Radar signal online sorting and matching method based on data stream clustering

Abstract

The invention discloses a radar signal online sorting and matching method based on data stream clustering, which is realized based on a main receiving station and N auxiliary receiving stations, the receiving stations keep time synchronization, electromagnetic signals emitted by radar radiation sources in the same reconnaissance area are reconnaissance intercepted to obtain pulse signals, and the pulse signals are processed to obtain PDW data; the method comprises the steps that N auxiliary receiving stations transmit PDW data to a main receiving station every set time T; the main receiving station processes the PDW data and calculates a pulse pair similarity factor by using the multi-parameter information of the pulse signal, and when the similarity factor is greater than or equal to a set threshold value, the corresponding pulse pair is successfully paired, and the time difference of arrival is calculated; performing data stream clustering processing, classifying newly-input data points into an existing target cluster or storing the newly-input data points into a cache region according to a discrimination condition, and performing online updating on the existing target cluster to obtain a category to which the newly-input data points belong; and repeating the steps until sorting and matching of all pulse signals are completed.

Description

Technical Field

[0001] The present invention belongs to the technical field of radar signal processing, and in particular relates to an online radar signal sorting and pairing method based on data stream clustering. Background Art

[0002] Radar signal sorting technology is also called radar signal deinterleaving. Its main purpose is to classify the randomly interleaved pulse streams received by the receiving station according to the radar to which the pulses belong. Radar signal matching technology mainly refers to matching the pulse streams captured by each receiving station according to the source of the pulses in a multi-station reconnaissance system, and matching the signals of the same pulse from the same radar. Radar signal sorting and matching is a key link in a multi-station reconnaissance system, and plays a decisive role in subsequent radar positioning and identification. The electromagnetic environment of the modern battlefield is highly complex. On the one hand, radar signal parameters are highly overlapped in the time domain, space domain, and frequency domain, and the inter-pulse and intra-pulse modulation methods are complex and changeable, showing certain non-stationary characteristics; on the other hand, there are a large number of radar radiation sources and a highly dense signal environment. In a typical environment, it can reach millions of pulses per second, which brings many difficulties to the processing of radar signals.

[0003] The sorting method based on pulse-to-pulse parameters mainly relies on the pulse descriptive word (PDW) of the signal. The template matching method compares the PDW with a pre-established characteristic parameter database to sort pulses with similar or identical parameters. However, it is often difficult to obtain the characteristic parameter database for non-cooperative radars. The pulse sorting method based on pulse repetition interval (PRI) utilizes the TOA information of the signal and sorts the signal by mining the PRI of the pulse sequence hidden under the TOA information. However, this method is only applicable to scenarios with low pulse density, few radiation sources, and relatively fixed PRI parameters. With the complexity of the electromagnetic environment and the increase of pulse density, the PRI-based sorting method has been difficult to adapt to the increasingly complex electromagnetic environment. The sorting method based on intra-pulse features mainly extracts the features of the signal in the time domain, frequency domain, time-frequency domain and other transform domains to complete the sorting of the pulse stream. These methods are greatly affected by the signal modulation type, have poor ability to distinguish signals of different modulation types, have limited generalization and low stability.

[0004] With the development of machine learning technology, more and more scholars use machine learning algorithms to sort radar emitter pulses, mainly including unsupervised clustering, neural network-based sorting, and radar signal sorting based on supervised classification models. For unsupervised clustering, it is usually limited by the shape and number of clusters or requires prior knowledge such as neighborhood radius. For neural networks, supervised classification models, and deep learning methods, prior knowledge and training are usually required. In electronic reconnaissance scenarios, prior knowledge of emitter signals is often not available in advance, so the effect is not good in practical applications. Summary of the invention

[0005] Aiming at the problems of low radar pulse signal sorting accuracy, low pulse pairing success rate, low processing efficiency of high-density pulse streams, and poor real-time performance in the multi-station reconnaissance system of the prior art, the purpose of the present invention is to overcome the above-mentioned defects of the prior art and propose a radar pulse signal sorting and pairing method based on data stream clustering.

[0006] In order to achieve the above object, the present invention proposes an online sorting and pairing method for radar signals based on data stream clustering, which is implemented based on a main receiving station and N secondary receiving stations. All receiving stations keep time synchronization, and conduct reconnaissance and interception of electromagnetic signals emitted by radar radiation sources in the same reconnaissance area to obtain pulse signals, and obtain PDW data after processing; the method comprises:

[0007] Step 1) N secondary receiving stations transmit the PDW data of the intercepted pulse signal to the primary receiving station at a set time interval T;

[0008] Step 2) The main receiving station processes the PDW data of the pulse signal and calculates the pulse pair similarity factor using the multi-parameter information of the pulse signal. When the similarity factor is greater than or equal to the set threshold, the corresponding pulse pair is successfully paired, and the arrival time difference of the successfully paired pulse pair is calculated;

[0009] Step 3) Perform data stream clustering on the arrival time difference of the successfully paired pulse pairs, classify the newly input data points into the existing target cluster or store them in the buffer area according to the discrimination conditions, update the existing target cluster online, and obtain the category to which the newly input data points belong;

[0010] Step 4) Repeat the above steps until all pulse signals are sorted and paired.

[0011] Preferably, the PDW data includes carrier frequency, pulse width, amplitude, arrival time and arrival direction, wherein the arrival time of the nth pulse received by the i-th receiving station is for:

[0012]

[0013] Where i=1 represents the primary receiving station, and the others represent the secondary receiving stations; c represents the speed of light, l i Represents the distance from the radar radiation source to the i-th receiving station.

[0014] Preferably, in step 2), the master receiving station processes the PDW data of the pulse signal and calculates the pulse pair similarity factor using the multi-parameter information of the pulse signal, including:

[0015] The arrival time of the nth pulse in the pulse stream received by the main receiving station is recorded as Calculate the time difference window Θ based on the distance between the pulse to be paired at the primary receiving station and the secondary receiving station i (n):

[0016]

[0017] Among them, ε represents the time difference measurement error, τ 1i It represents the maximum arrival time difference between the pulse arriving at the primary receiving station and the secondary receiving station i, where i≠1, and is calculated as R 1i is the distance between the primary receiving station and the secondary receiving station i, and c is the speed of light;

[0018] Calculate the pulse pair similarity factor α by comparing the pulses of the main receiving station with the pulses in the time difference window of each secondary receiving station j,k :

[0019]

[0020] Among them, ‖·‖ means to find the L2 norm of the matrix, T means transpose, represents the diagonal matrix consisting of the pulse description words of the nth pulse of the main receiving station, represents the diagonal matrix consisting of the pulse description words of the m-th pulse of the secondary receiving station i, f, A, γ, θ represent the carrier frequency, amplitude, pulse width and pulse arrival direction of the pulse respectively, σ is the weighting coefficient matrix, σ f ,σ A ,σ γ ,σ θ They respectively represent the measurement errors of pulse carrier frequency, amplitude, pulse width and pulse arrival direction.

[0021] Preferably, the arrival time difference of the successfully paired pulse pairs is calculated in step 2) as:

[0022]

[0023] in, represents the arrival time of the nth pulse at the primary receiving station, Indicates the arrival time of the mth pulse at secondary receiving station i.

[0024] Preferably, the step 3) performs data stream clustering processing on the arrival time differences of the successfully paired pulse pairs, comprising:

[0025] Read data points in real time from the arrival time difference data stream of the successfully paired pulse pairs, determine the relationship between the data points and the existing micro-clusters and core micro-clusters, merge the data points into micro-clusters or core micro-clusters, and update the weight of each micro-cluster or core micro-cluster, update and merge the existing micro-clusters and core micro-clusters, and transform the micro-clusters that meet the conditions into core micro-clusters;

[0026] Each interval time T P The core micro-clusters and outlier micro-clusters are checked, the core micro-clusters and outlier micro-clusters whose weights are less than the threshold parameter μ are deleted, and the data of the points in the cache area are maintained.

[0027] Preferably, the microcluster is represented by MC(w,c,r,p), which is a set of time-scaled T 1 ,T 2 ,…,T n Data point x 1 ,x 2 ,…,x n At the time t, w is the weight of the micro-cluster, indicating the activity of the micro-cluster, which decays exponentially over time; c is the center of the micro-cluster, r is the radius of the micro-cluster, and p is the number of current points in the cluster.

[0028] Preferably, the step of reading data points from the arrival time difference data stream of the successfully paired pulse pairs in real time, determining the relationship between the data points and the existing micro-clusters and core micro-clusters, merging the data points into the micro-clusters or core micro-clusters, and updating the weight of each micro-cluster or core micro-cluster comprises:

[0029] Read the data point x in real time from the arrival time difference data stream of the successfully paired pulse pair, and judge whether the Euclidean distance from the center of the core micro-cluster is less than the micro-cluster radius r. If it is, merge x into the core micro-cluster and update the center and weight of the core micro-cluster; otherwise, judge whether the Euclidean distance from the center of the micro-cluster is less than the micro-cluster radius r. If it is, merge x into the micro-cluster and update the center and weight of the micro-cluster;

[0030] If the Euclidean distance is not satisfied, the data point x is stored in the buffer.

[0031] Preferably, the updating and merging of the existing micro-clusters and core micro-clusters to transform the micro-clusters that meet the conditions into core micro-clusters includes:

[0032] When the number of points in a micro-cluster exceeds N and the weight w>μ, the micro-cluster is upgraded to a core micro-cluster;

[0033] When the weight w of a micro-cluster or core micro-cluster is less than μ, the micro-cluster is reduced to an outlier micro-cluster;

[0034] If the distance between the center positions of two core microclusters is less than 2 times the microcluster radius r, they are density-reachable and belong to the same category, otherwise they are classified into a new category.

[0035] Preferably, the step 3) also includes maintaining the cache area, including:

[0036] The weight w of the points in the buffer area is calculated according to the following formula:

[0037]

[0038] Where λ represents the time attenuation coefficient, T now represents the current time, T L Indicates the moment when the pairing of the Lth data point is completed;

[0039] Determine whether the weight is less than the threshold parameter μ. If it is, it is a noise point and is deleted from the buffer. Use the DBSCAN algorithm to generate new micro-clusters for the remaining points, and delete the points that can be described by the new micro-clusters from the buffer.

[0040] Compared with the prior art, the advantages of the present invention are:

[0041] 1. A radar pulse signal sorting and pairing method based on data stream clustering of the present invention uses the positional relationship between multiple stations and the characteristic parameters of the pulses to complete the sorting and pairing of multi-station pulse sequences compared to the traditional single-station sorting, thereby improving the signal sorting and pairing efficiency and accuracy between multi-station electronic reconnaissance systems;

[0042] 2. The present invention introduces the idea of ​​data stream clustering algorithm, regards pulse flow as a continuous pulse stream, and uses the data stream clustering method to complete signal sorting and pairing in dynamic scenarios, thereby improving the real-time performance of sorting and being able to effectively deal with the appearance and disappearance of radiation sources, and is more suitable for reconnaissance environments in real scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a schematic diagram of the time difference window corresponding to the nth pulse of the secondary receiving station i and the primary receiving station;

[0044] FIG2 is a schematic diagram of representing clusters by core microclusters, FIG2(a) represents original data points, and FIG2(b) represents the description of the distribution of data points by core microclusters;

[0045] Figure 3 It is a flow chart of density-based data stream clustering algorithm;

[0046] Figure 4is the micro-cluster update algorithm flow chart;

[0047] Figure 5 Flowchart of micro-cluster merging algorithm;

[0048] Figure 6 Flowchart of the algorithm for maintaining the cache area;

[0049] Figure 7 It is a three-dimensional distribution diagram of some pulse parameter data received by the receiving station;

[0050] Figure 8 It is a three-dimensional distribution diagram of pulse pairs that are successfully sorted and paired after being processed by the online sorting and pairing method proposed by the present invention. DETAILED DESCRIPTION

[0051] The online sorting and pairing method based on data stream clustering is implemented based on a main receiving station and N secondary receiving stations. All receiving stations conduct reconnaissance and interception of electromagnetic signals emitted by radar radiation sources in the same reconnaissance area. During the observation period, the receiver of each receiving station obtains a pulse sequence for the intercepted radar signal, and processes the pulse sequence to obtain the PDW data of the pulse. Each PDW data is used as a sample to describe the pulse signal. The parameters in the PDW data include but are not limited to carrier frequency, pulse width, amplitude, arrival time and arrival direction.

[0052] The online sorting and pairing method of radar signals based on data stream clustering includes the following steps:

[0053] Step 1. When the three receiving stations are conducting electronic reconnaissance, they maintain strict time synchronization. The two secondary stations intercept pulses and send them to the main station for unified processing, transmitting data to the main station once every time T.

[0054] Step 2. After receiving the data, the master station processes the T seconds of data and calculates the pulse pair similarity factor using the multi-parameter information of the pulse signal. For pulse pairs with a similarity factor greater than a certain threshold, they are considered to be paired successfully, and the arrival time difference of the successfully paired pulse pairs is calculated.

[0055] Step 3. Perform data stream clustering on the arrival time difference obtained after pairing in step 2. According to the discrimination criteria, the newly input data point is classified into the existing target cluster or stored in the temporary storage area. The existing target cluster is updated online to obtain the category to which the newly input data point belongs.

[0056] Step 4. Repeat steps 1 to 3 to continuously obtain radar pulse signal sorting results until the algorithm is terminated.

[0057] The technical solution of the present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0058] Example

[0059] The embodiment of the present invention provides a radar signal online sorting and pairing method based on data stream clustering. In this embodiment, three receiving stations are used as an example, but the method is not limited thereto. The method includes the following steps:

[0060] 1. When the three receiving stations are conducting electronic reconnaissance, they maintain strict time synchronization. The two secondary stations intercept pulses and send them to the main station for unified processing, transmitting data to the main station once every time T.

[0061] 2. After receiving the data, the master station processes the T-second data. When T is very small, it can be considered that the receiving station and the radar remain relatively stationary during this time interval. 1 , l 2 and l 3 , the distances between the receiving stations are L 12 , L 13 and L 23 The pulse arrival time of the radar received by each receiving station is:

[0062]

[0063] If there are N secondary receiving stations, the arrival time of the nth pulse received by the i-th receiving station is for:

[0064]

[0065] Where i=1 represents the primary receiving station, and the others represent the secondary receiving stations; c represents the speed of light, l i Represents the distance from the radar radiation source to the i-th receiving station.

[0066] From the relationship between the three sides of the triangle, it can be seen that the arrival time of a pulse of the radiation source intercepted by each receiving station should satisfy:

[0067]

[0068] The arrival time of the nth pulse in the pulse stream received by the master station is recorded as Then the arrival time ranges of the same pulse signal in the pulse stream received by the secondary stations 2 and 3, which may be emitted from the same radiation source as the nth pulse in the pulse stream received by the primary station, are:

[0069]

[0070] Θ i (n) is called receiving station 1 The pulse received at the time is the time difference window corresponding to the i-th receiving station, and ε represents the time difference measurement error. τ 1i(i≠1), which indicates the maximum arrival time difference between the pulse arriving at the primary receiving station and the secondary receiving station i, which is calculated as c is the speed of light; in the time difference window Θ i The pulses at receiving station i within range (n) may all come from the same radar as the nth pulse at receiving station 1. Figure 1 shown.

[0071] Assuming that there are M pulses in the time difference window of receiving station i, M groups of potential pulse pairings can be obtained through the time difference window. Since the intra-pulse and inter-pulse parameters of radar pulse signals are variable and random, the parameters between different pulses from different radiation sources received by the same receiving station are quite different, while the inter-pulse parameters other than TOA between the same pulses from the same radiation source received by different receiving stations are very small. Therefore, the remaining inter-pulse parameters of a single pulse can be used to distinguish the potential pulse pairing methods. In order to measure the similarity between different pulses at different receiving stations, the pulse pair similarity factor is defined as:

[0072]

[0073] Among them, ‖·‖ means to find the L2 norm of the matrix, σ is the weighting coefficient matrix, σ f ,σ A ,σ γ ,σ θ Respectively represent the measurement errors of pulse carrier frequency, amplitude, pulse width and pulse direction. After obtaining the pulse pair similarity factor, set the decision threshold β, β∈[0,1] to determine whether a pulse is received by two receiving stations at the same time. i,j >β, the two pulses at different receiving stations are considered to be the same pulse from the same radiation source, and the pulse pair matching is successful, otherwise the matching fails.

[0074] The arrival time difference of the successfully paired pulse pair is calculated as:

[0075]

[0076] in, represents the arrival time of the nth pulse of the master station, Indicates the arrival time of the mth pulse at the secondary station.

[0077] The pulse pairing method can be used to pair the same pulses received by three receiving stations from the same radiation source, and the successfully paired pulse pairs are recorded as The time differences of the pulse emitted by the unknown radiation source arriving at the three receiving stations are:

[0078]

[0079] The time differences of pulses emitted by radiation sources at different locations arriving at the three receiving stations are different. In a short period of time, it can be approximately considered that the time for the radiation source to arrive at the three receiving stations is a fixed value. From a longer time span, the time difference of pulses emitted by the radiation source arriving at the three receiving stations shows a continuous change pattern. Therefore, the time difference information of the paired pulses arriving at each receiving station can be used as the input of the data stream clustering algorithm below to sort the paired pulses.

[0080] 3. Use data stream clustering algorithm to sort radar pulses:

[0081] 3.1 Problem Analysis and Modeling

[0082] The pulse stream of the radar radiation source can be essentially regarded as a data stream. The pulse signals from the same radiation source are regarded as one type. The problem of radar signal sorting can also be abstracted into a problem of data stream clustering. In this paper, the time difference information of paired pulses is used as input to sort radar pulses. Assume that at time t, the time difference feature vector of the paired pulse is Then the data flow for data flow clustering can be expressed as And at time t there are k t There are classes, each class is represented by ε, then k t The class can be represented as

[0083] The radar radiation source pulse stream is a data stream with an evolving nature. With the relative movement between the reconnaissance station and the radiation source, new radiation sources continue to appear and old radiation sources disappear. Corresponding to this is the appearance and disappearance of classes. The appearance and disappearance of classes are defined as follows:

[0084] Class emergence: When a class ε satisfies And ε∈E t , we call the class ε appearing at time t

[0085] Class disappearance: When a class ε satisfies and We say that class ε vanishes at time t.

[0086] Therefore, the problem of online radar radiation source sorting can be abstracted into the following mathematical problem: Under the condition of a given data stream χ, at any time t, determine the class E t And for each data x t Assign a corresponding class ε i , and simultaneously realize the detection of the appearance and disappearance of the class.

[0087] 3.2 Density-based Data Stream Clustering Algorithm

[0088] Considering the large amount of data in radar pulse data stream, the shape of clusters generated by clustering changes over time and is irregular. In the problem of data stream clustering, the memory space is limited relative to the length of the data stream, so the concepts of micro-clusters, core micro-clusters and outlier micro-clusters are proposed. Micro-clusters are used to represent a large number of data points without storing all data points in memory, thereby reducing the resource consumption of the system. The following defines micro-clusters, core micro-clusters and outlier micro-clusters:

[0089] Definition 1: micro-cluster, for a set of time scales T 1 ,T 2 ,…,T n Data point x 1 ,x 2 ,…,x n At time t, it is defined as MC(w,c,r,n). w is the weight of the micro-cluster, c is the center of the microcluster, r is the radius of the micro-cluster, and n is the number of current points in the cluster. Among them, w is the time weight of the micro-cluster, which indicates the activity of the micro-cluster, and decays exponentially with time. The farther the time when the micro-cluster is generated is from the current time, the lower the weight of the cluster. r is the radius of the micro-cluster, which is used to judge the connectivity between micro-clusters and is related to the time interval of micro-cluster updates.

[0090] Definition 2: Core micro-cluster: When the number of points in a micro-cluster exceeds N and the weight w>μ, the micro-cluster can be upgraded to a core micro-cluster. 0<μ<1 is the threshold parameter used to determine the classification of micro-cluster.

[0091] Definition 3 Outline micro-cluster: When the weight w of a micro-cluster or core micro-cluster is less than μ, the cluster is reduced to an outlier micro-cluster.

[0092] As shown in Figure 2, assuming that Figure 2(a) represents a set of data points with certain distribution characteristics, using the core micro-cluster set {CMC 1 ,CMC 2 ,…,CMC 3} can describe clusters of any shape in the data stream. The description of clusters using core micro-clusters is shown in Figure 2(b).

[0093] The overall framework of the density-based data stream clustering algorithm is as follows: Figure 3 As shown. The algorithm reads data points from the data stream in real time, determines the relationship between the data points and the existing micro-clusters and core micro-clusters, and merges the data points into micro-clusters or core micro-clusters. It updates and merges the existing micro-clusters and core micro-clusters, converts the micro-clusters that meet the conditions into core micro-clusters, and updates the weight of each micro-cluster or core micro-cluster. Every interval TP The core micro-clusters and outlier micro-clusters are checked, and the core micro-clusters and outlier micro-clusters with weights less than μ are deleted. Finally, data maintenance is performed on the points in the cache area.

[0094] The algorithm for updating micro clusters is as follows Figure 4 As shown in the figure, when a new data point x arrives, we first try to classify point x into the nearest core micro-cluster or potential micro-cluster based on the Euclidean distance. After the point is classified into a cluster, the cluster description parameters are updated according to the incremental maintainability of the cluster. When the data points in a micro-cluster exceed the threshold N, it means that the micro-cluster grows into a core micro-cluster. If point x cannot be classified into an existing micro-cluster or core micro-cluster, point x is inserted into the cache, where dist(·,·) represents the Euclidean distance between two points.

[0095] The online updated micro-clusters can accurately describe the data points in the data stream. In order to obtain the final sorting results, it is necessary to apply the clustering algorithm to the core micro-clusters to obtain the final results. This paper refers to the concept of density connectivity in the DBSCAN algorithm to merge micro-clusters. The micro-cluster merging algorithm process is as follows: Figure 5 As shown, each core micro cluster CMC i As a point at its center c i The virtual point with weight w on the top is another core microcluster CMC j The center position c j With c i The distance dist(c i ,c j )≤2·r, then c is called the core microcluster CMC i With CMC j Density reachable. For multiple core microclusters with density reachable, they can be considered to belong to the same category.

[0096] In addition, considering the practical problem of using multi-station time difference information to sort and pair radar pulses, the time difference of the radar pulses emitted by the same radiation source arriving at each receiving station changes continuously. If the time difference information is plotted on a two-dimensional plane using the time difference of the pulse arriving at the three stations as the horizontal and vertical coordinates, the curve formed by the time difference from the same radiation source is differentiable. Therefore, for the radar pulse sorting problem, the merging of microclusters does not need to judge the connectivity relationship between the core microcluster to be classified and each core microcluster in a certain category, but only needs to judge the connectivity relationship between the core microcluster to be classified and the latest core microclusters merged into the category to determine the classification of the core microcluster to be classified.

[0097] In addition, considering the practical problem of using multi-station time difference information to sort and pair radar pulses, the time difference of the radar pulses emitted by the same radiation source arriving at each receiving station changes continuously. If the time difference information is plotted on a two-dimensional plane using the time difference of the pulse arriving at the three stations as the horizontal and vertical coordinates, the curve formed by the time difference from the same radiation source is differentiable. Therefore, for the radar pulse sorting problem, the merging of microclusters does not need to judge the connectivity relationship between the core microcluster to be classified and each core microcluster in a certain category, but only needs to judge the connectivity relationship between the core microcluster to be classified and the latest core microclusters merged into the category to determine the classification of the core microcluster to be classified.

[0098] The cache maintenance algorithm is as follows: Figure 6 As shown in , the point set {x} in the buffer area includes two categories, one is the noise points generated by incorrect pairing, and the other is the new data points generated with the emergence of new radiation sources. First, the points in the buffer area are calculated using the formula Calculate the weight of the point at the current moment. For points with weights less than the threshold μ, they are considered to be noise points that do not belong to any cluster and are deleted from the buffer. Secondly, if the DBSCAN algorithm is used for the remaining points, new micro-clusters are generated, and the points that can be described by the new micro-clusters are deleted from the buffer.

[0099] Simulation example:

[0100] The results of the present invention can be verified by the following simulation data:

[0101] Three moving receiving stations are set up, and the starting position and the ending position of each receiving station are shown in Table 1. The receiving station moves from the starting position to the ending position at a constant speed of 7.5 km / s, and the ground radar radiation source is detected during the movement. Seven radar radiation sources are set up, and their positions and radiation source parameters are shown in Table 2. The radar pulse description word data received by the receiving station is generated according to the parameter information provided in Table 1 and Table 2, and 10% of randomly generated false radar pulse description word data is added to the data to verify the actual sorting and pairing effect of the method of the present invention.

[0102] Table 1 Parameters of the starting and ending positions of the receiving station

[0103]

[0104] Table 2 Radar radiation source position and parameters

[0105]

[0106]

[0107] The three-dimensional distribution diagram of some of the data generated is as follows Figure 7As shown, after being processed by the online sorting and pairing method proposed in the present invention, the three-dimensional distribution of the successfully sorted and paired pulse pairs is as shown in FIG. Figure 8 The results of sorting and pairing each radiation source are shown in Table 3.

[0108] Table 3 Sorting pairing accuracy

[0109] Radiation source number Number of pulses received by the main receiving station Number of successfully sorted paired pulses Sorting and matching accuracy 1 861561 832205 96.59％ 2 791066 777197 98.25％ 3 330175 328214 99.41％ 4 485398 480260 98.94％ 5 582345 569711 97.83％ 6 673278 666402 98.98％ 7 64217 62978 98.07％

[0110] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention is described in detail with reference to the embodiments, it should be understood by those skilled in the art that any modification or equivalent replacement of the technical solutions of the present invention does not depart from the spirit and scope of the technical solutions of the present invention and should be included in the scope of the claims of the present invention.

Claims

1. A radar signal online sorting and pairing method based on data stream clustering, which is implemented based on a main receiving station and N secondary receiving stations, all receiving stations keep time synchronization, and conduct reconnaissance and interception of electromagnetic signals emitted by radar radiation sources in the same reconnaissance area to obtain pulse signals, and obtain PDW data after processing; the method comprises: Step 1) N secondary receiving stations transmit the PDW data of the intercepted pulse signal to the primary receiving station at a set time interval T; Step 2) The main receiving station processes the PDW data of the pulse signal and calculates the pulse pair similarity factor using the multi-parameter information of the pulse signal. When the similarity factor is greater than or equal to the set threshold, the corresponding pulse pair is successfully paired, and the arrival time difference of the successfully paired pulse pair is calculated; Step 3) Perform data stream clustering on the arrival time difference of the successfully paired pulse pairs, classify the newly input data points into the existing target cluster or store them in the buffer area according to the discrimination conditions, update the existing target cluster online, and obtain the category to which the newly input data points belong; Step 4) Repeat the above steps until all pulse signals are sorted and paired.

2. The radar signal online sorting and pairing method based on data stream clustering according to claim 1 is characterized in that: The PDW data includes carrier frequency, pulse width, amplitude, arrival time and arrival direction, where the arrival time of the nth pulse received by the i-th receiving station is for: Where i=1 represents the primary receiving station, and the others represent the secondary receiving stations; c represents the speed of light, l i Represents the distance from the radar radiation source to the i-th receiving station.

3. The method for online sorting and pairing of radar signals based on data stream clustering according to claim 1, characterized in that: The step 2) the main receiving station processes the PDW data of the pulse signal and calculates the pulse pair similarity factor using the multi-parameter information of the pulse signal, including: The arrival time of the nth pulse in the pulse stream received by the main receiving station is recorded as Calculate the time difference window Θ based on the distance between the pulse to be paired at the primary receiving station and the secondary receiving station i (n): Among them, ε represents the time difference measurement error, τ 1i It represents the maximum arrival time difference between the pulse arriving at the primary receiving station and the secondary receiving station i, where i≠1, and is calculated as R 1i is the distance between the primary receiving station and the secondary receiving station i, and c is the speed of light; Calculate the pulse pair similarity factor α by comparing the pulses of the main receiving station with the pulses in the time difference window of each secondary receiving station j,k : Among them, ‖·‖ means to find the L2 norm of the matrix, T means transpose, represents the diagonal matrix consisting of the pulse description words of the nth pulse of the main receiving station, represents the diagonal matrix consisting of the pulse description words of the m-th pulse of the secondary receiving station i, f, A, γ, θ represent the carrier frequency, amplitude, pulse width and pulse arrival direction of the pulse respectively, σ is the weighting coefficient matrix, σ f ,σ A ,σ γ ,σ θ They respectively represent the measurement errors of pulse carrier frequency, amplitude, pulse width and pulse arrival direction.

4. The method for online sorting and pairing of radar signals based on data stream clustering according to claim 1, characterized in that: The step 2) calculates the arrival time difference of the successfully paired pulse pairs as: in, represents the arrival time of the nth pulse at the primary receiving station, Indicates the arrival time of the mth pulse at secondary receiving station i.

5. The method for online sorting and pairing of radar signals based on data stream clustering according to claim 1, characterized in that: The step 3) performs data stream clustering processing on the arrival time differences of the successfully paired pulse pairs, including: Read data points in real time from the arrival time difference data stream of the successfully paired pulse pairs, determine the relationship between the data points and the existing micro-clusters and core micro-clusters, merge the data points into micro-clusters or core micro-clusters, and update the weight of each micro-cluster or core micro-cluster, update and merge the existing micro-clusters and core micro-clusters, and transform the micro-clusters that meet the conditions into core micro-clusters; Each interval time T P The core micro-clusters and outlier micro-clusters are checked, the core micro-clusters and outlier micro-clusters whose weights are less than the threshold parameter μ are deleted, and the data of the points in the cache area are maintained.

6. The method for online sorting and pairing of radar signals based on data stream clustering according to claim 5, characterized in that: The microcluster is denoted as MC(w,c,r,p), which is a set of time-labeled n The data points x1,x2,…,x n At the time t, w is the weight of the micro-cluster, indicating the activity of the micro-cluster, which decays exponentially over time; c is the center of the micro-cluster, r is the radius of the micro-cluster, and p is the number of current points in the cluster.

7. The method for online sorting and pairing of radar signals based on data stream clustering according to claim 6, characterized in that: The step of reading data points from the arrival time difference data stream of the successfully paired pulse pairs in real time, determining the relationship between the data points and the existing micro-clusters and core micro-clusters, merging the data points into the micro-clusters or core micro-clusters, and updating the weight of each micro-cluster or core micro-cluster includes: Read the data point x in real time from the arrival time difference data stream of the successfully paired pulse pair, and judge whether the Euclidean distance from the center of the core micro-cluster is less than the micro-cluster radius r. If it is, merge x into the core micro-cluster and update the center and weight of the core micro-cluster; otherwise, judge whether the Euclidean distance from the center of the micro-cluster is less than the micro-cluster radius r. If it is, merge x into the micro-cluster and update the center and weight of the micro-cluster; If the Euclidean distance is not satisfied, the data point x is stored in the buffer.

8. The method for online sorting and pairing of radar signals based on data stream clustering according to claim 7, characterized in that: The updating and merging of the existing micro-clusters and core micro-clusters to transform the micro-clusters that meet the conditions into core micro-clusters includes: When the number of points in a micro-cluster exceeds N and the weight w>μ, the micro-cluster is upgraded to a core micro-cluster; When the weight w of a micro-cluster or core micro-cluster is less than μ, the micro-cluster is reduced to an outlier micro-cluster; If the distance between the center positions of two core microclusters is less than 2 times the microcluster radius r, they are density-reachable and belong to the same category, otherwise they are classified into a new category.

9. The method for online sorting and pairing of radar signals based on data stream clustering according to claim 7, characterized in that: The step 3) also includes maintaining the buffer area, including: The weight w of the points in the buffer area is calculated according to the following formula: Where λ represents the time attenuation coefficient, T now represents the current time, T L Indicates the moment when the pairing of the Lth data point is completed; Determine whether the weight is less than the threshold parameter μ. If it is, it is a noise point and is deleted from the buffer. Use the DBSCAN algorithm to generate new micro-clusters for the remaining points, and delete the points that can be described by the new micro-clusters from the buffer.

Images