A radar signal sorting method and system based on a clustering effectiveness index DPDBI

Abstract

This invention provides a radar signal sorting method and system based on the Clustering Effectiveness Index (DPDBI), relating to the field of clustering sorting. The method includes: S1: acquiring a radar signal dataset and preprocessing it to obtain a distance matrix; S2: estimating the kernel density of the distance matrix to obtain a neighborhood radius; dividing the radar signal dataset by step size using the neighborhood radius to obtain a set of parameter pairs; S3: clustering the set of parameter pairs to obtain a set of clustering results; S4: scoring and ranking the set of clustering results using the DPDBI algorithm to obtain the optimal clustering result. This invention proposes a DPDBI algorithm for scoring and ranking the set of clustering results. By setting density cluster centers and weight coefficients in the DPDBI scoring calculation, the DPDBI algorithm improves the effectiveness of DPDBI calculation in the clustering process of radar signals and other high-dimensional data, avoiding the problem of traditional clustering effectiveness indices being difficult to identify when over-segmentation or abnormally small clusters occur.

Description

Technical Field

[0001] This invention relates to the field of clustering and sorting, and in particular to a radar signal sorting method and system based on the clustering effectiveness index DPDBI. Background Technology

[0002] Current radar signal sorting algorithms face the challenge that, with the increasingly diverse operating modes of advanced radar systems such as MIMO radar, the radar signals detected by the receiver are characterized by non-uniformity, high density, a large number of pulses, and severe pulse parameter overlap. Existing algorithms have not completely overcome their reliance on manually set parameters and struggle to handle situations with severe pulse parameter overlap.

[0003] Some classic clustering effectiveness metrics, such as the Calinski and Harabasz index (CH), the original Dunn index (DNo), and the silhouette coefficient, can be used on many classic public datasets, but they often perform poorly when dealing with overlapping data. The DNo index and its improvements, such as the generalized Dunn index (DNg) and the modified Dunn's index (DNs), are calculated by weighing the ratio of the minimum inter-cluster separation to the maximum. The silhouette coefficient measures the similarity of a sample point to its cluster, with values ​​ranging from -1 to 1, where values ​​close to 1 indicate a high degree of match between the sample point and its current cluster. The Davies-Bouldin index, proposed by Davies and Bouldin, considers both inter-cluster separation and intra-cluster compactness, but it does not account for the impact of uneven cluster densities and is also difficult to apply to high-dimensional data. Most existing algorithms for radar signal sorting are improvements on the K-means algorithm. Some algorithms can determine initial cluster centers, reducing dependence on them, but K-means still relies on prior information. Some improved fuzzy C-means (FCM) algorithms optimize the initial threshold setting, but FCM performs poorly on datasets with uneven density. In summary, existing algorithms cannot solve the challenge of sorting overlapping radar signals with parameters determined simultaneously without prior information. Summary of the Invention

[0004] In view of this, the purpose of this invention is to provide a radar signal sorting method based on the clustering effectiveness index DPDBI that can solve the problem of simultaneously determining parameters in the absence of prior information.

[0005] This invention provides a radar signal sorting method based on the clustering effectiveness index DPDBI, comprising:

[0006] S1: Obtain the radar signal dataset, preprocess the radar signal dataset, and obtain the range matrix;

[0007] S2: Perform kernel function density estimation on the distance matrix to obtain the neighborhood radius; divide the radar signal dataset by step size using the neighborhood radius to obtain the parameter pair set;

[0008] S3: Cluster the parameter pair set to obtain the clustering result set;

[0009] S4: The clustering result set is scored and ranked using the DPDBI algorithm to obtain the optimal clustering result.

[0010] Preferably, step S1 specifically includes:

[0011] S11: Extract the parameters of each radar signal in the radar signal dataset and construct the pulse descriptor matrix D, expressed as:

[0012] D = {p1, p2, p3, ..., p} i , ..., p n )

[0013] Where i is the radar signal number, n is the total number of radar signals, and p i p is the pulse descriptor for radar signal i. i =[cf i ,pw i doa i ], cf i Let pw be the carrier frequency of radar signal i. i Let i be the pulse width of the radar signal, doa i The direction of arrival of radar signal i;

[0014] S12: Normalize the pulse descriptor matrix to obtain a normalized matrix;

[0015] S13: Calculate the Euclidean distance between every two pulse descriptors in the normalized matrix, and construct a distance matrix using the Euclidean distances.

[0016] Preferably, step S2 specifically includes:

[0017] S21: Perform kernel function density estimation on each Euclidean distance in the distance matrix to obtain the probability density distribution of each Euclidean distance, and construct the probability density distribution curve through each probability density distribution;

[0018] S22: Use the x-coordinate of the first peak point in the probability density distribution curve as the neighborhood radius Eps;

[0019] S23: Traverse each radar signal in the radar signal dataset to obtain the minimum number of contained points (Minpts) within the neighborhood radius Eps for each radar signal. The formula for calculating the minimum number of contained points is:

[0020]

[0021] Where i is the radar signal number, n is the total number of radar signals, and p i N is the pulse descriptor for radar signal i. eps (p i ) represents the number of points contained in the neighborhood radius Eps of radar signal i;

[0022] S24: Divide the neighborhood radius Eps and the minimum number of contained points Minpts step size into multiple elements to construct the neighborhood radius list Eps_list and the minimum number of contained points list Minpts_list.

[0023] S25: Combine each element in the neighborhood radius list Eps_list with each element in the minimum number of contained points list Minpts_list to obtain multiple parameter pairs, and form a parameter pair set from these parameter pairs.

[0024] Preferably, step S4 specifically includes:

[0025] S41: Obtain the clustering result set C = {C1, C2, ..., C...} k}, where k is the total number of clusters, C k For the k-th cluster, the local density parameters of each pulse descriptor within the neighborhood radius Eps are calculated in the distance matrix, with each pulse descriptor as the center and the neighborhood radius Eps as the radius.

[0026] S42: Extract cluster C l and cluster C m Where l and m are cluster numbers, l≠m; the density clustering center A of cluster C1 is obtained by calculating the local density parameters. l ;

[0027] S43: Configure cluster C l Weighting coefficient γ l Through γ l and A l Cluster C was calculated l Cluster density S l ;

[0028] S44: Configure cluster C l and cluster C m Inter-cluster separation D lm via S1 and D lm Cluster C was calculated l and cluster Cm Clustering comprehensive measure Cs lm ;

[0029] S45: Repeat steps S42-S44 to calculate the DPDBI score for each cluster, and take the cluster with the smallest DPDBI score as the optimal clustering result.

[0030] Preferably, the formula for calculating the local density parameter in step S41 is:

[0031]

[0032] Where i and j are the radar signal numbers, ρ ij Let d be the local density parameter between radar signals i and j. ij Let x be the Euclidean distance between radar signals i and j; when x > 0, χ(x) = 0; otherwise, χ(x) = 1.

[0033] Preferably, in step S42, the density cluster center A l The calculation formula is:

[0034]

[0035] Where, ρ im Let be the local density parameter between radar signals i and m.

[0036] Preferably, in step S43, the weighting coefficient γ l The calculation formula is:

[0037]

[0038] Among them, |C l |For cluster C l The number of radar signals in the middle, where N is the total number of radar signals successfully sorted;

[0039] Cluster density S l The calculation formula is:

[0040]

[0041] Where j is the radar signal number, p j Let be the pulse description word for radar signal j.

[0042] Preferably, in step S44, the inter-cluster separation degree D lm The calculation formula is:

[0043]

[0044] Among them, A m For cluster C mThe density of cluster centers, where q is a set parameter, p il For cluster C l The i-th pulse description word within, p jm For cluster C m The j-th pulse description word within;

[0045] Clustering Comprehensive Measure Cs lm The calculation formula is:

[0046]

[0047] Preferably, the formula for calculating the DPDBI score in step S45 is as follows:

[0048]

[0049] A radar signal sorting system based on the clustering effectiveness index DPDBI includes:

[0050] The distance matrix acquisition module is used to acquire radar signal datasets, preprocess the radar signal datasets, and obtain the distance matrix.

[0051] The parameter pair set acquisition module is used to perform kernel function density estimation on the distance matrix to obtain the neighborhood radius; and to divide the radar signal dataset by step size using the neighborhood radius to obtain the parameter pair set.

[0052] The clustering result set acquisition module is used to cluster the parameter pair set and obtain the clustering result set;

[0053] The optimal clustering result acquisition module is used to score and rank the clustering result set using the DPDBI algorithm to obtain the optimal clustering result.

[0054] The present invention has the following beneficial effects:

[0055] A DPDBI algorithm is proposed to score and rank clustering results. By setting density cluster centers and weight coefficients to participate in the DPDBI score calculation, the DPDBI algorithm improves the effectiveness of DPDBI calculation in the clustering process of radar signals and other high-dimensional data. It avoids the problem that traditional clustering effectiveness indicators are difficult to identify when there is over-segmentation and abnormally small clusters, and realizes parameter-free adaptive clustering. Attached Figure Description

[0056] Figure 1 This is a flowchart of a method according to an embodiment of the present invention;

[0057] Figure 2 This is a schematic diagram of the probability density distribution curve;

[0058] Figure 3 This is a schematic diagram of clustering results on a high-dimensional dataset;

[0059] Figure 4 This is a schematic diagram of the clustering results on a two-dimensional dataset;

[0060] Figure 5 This is the first embodiment of the DPDBI algorithm for obtaining optimal clustering results;

[0061] Figure 6 This is a second embodiment of the DPDBI algorithm for obtaining optimal clustering results;

[0062] Figure 7 This is a third embodiment of the DPDBI algorithm for obtaining optimal clustering results;

[0063] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0064] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0065] Reference Figure 1 This invention provides a radar signal sorting method based on the clustering effectiveness index DPDBI, comprising:

[0066] S1: Obtain the radar signal dataset, preprocess the radar signal dataset, and obtain the range matrix;

[0067] S2: Perform kernel function density estimation on the distance matrix to obtain the neighborhood radius; divide the radar signal dataset by step size using the neighborhood radius to obtain the parameter pair set;

[0068] S3: Cluster the parameter pair set to obtain the clustering result set;

[0069] S4: The clustering result set is scored and ranked using the DPDBI algorithm to obtain the optimal clustering result.

[0070] Furthermore, step S1 specifically includes:

[0071] S11: Extract the parameters of each radar signal in the radar signal dataset and construct the pulse descriptor matrix D, expressed as:

[0072] D = {p1, p2, p3, ..., p} i , ..., p n )

[0073] Where i is the radar signal number, n is the total number of radar signals, and p i p is the pulse descriptor for radar signal i. i =[cf i ,pwi doa i ], cf i Let pw be the carrier frequency of radar signal i. i Let i be the pulse width of the radar signal, doa i The direction of arrival of radar signal i;

[0074] S12: Normalize the pulse descriptor matrix to obtain a normalized matrix;

[0075] S13: Calculate the Euclidean distance between every two pulse descriptors in the normalized matrix, and construct a distance matrix using the Euclidean distances.

[0076] Specifically, the formula for calculating Euclidean distance is:

[0077]

[0078] Furthermore, step S2 specifically involves:

[0079] S21: Perform kernel function density estimation on each Euclidean distance in the distance matrix to obtain the probability density distribution of each Euclidean distance, and construct the probability density distribution curve through each probability density distribution;

[0080] Specifically, the Epanechnikov kernel is used to estimate the kernel density of each Euclidean distance in the distance matrix, and the resulting probability density distribution curve is shown below. Figure 2 As shown, the kernel function density estimation is calculated as shown in formula (2):

[0081]

[0082] Where K(·) is the kernel function, and h is the bandwidth of the kernel function;

[0083] Figure 2 The curve shown is formed by kernel density estimation of the distance matrix of a dataset consisting of 1039 radar signal data points. This dataset contains three different radar signal sources, which means there are three cluster centers. First, we need to calculate the distance between all sample points and generate a distance matrix Dist. Using the kernel density estimation method, we plotted... Figure 2 The density curve in the image shows three peaks. The first peak represents the distance with higher density within a cluster, while the second and third peaks represent the distance with higher density between clusters. Therefore, we should consider selecting the value corresponding to the first peak as the candidate interval for Eps, i.e., Eps∈(0,0.31).

[0084] S22: Use the x-coordinate of the first peak point in the probability density distribution curve as the neighborhood radius Eps;

[0085] S23: Traverse each radar signal in the radar signal dataset to obtain the minimum number of contained points (Minpts) within the neighborhood radius Eps for each radar signal. The formula for calculating the minimum number of contained points is:

[0086]

[0087] Where i is the radar signal number, n is the total number of radar signals, and p i N is the pulse descriptor for radar signal i. eps (p i ) represents the number of points contained in the neighborhood radius Eps of radar signal i;

[0088] S24: Divide the neighborhood radius Eps and the minimum number of contained points Minpts step size into multiple elements to construct the neighborhood radius list Eps_list and the minimum number of contained points list Minpts_list.

[0089] S25: Combine each element in the neighborhood radius list Eps_list with each element in the minimum number of contained points list Minpts_list to obtain multiple parameter pairs, and form a parameter pair set from these parameter pairs.

[0090] Furthermore, step S3 specifically involves: using the set of parameter pairs as input parameters for the DBSCAN algorithm to cluster the radar signal dataset. Each parameter pair will yield a clustering result, ultimately resulting in a clustering result set C = {C1, C2, ..., C...}. k}

[0091] Furthermore, step S4 specifically involves:

[0092] S41: Obtain the clustering result set C = {C1, C2, ..., C...} k}, where k is the total number of clusters, C k For the k-th cluster, the local density parameters of each pulse descriptor within the neighborhood radius Eps are calculated in the distance matrix, with each pulse descriptor as the center and the neighborhood radius Eps as the radius.

[0093] S42: Extract cluster C l and cluster C m Where l and m are cluster numbers, l≠m; cluster C is obtained through local density parameters. l Density cluster center A l ;

[0094] S43: Configure cluster C l Weighting coefficient γ l Through γ l and A l Cluster C was calculated l Cluster density Sl ;

[0095] S44: Configure cluster C l and cluster C m Inter-cluster separation D lm Through S l and D lm Cluster C was calculated l and cluster C m Clustering comprehensive measure Cs lm ;

[0096] S45: Repeat steps S42-S44 to calculate the DPDBI score for each cluster, and take the cluster with the smallest DPDBI score as the optimal clustering result.

[0097] Furthermore, the formula for calculating the local density parameter in step S41 is as follows:

[0098]

[0099] Where i and j are the radar signal numbers, ρ ij Let d be the local density parameter between radar signals i and j. ij Let x be the Euclidean distance between radar signals i and j; when x > 0, χ(x) = 0; otherwise, χ(x) = 1.

[0100] Furthermore, based on the defined local density parameters, density cluster centers are defined, and the radar signal point with the largest local density is called the density cluster center;

[0101] In step S42, density cluster center A l The calculation formula is:

[0102]

[0103] Where, ρ im Let be the local density parameter between radar signals i and m.

[0104] Specifically, to determine the density cluster center of each cluster as the centroid, it is necessary to traverse all radar signals in each cluster and query its local density parameters.

[0105] Furthermore, before calculating the intra-cluster compactness and inter-cluster separation of each cluster, weighting coefficients need to be defined for use in subsequent calculations;

[0106] In step S43, the weighting coefficient Y l The calculation formula is:

[0107]

[0108] Among them, |C l |For cluster Cl The number of radar signals in the middle, where N is the total number of radar signals successfully sorted;

[0109] Specifically, by defining weight coefficients, radar signal datasets and other high-dimensional datasets in high-dimensional space can largely avoid the problem of segmentation anomalies, avoiding over-segmentation and abnormally small clusters;

[0110] Intra-cluster density is calculated based on the average distance between all other radar signals within the cluster and the defined density cluster center;

[0111] Cluster density S l The calculation formula is:

[0112]

[0113] Where j is the radar signal number, p j Let be the pulse description word for radar signal j.

[0114] Specifically, when q=1, the Manhattan distance between each radar pulse signal within the cluster and the density cluster center is calculated; when q=2, the average Euclidean distance between all other radar signals and the density cluster center is calculated; due to the typical spatial distribution of radar signals, q=2 is used to define the cluster density.

[0115] Furthermore, in step S44, the inter-cluster separation degree D lm The calculation formula is:

[0116]

[0117] Among them, A m For cluster C m The density of cluster centers, where q is a set parameter, p il For cluster C l The i-th pulse description word within, p jm For cluster C m The j-th pulse description word within;

[0118] Clustering Comprehensive Measure Cs lm The calculation formula is:

[0119]

[0120] Specifically, when q = 2, S l D is the standard deviation of the distance from each sample within a cluster to its respective density cluster center. lm It is the standard deviation of the distance between different clusters; in this case, Cs lm It is the reciprocal of the classic Fisher similarity measure between different clusters; therefore, we define it as a clustering comprehensive measure; Cslm The lower value indicates that the overall sample performance is characterized by smaller intra-cluster distances and larger inter-cluster distances.

[0121] Furthermore, based on the comprehensive clustering metric among all clusters, the maximum inter-cluster similarity between each cluster and other clusters is retained, and this operation is performed on all clusters, taking the average value.

[0122] The formula for calculating the DPDBI score in step S45 is as follows:

[0123]

[0124] Specifically, regardless of the effectiveness metric for clustering, we always hope that the clustering results tend to converge towards the clustering comprehensive measure Cs. lm Therefore, after clustering, we iterate through and calculate for each cluster, and take the average of the overall results. The goal is to ensure that the optimal cluster partitioning minimizes the clustering metric, and finally obtain the DPDBI score based on this.

[0125] In this embodiment, to evaluate the performance of the proposed method, the DPDBI algorithm is first used to cluster the radar signal dataset to demonstrate its effectiveness. Then, the algorithm is applied to six high-dimensional real datasets and six artificially synthesized datasets to prove that the proposed algorithm can solve a series of clustering problems in high-dimensional and two-dimensional datasets with overlapping data. Finally, the DPDBI metric used in the proposed method is compared with five commonly used metrics, including CH, BCVI, COP, Sil, and DBI, demonstrating the superiority of the proposed method.

[0126] Depend on Figure 3 As can be seen, the dimensionality of the six real datasets all exceeds 3, and the dataset sizes vary. These characteristics validate the performance of the DPDBI algorithm on high-dimensional datasets of different scales. Furthermore, in datasets such as seed and wine, there is significant overlap between their respective clusters. This example aims to comprehensively verify whether the proposed DPDBI algorithm can effectively handle data overlap situations similar to those encountered in radar signal datasets. The figure shows good clustering results, and the specific performance comparison is shown in Table 1.

[0127] according to Figure 4 It can be observed that all six synthetic datasets are two-dimensional. Furthermore, most of the selected datasets exhibit an imbalanced cluster distribution, and their sizes vary considerably. These features were chosen to comprehensively evaluate the performance of the proposed DPDBI algorithm on two-dimensional datasets, and the figure shows that the clustering effect is good.

[0128] The method's sorting effect on radar signal datasets is... Figure 5 , 67. Display.

[0129] Table 1 presents the optimal number of clusters determined for different clustering effectiveness metrics for each dataset. The second column of the table provides the actual number of clusters, K. opt This represents the optimal number of clusters for 12 datasets. Bold numbers indicate that the COP (Clustering Validity Index) correctly determined the number of clusters for that dataset. Furthermore, the value in parentheses below each number represents the corresponding index value. For example, entry 3 (0.304) in the fifth row, fourth column indicates that for the seed dataset, the COP reached its minimum value of 0.304 when K=3, correctly determining the number of clusters for that dataset. The data in the table demonstrates that our method has a significant advantage in handling high-dimensional overlapping datasets, clearly outperforming other methods, and is also capable of handling two-dimensional datasets.

[0130] Table 1

[0131]

[0132]

[0133] A radar signal sorting system based on the clustering effectiveness index DPDBI includes:

[0134] The distance matrix acquisition module is used to acquire radar signal datasets, preprocess the radar signal datasets, and obtain the distance matrix.

[0135] The parameter pair set acquisition module is used to perform kernel function density estimation on the distance matrix to obtain the neighborhood radius; and to divide the radar signal dataset by step size using the neighborhood radius to obtain the parameter pair set.

[0136] The clustering result set acquisition module is used to cluster the parameter pair set and obtain the clustering result set;

[0137] The optimal clustering result acquisition module is used to score and rank the clustering result set using the DPDBI algorithm to obtain the optimal clustering result.

[0138] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0139] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments. In the unit claims listing several devices, several of these devices may be embodied by the same hardware item. The use of the terms first, second, and third, etc., does not indicate any order and can be interpreted as identifiers.

[0140] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A radar signal sorting method based on the clustering effectiveness index DPDBI, characterized in that, include: S1: Obtain the radar signal dataset, preprocess the radar signal dataset, and obtain the range matrix; Step S1 is as follows: S11: Extract the parameters of each radar signal in the radar signal dataset and construct the pulse descriptor matrix D, expressed as: Where i is the radar signal number, and n is the total number of radar signals. For the pulse description word of radar signal i, , Let i be the carrier frequency of the radar signal. Let i be the pulse width of the radar signal. The direction of arrival of radar signal i; S12: Normalize the pulse descriptor matrix to obtain a normalized matrix; S13: Calculate the Euclidean distance between every two pulse descriptors in the normalized matrix, and construct a distance matrix using the Euclidean distances; S2: Perform kernel function density estimation on the distance matrix to obtain the neighborhood radius; divide the radar signal dataset by step size using the neighborhood radius to obtain the parameter pair set; Step S2 is as follows: S21: Perform kernel function density estimation on each Euclidean distance in the distance matrix to obtain the probability density distribution of each Euclidean distance, and construct the probability density distribution curve through each probability density distribution; S22: Use the x-coordinate of the first peak point in the probability density distribution curve as the neighborhood radius. ; S23: Traverse the radar signals in the radar signal dataset to obtain the neighborhood radius of each radar signal. Minimum number of contained points The formula for calculating the minimum number of contained points is: Where i is the radar signal number, and n is the total number of radar signals. For the pulse description word of radar signal i, Let the radar signal i have a neighborhood radius The number of points contained within; S24: Set the neighborhood radius and minimum number of contained points Divide the step size into multiple elements and construct a list of neighborhood radii. and the minimum list of points , S25: List the neighborhood radii The list of elements and the minimum number of contained points The elements in the table are combined in pairs to obtain multiple parameter pairs, and the parameter pairs form a parameter pair set. S3: Cluster the parameter pair set to obtain the clustering result set; S4: The clustering results set is scored and ranked using the DPDBI algorithm to obtain the optimal clustering result; Step S4 is as follows: S41: Obtain the clustering result set Where k is the total number of clusters, For the k-th cluster; in the distance matrix, with each pulse descriptor as the center and the neighborhood radius... Using the radius as the local density parameter, the local density parameter of each pulse descriptor within the neighborhood radius is calculated; S42: Extract Clusters and cluster Where l and m are cluster numbers, l≠m; clusters are obtained through local density parameters. Density cluster centers ; S43: Configure Cluster Weighting coefficients ,pass and Calculate and obtain clusters Intra-cluster density ; S44: Configure Cluster and cluster Inter-cluster separation ,pass and Calculate and obtain clusters and cluster Clustering comprehensive measure ; S45: Repeat steps S42-S44 to calculate the DPDBI score for each cluster, and take the cluster with the smallest DPDBI score as the optimal clustering result.

2. The radar signal sorting method based on the clustering effectiveness index DPDBI according to claim 1, characterized in that, The formula for calculating the local density parameter in step S41 is: Where i and j are the radar signal numbers, Let i be the local density parameter between radar signals i and j. Let i be the Euclidean distance between radar signals i and j; when hour, ;otherwise .

3. The radar signal sorting method based on the clustering effectiveness index DPDBI according to claim 1, characterized in that, Density cluster centers in step S42 The calculation formula is: in, Let be the local density parameter between radar signals i and m.

4. The radar signal sorting method based on the clustering effectiveness index DPDBI according to claim 1, characterized in that, Weighting coefficients in step S43 The calculation formula is: in, For clusters The number of radar signals in the middle, where N is the total number of radar signals successfully sorted; Cluster density The calculation formula is: Where j is the radar signal number, Let be the pulse description word for radar signal j.

5. The radar signal sorting method based on the clustering effectiveness index DPDBI according to claim 1, characterized in that, Inter-cluster separation in step S44 The calculation formula is: in, For clusters The density of cluster centers, where q is a set parameter. For clusters The i-th pulse description word within, For clusters The j-th pulse description word within; Clustering Comprehensive Measure The calculation formula is: 。 6. The radar signal sorting method based on the clustering effectiveness index DPDBI according to claim 1, characterized in that, The formula for calculating the DPDBI score in step S45 is as follows: 。 7. A radar signal sorting system based on the clustering effectiveness index DPDBI, used to implement the radar signal sorting method based on the clustering effectiveness index DPDBI as described in any one of claims 1-6, characterized in that, include: The distance matrix acquisition module is used to acquire radar signal datasets, preprocess the radar signal datasets, and obtain the distance matrix. Specifically, it includes: S11: Extract the parameters of each radar signal in the radar signal dataset and construct the pulse descriptor matrix D, expressed as: Where i is the radar signal number, and n is the total number of radar signals. For the pulse description word of radar signal i, , Let i be the carrier frequency of the radar signal. Let i be the pulse width of the radar signal. The direction of arrival of radar signal i; S12: Normalize the pulse descriptor matrix to obtain a normalized matrix; S13: Calculate the Euclidean distance between every two pulse descriptors in the normalized matrix, and construct a distance matrix using the Euclidean distances; The parameter pair set acquisition module is used to perform kernel function density estimation on the distance matrix to obtain the neighborhood radius; and to divide the radar signal dataset by step size using the neighborhood radius to obtain the parameter pair set, specifically including: Step S2 is as follows: S21: Perform kernel function density estimation on each Euclidean distance in the distance matrix to obtain the probability density distribution of each Euclidean distance, and construct the probability density distribution curve through each probability density distribution; S22: Use the x-coordinate of the first peak point in the probability density distribution curve as the neighborhood radius. ; S23: Traverse the radar signals in the radar signal dataset to obtain the neighborhood radius of each radar signal. Minimum number of contained points The formula for calculating the minimum number of contained points is: Where i is the radar signal number, and n is the total number of radar signals. For the pulse description word of radar signal i, Let the radar signal i have a neighborhood radius The number of points contained within; S24: Set the neighborhood radius and minimum number of contained points Divide the step size into multiple elements and construct a list of neighborhood radii. and the minimum list of points , S25: List the neighborhood radii The list of elements and the minimum number of contained points The elements in the table are combined in pairs to obtain multiple parameter pairs, and the parameter pairs form a parameter pair set. The clustering result set acquisition module is used to cluster the parameter pair set and obtain the clustering result set; The optimal clustering result acquisition module is used to score and rank the clustering result set using the DPDBI algorithm to obtain the optimal clustering result, specifically including: Step S4 is as follows: S41: Obtain the clustering result set Where k is the total number of clusters, For the k-th cluster; in the distance matrix, with each pulse descriptor as the center and the neighborhood radius... Using the radius as the local density parameter, the local density parameter of each pulse descriptor within the neighborhood radius is calculated; S42: Extract Clusters and cluster Where l and m are cluster numbers, l≠m; clusters are obtained through local density parameters. Density cluster centers ; S43: Configure Cluster Weighting coefficients ,pass and Calculate and obtain clusters Intra-cluster density ; S44: Configure Cluster and cluster Inter-cluster separation ,pass and Calculate and obtain clusters and cluster Clustering comprehensive measure ; S45: Repeat steps S42-S44 to calculate the DPDBI score for each cluster, and take the cluster with the smallest DPDBI score as the optimal clustering result.

Images