Multifunctional radar signal sorting method based on weighted undirected graph features

Abstract

The invention discloses a multifunctional radar signal sorting method based on weighted undirected graph features, and the method comprises the steps: receiving a radiation source pulse sequence of a multifunctional radar, carrying out the normalization processing of the radiation source pulse sequence, dividing the normalized radiation source pulse sequence into subsequences through a sliding window, and enabling each subsequence to comprise a plurality of PDW data, each PDW data comprises a plurality of characteristic parameters; constructing a complex network with unweighted undirected graph features by using PDW data in the subsequences; on the basis of the complex network, community detection is achieved through a label propagation algorithm, and nodes with the same label are divided into the same sub-community; determining the local density and the relative distance of the sub-communities; and based on density peak clustering, determining a decision value by using the local density and the relative distance of the sub-communities, calculating weighted slope measurement, determining the number of clustering centers by using the slope measurement, and clustering the sub-communities based on the decision value and the number of the clustering centers, thereby realizing signal sorting of the radiation source pulse sequence of the multifunctional radar.

Description

Technical Field

[0001] The invention relates to the field of radar electronic reconnaissance and is a multifunctional radar signal sorting method based on weighted undirected graph features. Background Art

[0002] Radar emitter pulse signal sorting is an important part of the electronic reconnaissance field. It is an important premise and basis for radar signal feature extraction, identification and threat assessment. However, with the widespread application of multi-function radars, radar signal sorting faces new challenges. Multi-function radars will switch different working modes according to actual tasks and dynamically adjust parameter ranges and modulation types. This makes it easy for traditional pulse repetition interval (PRI)-based sorting methods to identify different working modes of a single multi-function radar as multiple radars or to identify two radars with similar parameters as one, resulting in "increased batches" and "missed batches". The existing multi-function radar signal sorting algorithm based on complex networks has achieved certain results by constructing pulse sequences into complex networks and using community detection algorithms to realize signal sorting. However, when using radar pulse sequences to construct complex networks, the correlation between pulse sequences of the same radar emitter is not considered, resulting in a sharp drop in sorting performance in the absence of PRI feature parameters, which has certain limitations. Summary of the invention

[0003] The purpose of the present invention is to provide a multifunctional radar signal sorting method based on weighted undirected graph features to overcome the problems existing in the prior art.

[0004] In order to achieve the above tasks, the present invention adopts the following technical solutions:

[0005] A multifunctional radar signal sorting method based on weighted undirected graph features, comprising:

[0006] The radiation source pulse sequence of the multi-function radar is received and normalized, and the normalized radiation source pulse sequence is divided into subsequences using a sliding window, each subsequence contains multiple PDW data, and each PDW data contains multiple characteristic parameters; the PDW data in the subsequence is used to construct a complex network of unweighted undirected graph features;

[0007] Based on the complex network, a label propagation algorithm is used to implement community detection, and nodes with the same label are divided into the same subcommunity;

[0008] Determine the local density and relative distance of the subcommunity; based on density peak clustering, use the local density and relative distance of the subcommunity to determine the decision value, and then calculate the weighted slope metric on this basis, use the slope metric to determine the number of cluster centers, cluster the subcommunities based on the decision value and the number of cluster centers, thereby realizing signal sorting of the radiation source pulse sequence of the multi-function radar.

[0009] Furthermore, the radiation source pulse sequence is normalized and expressed as:

[0010]

[0011] in, represents the t-th characteristic parameter in the i-th PDW data in the radiation source pulse sequence, t=1,2,...,T, T is the dimension of the characteristic parameter, N is the length of the radiation source pulse sequence, Represents the normalized

[0012] Furthermore, the normalized radiation source pulse sequence is divided into subsequences using a sliding window, each subsequence contains multiple PDW data, and each PDW data contains multiple characteristic parameters; the PDW data in the subsequence is used to construct a complex network of unweighted undirected graph features, including:

[0013] For the normalized radiation source pulse sequence, each PDW data is used as a node in the complex network, and the radiation source pulse sequence is divided into multiple subsequences using a sliding window; W and L are defined as the length of the sliding window and the sliding step, respectively. The entire sequence can be divided into N-W+1 sliding windows, each window contains W nodes, and these PDW data constitute a subsequence;

[0014] For the tth characteristic parameter in the i-th PDW data in the window calculate and the tth characteristic parameter of the remaining W-1 nodes The distance is sorted from small to large, and the K feature parameters with the smallest distance are selected. Connect the edges;

[0015] Using the adjacency matrix A t To represent the complex network of the tth characteristic parameter, if two nodes x i ,x j The tth characteristic parameter If there is an edge between them, then the corresponding element in the i-th row and j-th column of the adjacency matrix Thus, a complex network for the tth characteristic parameter is constructed;

[0016] Each characteristic parameter of the PDW data node in the subsequence can construct a complex network by superimposing the adjacency matrix of all characteristic parameters, that is, This can lead to a complex network of weighted undirected graph features for the nodes in each subsequence.

[0017] Furthermore, the calculation and the tth characteristic parameter of the remaining W-1 nodes The distance is sorted from small to large, and the K feature parameters with the smallest distance are selected. To connect the edges, specifically expressed as:

[0018] And|S K |=K

[0019] Where V represents the set of the tth feature parameters of all nodes in the current sliding window, Represents two nodes x i ,x j The tth characteristic parameter The distance between K Represents the node x i The tth characteristic parameter The set of the K t-th feature parameters that are closest to each other.

[0020] Furthermore, based on the complex network, the community detection is realized by using a label propagation algorithm to divide nodes with the same label into the same subcommunity, including:

[0021] Randomly assign a unique label to each node in the complex network, and sort all nodes in descending order according to the size of the clustering coefficient of each node;

[0022] Update the labels of the nodes sorted in descending order and propagate the node labels to the nodes with the greatest similarity;

[0023] Repeat the above steps until the label of each node no longer changes, then stop the iteration process, and then divide the nodes with the same label into the same subcommunity.

[0024] Furthermore, for each node x i The clustering coefficient ε i The calculation formula is:

[0025]

[0026] Among them, s i Represents node x i The strength of k i Represents node x i The degree, w ij Represents node x i ,x j The weight of the edge is if and only if x i ,x j ,x k When three nodes can form a triangle, a ij a ik a jk=1, otherwise a ij a ik a jk =0; where a ij For node x i ,x j The connection relationship in a complex network, with a value of 1 or 0.

[0027] Furthermore, two nodes x i ,x j The similarity between them is expressed as:

[0028]

[0029] where d ij Represents node x i ,x j The Euclidean distance between .

[0030] Furthermore, the local density of each subcommunity It is expressed as:

[0031]

[0032] in, represents subcommunity c m and another subcommunity c n The average distance between the node pairs x and y, and d(x,y) represents the subcommunity c m Node x and subcommunity c in n The Euclidean distance between nodes y in , d c Represents the neighborhood cutoff distance.

[0033] Furthermore, before calculating the relative distance of subcommunities, the local density of each subcommunity is sorted; for the subcommunity with the highest local density Its relative distance Defined as:

[0034]

[0035] For the remaining sub-communities Its relative distance Defined as:

[0036]

[0037] in, Subcommunities Subcommunity c n The local density of .

[0038] Furthermore, setting the decision value in represents subcommunity c m The local density of all subcommunities c m The decision value of And sort them in descending order, take the decision values ​​of the first P subcommunities, and calculate the weighted slope metric, which is expressed as:

[0039]

[0040] Where (l-1) represents the weight of the slope metric, represents the decision value of the lth subcommunity among the first P subcommunities, γ max , γ min Represents the maximum and minimum values ​​of the decision values ​​of the first P subcommunities;

[0041] Select the largest k l The corresponding parameter l is the number of cluster centers, and the decision value is selected The first l largest subcommunities are taken as cluster centers, and the remaining subcommunities are clustered according to their density peaks.

[0042] A terminal device comprises a processor, a memory and a computer program stored in the memory; when the processor executes the computer program, the multifunctional radar signal sorting method based on weighted undirected graph features is implemented.

[0043] A computer-readable storage medium stores a computer program; when the computer program is executed by a processor, the multifunctional radar signal sorting method based on weighted undirected graph features is implemented.

[0044] Compared with the prior art, the present invention has the following technical features:

[0045] 1. The present invention proposes a sliding window nearest neighbor connection method, which increases the number of pulse connections of the same radar radiation source, reduces the number of connections of pulses of different radar radiation sources, and speeds up the construction of complex networks.

[0046] 2. The present invention proposes a method for constructing a weighted undirected graph feature complex network with multi-PDW parameter fusion, which improves the accuracy of sorting multi-function radar signals.

[0047] 3. The present invention adopts an improved density peak clustering algorithm, which can automatically select cluster centers and improve the stability and reliability of the algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic diagram of the process of the present invention;

[0049] Figure 2 A model for building a complex network based on radar pulse sequences in the present invention;

[0050] Figure 3 The sample distribution diagram of radar simulation parameters in the feature space, where (a) is the original data and (b) is the normalized data;

[0051] Figure 4 These are the sorting results corresponding to four different methods, where (a) is the sorting result of the k-means algorithm, (b) is the sorting result of the DBSCAN algorithm, (c) is the sorting result of the C-Net algorithm, and (d) is the sorting result of the method of the present invention. DETAILED DESCRIPTION

[0052] The present invention considers the correlation between radar radiation source pulse sequences, proposes a multifunctional radar signal sorting method based on weighted undirected graph features, uses a sliding window nearest neighbor connection method to process radar pulse description words (PDWs), constructs a weighted undirected graph feature complex network of multi-PDW parameter fusion, and combines an improved density peak clustering algorithm to achieve radar signal sorting, thereby improving the sorting performance of multifunctional radar signals. The method of the present invention comprises the following steps:

[0053] Step 1, receive the radiation source pulse sequence of the multi-function radar and normalize it, use the sliding window to divide the normalized radiation source pulse sequence into subsequences, each subsequence contains multiple PDW data, and each PDW data contains multiple feature parameters; use the PDW data in the subsequence to construct a complex network of unweighted undirected graph features.

[0054] In this scheme, the received radiation source pulse sequence is expressed as X=(x 1 ,x 2 ,...,x N ), where x i represents the i-th PDW data, x i =(PW i ,RF i ,DOA i ), PW i ,RF i ,DOA i They represent the three characteristic parameters of the pulse width, repetition rate and arrival angle of the i-th PDW data respectively; N is the length of the radiation source pulse sequence, that is, the number of PDW data.

[0055] Step 1.1, normalize the radiation source pulse sequence. The normalization method used here is maximum and minimum value normalization, which is specifically expressed as follows:

[0056]

[0057] in, represents the t-th characteristic parameter in the i-th PDW data, t=1, 2, ..., T, T is the dimension of the characteristic parameter, which is 3 in this embodiment; Represents the normalized

[0058] Step 1.2, for the normalized radiation source pulse sequence, take each PDW data as a node in the complex network, and use the sliding window to divide the radiation source pulse sequence into multiple subsequences; define W and L as the length of the sliding window and the sliding step, respectively, usually satisfying W<<N, L=1; therefore, the entire sequence can be divided into N-W+1 sliding windows, each window contains W nodes, and these PDW data constitute a subsequence.

[0059] For the tth characteristic parameter in the i-th PDW data in the window calculate and the tth characteristic parameter of the remaining W-1 nodes The distance is sorted from small to large, and the K feature parameters with the smallest distance are selected. To connect the edges, specifically expressed as:

[0060]

[0061] Where V represents the set of the tth feature parameters of all nodes in the current sliding window, Represents two nodes x i ,x j The tth characteristic parameter The distance between K Represents the node x i The tth characteristic parameter The set of the K t-th feature parameters that are closest to each other.

[0062] Using the adjacency matrix A t To represent the complex network of the tth characteristic parameter, if two nodes x i ,x j The tth characteristic parameter If there is an edge between them, then the corresponding element in the i-th row and j-th column of the adjacency matrix In this way, a complex network for the tth characteristic parameter is constructed.

[0063] Step 1.3, each characteristic parameter of the PDW data node in the subsequence can construct a complex network by superimposing the adjacency matrix of all characteristic parameters, that is, This can lead to a complex network of weighted undirected graph features for the nodes in each subsequence.

[0064] Step 2: Based on the complex network, community detection is implemented using a label propagation algorithm to divide nodes with the same label into the same subcommunity.

[0065] Step 2.1, randomly assign a unique label to each node in the complex network, and sort all nodes in descending order according to the size of the clustering coefficient of each node; where each node x i The clustering coefficient ε i The calculation formula is:

[0066]

[0067] Among them, s i Represents node x i The strength of k i Represents node x i The degree, w ij Represents node x i ,x j The weight of the edge is if and only if x i ,x j ,x k When three nodes can form a triangle, a ij a ik a jk =1, otherwise a ij a ik a jk =0; where a ij For node x i ,x j The connection relationship in a complex network, with a value of 1 or 0.

[0068] Step 2.2, update the labels of the nodes after descending sorting, and propagate the labels of the nodes to the nodes with the greatest similarity; among which two nodes x i ,x j The similarity between them is expressed as:

[0069]

[0070] where d ij Represents node x i ,x j The Euclidean distance between .

[0071] Step 2.2: Repeat step 2.2 until the label of each node no longer changes, then stop the iteration process, and then divide the nodes with the same label into the same subcommunity.

[0072] Step 3, determining the local density and relative distance of the subcommunity; based on density peak clustering, using the local density and relative distance of the subcommunity to determine the decision value, and then calculating the weighted slope metric on this basis, using the slope metric to determine the number of cluster centers, clustering the subcommunities based on the decision value and the number of cluster centers, thereby realizing signal sorting of the radiation source pulse sequence of the multi-function radar.

[0073] Step 3.1, calculate the local density of each subcommunity It is expressed as:

[0074]

[0075] in, represents subcommunity c m and another subcommunity c n The average distance between the node pairs x and y, and d(x,y) represents the subcommunity c m Node x and subcommunity c in n The Euclidean distance between nodes y in the neighborhood. The larger the Euclidean distance between node pairs, the smaller the local density. c The selection of should make the average number of neighbors of each node about 1% to 2% of the total number of nodes. The Euclidean distance between nodes within the subcommunity is set to 0.

[0076] Step 3.2, calculate the relative distance of each subcommunity.

[0077] Before calculating the relative distance of subcommunities, the local density of each subcommunity is sorted; for the subcommunity with the highest local density Its relative distance Defined as:

[0078]

[0079] For the remaining sub-communities Its relative distance Defined as:

[0080]

[0081] in, Subcommunities Subcommunity c n The local density of .

[0082] Step 3.3, set decision value in represents subcommunity c m The local density of all subcommunities c m The decision value of And sort them in descending order, take the decision values ​​of the first P subcommunities, and calculate the weighted slope metric, which is expressed as:

[0083]

[0084] Where (l-1) represents the weight of the slope metric, represents the decision value of the lth subcommunity among the first P subcommunities, γ max , γ min Indicates the maximum and minimum values ​​of the decision values ​​of the first P subcommunities; in this embodiment, the value of P is 40.

[0085] Step 4.4, select the largest k l The corresponding parameter l is the number of cluster centers, and the decision value is selected The first l largest subcommunities are used as cluster centers, and the remaining subcommunities are clustered according to density peaks, thus completing the signal sorting process.

[0086] Example:

[0087] In order to verify the effectiveness of the algorithm, this example simulated the PDW parameters of four radars, as shown in Table 1. Specifically, two conventional radars and two multi-function radars were simulated, including four parameters of PW, RF, DOA and PRI with multiple modulation modes, and the parameters of the four radars overlapped to varying degrees. In addition, PRI is used to count the number of pulses within the simulation time, which is difficult to accurately estimate in the measured data, so the three characteristic parameters of PW, RF, and DOA are selected for sorting.

[0088] Table 1 Radar signal simulation parameters

[0089]

[0090] Based on the comparison between this method and the k-means clustering algorithm, DBSACN algorithm and the original complex network-based multifunctional radar signal sorting algorithm (C-Net), the parameters set by the four methods are:

[0091] k-means algorithm: the initial number of clusters is set to 4, and the maximum number of iterations is set to 100;

[0092] DBSCAN algorithm: The neighborhood radius is set to 0.1, and the minimum number of neighborhood sample points is set to 100;

[0093] C-Net algorithm: The sliding window length is set to 10 and the penetration distance is set to 2;

[0094] The method of the present invention: the sliding window length W is set to 10, and the number of connection nodes K is set to 2.

[0095] The sorting results of the proposed method and the other three algorithms on the radar PDW data set are shown in the following figure: Figure 4 As shown, (a) is the sorting result of the k-means algorithm, (b) is the sorting result of the DBSCAN algorithm, (c) is the sorting result of the C-Net algorithm, and (d) is the sorting result of the method of the present invention.

[0096] By analyzing Figure 4 It is found that there are two defects in the k-means algorithm. One is that the initial number of clusters needs to be specified, and the other is that it cannot process clusters of arbitrary shapes. The multifunctional radar signals with overlapping parameters cannot be accurately sorted. For the DBSCAN algorithm, it is greatly affected by the two parameters of neighborhood radius and minimum number of neighborhood samples. In the parameters set in this experiment, radar 4 is mistakenly removed as a noise point. For the C-Net algorithm, the correlation between radar signals is not considered when constructing a complex network, and radars 2, 3 and some pulses under working mode Ⅰ of radar 4 are mistakenly divided into the same radar. For the method of the present invention, the multifunctional radar signal sorting can be well realized, and only a small number of pulses under working mode Ⅰ of radar 4 are identified as conventional radar 3. The overall effect is better than the other three methods.

[0097] The sorting performance of the four algorithms is shown in Table 2. Compared with other methods, the method of the present invention has a higher sorting accuracy. This is because the method tries to connect the pulse signals of the same radiation source to avoid the connection between the signals of different radiation sources. At the same time, the weighting coefficient is considered to construct a complex network with weighted undirected graph characteristics, which effectively improves the accuracy of multi-function radar signal sorting. For the k-means algorithm, the number of cluster centers needs to be confirmed in advance and clusters of arbitrary shapes cannot be processed. The sorting of multi-function radar signals with overlapping parameters is not accurate enough; for the DBSCAN algorithm, it is greatly affected by the two parameters of neighborhood radius and the minimum number of neighborhood samples, and is prone to "increased batches" and "missed batches" problems; for the C-Net algorithm, when the PRI parameter is missing as an input feature, the "missed batch" phenomenon occurs, and the signal cannot be effectively sorted;

[0098] Table 2 Comparison of sorting performance of four algorithms

[0099] algorithm Radar 1 Radar 2 Radar 3 Radar 4 Total correct rate k-means 100％ 28.57％ 100％ 87％ 75.38％ DBSACN 100％ 43.14％ 100％ 0％ 54.25％ C-Net 100％ 100％ 0％ 90.57％ 76.417％ The present invention 100％ 100％ 100％ 92.57％ 97.83％

[0100] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A multifunctional radar signal sorting method based on weighted undirected graph features, characterized in that: include: A radiation source pulse sequence of a multifunctional radar is received and normalized, and a sliding window is used to divide the normalized radiation source pulse sequence into subsequences, each subsequence contains a plurality of PDW data, and each PDW data contains a plurality of characteristic parameters; The PDW data in the subsequence is used to construct a complex network of unweighted and undirected graph features; Based on the complex network, a label propagation algorithm is used to implement community detection, and nodes with the same label are divided into the same subcommunity; Determine the local density and relative distance of the subcommunity; based on density peak clustering, use the local density and relative distance of the subcommunity to determine the decision value, and then calculate the weighted slope metric on this basis, use the slope metric to determine the number of cluster centers, cluster the subcommunities based on the decision value and the number of cluster centers, thereby realizing signal sorting of the radiation source pulse sequence of the multi-function radar.

2. The multifunctional radar signal sorting method based on weighted undirected graph features according to claim 1 is characterized in that: The radiation source pulse sequence is normalized and expressed as: in, represents the t-th characteristic parameter in the i-th PDW data in the radiation source pulse sequence, t=1,2,...,T, T is the dimension of the characteristic parameter, N is the length of the radiation source pulse sequence, Represents the normalized 3. The multifunctional radar signal sorting method based on weighted undirected graph features according to claim 1 is characterized in that: The normalized radiation source pulse sequence is divided into subsequences using a sliding window, each subsequence contains multiple PDW data, and each PDW data contains multiple characteristic parameters; The PDW data in the subsequence is used to construct a complex network of unweighted undirected graph features, including: For the normalized radiation source pulse sequence, each PDW data is used as a node in the complex network, and the radiation source pulse sequence is divided into multiple subsequences using a sliding window; W and L are defined as the length of the sliding window and the sliding step, respectively. The entire sequence can be divided into N-W+1 sliding windows, each window contains W nodes, and these PDW data constitute a subsequence; For the tth characteristic parameter in the i-th PDW data in the window calculate and the tth characteristic parameter of the remaining W-1 nodes The distance is sorted from small to large, and the K feature parameters with the smallest distance are selected. Connect the edges; Using the adjacency matrix A t To represent the complex network of the tth characteristic parameter, if two nodes x i ,x j The tth characteristic parameter If there is an edge between them, then the corresponding element in the i-th row and j-th column of the adjacency matrix Thus, a complex network for the tth characteristic parameter is constructed; Each characteristic parameter of the PDW data node in the subsequence can construct a complex network by superimposing the adjacency matrix of all characteristic parameters, that is, This can lead to a complex network of weighted undirected graph features for the nodes in each subsequence.

4. The multifunctional radar signal sorting method based on weighted undirected graph features according to claim 1 is characterized in that: The calculation and the tth characteristic parameter of the remaining W-1 nodes The distance is sorted from small to large, and the K feature parameters with the smallest distance are selected. To connect the edges, specifically expressed as: Where V represents the set of the tth feature parameters of all nodes in the current sliding window, Represents two nodes x i ,x j The tth characteristic parameter The distance between K Represents the node x i The tth characteristic parameter The set of the K t-th feature parameters that are closest to each other.

5. The multifunctional radar signal sorting method based on weighted undirected graph features according to claim 1 is characterized in that: The method of implementing community detection based on the complex network and using a label propagation algorithm to divide nodes with the same label into the same subcommunity includes: Randomly assign a unique label to each node in the complex network, and sort all nodes in descending order according to the size of the clustering coefficient of each node; Update the labels of the nodes sorted in descending order and propagate the node labels to the nodes with the greatest similarity; Repeat the above steps until the label of each node no longer changes, then stop the iteration process, and then divide the nodes with the same label into the same subcommunity.

6. The multifunctional radar signal sorting method based on weighted undirected graph features according to claim 1 is characterized in that: Each node x i The clustering coefficient ε i The calculation formula is: Among them, s i Represents node x i The strength of k i Represents node x i The degree, w ij Represents node x i ,x j The weight of the edge is if and only if x i ,x j ,x k When three nodes can form a triangle, a ij a ik a jk =1, otherwise a ij a ik a jk =0; where a ij For node x i ,x j The connection relationship in a complex network, with a value of 1 or 0.

7. The multifunctional radar signal sorting method based on weighted undirected graph features according to claim 1 is characterized in that: The local density of each subcommunity It is expressed as: in, represents subcommunity c m and another subcommunity c n The average distance between the node pairs x and y, and d(x,y) represents the subcommunity c m Node x and subcommunity c in n The Euclidean distance between nodes y in , d c Represents the neighborhood cutoff distance.

8. The multifunctional radar signal sorting method based on weighted undirected graph features according to claim 1 is characterized in that: Before calculating the relative distance of subcommunities, the local density of each subcommunity is sorted; for the subcommunity with the highest local density Its relative distance Defined as: For the remaining sub-communities Its relative distance Defined as: in, Subcommunities Subcommunity c n The local density of .

9. The multifunctional radar signal sorting method based on weighted undirected graph features according to claim 1 is characterized in that: Setting decision values in represents subcommunity c m The local density of all subcommunities c m The decision value of And sort them in descending order, take the decision values ​​of the first P subcommunities, and calculate the weighted slope metric, which is expressed as: Where (l-1) represents the weight of the slope metric, represents the decision value of the lth subcommunity among the first P subcommunities, γ max , γ min Represents the maximum and minimum values ​​of the decision values ​​of the first P subcommunities; Select the largest k l The corresponding parameter l is the number of cluster centers, and the decision value is selected The first l largest subcommunities are taken as cluster centers, and the remaining subcommunities are clustered according to their density peaks.

10. A terminal device comprising a processor, a memory and a computer program stored in the memory; characterized in that: When the processor executes the computer program, the multifunctional radar signal sorting method based on weighted undirected graph features according to any one of claims 1 to 9 is implemented.

Images