Single station multi-target passive location method based on data stream clustering algorithm

Abstract

This invention discloses a single-station multi-target passive localization method based on a data flow clustering algorithm, belonging to the field of passive localization technology. The invention employs a long baseline interferometer and includes the following steps: calculating historical line-of-sight angles; storing the historical line-of-sight angle data in memory using a tilted time window model; performing cross-location to obtain line-of-sight intersection points; treating these intersection points as a real-time spatial data stream, and using a data flow clustering method to remove interference and erroneous information from the data stream, obtaining one or more clusters at different time granularities; and extracting the effective spatial information of the target point from these clusters. This method utilizes a long baseline interferometer, resulting in simple hardware; it automatically removes erroneous intersection points during clustering, exhibiting strong anti-interference performance. Furthermore, this method is also applicable to multi-target passive localization, improving the passive localization capability of observation stations in complex electromagnetic environments and possessing high application value.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of passive location technology, and particularly relates to a single-station multi-target passive location method based on a data stream clustering algorithm. BACKGROUND

[0002] Passive location technology realizes location tracking through passive measurement of electromagnetic target radiation signals, has the advantages of good concealment, long action distance and anti-electronic interference, and thus attracts many researchers. According to the number of base stations, passive location technology can be divided into single-station passive location technology and multi-station passive location technology, wherein the single-station passive location technology can complete the location of a target without other auxiliary platforms, and is one of the current research hotspots. The traditional single-station passive location technology uses a single moving observer to measure the direction of arrival (DOA) of the incoming signal at multiple different times to realize cross location. The location convergence time and location accuracy are strongly related to the angle measurement accuracy. For the interferometer phase comparison angle measurement method commonly used in modern electronic reconnaissance systems, the longer the baseline length, the higher the angle measurement accuracy. However, the phase of the long baseline interferometer (LBI) has a phase ambiguity phenomenon. In order to solve the phase ambiguity problem, the interferometer adopts a long-short baseline combination method.

[0003] On the basis of the traditional location and tracking method, new observation quantities such as phase difference rate of change, angle of arrival rate of change, Doppler frequency rate of change, etc. can also greatly improve the location accuracy and accelerate the convergence speed, but accurate angle information is still needed to realize location. Professor Guo Fucheng of the National University of Defense Technology proposed a new method of single-station passive location using only the phase difference rate of change of the long baseline interferometer (LBI). Although this method does not require angle information, the nonlinear changes in the attitude and phase difference during the movement of the observer will cause large errors in the linearization estimator of the phase difference rate of change. In addition, a high-precision measurement platform speed and attitude rate of change are also required.

[0004] The prior art has the following disadvantages:

[0005] The existing location and tracking method requires accurate angle information to realize location, that is, before location and tracking, the phase ambiguity problem and the mirror ambiguity problem must be solved. In order to solve the phase ambiguity and mirror ambiguity problems, the interferometer adopts a long-short baseline combination method, so that the location system must have multiple antenna receiving channels, and the system has a large volume and high complexity. SUMMARY

[0006] In view of the above technical problems in the prior art, the present application provides a single-station multi-target passive location method based on a data stream clustering algorithm, which is reasonable in design, overcomes the shortcomings of the prior art, and has good effects.

[0007] In order to achieve the above object, the present application adopts the following technical solutions:

[0008] A single-station multi-target passive positioning method based on a data stream clustering algorithm adopts a long baseline interferometer and comprises the following steps:

[0009] Step 1: The phase difference between channels is collected by the long baseline interferometer, and a plurality of line-of-sight angles containing ambiguous angles or a single line-of-sight angle is obtained by decoupling;

[0010] Step 2: The line-of-sight angle data in step 1 is stored in the memory in the form of a tilt time window model;

[0011] Step 3: The line-of-sight intersection point is obtained by cross positioning the line-of-sight angle calculated in step 1 and the historical line-of-sight angle stored in the memory in step 2;

[0012] Step 4: Without resolving ambiguity, the line-of-sight intersection point is regarded as real-time spatial data stream, and the real target point information is hidden in the real-time spatial data stream; the interference and error information in the data stream are removed by a data stream clustering method to obtain a single or multiple clustering clusters of different time granularity layers, and the effective spatial information of the target point is extracted in the clustering cluster.

[0013] Preferably, in step 1, the phase difference is represented according to the phase interferometer angle measurement principle:

[0014]

[0015]

[0016] In the formula, is the phase difference; α is the apparent line-of-sight angle; λ is the wavelength of the incoming wave; d is the baseline of the interferometer antenna; k is an unknown integer, which is taken as n according to the relationship between the baseline length and the wavelength; n incoming wave apparent line-of-sight angles are calculated through the phase difference between channels at T time, of which only one is the real incoming wave apparent line-of-sight angle and the other n-1 are ambiguous incoming wave apparent line-of-sight angles; the attitude and inertial system position information of the interferometer are obtained by using the inertial navigation data of the long baseline interferometer platform; the n inertial system ray data are obtained at T time by combining the attitude and inertial system position information of the interferometer with all the incoming wave apparent line-of-sight angles at T time, and the data are stored in the memory.

[0017] Preferably, in step 2, the feature of the inclined time window model is that the data closer to the current time point is stored on a finer granularity layer, and the data farther from the current time is stored on a coarser granularity layer; the pyramid time structure, namely the progressive logarithmic inclined time window model, is a special inclined time window model; first, the data is stored in the form of snapshots in the pyramid time structure, and the snapshot refers to a set of statistical information of the data stored at a specific time in the real-time data stream; then the pyramid time structure can store the snapshots on different granularity layers according to time, that is, the data closer to the current time point is stored on a finer granularity layer, and the data farther from the current time is stored on a coarser granularity layer.

[0018] Preferably, in step 4, the data stream clustering method is as follows:

[0019] Online micro-cluster maintenance

[0020] Step 4.1: Micro-cluster initialization is performed by using a density clustering algorithm;

[0021] Step 4.2: When a data point arrives, the nearest potential core micro-cluster to the data point is obtained by calculating the distance between the data point and the potential core micro-cluster center, and the data point is attempted to be merged into the nearest potential core micro-cluster;

[0022] Step 4.3: The core micro-cluster radius r p after the data point is merged is judged to be less than the radius threshold ε;

[0023] If the judgment result is that the core micro-cluster radius is less than the radius threshold, step 4.8 is performed;

[0024] Or the judgment result is that the core micro-cluster radius is greater than or equal to the radius threshold, and step 4.4 is performed;

[0025] Step 4.4: The nearest outlier micro-cluster to the data point is obtained, and the data point is attempted to be merged into the nearest outlier micro-cluster;

[0026] Step 4.5: The outlier micro-cluster radius r o is judged to be less than the radius threshold ε;

[0027] If the judgment result is that the outlier micro-cluster radius is less than the radius threshold, step 4.6 is performed;

[0028] Or the judgment result is that the outlier micro-cluster radius is greater than or equal to the radius threshold, and step 4.7 is performed;

[0029] Step 4.6: The weight ω of the outlier micro-cluster is checked, and it is judged whether the outlier micro-cluster weight ω is greater than the threshold βμ;

[0030] If the judgment result is that the weight ω is greater than the threshold βμ; that is, ω>βμ; the outlier micro-cluster evolves into a core micro-cluster, and then step 4.8 is performed;

[0031] or the result of the judgment is that the weight ω is less than or equal to the threshold βμ; the property of the outlier micro-cluster does not change, and then step 4.8 is executed;

[0032] Step 4.7: the outlier micro-cluster is established with the data point;

[0033] Step 4.8: the state of the potential core micro-cluster and the outlier micro-cluster is detected periodically, all the potential core micro-clusters are checked, the micro-cluster with the weight less than βμ is deleted and added to the outlier micro-cluster, and all the outlier micro-clusters are checked, the outlier micro-cluster with the weight less than ξ is deleted completely;

[0034] Step 4.9: the micro-cluster data is stored according to the tilt time window model, including the potential core micro-cluster and the outlier micro-cluster;

[0035] Offline generation of clustering cluster

[0036] Step 4.10: the potential core micro-cluster in the current state is aggregated by using the density clustering algorithm;

[0037] Step 4.11: the clustering result of different time granularity layers is obtained, and the geometric center of the class is calculated.

[0038] The beneficial technical effects brought by the present application are as follows:

[0039] The method of the present application only uses a long baseline interferometer, and the hardware device is simple, small in volume and low in complexity; in the clustering process, the wrong intersection points can be automatically removed, so that the method has strong anti-interference performance, and the method is also applicable to multi-target passive positioning, and the passive positioning ability of the observation station in a complex electromagnetic environment can be greatly improved. The present application has high popularization and application value in the field of passive positioning. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 It is a flow chart of the single-station passive positioning method based on the data stream clustering algorithm;

[0041] Figure 2 It is a flow chart of the data stream clustering algorithm. DETAILED DESCRIPTION

[0042] The present application will be further described in detail below in combination with the drawings and specific embodiments:

[0043] The present application uses only a long baseline interferometer, uses a single moving observation station to measure the angle of arrival (DOA) of the incoming wave signal at different times to perform intersection positioning, obtains multiple line-of-sight intersection points, takes the intersection point coordinates as real-time data stream, combines the data stream clustering algorithm, and proposes a single-station passive positioning method based on the data stream clustering algorithm. The algorithm flow chart of the present application is shown in Figure 1 .

[0044] The technical scheme of the present application is:

[0045] 1) Line of sight angle calculation

[0046] According to the principle of phase interferometer angle measurement, the phase difference can be expressed as

[0047]

[0048]

[0049] In the formula, is the phase difference; α is the apparent line of sight angle of the incoming wave; λ is the wavelength of the incoming wave; d is the baseline of the interferometer antenna; k is an unknown integer, which is taken as n according to the relationship between the baseline length and the wavelength. Through the phase difference between the channels at T time, n apparent line of sight angles of the incoming wave can be calculated, of which only one is the true line of sight angle of the incoming wave, and the other n-1 are ambiguous line of sight angles of the incoming wave. The inertial navigation data of the interferometer carrying platform can be used to know the attitude and inertial system position information of the interferometer. Combined with the apparent line of sight angles of the incoming wave at T time, the attitude and inertial system position information of the interferometer, at T time, n ray data in the inertial system can be obtained, which is stored in the memory.

[0050] 2) History data storage

[0051] Due to the limitation of storage space and computing power, it is impossible to store all the time data, and it is necessary to store according to a specific time interval (or called window). Common time window models include sliding window model, decay function window model and tilt time window model. The sliding window model and the decay function window model can only store data at a single time granularity. However, fine time granularity data affects the real-time performance of clustering results, and coarse time granularity data affects the accuracy of clustering results, so data at different time granularity levels need to be analyzed and mined. The tilt time window model can store data at multiple time granularities, so the tilt time window model is used to store ray data in the inertial system.

[0052] The tilt time window model is characterized in that the data closer to the current time point is stored on a finer granularity layer, and the data farther from the current time is stored on a coarser granularity layer. Using the tilt time window model, not only memory is saved, but also data can be observed on different granularity layers. The pyramid time structure (progressive logarithmic tilt time window model) is a special tilt time window model. Firstly, data is stored in the form of snapshots in the pyramid time structure, and the snapshot refers to a set of statistical information of data stored at a specific time in a real-time data stream. Then the pyramid time structure can store the snapshots on different granularity layers according to time, that is, the data closer to the current time point is stored on a finer granularity layer, and the data farther from the current time is stored on a coarser granularity layer.

[0053] 3) Cross point calculation and data stream clustering

[0054] Crossing n rays in the inertial system at T time with the line of sight stored in the memory at the historical time can obtain m cross points. Taking the cross point generated at each time as a real-time data stream of numerical attributes, using a suitable data stream clustering algorithm and clustering condition, a suitable clustering cluster can be obtained. Taking each clustering cluster as a target, and the geometric center of the clustering cluster as the position target. The data stream clustering algorithm is as shown in the formula (1). Figure 2

[0055] A plurality of interferometers can obtain the angle measurement information of a plurality of electromagnetic targets at different times. This method does not need to distinguish the angle measurement information at a specific time from a certain electromagnetic target in the angle measurement stage, but only needs to view the number of clustering clusters in the final clustering result, so that the number of targets can be estimated.

[0056] Key points of the present application:

[0057] The historical line of sight angle data is stored in the memory in the form of a tilt time window model;

[0058] The line of sight angle calculated at this time is crossed with the historical line of sight angle stored in the memory to obtain a line of sight cross point;

[0059] Without demisting, the line of sight cross point is regarded as a real-time data stream, and data stream clustering processing is performed to obtain a single or multiple clustering clusters, and the spatial information of the target point is extracted in the clustering cluster.

[0060] Protection points of the present application:

[0061] The line of sight cross point is regarded as a real-time spatial data stream, and the real target point information is hidden in the spatial data stream. Through the data stream clustering algorithm, the interference and error information in the data stream are removed to obtain a clustering cluster of different time granularity layers, and the clustering cluster is effective information of the target point.

[0062] ​Of course, the above description is not a limitation of the present application, and the present application is not limited to the above examples. Changes, modifications, additions or substitutions made by those skilled in the art within the spirit and scope of the present application should also be included in the protection scope of the present application.

Claims

1. A single station multi-target passive location method based on a data stream clustering algorithm, characterized in that: The long baseline interferometer comprises the following steps: Step 1: the phase difference between channels is collected by the long baseline interferometer, and the line-of-sight angle containing the ambiguity angle or the single line-of-sight angle is obtained by decoupling; Step 2: the line-of-sight angle data in step 1 is stored in the memory in the form of a tilt time window model; Step 3: the line-of-sight intersection point is obtained by cross positioning the line-of-sight angle calculated in step 1 and the historical line-of-sight angle stored in the memory in step 2; Step 4: without ambiguity, the line-of-sight intersection point is regarded as a real-time spatial data stream, the real target point information is hidden in the real-time spatial data stream, the interference and error information in the data stream are removed by a data stream clustering method, and a single or multiple clustering clusters of different time granularity layers are obtained, and the effective spatial information of the target point is extracted in the clustering cluster; The data stream clustering method is as follows: Online micro-cluster maintenance Step 4.1: micro-cluster initialization is performed by using a density clustering algorithm; Step 4.2: when a data point arrives, the distance between the data point and the potential core micro-cluster center is calculated to obtain the potential core micro-cluster closest to the data point, and the potential core micro-cluster is tried to be merged; Step 4.3: Determine the core micro-cluster radius after data point fusion whether it is less than the radius threshold ; If: the judgment result is that the core micro-cluster radius is less than the radius threshold value; then step 4.8 is executed; Or the judgment result is that the core micro-cluster radius is greater than or equal to the radius threshold value, then step 4.4 is executed; Step 4.4: the outlying micro-cluster closest to the data point is obtained, and the outlying micro-cluster is tried to be merged; Step 4.5: Determine Outlier Microcluster Radius Less than radius threshold ; If: the judgment result is that the outlying micro-cluster radius is less than the radius threshold value; then step 4.6 is executed; Or the judgment result is that the outlying micro-cluster radius is greater than or equal to the radius threshold value, then step 4.7 is executed; Step 4.6: Check the weight of the outlier micro-cluster , determine whether the weight of the outlier micro-cluster is greater than a threshold ;​ If: the result of the judgment is the weight is greater than the threshold value ; that is ; the outlier micro-cluster evolves into the core micro-cluster, and then step 4.8 is executed; Or the judgment result is the weight Less than or equal to the threshold value ; the property of the outlier micro-cluster does not change, and then step 4.8 is executed; Step 4.7: an outlying micro-cluster is established with the data point; Step 4.8: Periodically, check the status of potential core micro-clusters and outlier micro-clusters, check all potential core micro-clusters, delete micro-clusters with weight less than and add them to the outlier micro-clusters, check all outlier micro-clusters, and delete outlier micro-clusters with weight less than completely. Step 4.9: the micro-cluster data including the potential core micro-cluster and the outlying micro-cluster are stored according to the tilt time window model; Offline generation of clustering clusters Step 4.10: the potential core micro-cluster in the current state is aggregated by using a density clustering algorithm; Step 4.11: the clustering results of different time granularity layers are obtained, and the geometric center of the class is calculated.

2. The method of claim 1, wherein the data stream clustering algorithm is based on: In step 1, the phase difference is represented according to the principle of phase interferometer angle measurement: ； ； In the formula, Phase difference; The angle of view of the incoming wave body; The wavelength of the incoming wave; The baseline of the interferometer antenna; For an unknown integer, based on the relationship between baseline length and wavelength, take... All possible values, set as ;pass The phase difference between time channels is calculated. Of the incoming wave body's line-of-sight angles, only one is the true line-of-sight angle of the incoming wave body; the others... The line-of-sight angle of the incoming wave body is ambiguous; the attitude and inertial frame position information of the interferometer are known using inertial navigation data from the long-baseline interferometer platform; combined with... The entire line-of-sight angle of the incoming wave body at any given moment, the attitude of the interferometer, and the position information of the inertial frame, in At that moment, we received The ray data in the inertial frame is stored in memory.

3. The method of claim 1, wherein the data stream clustering algorithm is based on: In step 2, the tilt time window model is characterized in that the data closer to the current time point is stored in a finer granularity layer, and the data farther from the current time is stored in a coarser granularity layer; the pyramid time structure, that is, the progressive logarithmic tilt time window model, is a special tilt time window model; first, the data is stored in the form of a snapshot in the pyramid time structure, and the snapshot refers to a set of statistical information of the data stored at a specific time in the real-time data stream; then the pyramid time structure can store the snapshots in different granularity layers according to time, that is, the data closer to the current time point is stored in a finer granularity layer, and the data farther from the current time is stored in a coarser granularity layer.

Images