An anti-radiation target sorting method based on a data stream clustering algorithm

Abstract

The present application belongs to the technical field of anti-radiation target sorting, and particularly relates to an anti-radiation target sorting method based on a data stream clustering algorithm. The anti-radiation target sorting method comprises: direction finding, direction finding is performed on full pulses, and a target line-of-sight angle calculated by an anti-radiation seeker is taken as real-time angle data stream; signal feature estimation, the real-time angle data stream is clustered to obtain pulse clusters with similar spatial relationships, signal feature estimation and threat determination of electromagnetic targets are performed on the pulse clusters after clustering, and target sorting is realized. The target sorting method provided by the present application can obtain high-quality and stable target sorting results, so that the anti-radiation seeker has stronger adaptability to interference and deception in complex environments, thereby improving the combat effectiveness of anti-radiation weapons.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of anti-radiation target sorting, and particularly relates to an anti-radiation target sorting method based on a data stream clustering algorithm. BACKGROUND

[0002] As one of the core components of anti-radiation weapons, the anti-radiation seeker works in a battlefield environment with a complex electromagnetic scene, and provides guidance information for anti-radiation weapons through the reception, identification and direction finding of electromagnetic signals in space. The output result directly affects the combat effectiveness of anti-radiation weapons. In the modern electronic countermeasure battlefield environment, the typical signal environment faced by the anti-radiation seeker is a dense pulse stream formed by the overlapping of pulse sequences emitted by many radiation sources in the time domain. The above pulse stream can be described in six dimensions in the frequency domain, including frequency, repetition frequency, pulse width, amplitude, phase and polarization. The frequency, repetition frequency, pulse width and polarization are related to the specific working mode of the radar and have the ability to change quickly. The spatial relationship between the target radiation source and the seeker is directly reflected in the relative amplitude and relative phase of the signals received by each antenna array unit of the seeker. The target line-of-sight angle can be obtained through direction finding algorithms such as correlation interferometer, phase interferometer, correlation interferometer or spatial spectrum, and even the target distance can be obtained through passive positioning algorithms. Therefore, in the spatial domain, the electromagnetic target characteristics can be described in two dimensions of line-of-sight angle and distance. Target sorting technology is a key technology in the design of anti-radiation seekers, which utilizes the correlation of electromagnetic target signals in time domain, frequency domain and spatial domain.

[0003] At present, the traditional target sorting mode is a multi-layer screening mode that starts from the working state of the electromagnetic target, mainly sorts by repetition frequency, and pre-sorts by frequency, pulse width and amplitude. This mode can ignore the pulse streams that do not meet the sorting conditions, thereby reducing the data amount of signal processing and the computing pressure of the signal processing board, which is very suitable for engineering applications when the computing capacity of the early embedded processor is very limited.

[0004] However, the traditional target sorting mode has the following disadvantages: while ignoring the pulse streams that do not meet the sorting conditions, this mode may also ignore the target pulses with higher threat degree, resulting in target loss; and this mode is heavily dependent on whether the electrical modulation pattern of the target radar signal meets the expectation. Since the electrical modulation pattern of the electromagnetic target has the ability to change quickly, when the electromagnetic target changes its modulation pattern or enables an unknown modulation pattern, the traditional target sorting mode may fail to complete signal sorting due to the inability to match, and thus lose the target.

[0005] With the development of embedded processors, the computing power and memory of embedded processors have been significantly improved, making it possible to apply some big data technologies to target sorting. In recent years, a space angle clustering priority anti-radiation target sorting method is proposed in the literature (Qin Wan- zhi. A space angle clustering priority anti-radiation target sorting method [J]. Electronic information countermeasure technology, 2022, 37(05): 19-23.) Breakthrough the traditional radar signal sorting on the frequency, pulse width, amplitude, frequency and other electromagnetic parameters, not subject to the matching degree between the actual working mode of the radar and the embedded sorting logic of the sorting algorithm; With the guidance data reporting cycle specified by the flight control system as the beat, all pulse direction finding within one beat is converted into the form of "azimuth-elevation angle pair (AZ-EL)", and the overall processing is carried out in the two-dimensional angle plane, so as to expand the data analysis idea from the traditional pulse timing characteristics to the two-dimensional angle image characteristics, which is more intuitive and more essential, and has stronger adaptability to interference and deception under complex battlefield electromagnetic environment conditions.

[0006] However, the space angle clustering priority anti-radiation target sorting mode has the following disadvantages: considering the limitation of the computing power and memory space of the signal processing board, the space angle clustering priority anti-radiation target sorting mode regards the data within one guidance data reporting cycle specified by the flight control system as static data object, and only clusters the data within one guidance data reporting cycle specified by the flight control system. This method completely discards the historical data before the cycle, and the number of clustered data elements is quite limited, so that the clustering result is distorted, and the target sorting cannot be effectively realized. SUMMARY

[0007] In view of the above technical problems, the present application provides an anti-radiation target sorting method based on data stream clustering algorithm, which regards the line-of-sight angle data calculated by the anti-radiation seeker as a fast-arriving, unbounded, time-varying and unpredictable data stream on the basis of the space line-of-sight angle clustering priority target sorting process, and realizes real-time high-quality target sorting by combining data stream mining technology.

[0008] The present application is realized by the following technical solutions:

[0009] An anti-radiation target sorting method based on data stream clustering algorithm, the anti-radiation target sorting method comprising:

[0010] Direction finding: direction finding is performed on all pulses, and the target line-of-sight angle calculated by the anti-radiation seeker is regarded as real-time angle data stream;

[0011] Signal feature estimation: clustering the real-time angle data stream to obtain pulse clusters with similar spatial relationships, and performing signal feature estimation and threat determination of electromagnetic targets on the clustered pulse clusters to realize target sorting.

[0012] Further, the step of direction finding specifically comprises:

[0013] (1) the preselector of the multi-channel signal receiver mounted on the anti-radiation seeker selects certain passband electromagnetic signals from the dense signal environment and sends them into the mixer, the mixer mixes the electromagnetic signals with the local oscillator signal to generate beat signals, i.e. intermediate frequency signals; the intermediate frequency signals are amplified by the intermediate frequency amplifier and sent into the pulse acquisition card, in which the pulse data acquisition is completed;

[0014] (2) the direction finding is performed on all the pulse data by using the pulse relative amplitude and pulse relative phase among the channels of the multi-channel signal receiver, the full-pulse direction finding is completed, and the target line-of-sight angle is obtained, which is expressed in the form of azimuth-elevation angle pair;

[0015] (3) the target line-of-sight angle is taken as a real-time angle data stream, and the real-time angle data stream is taken as the data input of the online micro-cluster maintenance stage in the data stream clustering algorithm, and the potential core micro-cluster is obtained.

[0016] Further, in step (2), the direction finding is performed on all the pulse data by using a correlation interferometer, a phase interferometer, a correlation interferometer or a spatial spectrum direction finding algorithm.

[0017] Further, in step (3), the method of taking the real-time angle data stream as the data input of the online micro-cluster maintenance stage in the data stream clustering algorithm to obtain the potential core micro-cluster specifically comprises:

[0018] (1) micro-cluster initialization is performed by using a density clustering algorithm;

[0019] (2) the real-time angle data stream is taken as the data input, when a data point arrives, the distance between the data point and the center of the potential core micro-cluster is calculated to obtain the potential core micro-cluster closest to the data point, and the data point is merged into the closest potential core micro-cluster;

[0020] (3) it is judged whether the core micro-cluster radius r of the potential core micro-cluster after the data point is merged is smaller than a radius threshold ε; if the core micro-cluster radius r is smaller than the radius threshold ε, step (8) is performed; if the core micro-cluster radius r is greater than or equal to the radius threshold ε, step (4) is performed. p p p

[0021] (4) the outlying micro-cluster closest to the data point is obtained, and the data point is merged into the outlying micro-cluster;

[0022] (5) it is judged whether the outlying micro-cluster radius r of the outlying micro-cluster after the data point is merged is smaller than a radius threshold ε; if the outlying micro-cluster radius r is smaller than the radius threshold ε, step (8) is performed; if the outlying micro-cluster radius r is greater than or equal to the radius threshold ε, step (6) is performed. o ​​​whether the radius r of the outlier micro-cluster is less than a radius threshold ε; if the radius r of the outlier micro-cluster is less than the radius threshold ε, step (6) is executed; if the radius r of the outlier micro-cluster is greater than or equal to the radius threshold ε, step (7) is executed. o whether the radius r of the outlier micro-cluster is less than a radius threshold ε; if the radius r of the outlier micro-cluster is less than the radius threshold ε, step (6) is executed; if the radius r of the outlier micro-cluster is greater than or equal to the radius threshold ε, step (7) is executed. o whether the radius r of the outlier micro-cluster is less than a radius threshold ε; if the radius r of the outlier micro-cluster is less than the radius threshold ε, step (6) is executed; if the radius r of the outlier micro-cluster is greater than or equal to the radius threshold ε, step (7) is executed.

[0023] (6) checking the weight ω of the outlier micro-cluster, judging whether the weight ω of the outlier micro-cluster is greater than a threshold βμ; if the weight ω of the outlier micro-cluster is greater than the threshold βμ, the outlier micro-cluster evolves into a core micro-cluster, and then step (8) is executed; if the weight ω of the outlier micro-cluster is less than or equal to the threshold βμ, the property of the outlier micro-cluster does not change, and step (8) is executed.

[0024] (7) establishing an outlier micro-cluster with the data points.

[0025] (8) periodically detecting the states of the potential core micro-cluster and the outlier micro-cluster: checking all potential core micro-clusters, and converting a potential core micro-cluster with a weight less than βμ into an outlier micro-cluster; checking all outlier micro-clusters, and completely deleting an outlier micro-cluster with a weight less than ξ.

[0026] (9) storing micro-cluster data according to a slanted time window model, the micro-cluster data including the potential core micro-cluster and the outlier micro-cluster.

[0027] Further, the step of estimating the signal feature specifically includes:

[0028] (1) establishing a target management database, and periodically reading the potential core micro-cluster; a data element of a target in the target management database includes: a unique ID number, and frequency domain and space domain feature parameters of the target;

[0029] (2) offline generating an angle clustering cluster, i.e., a pulse cluster, and taking the pulse cluster as a pulse captured by a countermeasure head on a same electromagnetic target;

[0030] (3) estimating an electromagnetic signal feature according to the pulse cluster after clustering, and taking a geometric center of the pulse cluster as a line of sight angle of an estimation operation period of the electromagnetic signal feature; a plurality of electromagnetic targets are sorted out in each operation period, and frequency domain and space domain features of the electromagnetic targets are obtained;

[0031] (4) according to steps (1)-(3), a conversion of an electromagnetic target from a time domain pulse stream to frequency domain and space domain features is completed; according to the frequency domain and space domain features of the electromagnetic target, a target sorting result of an operation period is data-associated with a target in a target management library, finally the target management database is updated, and a threat degree of the electromagnetic target in the target management database is determined; sorting and management of a countermeasure target are completed.

[0032] Further, in step (2), the method of offline generating an angle clustering cluster includes:

[0033] (1) using a density-based clustering algorithm to cluster the potential core micro-cluster of the current state, generating an angle cluster, i.e. a pulse cluster;

[0034] (2) calculating the geometric center of the pulse cluster.

[0035] Further, in step (3), the frequency domain and spatial domain features include line-of-sight angle, frequency, pulse repetition frequency, pulse width, amplitude

[0036] The beneficial technical effects of the present application are:

[0037] The anti-radiation target sorting method provided by the present application first clusters the line-of-sight angle in priority, performs direction finding on the full pulse, takes the line-of-sight angle as a real-time data stream, obtains a pulse cluster with similar spatial relationship, and then estimates the signal features of the electromagnetic target of the pulse cluster, thereby breaking the severe dependence on the electromagnetic signal features such as frequency, pulse width, pulse repetition frequency and amplitude in the traditional target sorting technology, and unknown electromagnetic target features can be estimated to provide a reference for the subsequent target attack process.

[0038] The anti-radiation target sorting method provided by the present application takes the line-of-sight angle as a fast-arriving, unbounded, time-varying and unpredictable real-time data stream, and the influence of each data element on the final result gradually decreases over time after the action of the attenuation function, instead of being directly discarded. While considering the computing power and memory space limitations of the signal processing board, the line-of-sight angle of the real-time pulse stream is clustered, and data correlation is performed through data in different running periods to obtain a high-quality and stable target sorting result, so that the anti-radiation seeker has stronger adaptability to interference and deception in complex environments, thereby improving the combat effectiveness of the anti-radiation weapon.

[0039] The anti-radiation target sorting method provided by the present application breaks the severe dependence on the electromagnetic signal features such as frequency, pulse width, pulse repetition frequency and amplitude in the traditional target sorting technology, and unknown electromagnetic target features can be estimated to provide a reference for the subsequent target attack process, realizes clustering of the real-time pulse stream, breaks the limitation of insufficient number of clustered elements in one running period, and performs data correlation through data in different running periods to obtain a high-quality and stable target sorting result, so that the anti-radiation seeker has stronger adaptability to interference and deception in complex environments, thereby improving the combat effectiveness of the anti-radiation weapon. The anti-radiation target sorting method provided by the present application has high popularization and application value in the field of target sorting. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 The flowchart of the anti-radiation target sorting method based on the data stream clustering algorithm in the embodiments of the present application;

[0041] Figure 2 Figure 1 is a flow chart of a data stream clustering algorithm in an embodiment of the present application. DETAILED DESCRIPTION

[0042] In order to make the objects, technical solutions and advantages of the present application clearer, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0043] On the contrary, the present application covers any substitution, modification, equivalent method and solution defined by the claims on the essence and scope of the present application. Further, in order to make the public have a better understanding of the present application, some specific details are described in detail in the following detailed description of the present application. The present application can also be completely understood without the description of these details by those skilled in the art.

[0044] In the process of anti-radiation attack, the user pays more attention to the spatial position of the anti-radiation target to achieve the position tracking and even the destruction of the anti-radiation target, and the spatial position of the target is difficult to change rapidly within the signal processing period of the anti-radiation seeker, which is different from the fast change ability of the electrical modulation pattern of the target electromagnetic signal. Therefore, the present application retains the advantage of the anti-radiation target sorting mode with spatial angle clustering priority, considers the line-of-sight angle data calculated by the anti-radiation seeker as fast-arriving, unbounded, time-varying and unpredictable data stream, combines the data stream mining technology to obtain real-time high-quality angle clustering results, then estimates the feature parameters of the clustered pulse cluster and determines the threat, so as to achieve high-quality target sorting.

[0045] The present application provides an anti-radiation target sorting method based on a data stream clustering algorithm, as shown in Figure 1 The anti-radiation target sorting method comprises the following steps:

[0046] Direction finding: direction finding is performed on all pulses, and the line-of-sight angle of the target calculated by the anti-radiation seeker is taken as real-time angle data stream;

[0047] Signal feature estimation: the real-time angle data stream is clustered to obtain pulse clusters with similar spatial relationships, and the signal feature estimation and threat determination of the electromagnetic target are performed on the clustered pulse clusters to achieve target sorting.

[0048] In the present embodiment, the step of direction finding specifically comprises:

[0049] (1) The preselector of the multi-channel signal receiver mounted on the anti-radiation seeker selects electromagnetic signals within a certain passband from the dense signal environment and sends them to the mixer. The mixer mixes the electromagnetic signals with the local oscillator signal to generate a beat signal, i.e., an intermediate frequency signal. The intermediate frequency signal is amplified by the intermediate frequency amplifier and sent to the pulse acquisition card. The pulse data is acquired in the pulse acquisition card to obtain the pulse center frequency, azimuth and amplitude.

[0050] (2) Using the relative amplitude and relative phase of the pulses between the channels of the multi-channel signal receiver, the direction of all pulse data is determined to complete the full pulse direction finding and the target line of sight angle is obtained. The target line of sight angle is expressed in the form of azimuth-elevation angle pairs.

[0051] (3) The target line-of-sight angle is used as a real-time angle data stream, and this real-time angle data stream is used as the data input in the online micro-cluster maintenance stage of the data stream clustering algorithm to obtain potential core micro-clusters. Specifically, the real-time angle data stream is a fast-arriving, unbounded, time-varying, and unpredictable data stream.

[0052] In step (2) of this embodiment, a radiation ratio interferometer, a phase interferometer, a correlation interferometer, or a spatial spectrum direction finding algorithm is used to determine the direction of all pulse data.

[0053] In step (3) of this embodiment, as Figure 2 As shown, the method for obtaining potential core microclusters by using the real-time angle data stream as data input in the online microcluster maintenance stage of the data stream clustering algorithm is as follows:

[0054] Micro-cluster initialization is performed using a density-based clustering algorithm;

[0055] (2) The real-time angle data stream is used as data input. When the data point arrives, the distance between the data point and the center of the potential core micro-cluster is calculated to obtain the potential core micro-cluster closest to the data point. The data point is then integrated into the potential core micro-cluster closest to the data point.

[0056] (3) Determine the core cluster radius r of the potential core microcluster after the data points are integrated. p Is it less than the radius threshold ε? If the core microcluster radius r p If the radius is less than the radius threshold ε, then proceed to step (8); if the core microcluster radius r p If the radius is greater than or equal to the radius threshold ε, then proceed to step (4);

[0057] (4) Obtain the outlier cluster closest to the data point and integrate the data point into the outlier cluster;

[0058] (5) Determine the outlier radius r of the outlier cluster after the data points are integrated. owhether the radius r of the outlier micro-cluster is less than a radius threshold value ε; if the radius r of the outlier micro-cluster is less than the radius threshold value ε, step (6) is performed; if the radius r of the outlier micro-cluster is greater than or equal to the radius threshold value ε, step (7) is performed. o whether the radius r of the outlier micro-cluster is less than a radius threshold value ε; if the radius r of the outlier micro-cluster is less than the radius threshold value ε, step (6) is performed; if the radius r of the outlier micro-cluster is greater than or equal to the radius threshold value ε, step (7) is performed. o whether the radius r of the outlier micro-cluster is less than a radius threshold value ε; if the radius r of the outlier micro-cluster is less than the radius threshold value ε, step (6) is performed; if the radius r of the outlier micro-cluster is greater than or equal to the radius threshold value ε, step (7) is performed.

[0059] (6) checking the weight ω of the outlier micro-cluster, judging whether the weight ω of the outlier micro-cluster is greater than a threshold value βμ; if the weight ω of the outlier micro-cluster is greater than the threshold value βμ, the outlier micro-cluster evolves into a core micro-cluster, and then step (8) is performed; if the weight ω of the outlier micro-cluster is less than or equal to the threshold value βμ, the property of the outlier micro-cluster does not change, and step (8) is performed.

[0060] (7) establishing an outlier micro-cluster with the data points.

[0061] (8) periodically detecting the states of the potential core micro-cluster and the outlier micro-cluster: checking all potential core micro-clusters, and converting a potential core micro-cluster with a weight less than βμ into an outlier micro-cluster; checking all outlier micro-clusters, and completely deleting an outlier micro-cluster with a weight less than ξ.

[0062] (9) storing micro-cluster data including the potential core micro-cluster and the outlier micro-cluster according to a slanted time window model. In the present application, the threshold values βμ, ε and ξ are conventional techniques, and thus will not be described herein.

[0063] In the present embodiment, the step of estimating the signal feature specifically includes:

[0064] (1) establishing a target management database, and the data elements of the target include: a unique ID number, and frequency domain and space domain feature parameters of the target; periodically reading the potential core micro-cluster;

[0065] (2) generating an angle clustering cluster, i.e. a pulse cluster, offline, and taking the pulse cluster as a pulse captured by a countermeasure radar seeker for the same electromagnetic target;

[0066] (3) estimating the electromagnetic signal feature according to the clustered pulse cluster, taking the geometric center of the pulse cluster as the line-of-sight angle of the estimation operation beat of the electromagnetic signal feature; sorting a plurality of electromagnetic targets from each operation beat, and obtaining the frequency domain and space domain features of the electromagnetic targets;

[0067] (4) completing the conversion of the electromagnetic target from the time domain pulse stream to the frequency domain and space domain features according to steps (1)-(3); data correlating the target sorting result of the operation beat with the target in the target management library according to the frequency domain and space domain features of the electromagnetic target, finally updating the target management database, and judging the threat degree of the electromagnetic target in the target management database; completing the sorting and management of the countermeasure target.

[0068] In the embodiment, in step (2), as shown in Figure 2 The method for generating angle clustering clusters offline includes:

[0069] (1) using a density-based clustering algorithm (DBSCAN algorithm) to cluster the potential core micro-cluster of the current state, generating an angle clustering cluster, i.e., a pulse cluster;

[0070] (2) calculating the geometric center of the pulse cluster.

[0071] The method provided by the application divides the anti-radiation target sorting process into two different running periods of parallel tasks of direction finding and signal feature estimation; and places the two parts of "online micro-cluster maintenance" and "offline generation of clustering clusters" of data stream clustering in the two different running periods of parallel tasks of direction finding and signal feature estimation; deeply integrates the anti-radiation target sorting and the data stream clustering algorithm, realizes a clear process, and can be modularized when software is realized, and the two parallel tasks are placed in two threads or two cores, or even two processors.

[0072] The method provided by the application gives priority to spatial line-of-sight angle clustering, performs direction finding on the full pulse, regards the line-of-sight angle as real-time data stream, obtains a pulse cluster with similar spatial relationships, then performs signal feature estimation of the electromagnetic target, finally correlates the data in different running periods, and thus obtains a high-quality and stable target sorting result.

[0073] The above only describes the preferred embodiments of the application and is not used to limit the application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the application should be included in the protection scope of the application.

Claims

1. A method for anti-radiation target sorting based on data stream clustering algorithm, characterized in that, The anti-radiation target sorting method comprises: Direction finding: direction finding is performed on all pulses, and a target line-of-sight angle calculated by an anti-radiation seeker is taken as a real-time angle data stream; Signal feature estimation: the real-time angle data stream is clustered to obtain pulse clusters with similar spatial relationships, signal feature estimation and threat determination of electromagnetic targets are performed on the pulse clusters after clustering, and target sorting is realized; The direction finding step specifically comprises: (1) A preselector of a multi-channel signal receiver selects electromagnetic signals from a dense signal environment and sends the electromagnetic signals to a mixer, the mixer mixes the electromagnetic signals with a local oscillator signal to generate beat signals, i.e. intermediate frequency signals; the intermediate frequency signals are amplified by an intermediate frequency amplifier and sent to a pulse acquisition card, and pulse data acquisition is completed in the pulse acquisition card; (2) The pulse relative amplitude and pulse relative phase among channels of the multi-channel signal receiver are used to direction find all pulse data, complete full-pulse direction finding, and obtain a target line-of-sight angle, the target line-of-sight angle is expressed in the form of a bearing-elevation angle pair; (3) The target line-of-sight angle is taken as a real-time angle data stream, and the real-time angle data stream is taken as data input in an online micro-cluster maintenance stage of a data stream clustering algorithm to obtain a potential core micro-cluster.

2. The method of claim 1, wherein the data stream clustering algorithm is a DBSCAN algorithm. In step (2), a correlation interferometer, a phase interferometer, a correlation interferometer or a spatial spectrum direction finding algorithm is used to direction find all pulse data.

3. The method of claim 2, wherein the data stream clustering algorithm is a DBSCAN algorithm. In step (3), the method for taking the real-time angle data stream as data input in the online micro-cluster maintenance stage of the data stream clustering algorithm to obtain the potential core micro-cluster specifically comprises: (1) A density clustering algorithm is used for micro-cluster initialization; (2) The real-time angle data stream is taken as data input, when a data point arrives, the distance between the data point and a potential core micro-cluster center is calculated to obtain a potential core micro-cluster closest to the data point, and the data point is merged into the potential core micro-cluster closest to the data point; (3) Whether a core micro-cluster radius r p of the potential core micro-cluster after the data point is merged is smaller than a radius threshold ε is judged; if the core micro-cluster radius r p is smaller than the radius threshold ε, step (8) is performed; if the core micro-cluster radius r p is greater than or equal to the radius threshold ε, step (4) is performed; (4) A potential outlier micro-cluster closest to the data point is obtained, and the data point is merged into the potential outlier micro-cluster; (5) Whether an outlier micro-cluster radius r o of the potential outlier micro-cluster after the data point is merged is smaller than the radius threshold ε is judged; if the outlier micro-cluster radius r o is smaller than the radius threshold ε, step (6) is performed; if the outlier micro-cluster radius r o is greater than or equal to the radius threshold ε, step (7) is performed; (6) The weight ω of the outlier micro-cluster is checked, and whether the weight ω is greater than a threshold βμ is judged; if the weight ω is greater than the threshold βμ, the outlier micro-cluster evolves into a core micro-cluster, and then step (8) is performed; if the weight ω is smaller than or equal to the threshold βμ, the property of the outlier micro-cluster does not change, and step (8) is performed; (7) An outlier micro-cluster is established with the data point; (8) Periodically detecting the state of potential core micro-cluster and outlier micro-cluster: checking all potential core micro-clusters, converting potential core micro-cluster with weight less than βμ into outlier micro-cluster; checking all outlier micro-clusters, completely deleting outlier micro-cluster with weight less than ξ; (9) Storing micro-cluster data according to the tilt time window model, wherein the micro-cluster data includes potential core micro-cluster and outlier micro-cluster.

4. The method of claim 3, wherein the data stream clustering algorithm is a DBSCAN algorithm. The step of signal feature estimation specifically includes: (1) Establishing a target management database, periodically reading the potential core micro-cluster; the data elements of the target in the target management database include: unique ID number, frequency domain and spatial domain feature parameters of the target; (2) Offline generating angle clustering cluster, i.e. pulse cluster, taking the pulse cluster as the pulse captured by the anti-radiation seeker for the same electromagnetic target; (3) Estimating the electromagnetic signal feature according to the clustered pulse cluster, taking the geometric center of the angle clustering cluster as the line of sight angle of the electromagnetic target; sorting out several electromagnetic targets in each running beat, and obtaining the frequency domain and spatial domain feature parameters of the electromagnetic target; (4) According to steps (1)-(3), the conversion of the electromagnetic target from the time domain pulse stream to the frequency domain and spatial domain feature is completed; according to the frequency domain and spatial domain feature parameters of the electromagnetic target, data association is performed between the target sorting result of the running beat and the target in the target management library, finally the target management database is updated, and the threat degree of the electromagnetic target in the target management database is determined; the sorting and management of the anti-radiation target are completed.

5. The method of claim 4, wherein the data stream clustering algorithm is a DBSCAN algorithm. In step (2), the method for offline generating angle clustering cluster includes: (1) Using a density-based clustering algorithm to cluster the potential core micro-cluster in the current state, generating an angle clustering cluster, i.e. a pulse cluster; (2) Calculating the geometric center of the pulse cluster.

6. The method of claim 4, wherein the data stream clustering algorithm is a DBSCAN algorithm. In step (3), the frequency domain and spatial domain features include line of sight angle, frequency, repetition frequency, pulse width and amplitude.

Images