Parallel adaptive clustering radar pulse signal sorting method

Abstract

The invention discloses a parallel adaptive clustering radar pulse signal sorting method. The method comprises the following implementation steps of: performing grid division on a three-dimensional data space formed by radar pulse description word carrier frequency, pulse width and pulse arrival angle; adaptively adjusting clustering algorithm parameters in a sub-data space in parallel; performing adaptive Birch clustering on the sub-data space; measuring the difference of different clusters after clustering; according to the radar pulse signal sorting method provided by the invention, the problems that the sorting task cannot be quickly completed when the data volume of the pulse signal is large, the precision of the sorting method is limited by parameter selection, and the requirement of a dense and complex electromagnetic environment on the measurement precision cannot be met can be solved; the method is suitable for a modern electromagnetic environment with serious aliasing and complexity, improves the radar signal analysis technology level, and can be widely applied to sorting of modern radar pulse signals.

Description

Technical Field

[0001] The present invention belongs to the field of communication technologies, and more particularly relates to a method for sorting radar pulse signals with parallel adaptive clustering in the field of radar communication technologies. The present invention can be used to separate pulse signals belonging to the same radar radiation source, facilitating the analysis of the separated pulse signals. Background Art

[0002] Radar radiation source signal sorting, also known as radar radiation source signal deinterleaving, refers to the process of separating each radar pulse sequence from a randomly interleaved pulse stream. Radar pulse signal sorting mainly differentiates based on the characteristics of the pulse sequences emitted by different radar radiation sources. By performing data processing on the radar pulse signals received by a radar receiver, information capable of distinguishing different radar radiation sources is extracted, and then the pulse signals belonging to the same radar radiation source are classified into the same category to achieve the sorting of radar pulse signals. Due to the complex electromagnetic environment, the radar often receives a radar pulse signal sequence with high density and severe overlapping. The number of pulses to be sorted is very large, resulting in a large amount of calculation. It is very difficult to deinterleave the severely overlapping radar pulse sequences received by the radar receiver, leading to problems such as low sorting accuracy.

[0003] Southwest Jiaotong University disclosed a method for sorting radar signals based on a bionic pattern recognition algorithm with spherical coverage in its patent document "A Method for Sorting Radar Signals Based on a Bionic Pattern Recognition Algorithm with Spherical Coverage" (application number: 202011187495.5, publication number: CN 112014804 A). The implementation steps of this method are as follows: (1) Obtain various types of radar pulse signal data as sample data, clean the false data in the sample data, then fill in the missing values in the sample data after data cleaning, and finally perform normalization processing on the sample data; (2) Perform an upsampling process from n dimensions to n + 1 dimensions on the preprocessed sample data, so that the Euclidean distances of all sample points to the origin of the coordinate axes in the n + 1 - dimensional space are equal, that is, map the data onto the spherical surface S n+1 and divide the samples into a training set and a test set according to a preset ratio; (3) On the spherical surface S n+1 , find the covering center and covering radius, construct spherical coverings one by one until all the training set sample points are trained to obtain a classification covering model; (4) Input all the test sets into the classification covering model, and judge the category of the test sample according to the category of the spherical region where the sample falls. The disadvantage of this method is that there are many processing steps for sample data. When the data volume is large, the processing time and sorting time will increase, and it is impossible to quickly complete the sorting task when the amount of radar pulse signal data is large.

[0004] Southeast University disclosed a radar pulse sequence sorting method based on image features in its patent document application "A Radar Pulse Sequence Sorting Method Based on Image Features" (application number: 202110136843.4, publication number: CN 112986925 A). The implementation steps of this method are as follows: First, pre-sort the radar pulse descriptor set based on frequency and histogram statistics method, and divide the original pulse descriptor set into several pulse descriptor subsets; Second, use arrival time, frequency, amplitude, phase, and pulse width, combined with the serial number of this pulse, to jointly form a pulse descriptor; Third, for each pulse descriptor subset, perform image processing based on arrival time and phase and use the Hough transform method to extract Hough line segments from it, and then use the DBSCAN clustering method to obtain several Hough line segment clusters; Fourth, perform phase cycle extension on the obtained image data and Hough line segment clusters, perform the second DBSCAN clustering to obtain Hough line segment cluster classification labels, and then obtain radar radiation source information based on hierarchical clustering method and perform pulse sequence classification; Fifth, according to the obtained hierarchical clustering result, perform signal capture for each original pulse signal point, and classify each point into a certain radar classification or as a noise signal that does not belong to any radar category; Sixth, analyze the classification result, obtain the radar radiation source pulse repetition period type and optimize the redundant sorting result. The disadvantage of this method is that the parameters of the DBSCAN algorithm used are fixed, and it is impossible to dynamically adjust the algorithm parameters according to the data distribution of the divided pulse descriptor subsets. The final sorting accuracy will be limited by the parameter selection and cannot meet the measurement accuracy requirements of a dense and complex electromagnetic environment.

[0005] In summary, for the existing radar pulse signal sorting methods in the prior art, the existing sorting methods have unsatisfactory effects. Due to the large number of data processing steps, the processing time and sorting time will increase when the data volume is large, and there is a problem that it is impossible to quickly complete the sorting task when the pulse signal data volume is large; at the same time, limited by the fixed sorting algorithm parameters, there is a problem that the sorting accuracy is limited by the parameter selection and cannot meet the measurement accuracy requirements of a dense and complex electromagnetic environment. Summary of the Invention

[0006] The purpose of the present invention is to propose a radar pulse signal sorting method based on parallel adaptive clustering in view of the above deficiencies of the prior art, to solve the problems that it is impossible to quickly complete the sorting task when the pulse signal data volume is large in the existing technology, and the sorting method accuracy is limited by the parameter selection and cannot meet the measurement accuracy requirements of a dense and complex electromagnetic environment.

[0007] The technical idea for achieving the object of the present invention is as follows: Since the present invention uses the grid division and parallel processing technology for the pulse descriptor data space, this technology can perform data processing and clustering algorithms for multiple data subspaces simultaneously, and can solve the problem that the processing time and sorting time will increase when the data volume in the prior art is large, resulting in the inability to quickly complete the sorting task when the pulse signal data volume is large. At the same time, since the present invention uses the adaptive Birch clustering algorithm, this algorithm can adaptively adjust the usage parameters of the Birch clustering algorithm according to the data distribution of the pulse descriptors in the divided data subspaces, and can solve the problem that the accuracy of the existing sorting methods is limited by the parameter selection and cannot meet the measurement accuracy requirements of dense and complex electromagnetic environments.

[0008] To achieve the above object, the technical solution adopted by the invention includes the following steps:

[0009] Step 1, perform grid division based on the three-dimensional feature space composed of carrier frequency, pulse width, and pulse arrival angle, and divide the original pulse descriptor feature space into multiple pulse descriptor subspaces;

[0010] Step 2, perform data division on the pulse descriptor subspaces in parallel to obtain clusters, and calculate the spans of the carrier frequency, pulse width, and pulse arrival angle values of each cluster within the sub-blocks in parallel;

[0011] Step 3, take the median values of the spans of the carrier frequency, pulse width, and pulse arrival angle in each sub-block in parallel as the characteristic unit lengths of the corresponding pulse descriptors in the sub-block;

[0012] Step 4, use the comprehensive scaling factor of the pulse descriptor subspaces to adaptively adjust the radius parameter and the maximum number of clustering clusters parameter of the Birch clustering algorithm; perform the Birch clustering algorithm on each pulse descriptor subspace in parallel using the adaptive parameters, and generate pseudo-labels for the resulting clusters;

[0013] Step 5, measure the feature similarity of each resulting cluster after sorting, unify the corresponding pseudo-labels of the clusters that meet the merging conditions into the same label value, achieve the merging of the resulting clusters, and complete the sorting of radar pulse signals.

[0014] Further, the step of performing grid division based on the three-dimensional feature space composed of carrier frequency, pulse width, and pulse arrival angle is as follows:

[0015] The first step, receive the pulse descriptor data composed of five pulse descriptors: pulse arrival time, carrier frequency, pulse width, pulse arrival angle, and pulse amplitude;

[0016] The second step, according to the preset granularity N = [n 载频 ,n 脉宽 ,n 脉冲到达角Divide the three-dimensional feature space into multiple sub-blocks, each sub-block corresponding to a three-dimensional grid cell, and its data set can be expressed as the following formula:

[0017] D ijk ={x(1),x(2),x(3),…x(t),…x(n)}

[0018] where D ijk represents the divided pulse descriptor sub-block data space, i, j, k respectively represent the indices of carrier frequency, pulse width, and pulse arrival angle; n represents the number of received pulses, and x(t) represents the pulse descriptor set corresponding to the t-th pulse data point;

[0019] In the third step, calculate the density of each sub-block according to the following formula:

[0020]

[0021] where ρ ijk represents the data density of a sub-block, |D ijk | represents the sum of the absolute values of each element in this sub-block, and V ijk represents the volume of the pulse descriptor sub-block data space D ijk ;

[0022] In the fourth step, set the upper threshold and lower threshold of the density according to prior knowledge. The sub-blocks with density greater than the set upper density threshold are further subdivided, the sub-blocks with density less than the set lower density threshold are merged with adjacent sub-blocks, and the granularity of the sub-blocks with density not less than the set lower density threshold and not greater than the upper density threshold remains unchanged.

[0023] Furthermore, the steps to calculate the spans of the carrier frequency, pulse width, and pulse arrival angle values of each cluster in the sub-block are as follows:

[0024] In the first step, perform data partitioning on each sub-block in parallel to obtain m clusters, that is, divide the sub-block into D ijk =[C 1 ,C 2 ,…C e ,…C m , where C represents a cluster, and C e represents the e-th cluster after data partitioning;

[0025] In the second step, for each cluster after partitioning within each sub-block, calculate the difference between the maximum and minimum values of the carrier frequency, pulse width, and pulse arrival angle of this cluster, and obtain the spans of the carrier frequency, pulse width, and pulse arrival angle of this cluster within the sub-block.

[0026] Further, the median values of the carrier frequency, pulse width, and pulse arrival angle span of each sub-block are selected to form the characteristic unit length of the pulse descriptor corresponding to the sub-block, which means that according to the spans of the carrier frequency, pulse width, and pulse arrival angle of m clusters in each sub-block, the median values of the spans of the carrier frequency, pulse width, and pulse arrival angle are respectively selected as the characteristic unit lengths corresponding to the carrier frequency, pulse width, and pulse arrival angle in the sub-block.

[0027] Further, the comprehensive scaling factor means that the characteristic unit lengths corresponding to the three pulse descriptors of each sub-block are divided by the span of the corresponding pulse descriptor in the sub-block to obtain the scaling factors of the three pulse descriptors in the sub-block; the cube root of the product of the scaling factors of the three pulse descriptors of each sub-block is taken as the comprehensive scaling factor of the sub-block data.

[0028] Further, the adaptive adjustment of the radius parameter and the maximum number of clustering clusters parameter of the Birch clustering algorithm means that for each sub-block, the initial radius parameter set is divided by the comprehensive scaling factor of the sub-block to obtain the radius parameter used for the Birch algorithm in the sub-block; the initial maximum number of clustering clusters parameter set is divided by the comprehensive scaling factor of the sub-block to obtain the maximum number of clustering clusters parameter used for the Birch algorithm in the sub-block.

[0029] Further, the use of the adaptive parameter to perform the Birch clustering algorithm on each pulse descriptor subspace in parallel means that the Birch clustering algorithm of each sub-block is performed in parallel on each sub-block using the adjusted adaptive Birch clustering algorithm radius parameter and the maximum number of clustering clusters parameter corresponding to the sub-block.

[0030] Further, the generation of the pseudo-labels of the result clusters means that according to the results of the adaptive Birch clustering algorithm for each sub-block, non-negative integer labels are added to the pulse data clustered into the same cluster within the sub-block.

[0031] Further, the steps for calculating the feature similarity of each result cluster are as follows:

[0032] In the first step, according to the following formula, the pulse repetition interval (PRI, Pulse Repetition Interval) of different clusters after each adaptive Birch clustering is calculated:

[0033] PRI r =TOA r+1 -TOA r

[0034] where, PRI r represents the calculated r-th PRI, TOA rIt represents the pulse arrival time value corresponding to the r-th pulse of a cluster obtained after the Birch algorithm, where the range of r is [1, s - 1], and s represents the number of pulses in the cluster;

[0035] In the second step, by comparing the differences in the histogram distributions of PRIs of different clusters, the similarity of PRIs of different clusters is measured. If the difference in the histogram distributions of PRIs of two clusters is less than the set PRI difference threshold, it is determined that the radar pulses corresponding to these two clusters belong to the same radar radiation source, and these two clusters are merged, that is, the labels of the pulses in these two clusters are modified to the same value;

[0036] In the third step, the mean values of the carrier frequency, pulse width, and pulse arrival angle of the data within the cluster after the PRI difference-based merging are calculated, and the differences in the mean values of the corresponding pulse descriptor of different clusters are calculated to measure the differences in the carrier frequency, pulse width, and pulse arrival angle of different clusters. If the differences in the mean values of the carrier frequency, pulse width, and pulse arrival angle of two clusters are all less than their respective set merging thresholds, namely the carrier frequency difference threshold, pulse width difference threshold, and pulse arrival angle difference threshold, it is determined that the radar pulses corresponding to these two clusters belong to the same radar radiation source, and these two clusters are merged, that is, the labels of the pulses in these two clusters are modified to the same value.

[0037] Furthermore, the merging thresholds, namely the PRI difference threshold, carrier frequency difference threshold, pulse width difference threshold, and pulse arrival angle difference threshold, refer to that the PRI difference threshold, carrier frequency difference threshold, and pulse width difference threshold are non-negative values defined according to the prior knowledge of the numerical ranges of the PRI, carrier frequency, and pulse width of the detected radar, and the pulse arrival angle difference threshold is a four-decimal value limited within the range [0°, 4°].

[0038] The present invention has the following advantages compared with the prior art:

[0039] First, the present invention performs grid division on the pulse descriptor space. By performing grid division on the three-dimensional data space composed of the carrier frequency, pulse width, and pulse arrival angle, it can simultaneously perform data processing and cluster clustering division on the data in multiple data sub-modules, and perform clustering tasks in parallel, solving the problem in the prior art that when the data volume is large, the processing time and sorting time will increase, resulting in the inability to quickly complete the sorting task when there is a large amount of pulse signal data. This enables the present invention to improve the parallelism and calculation speed of the sorting algorithm, support the requirement for a faster sorting speed of radar pulse signals in a complex electromagnetic environment, and improve the efficiency of radar pulse signal sorting.

[0040] Second, the present invention uses an adaptive Birch clustering algorithm. By calculating the scaling factors of the three-dimensional data space composed of carrier frequency, pulse width, and pulse arrival angle for different sub-blocks, and adjusting the parameters when each sub-block performs the Birch clustering algorithm, the problem that the accuracy of the existing sorting method is limited by the parameter selection and cannot meet the measurement accuracy requirements in a dense and complex electromagnetic environment is solved. The sorting accuracy of the radar pulse signal of the present invention is improved, which can support the requirements for the sorting accuracy of radar pulse signals in a complex electromagnetic environment, and the application scenario is more extensive. Brief Description of the Drawings

[0041] Figure 1 is a flowchart of the present invention;

[0042] Figure 2 is a simulation result diagram of the present invention. Detailed Embodiment

[0043] To more clearly illustrate the present invention, the following further detailed description is made in conjunction with embodiments and drawings. Obviously, the drawings and embodiments in the following description are only a part of the embodiments of the present invention, rather than all embodiments. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

[0044] Refer to Figure 1 , and make a further detailed description of the specific implementation steps of the present invention.

[0045] Step 1: Receive the pulse description word data transmitted by the radar receiving device, and divide the data space composed of carrier frequency, pulse width, and pulse arrival angle into grids.

[0046] Step 1: The received pulse description word data is data composed of five pulse description words: pulse arrival time, carrier frequency, pulse width, pulse amplitude, and pulse arrival angle, in the data form shown in Table 1.

[0047] Table 1 List of Compositions of Pulse Description Word Data

[0048] Pulse arrival time Carrier frequency Pulse width Pulse amplitude Pulse arrival angle Pulse arrival time 1 Carrier frequency 1 Pulse width 1 Pulse amplitude 1 Pulse arrival angle 1 Pulse arrival time 2 Carrier frequency 2 Pulse width 2 Pulse amplitude 1 Pulse arrival angle 2 Pulse arrival time 3 Carrier frequency 3 Pulse width 3 Pulse amplitude 1 Pulse arrival angle 3 Pulse arrival time 4 Carrier frequency 4 Pulse width 4 Pulse amplitude 1 Pulse arrival angle 4

[0049] Step 2: Divide the three-dimensional feature space into multiple sub-blocks according to the preset granularity N = [n 载频 , n 脉宽 , n 脉冲到达角 , where n is the number of received pulses, and each sub-block corresponds to a three-dimensional grid cell. Its data set can be expressed as the following formula:

[0050] Dijk = {x(1), x(2), x(3), … x(t) … x(n)}

[0051] Among them, x(t) represents the set of pulse description words corresponding to the t-th pulse data point, and i, j, and k respectively correspond to the indexes of the carrier frequency, pulse width, and pulse arrival angle of the pulse description word. Then, according to the following formula, calculate the density of each sub-block:

[0052]

[0053] Among them, ρ ijk represents the data density of a certain sub-block among the carrier frequency, pulse width, and pulse arrival angle of the corresponding pulse description word, |D ijk | represents the sum of the absolute values of each element in this sub-block, and V ijk represents the volume of this sub-block D ijk . The sub-blocks with a density greater than the set upper density threshold are further subdivided, the sub-blocks with a density less than the set lower density threshold are merged with adjacent sub-blocks, and the granularity of the sub-blocks with a density not less than the set lower density threshold and not greater than the upper density threshold remains unchanged.

[0054] Step 2, calculate the comprehensive scaling factor of each sub-block in parallel.

[0055] Step 1, perform data partitioning in parallel within each sub-block divided in Step 1 to obtain m clusters, that is, divide the sub-block into D ijk = [C 1 , C 2 , … C e …, C m , where C represents a cluster, and C e represents the e-th cluster obtained after the preliminary partitioning.

[0056] Step 2, for each cluster after partitioning within each sub-block, calculate the difference between the maximum and minimum values of the carrier frequency, pulse width, and pulse arrival angle of this cluster respectively, to obtain the span of the carrier frequency, pulse width, and pulse arrival angle of this cluster within the sub-block.

[0057] Step 3, divide the characteristic unit length corresponding to the three pulse description words of each sub-block by the span of the corresponding pulse description word within this sub-block to obtain the scaling factor of the three pulse description words within this sub-block; take the cube root of the product of the scaling factors of the three pulse description words of each sub-block as the comprehensive scaling factor of the data of this sub-block.

[0058] Step 3, perform the adaptive Birch clustering algorithm in parallel for different sub-blocks.

[0059] Step 1: For each sub-block, divide the set initial radius parameter by the comprehensive scaling factor of the sub-block to obtain the radius parameter used for the Birch algorithm in this sub-block; divide the set initial maximum number of clustering clusters parameter by the comprehensive scaling factor of the sub-block to obtain the maximum number of clustering clusters parameter used for the Birch algorithm in this sub-block.

[0060] Step 2: Parallelly perform the Birch clustering algorithm for each sub-block using the adjusted adaptive Birch clustering algorithm radius parameter and maximum number of clustering clusters parameter corresponding to this sub-block. According to the results of the adaptive Birch clustering algorithm for each sub-block, add non-negative integer labels to the pulse data clustered into the same cluster within this sub-block.

[0061] Step 4: Measure the similarity of different clusters and complete sorting through merging.

[0062] Step 1: For different clusters after adaptive Birch clustering in Step 3, calculate the PRI of this cluster according to the following formula:

[0063] PRI r = TOA r+1 - TOA r

[0064] where, PRI r represents the calculated r-th PRI, and TOA r represents the pulse arrival time value corresponding to the r-th pulse of a cluster obtained after the Birch algorithm. The range of r is [1, s - 1], and s represents the number of pulses in this cluster.

[0065] Step 2: By comparing the differences in the histogram distributions of the PRIs of different clusters, measure the similarity of the PRIs of different clusters. If the difference in the histogram distributions of the PRIs of two clusters is less than the set PRI difference threshold, it is determined that the radar pulses corresponding to these two clusters belong to the same radar radiation source, and these two clusters are merged, that is, the labels of the pulses in these two clusters are modified to the same value.

[0066] Step 3: Calculate the mean values of the carrier frequency, pulse width, and pulse arrival angle of the data within the cluster after the PRI difference merging. Calculate the differences in the mean values of the corresponding pulse description words of different clusters to measure the differences in the carrier frequency, pulse width, and pulse arrival angle of different clusters. If the differences in the mean values of the carrier frequency, pulse width, and pulse arrival angle of two clusters are all less than their respective set merging threshold carrier frequency difference threshold, pulse width difference threshold, and pulse arrival angle difference threshold, it is determined that the radar pulses corresponding to these two clusters belong to the same radar radiation source, and these two clusters are merged, that is, the labels of the pulses in these two clusters are modified to the same value.

[0067] Step 4: Write the cluster labels of the sorted pulse data into the data file to complete the parallel adaptive clustering and sorting based on pulse descriptor words.

[0068] The effects of the present invention will be further described below in conjunction with simulation experiments:

[0069] 1. Simulation experiment conditions.

[0070] The hardware platform for the simulation experiment of the present invention is: the processor is an Intel i5 9300H CPU with a main frequency of 2.4 GHz and 16 GB of memory.

[0071] The software platform for the simulation experiment of the present invention is: Windows 10 operating system and Pycharm2023.3.5.

[0072] The data files and set parameters used in the simulation experiment of the present invention are as follows:

[0073] The file used in this simulation is the data file of radar pulse descriptor words provided by the user, and its content is shown in Table 2.

[0074] Table 2 Parameter table of the radar pulse descriptor word data file provided by the user

[0075] Pulse arrival time Carrier frequency Pulse width Pulse amplitude Pulse arrival angle Label Pulse arrival time 1 Carrier frequency 1 Pulse width 1 Pulse amplitude 1 Pulse arrival angle 1 Label 1 Pulse arrival time 2 Carrier frequency 2 Pulse width 2 Pulse amplitude 1 Pulse arrival angle 2 Label 2 Pulse arrival time 3 Carrier frequency 3 Pulse width 3 Pulse amplitude 1 Pulse arrival angle 3 Label 3 Pulse arrival time 4 Carrier frequency 4 Pulse width 4 Pulse amplitude 1 Pulse arrival angle 4 Label 4

[0076] In the simulation experiment of the present invention, the initial radius parameter is set to 0.5, and the initial maximum number of clustering clusters parameter is set to 12. The neighborhood radius parameter of the comparative experiment DBSCAN algorithm is set to 0.6, and the minimum number of points in the neighborhood parameter is set to 150.

[0077] 2. Simulation content and its result analysis.

[0078] The simulation experiment of the present invention uses the method of the present invention and the DBSCAN algorithm to conduct a sorting experiment on radar pulse signals based on the pulse descriptor word data file provided by the user. The results of the simulation experiment are as Figure 2 shown.

[0079] The following will further describe the effects of the present invention in conjunction with Figure 2 the result graph.

[0080] Figure 2 (a) is a three-dimensional data graph of the original unsorted data plotted based on some data of the carrier frequency, pulse width, and pulse arrival angle in the radar pulse descriptor word file provided by the user. Figure 2 "3D Scatter plot of RF,PW,DOAData" in (a) represents Figure 2(a) is a three-dimensional data graph plotted based on carrier frequency, pulse width, and pulse arrival angle data. The RF axis represents the numerical range of the carrier frequency in the pulse descriptor word, the PW axis represents the numerical range of the pulse width in the pulse descriptor word, and the DOA axis represents the numerical range of the pulse arrival angle in the pulse descriptor word.

[0081] Figure 2 (b) is an original unsorted three-dimensional data graph plotted based on partial data of the carrier frequency, pulse width, and pulse arrival angle of the radar pulse descriptor word file provided by the user and their corresponding original labels. Figure 2 (b) "3DScatter Plotby Label" means Figure 2 (b) is an original data partitioned cluster three-dimensional graph plotted based on carrier frequency, pulse width, pulse arrival angle data, and their corresponding original labels. The RF axis represents the numerical range of the carrier frequency in the pulse descriptor word, the PW axis represents the numerical range of the pulse width in the pulse descriptor word, and the DOA axis represents the numerical range of the pulse arrival angle in the pulse descriptor word. Figure 2 In the legend of (b), different colors and the numbers next to them correspond to three-dimensional radar pulse descriptor word data with the same original label.

[0082] Figure 2 (c) is a three-dimensional data graph of the cluster distribution of the carrier frequency, pulse width, and pulse arrival angle sorted by the method of the present invention based on partial data of the carrier frequency, pulse width, and pulse arrival angle of the radar pulse descriptor word file provided by the user. Figure 2 (c) "3D Scatter Plot by pred" means Figure 2 (c) is a predicted sorted cluster three-dimensional graph plotted based on carrier frequency, pulse width, pulse arrival angle data, and the predicted labels sorted by the method of the present invention. The RF axis represents the numerical range of the carrier frequency in the pulse descriptor word, the PW axis represents the numerical range of the pulse width in the pulse descriptor word, and the DOA axis represents the numerical range of the pulse arrival angle in the pulse descriptor word. Figure 2 In the legend of (c), different colors and the numbers next to them correspond to three-dimensional radar pulse descriptor word data with the same sorted prediction label.

[0083] Figure 2 (d) is a three-dimensional data graph of the cluster distribution of the carrier frequency, pulse width, and pulse arrival angle after using the DBSCAN clustering algorithm based on partial data of the carrier frequency, pulse width, and pulse arrival angle of the radar pulse descriptor word file provided by the user. Figure 2 (d) "3D Scatter Plot by DBSCAN" means Figure 2(d) is a three-dimensional DBSCAN prediction sorting cluster diagram drawn based on carrier frequency, pulse width, pulse arrival angle data, and the corresponding DBSCAN prediction labels after sorting using the DBSCAN clustering algorithm. The RF axis represents the numerical range of the carrier frequency in the pulse descriptor, the PW axis represents the numerical range of the pulse width in the pulse descriptor, and the DOA axis represents the numerical range of the pulse arrival angle in the pulse descriptor. Figure 2 In the legend of (d), the different colors and the numbers beside them correspond to the three-dimensional radar pulse descriptor data with the same DBSCAN sorting prediction label.

[0084] Figure 2 (e) is a result diagram of data division and the sorting accuracy rate of the corresponding subspace data sorted by the method of the present invention based on a complete data set of pulse descriptors provided by the user, where Accuracy represents the sorting accuracy rate after sorting the pulse descriptor data in the corresponding divided subspace.

[0085] From Figure 2 (a), Figure 2 (b), it can be seen that the numerical aliasing of the radar pulse descriptor data set provided by the user is serious, and it is easy to divide the radar pulse descriptor data belonging to different radars into the data clusters belonging to the same radar, making sorting difficult, indicating the current serious and complex electromagnetic environment.

[0086] From Figure 2 (c), Figure 2 (d), it can be seen that under the condition of using the same data of carrier frequency, pulse width, and pulse arrival angle of the radar pulse descriptor, compared with the DBSCAN method of the existing technology, the method of the present invention can distinguish the pulse descriptor data that do not belong to the same kind of radar, indicating that the method of the present invention has improved the sorting performance of the radar pulse descriptor compared with the existing method.

[0087] From Figure 2 (e), it can be seen that by using the method of the present invention for a complete data set of pulse descriptors provided by the user, the pulse descriptor data can be divided into multiple data subspaces for sorting, and the sorting accuracy rate is relatively high, indicating that the method of the present invention has improved the sorting performance of the radar pulse descriptor compared with the existing method.

[0088] The above simulation experiment results show that under the condition of the same received radar pulse descriptor data, the method of the present invention can determine the pulse descriptor data belonging to different radars, and the sorting accuracy rate of each pulse descriptor data subspace is also relatively high, which is a radar pulse descriptor sorting method with higher sorting efficiency and accuracy.

Claims

1. A radar pulse signal sorting method based on parallel adaptive clustering, characterized in that: The three-dimensional data space composed of carrier frequency, pulse width and pulse arrival angle is grid-divided, the scaling factor of the three-dimensional data space composed of carrier frequency, pulse width and pulse arrival angle is calculated for different sub-blocks, and the parameters of each sub-block when performing the Birch clustering algorithm are adjusted; the steps of the sorting method include the following: Step 1, grid division is performed based on the three-dimensional feature space composed of carrier frequency, pulse width, and pulse arrival angle, and the original pulse description word feature space is divided into multiple pulse description word subspaces; Step 2, dividing the pulse description word subspace into clusters in parallel, and calculating the span of the carrier frequency, pulse width, and pulse arrival angle values ​​of each cluster in the subblock in parallel; Step 3, taking the median value of the carrier frequency, pulse width, and pulse arrival angle span in each sub-block in parallel as the characteristic unit length of the corresponding pulse description word in the sub-block; Step 4, using the comprehensive scaling factor of the pulse descriptor subspace, adaptively adjusting the radius parameter and the maximum cluster number parameter of the Birch clustering algorithm; using the adaptive parameters to perform the Birch clustering algorithm on each pulse descriptor subspace in parallel, and generating pseudo labels for the result clusters; Step 5: measure the feature similarity of each result cluster after sorting, unify the corresponding pseudo labels of the clusters that meet the merging conditions into the same label value, realize the merging of the result clusters, and complete the sorting of radar pulse signals.

2. The radar pulse signal sorting method of parallel adaptive clustering according to claim 1 is characterized in that: The steps of meshing the three-dimensional feature space based on carrier frequency, pulse width, and pulse arrival angle described in step 1 are as follows: The first step is to receive the pulse description word data consisting of five pulse description words: pulse arrival time, carrier frequency, pulse width, pulse amplitude, and pulse arrival angle; The second step is to use the preset granularity N = [n 载频 ,n 脉宽 ,n 脉冲到达角 ] The three-dimensional feature space is divided into multiple sub-blocks, each sub-block corresponds to a three-dimensional grid unit, and its data set can be expressed as follows: D ijk ={x(1),x(2),x(3),…x(t),…x(n)} Among them, D ijk represents the divided pulse description word sub-block data space, i, j, k represent the indexes of carrier frequency, pulse width, and pulse arrival angle respectively; n represents the number of received pulses, and x(t) represents the pulse description word set corresponding to the t-th pulse data point; The third step is to calculate the density of each sub-block according to the following formula: Among them, ρ ijk Indicates the data density of a sub-block, |D ijk | represents the sum of the absolute values ​​of each element in the sub-block, V ijk Indicates the pulse description word sub-block data space D ijk Volume; In the fourth step, the upper and lower density thresholds are set according to prior knowledge. Sub-blocks with a density greater than the set upper density threshold are further subdivided, sub-blocks with a density less than the set lower density threshold are merged with adjacent sub-blocks, and the granularity of sub-blocks with a density not less than the set lower density threshold and not greater than the upper density threshold remains unchanged.

3. The radar pulse signal sorting method of parallel adaptive clustering according to claim 1 is characterized in that: The steps of calculating the span of the carrier frequency, pulse width, and pulse arrival angle values ​​of each cluster in the sub-block in step 2 are as follows: The first step is to divide the data in each sub-block in parallel to obtain m clusters, that is, to divide the sub-block into D ijk =[C1,C2,…C e ,…C m ], where C represents a cluster, C e Represents the e-th cluster after data partitioning; In the second step, for each cluster divided in each sub-block, the difference between the maximum and minimum values ​​of the carrier frequency, pulse width, and pulse arrival angle of the cluster is calculated to obtain the span of the carrier frequency, pulse width, and pulse arrival angle of the cluster in the sub-block.

4. The radar pulse signal sorting method of parallel adaptive clustering according to claim 1, characterized in that: The step of selecting the median value of the span of the carrier frequency, pulse width, and pulse arrival angle of each sub-block in step 3 to form the characteristic unit length of the pulse description word corresponding to the sub-block means that, according to the respective spans of the carrier frequency, pulse width, and pulse arrival angle of the m clusters in each sub-block, the median of the respective spans of the carrier frequency, pulse width, and pulse arrival angle are selected as the characteristic unit length corresponding to the carrier frequency, pulse width, and pulse arrival angle in the sub-block.

5. The radar pulse signal sorting method of parallel adaptive clustering according to claim 1, characterized in that: The comprehensive scaling factor described in step 4 refers to the scaling factor of the three pulse description words in each sub-block obtained by dividing the characteristic unit length corresponding to the three pulse description words in each sub-block by the span of the corresponding pulse description words in the sub-block; and taking the cube root of the product of the scaling factors of the three pulse description words in each sub-block as the comprehensive scaling factor of the sub-block data.

6. The radar pulse signal sorting method of parallel adaptive clustering according to claim 1, characterized in that: The adaptive adjustment of the radius parameter and the maximum number of clusters parameter of the Birch clustering algorithm in step 4 means that, for each sub-block, the radius parameter used by the Birch algorithm for the sub-block is obtained by dividing the set initial radius parameter by the comprehensive scaling factor of the sub-block; and the maximum number of clusters parameter used by the Birch algorithm for the sub-block is obtained by dividing the set initial maximum number of clusters parameter by the comprehensive scaling factor of the sub-block.

7. The radar pulse signal sorting method of parallel adaptive clustering according to claim 1, characterized in that: The step 4 of using the adaptive parameters to perform the Birch clustering algorithm on each pulse description word subspace in parallel means that the Birch clustering algorithm of each subblock is performed in parallel using the adjusted adaptive Birch clustering algorithm radius parameter and the maximum cluster number parameter corresponding to the subblock.

8. The radar pulse signal sorting method of parallel adaptive clustering according to claim 1, characterized in that: Generating pseudo labels of result clusters in step 4 refers to adding non-negative integer labels to the pulse data clustered into the same cluster in the sub-block according to the results of the adaptive Birch clustering algorithm for each sub-block.

9. The radar pulse signal sorting method of parallel adaptive clustering according to claim 1, characterized in that: The steps for calculating the feature similarity of each result cluster described in step 5 are as follows: The first step is to calculate the pulse repetition interval (PRI) of each cluster after adaptive Birch clustering according to the following formula: PRI r =FIGHT r+1 -TOA r Among them, PRI r Indicates the calculated rth PRI, TOA r It represents the pulse arrival time value corresponding to the rth pulse of a cluster obtained by the Birch algorithm. The range of r is [1, s-1], and s represents the number of pulses in the cluster. The second step is to measure the similarity of PRIs of different clusters by comparing the differences in their histograms. If the difference in their histograms is less than the set PRI difference threshold, it is determined that the radar pulses corresponding to the two clusters belong to the same radar radiation source, and the two clusters are merged, that is, the labels of the pulses of the two clusters are modified to the same value. The third step is to calculate the mean of the carrier frequency, pulse width and pulse arrival angle of the data within the cluster after the PRI difference merging, and calculate the difference of the mean of the corresponding pulse description words of different clusters to measure the difference of the carrier frequency, pulse width and pulse arrival angle of different clusters. If the difference of the mean of the carrier frequency, pulse width and pulse arrival angle of two clusters is less than the respective set merging thresholds of the carrier frequency difference threshold, the pulse width difference threshold and the pulse arrival angle difference threshold, then it is determined that the radar pulses corresponding to the two clusters belong to the same radar radiation source, and the two clusters are merged, that is, the labels of the pulses of the two clusters are modified to the same value.

10. The radar pulse signal sorting method of parallel adaptive clustering according to claim 9, characterized in that: The combined threshold PRI difference threshold, carrier frequency difference threshold, pulse width difference threshold, and pulse arrival angle difference threshold refer to that the PRI difference threshold, carrier frequency difference threshold, and pulse width difference threshold are non-negative values ​​limited according to the prior knowledge of the numerical range of PRI, carrier frequency, and pulse width of the detected radar, and the pulse arrival angle difference threshold is a four-digit decimal value limited to the range [0°, 4°].

Images