088] Example 2: A signal sorting system based on the improved mutual nearest neighbor DBSCAN.

[0089] Reference Figure 4 This example includes: a distance calculation module 1, a mutual neighbor determination module 2, a core pulse module 3, a density direct pulse module 4, a density reachable pulse module 5, and a sorting module 6. The sorting module 6 includes a core pulse sorting submodule 61, a density direct pulse sorting submodule 62, and a density reachable pulse sorting submodule 63. The working principle of the entire system is as follows:

[0090] The distance calculation module 1 is used to calculate the distance between each pulse in the P-dimensional space and send the obtained distance to the mutual neighbor determination module 2, the density direct pulse module 4, and the density reachable pulse module 5 respectively;

[0091] The mutual neighbor module 2 is used to perform mutual neighbor determination on each pulse after receiving the distance between each pulse and send the determination result to the core pulse module 3;

[0092] The core module 3 is used to send the pulses that meet the mutual neighbor determination as core pulses to the density direct pulse module 4 and the sorting module 6 respectively after receiving the mutual neighbor determination result;

[0093] The density direct pulse module 4 is used to obtain K density direct pulses of each core pulse according to the distances between each pulse after receiving the core pulse sent by the core pulse module 3 and the distances between each pulse sent by the distance calculation module 1, and send the density direct pulses of each pulse to the density direct pulse module 5 and the sorting module 6 respectively;

[0094] The density-reachable pulse module 5 is configured to obtain the density-reachable pulse of each core point based on the distance between each pulse and the proportional factor after receiving the density-reachable pulse sent by the density-reachable pulse module 4 and the distance between each pulse sent by the distance calculation module 1, and send the density-reachable pulse to the sorting module 6;

[0095] The sorting module 6 is used to perform pulse sorting after receiving the core pulse sent by the core pulse module 3, the density direct pulse sent by the density direct pulse module 4, and the density reachable pulse sent by the pulse reachable pulse 5, wherein the core pulse sorting submodule 61 sorts the core pulse sent by the core pulse module 3, the density direct pulse sorting submodule 62 sorts the density direct pulse sent by the density direct pulse module 4, and the density reachable pulse sorting submodule 63 sorts the density reachable pulse sent by the density reachable pulse block 5.

[0096] It should be noted that the above-mentioned functional modules can be implemented in whole or in part through software, hardware, firmware, or any combination thereof. When implemented using software, they can be implemented in whole or in part in the form of a program instruction product. The program instruction product includes one or a group of program instructions. When the program instructions are loaded and executed on a computer, the process or function described is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The program instructions can be stored in a computer-readable and writable storage medium, or transferred from a computer-readable and writable storage medium to another computer-readable and writable storage medium.

[0097] The direct coupling or communication connection between the modules shown or discussed in this embodiment can be achieved through indirect coupling or communication connection of some interfaces, devices or modules. The various functional modules and submodules in this embodiment can be dynamically located in a processing component, or each module can exist physically separately, or two or more modules can be dynamically located in a processing component. When the above-mentioned dynamic components are implemented in the form of software functional modules and sold or used as independent products, they can also be stored in a computer-readable and writable storage medium. The storage medium can be a memory, a magnetic disk, an optical disk, etc.

[0098] The technical effects of the present invention can be further illustrated by the following simulation experiments.

[0099] 1. Simulation conditions and parameters:

[0100] Software environment: Intel(R) Core(TM) i9-14900HX 4.1GHz, Matlab R2024b simulation software under Windows 11 Home Chinese version 64-bit operating system.

[0101] The two data sets used in the simulation are shown in Table 1 and Table 2.

[0102] surface . Parameters of Dataset 1

[0103] category CF / MHz PW / μs DOA / (°) Radar 1 97~103 41~43 5~7 Radar 2 147~153 36~38 9~11 Radar 3 177~183 41~43 4~6 Radar 4 197~203 34~37 8~10 Radar 5 297~303 36~38 15~17

[0104] surface . Parameters of Dataset 2

[0105] category PW / μs CF / MHz DOA(°) Number of pulses Radar 1 20~30 135~145 32~34 100 Radar 2 27~37 120~130 59~61 170 Radar 3 17~27 135~145 59~61 150 Radar 4 18~28 165~175 40~42 90 Radar 5 27~37 105~115 67~69 60 Radar 6 15~25 150~160 67~69 45 Radar 7 20~30 145~155 40~42 300 Radar 8 15~25 165~175 55~57 100

[0106] In the above table, CF / MHz indicates that the parameter in this column is the carrier frequency CF, and the unit of carrier frequency is megahertz MHZ; PW / μs indicates that the parameter in this column is the bandwidth PW, and the unit of bandwidth is microsecond μs; DOA(°) indicates that the parameter in this column is the angle of arrival DOA, and the unit of arrival angle is degree (°).

[0107] 2. Simulation content and results:

[0108] Simulation 1: Under the above simulation conditions and parameters, radar pulses are sorted in dataset 1 using the present invention, the existing DBSCAN, and RNN-DBSCAN sorting methods. The results are as follows: Figure 5 As shown. Among them:

[0109] Figure 5 (a) is the simulation result of DBSCAN sorting method for pulse sorting of the first three radars in dataset 1;

[0110] Figure 5 (b) is the simulation result of sorting the pulses of the first three radars in dataset 1 using the RNN-DBSCAN sorting method;

[0111] Figure 5 (c) is the simulation result of the present invention for sorting the pulses of the first three radars in data set 1;

[0112] Figure 5 (d) is the simulation result of sorting the pulses of all five radars in dataset 1 using the DBSCAN sorting method;

[0113] Figure 5 (e) is the simulation result of sorting the pulses of all five radars in dataset 1 using the RNN-DBSCAN sorting method;

[0114] Figure 5 (f) is the simulation result of sorting the pulses of all five radars in data set 1 by the present invention.

[0115] Depend on Figure 5 (a) It can be seen that when sorting the first three radars, the DBSCAN sorting method can sort out three sorting clusters, but cluster 3 only contains one pulse.

[0116] Depend on Figure 5 (b) It can be seen that when sorting the first three radars, the RNN-DBSCAN sorting method can sort out three sorting clusters;

[0117] Depend on Figure 5 (c) It can be seen that when sorting the first three radars, the proposed method can successfully sort out three sorting clusters, and the number of sorting clusters is consistent with the number of radars;

[0118] Depend on Figure 5 (d) It can be seen that when sorting all five radars, the DBSCAN sorting method can only sort out four sorting clusters, among which cluster 3 contains only one pulse;

[0119] Depend on Figure 5 (e) It can be seen that when sorting all 5 radars, the RNN-DBSCAN sorting method can only sort out 4 sorting clusters;

[0120] Depend on Figure 5 (f) It can be seen that when sorting all five radars, the present invention can successfully sort out five sorting clusters, and the number of sorting clusters is consistent with the number of radars;

[0121] Depend on Figure 5 It can be seen from the six subgraphs that when the value of K is the same, the radar pulse sorting accuracy of the present invention is improved compared with the DBSCAN sorting method and the RNN-DBSCAN sorting method.

[0122] Simulation 2: Under the above simulation conditions and parameters, radar pulses are sorted in dataset 2 using the present invention, the existing DBSCAN, and RNN-DBSCAN sorting methods. The results are as follows: Figure 6 As shown. Among them:

[0123] Figure 6 (a) is the simulation result of DBSCAN sorting method for pulse sorting of the first three radars in dataset 2;

[0124] Figure 6 (b) is the simulation result of sorting the pulses of the first three radars in dataset 2 using the RNN-DBSCAN sorting method;

[0125] Figure 6 (c) is the simulation result of the present invention for sorting the pulses of the first three radars in data set 2;

[0126] Figure 6 (d) is the simulation result of sorting the pulses of the first 6 radars in dataset 2 using the DBSCAN sorting method;

[0127] Figure 6 (e) is the simulation result of sorting the pulses of the first 6 radars in dataset 2 using the RNN-DBSCAN sorting method;

[0128] Figure 6(f) is the simulation result of sorting the pulses of the first 6 radars in data set 2 by the present invention.

[0129] Figure 6 (g) is the simulation result of sorting the pulses of all 8 radars in dataset 2 using the DBSCAN sorting method;

[0130] Figure 6 (h) is the simulation result of sorting the pulses of all 8 radars in dataset 2 using the RNN-DBSCAN sorting method;

[0131] Figure 6 (i) is the simulation result of sorting the pulses of all 8 radars in data set 2 by the present invention.

[0132] from Figure 6 (a) It can be seen that when sorting the first three radars, the DBSCAN sorting method can only sort out two sorting clusters;

[0133] Depend on Figure 6 (b) It can be seen that when sorting the first three radars, the RNN-DBSCAN sorting method can only sort out two sorting clusters;

[0134] Depend on Figure 6 (c) It can be seen that when sorting the first three radars, the proposed method can successfully sort out three sorting clusters, and the number of sorting clusters is consistent with the number of radars;

[0135] Depend on Figure 6 (d) It can be seen that when sorting the first 6 radars, the DBSCAN sorting method can only sort out 3 sorting clusters;

[0136] Depend on Figure 6 (e) It can be seen that when sorting the first 6 radars, the DBSCAN sorting method can only sort out 5 sorting clusters;

[0137] Depend on Figure 6 (f) It can be seen that when sorting the first six radars, the proposed method can successfully sort out six sorting clusters, and the number of sorting clusters is consistent with the number of radars;

[0138] Depend on Figure 6 (g) It can be seen that when sorting all 8 radars, the DBSCAN sorting method can only sort out 4 sorting clusters;

[0139] Depend on Figure 6 (h) It can be seen that when sorting all 8 radars, the RNN-DBSCAN sorting method can only sort out 7 sorting clusters;

[0140] Depend on Figure 6(i) It can be seen that when sorting all 8 radars, the proposed method can successfully sort out 8 sorting clusters, and the number of sorting clusters is consistent with the number of radars;

[0141] Depend on Figure 6 It can be seen from the 9 subgraphs that when the value of K is the same, the radar pulse sorting accuracy of the present invention is improved compared with the DBSCAN sorting method and the RNN-DBSCAN sorting method.

[0142] The above descriptions are merely a few specific examples of the present invention and do not constitute any limitation to the present invention. It is obvious that for professionals in this field, after understanding the content and principles of the present invention, they may make various modifications and changes in form and details without departing from the principles and structure of the present invention. However, these modifications and changes based on the ideas of the present invention are still within the scope of protection of the claims of the present invention.

[0143] It should be noted that the step numbers in the specification and claims of the present invention are only for the purpose of clearly describing the embodiments of the present invention and facilitating understanding, and the order of the step numbers is not limited.

Claims

1. A signal sorting method based on improved mutual nearest neighbor DBSCAN, characterized by: include: 1) Use K nearest neighbors to construct the distance matrix of each pulse with respect to the pulse description word PDW; 2) Obtain the K nearest neighbor pulses of each pulse through the distance matrix and perform mutual neighbor determination: If the K nearest neighbors of each neighboring pulse all contain the pulse to be determined, the pulse to be determined is marked as a core point and step 3 is executed); Otherwise, the pulse to be determined is temporarily marked as a noise pulse; 3) Construct a core pulse matrix for all pulses that meet the core judgment and initialize the sorting labels; 4) Select any unsorted core pulse from the core pulse matrix, select the K neighboring pulses of the core pulse as the density direct pulses of the current core pulse and perform cluster division; 5) Construct a temporary pulse vector for the K density direct pulses of the core pulse, and according to the set scaling factor , select K of each density direct pulse The neighboring pulses are regarded as density-reachable pulses, and then they are clustered; 6) Repeat 4) to 5) for the core points and boundary points that have not been sorted until all core pulses are sorted.

2. The method according to claim 1, wherein: In 1), the distance matrix of each pulse with respect to the pulse description word PDW is constructed using K nearest neighbors, and its implementation includes: 1a) Assume there are M pulses, each pulse contains P pulse description word PDW parameters, and each pulse is marked with a sequence number; 1b) Build a Dimensional matrix and normalize it; 1c) Treat each normalized pulse as a point in P-dimensional space, use the pulse descriptor word (PDW) parameter of each normalized pulse as the coordinate of the pulse in P-dimensional space, and calculate the distance between pulse X and other pulses in P-dimensional space; 1d) Arrange the remaining M-1 pulses in order from nearest to farthest distance and construct a dimensional matrix, the first column of the matrix represents the sequence number of pulse X in M ​​pulses, and the subsequent M-1 columns are the sequence numbers of each pulse; 1e) For each pulse, construct the dimensional matrix, the final combination is dimensional matrix; 1f) Select according to parameter K Dimensional matrix Column, composition dimensional matrix, ignore dimensional matrix, reducing its dimensionality to dimensional distance matrix.

3. The method according to claim 2, wherein: In 1b) above, a dimensional matrix and normalize it. Its implementation includes: Assume that the pulse description word PDW parameters include carrier frequency CF, bandwidth PW, and arrival angle DOA parameters. Assume that the PDW parameters of the pulse X to be normalized are carrier frequency ,bandwidth , arrival angle ; Find the maximum and minimum values ​​of each pulse description word PDW parameter in all pulses, and subtract the minimum value of the pulse description word PDW parameter in all pulses from the current pulse description word PDW parameter of the current pulse to be normalized to obtain the carrier frequency CF parameter after pulse X normalization : ， in, and They are the minimum carrier frequency and the maximum carrier frequency among all pulse carrier frequency CF parameters respectively; Normalize the bandwidth PW parameter in the pulse description word PDW to obtain the normalized bandwidth PW parameter of pulse X : ， in, and They are the minimum bandwidth and maximum bandwidth of all pulse carrier frequency PW parameters respectively; Normalize the arrival angle DOA parameter in the pulse description word PDW to obtain the normalized arrival angle DOA parameter of pulse X : ， in, and They are the minimum arrival angle and the maximum arrival angle among all pulse carrier frequency DOA parameters respectively.

4. The method according to any one of claims 2 to 3, characterized in that: In 1c), the distance between pulse X and other pulses in P-dimensional space is calculated. , the formula is as follows: ， in, is the parameter carrier frequency CF in the pulse description word PDW after the normalization of pulse X, is the parameter carrier frequency CF in the normalized pulse description word PDW of pulse Y, is the angle of arrival (DOA) of the pulse description word (PDW) after normalization of pulse X. is the angle of arrival (DOA) parameter in the normalized pulse description word (PDW) of pulse Y, is the parameter bandwidth PW in the pulse description word PDW after the normalization of pulse X, is the parameter bandwidth PW in the normalized pulse description word PDW of pulse Y.

5. The method according to claim 1, wherein: In step 3), a core pulse matrix is ​​constructed for all pulses that meet the core judgment, and sorting labels are initialized. The implementation includes: 3a) Extract all Z core pulses that meet the mutual neighbor judgment and the K neighbor pulses of each core pulse in the P-dimensional space from the M pulses, and merge these core pulses that meet the mutual neighbor judgment and the K neighbor pulses of each core pulse into one dimensional core pulse matrix, and the Z pulses in the matrix that meet the mutual neighbor judgment are used as the core points during sorting; 3b) The sorting cluster labels of all core points are recorded as -1, that is, the pulse is not classified. The sorting cluster labels of the remaining pulses are recorded as 0, that is, these pulses are temporarily treated as noise points, and the initial number of sorting clusters is set to 0.

6. The method according to claim 1, wherein: In the above 4), the density of direct pulses of the current core pulse is clustered, and its implementation includes: 4a) Add 1 to the number of sorting clusters and let the sorting cluster at this time be numbered i; 4b_Among all core pulses, that is, among all elements in the first column of the core pulse matrix, randomly select a core pulse with a sorting cluster label of -1, and change the sorting cluster label of the current core pulse to be sorted to the number i of the current sorting cluster; 4c) Select the density direct pulses of the current pulse to be sorted, that is, the second element to the K+1th element in the row of the core pulse matrix where the current pulse to be sorted is located, and change the sorting cluster label corresponding to each density direct pulse to the current sorting cluster number i.

7. The method according to claim 1, wherein: In the above 5), a temporary pulse vector is constructed for the K dense direct pulses of the core pulse, and its implementation includes: 5a) In Find the current core pulse in the first column of the dimensional core pulse matrix; 5b) Assume that the current core pulse is located at The first column and the Yth row of the dimensional core pulse matrix are selected, and a total of K pulses from the 2nd column to the K+1th column of the Yth row are selected as temporary pulse vectors.

8. The method according to claim 1, characterized in that In 5), the density-reachable pulses are clustered, and the implementation thereof includes: 5c) taking the first pulse in the temporary vector with a sorting cluster label of 0 or -1 as the pulse to be sorted; 5d) In Find the pulse to be sorted in the first column of the dimensional matrix, and assume that the pulse to be sorted is located in The first column of the Y-th row in the dimensional matrix; 5e) In Select the first row from the Yth row and the 2nd column to the Kth column in the matrix pulses, merge them into a temporary pulse vector and remove duplicate pulses; 5f) Change the sorting cluster label corresponding to the pulse to be sorted to the current sorting cluster number i; 5g) Repeat 5c) to 5f) until all pulses in the temporary pulse vector are over-sorted.

9. A signal sorting system based on improved mutual nearest neighbor DBSCAN, characterized by: include: A distance calculation module is used to calculate the distance between each pulse and the rest of the pulses in the P-dimensional space; A mutual neighbor determination module is used to determine whether a pulse satisfies the mutual neighbor determination; A core pulse module is used to obtain core pulses that meet the mutual neighbor judgment among all pulses; Density direct access module, used to obtain density direct access pulses of each core pulse; Density reachable module, used to obtain density reachable pulses of each core pulse; The sorting module is used to determine which sorting cluster each pulse belongs to.

10. The system according to claim 7, wherein: The sorting module comprises: The core pulse sorting submodule is used to determine which sorting cluster each core pulse belongs to; The density direct pulse sorting submodule is used to determine to which sorting cluster each density direct pulse belongs; The density-reachable pulse sorting submodule is used to determine to which sorting cluster each density-reachable pulse belongs.

Images