A Method for Sorting Time Difference of Arrival Parameters of a Passive Location System for Low Earth Orbit Satellites

Abstract

The present invention relates to a method for sorting time difference of arrival parameters of a low-earth orbit satellite passive positioning system, belonging to the fields of satellite positioning and data mining. The method includes: based on multi-target time difference of arrival data, performing parameter modeling to fully extract the characteristics of the target data set; dividing the observation data into grids; based on the basic principle of hierarchical clustering, sorting data clusters for grid cells in unit time according to density characteristics; for clusters in different unit times, using the principle of position connection to perform inter-cluster connection; according to the density threshold and fitting parameters in the data clusters, selecting clusters that meet the requirements as the clustering result. The present invention constructs a multi-target time difference of arrival parameter clustering and sorting model integrating hierarchical, grid and density clustering, specifically solving the situation where target data in the low-earth orbit satellite passive positioning scenario has multiple complex characteristics. Through this sorting model, the time difference of arrival parameters of each target can be effectively sorted.

Description

Technical Field

[0001] The present invention belongs to the field of satellite positioning and data mining, and relates to a method for sorting time difference of arrival parameters of a low-orbit satellite passive positioning system. Background Art

[0002] A low-orbit satellite system generally refers to a large satellite system composed of multiple satellites that can perform real-time information processing. Low-orbit satellites are mainly used for military target detection, and it is easy to obtain the time difference of arrival parameters of the target to determine the target position. In satellite positioning, it is mainly judged based on whether the observation station itself radiates signals outward, and it is divided into two forms: passive positioning and active positioning. However, for active positioning, the device itself needs to radiate electromagnetic waves outward and its frequency band is relatively fixed. Therefore, it is very easy to be detected or tracked by the enemy, and then be subjected to targeted electronic interference or precision-guided weapon strikes by the enemy. This not only greatly affects the accuracy of accurate target positioning, but also endangers the safety of the positioning system to a large extent. Therefore, in view of the deficiencies of active positioning, passive positioning technology is a currently widely used satellite positioning method. Among them, low-orbit satellites have the advantages of fast running speed, small path loss, short transmission delay, etc. Therefore, a large amount of signals radiated outward by the radiation source can be obtained in a short time. By further processing the signals, information such as the radial velocity and azimuth of the radiation source can be obtained.

[0003] However, during the reconnaissance process of electronic countermeasures, signal sorting has always been a major problem in receiving technology. With the development of signal processing technology, although the ability to sort multiple signals seems to be improving, due to the complexity of radar signals themselves and the even faster increase in the density of the signal environment, the actual effect of signal sorting has become worse and worse. Signal sorting depends to a certain extent on a certain invariance of the signal. Under the condition of increasingly complex signals, almost all parameters of the signal are variable, even agile and pseudo-randomly variable, making it almost impossible for us to find a basis for sorting signals. Fortunately, the position of the target is stable and it cannot change suddenly. Even for a target moving at twice the speed of sound, within a time scale of 10 ms, this change is only about 7 m. That is to say, it can be considered that the position of the target remains fixed within a short period of time. Thus, the coordinates of the target will actually become the most reliable basis for signal sorting. Continuous observations by the positioning system will obtain a large number of observation parameters about the positioned target, and obtaining the target positioning parameters through continuous tracking and processing of the radiation source by the observation station can significantly improve the positioning accuracy. However, when multiple targets exist simultaneously, it is first necessary to sort the observation data in order to obtain the accurate positioning of each target separately. In the case where the number of targets is unknown and time-frequency aliasing occurs, observing the positioning parameters of multiple targets simultaneously will cause interference between different target radiation sources, making the problem of target parameter sorting very challenging. For example, there are difficulties in data pairing, inaccurate sorting is prone to fuzzy positioning or false positioning points, and there are difficulties in target positioning and tracking. Inaccurate sorting makes it difficult to achieve long-term accumulation or track fitting. From the above simple statement, it can be seen that how to effectively cluster and sort the target positioning parameters is the key to accurately obtaining the target positioning information.

[0004] In recent years, data mining has been widely applied in various fields, and there are many complex scenarios where the data set has multiple characteristics, such as those addressed in the present invention, where the target data curve is approximately linear, there are multiple targets, and there are density and accuracy differences among the target data. The limitations of traditional clustering algorithms are obvious, with both advantages and disadvantages, and they can only solve the clustering problem of a certain characteristic in a targeted manner. Currently, the main clustering analysis algorithms can be roughly divided into five categories: partitioning-based methods, hierarchical-based methods, density-based methods, grid-based methods, and model-based methods, etc.

[0005] Among them, partitioning is a relatively common method, and its most typical representative algorithm is the K-means clustering algorithm. Since this method requires setting the K value in advance and is very sensitive to the selection of the first K points, it cannot effectively solve the situation where the number of targets is unknown. For classical hierarchical clustering algorithms, such as the AGNES algorithm, each data point is used as an initial cluster, and clustering is performed from bottom to top, with large computational and time complexities, and generally requires manual specification of the clustering termination condition. At the same time, due to the density differences in the target dataset, the density-based DBSCAN clustering algorithm cannot obtain ideal results either. The same determination criteria may destroy the natural structure of the clustering, that is, sparser curves may be divided into multiple curves, while curves with high density and close proximity will be merged into one curve. Grid-based clustering methods are sensitive to parameters and cannot handle irregularly distributed data that sometimes exists; in addition, the accuracy of the clustering results of this algorithm is not easy to guarantee. Too large a grid granularity is likely to include data of different classes in the same unit, and too small a grid granularity is likely to split the interior of the same class of relatively sparse data.

[0006] Therefore, in the case where the number of targets to be observed is unknown, time-frequency aliasing exists, the target data distribution is approximately linear, there is a certain proportion of noise in the data, there are high-precision and low-precision curves in terms of parameter accuracy, and there are high-density and low-density curves in terms of density, simultaneously observing the time difference of arrival parameters of multiple targets will cause interference between different target radiation sources, making the sorting problem of the time difference parameters of each target very challenging. How to construct an effective multi-target time difference of arrival parameter clustering and sorting model in this satellite positioning scenario to achieve accurate sorting of the time difference parameters of each target is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0007] In view of this, the purpose of the present invention is to provide a method for sorting time difference of arrival parameters of a low-earth orbit satellite passive positioning system to achieve accurate sorting of multi-target time difference of arrival parameters in a complex environment.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A method for sorting time difference of arrival parameters of a low-earth orbit satellite passive positioning system includes the following steps:

[0010] S1. The data processing center performs data preprocessing based on the time difference of arrival parameters of multiple targets, and then performs parameter modeling to fully extract the characteristics of the target radiation source dataset;

[0011] S2. Partition the observation data by grids, and based on the basic principle of hierarchical clustering, perform data cluster sorting on the grid cells in a unit time according to density characteristics;

[0012] S3. Connect the clusters in different unit time periods according to the principle of position connection, and select the clusters that meet the requirements as the final clustering result based on the density threshold and fitting parameters in the data clusters.

[0013] Furthermore, step S1 specifically includes the following steps:

[0014] S11. According to the dataset obtained by the satellite, which includes the observation time series and the corresponding time difference series; at discrete time index m, let X m represent the relative time point of the m-th observation parameter, in seconds (s), and let Y m represent the corresponding time difference value, in microseconds (μs);

[0015] S12. On a two-dimensional plane, with X m as the abscissa and Y m as the ordinate, draw a plane graph composed of the time axis and the time difference axis;

[0016] S13. Through simulation experiment comparison and analysis, fully extract the basic dataset characteristics of the multi-target time difference of arrival parameters in the satellite positioning scenario.

[0017] Furthermore, the characteristics of the target radiation source dataset include:

[0018] (1) Each target on this plane corresponds to a curve; high-precision curves and low-precision curves may exist simultaneously, where the amplitude jitter of the high-precision time difference of arrival is small, and the amplitude jitter of the low-precision time difference of arrival is large;

[0019] (2) Fluctuating by several microseconds in unit time; the target data density is different; within the same time period, different curves show a separated state in terms of the time difference amplitude;

[0020] (3) Two adjacent curves are separated by obvious blank or sparse areas, and the larger the blank or sparse area, the more distinguishable the boundary between different curves.

[0021] Furthermore, step S2 specifically includes the following steps:

[0022] For the time difference parameter dataset obtained by the satellite, due to the large size of the dataset, if each data object is processed individually, the algorithm complexity is relatively high. Therefore, borrowing the idea of grid clustering, the data space is divided into grid cells, and the data objects are mapped into the grid cells. Only the grid cells are further processed. Since its processing speed is independent of the data objects, the algorithm only considers the number of data points on each grid. Therefore, the data processing efficiency can be improved. Then, according to the positions of the grid cells without data and the hierarchical clustering idea, for the grid cells in each unit time, data cluster sorting is performed according to the density characteristics;

[0023] The specific process is as follows:

[0024] S21. According to the value ranges of the discrete-time data X m and the time-difference data Y m in the observation sequence, divide the two-dimensional plane formed by the time axis (horizontal and vertical) and the time-difference axis (vertical) into grid cell structures with equal length and width, and set the time length of each grid to 1 second and the time-difference width to 1 microsecond, and record the observation parameter data points corresponding to each grid;

[0025] S22. Initialize the clustering iteration number l = 1 and the number of clusters k = 1, k ∈ [1, K], where K is the total number of clusters in the unit time, and l = 1 represents the unit time from 0 to 1 s, that is, within the l-th second;

[0026] S23. Obtain all the grid cell sets G l = {g l1 , g l2 ,..., g li}, i ∈ N, where g li represents the grid cell with the serial number i in the l-th second, i represents the grid number serial number, and N represents a positive integer;

[0027] S24. According to the positions of the grid cells without data, divide the grid cells in this second into several clusters SC lk , each cluster contains one or more grid cells and the time-difference amplitudes of the multiple grid cells are continuous, that is, there are no grid cells without data among them. Among them, k = 1, 2,...;

[0028] S25. According to all the clusters SC lk obtained in the l-th second, sequentially determine whether each cluster needs to be further split;

[0029] S26. After the processing of the l-th second is completed, let l = l + 1, k = 1 and repeat steps S22 to S25 until l = L, and the clustering ends, where L represents the last second, that is, the maximum observation time.

[0030] Furthermore, in step S25, for all the clusters SC lk in this second, calculate their root mean square errors in sequence to detect whether all the data in the data cluster can be fitted into a straight line with a small error. If the root mean square error is lower than the set threshold, the cluster can be fitted into a straight line with a small error and no further processing is performed on it; if the root mean square error is higher than the set threshold, it is determined that the data cluster cannot be fitted into a straight line with a small error and needs to be further split, then detect the density of each grid cell in the cluster, determine the peak and valley positions of its density, and split the cluster into two new clusters at the valley position. The specific process is as follows:

[0031] S251. Obtain the number N of grid cells included in each cluster in the l-th secondlk ;

[0032] S252. If N in a certain cluster lk ≥ 3, then calculate the root mean square error of the data in this cluster:

[0033]

[0034] where h(·) is the fitting function generated by the least squares method for the data in this cluster, represents the Euclidean distance between the fitting arrival time difference and the actual arrival time difference at this time point, x i is the time value within this cluster, y i is its corresponding arrival time difference value, and m is the total amount of data;

[0035] S253. If the root mean square error is lower than the set threshold parameter, that is:

[0036] RMSE(x, h) < RMSE_thr

[0037] then do not split this cluster, where RMSE_thr represents the root mean square threshold; otherwise, further split it.

[0038] Furthermore, the method for splitting the cluster includes the following steps:

[0039] A. Statistically record the density of each grid cell in this cluster as D lk (g li );

[0040] B. By finding the positions of the density peaks and valleys within this cluster, at the density valley position, that is, satisfying D lk (g li ) < D lk (g l(i-1) ) and D lk (g li ) < D lk (g l(i+1) ), further split it and generate new data clusters within the l-th second;

[0041] Furthermore, step S3 specifically includes the following steps:

[0042] For the clusters generated in each unit time, according to the principle of position connection, perform inter-cluster matching and connection, and select the clusters that meet the requirements as the final clustering result based on the density threshold and fitting parameters in the data clusters. The specific process is as follows:

[0043] S31. According to the clusters generated in each unit time, store the sorting results in the first second into the cluster set Cluster and set J = 2;

[0044] S32. For all the clusters SC in the J-th secondJk Perform inter-cluster matching connection with cluster C in Cluster; where the positions of each cluster are connected, which means that there is at least one edge or two grids with a grid vertex adjacent to each other among the grid cells g (J-1)n contained in each, that is, there exists Y(C Ji ) ≤ Y (J-1)n max (SC(SC Jk ) + 1 and Y(C (J-1)n ) ≥ Y min (SC Jk ) - 1. Where Y(·) represents the arrival time difference amplitude, and C (J-1)n represents the cluster existing in the unit time on the left side of the Jth second. n = 1, 2,..., k = 1, 2,..., where n represents the sub-cluster number index in C (J-1)n and k represents the sub-cluster number index in SC Jk ;

[0045] If a confluence situation is encountered, that is, multiple clusters in Cluster are connected to the same cluster in the Jth second, calculate the root mean square error of each cluster in Cluster and its corresponding fitting function, and calculate the fitting residuals of each cluster in Cluster with the time mean value and arrival time difference mean value of the cluster in the Jth second. Then select the cluster that best matches the cluster in the Jth second among each cluster in Cluster for inter-cluster connection;

[0046] When a cluster in Cluster_buffer is connected to multiple clusters within the Jth second, that is, a bifurcation situation, also select the cluster that best matches this cluster for connection according to the linear fitting parameters;

[0047] The clusters that fail to be connected to any cluster in Cluster in the Jth second are stored as new clusters in the cluster set Cluster, and the connected ones are merged into the corresponding clusters in Cluster;

[0048] S33. After the processing of the Jth second is completed, let J = J + 1, and repeat step S32 until J = L, where L represents the last second, that is, the maximum observation time;

[0049] S34. In the cluster set Cluster, judge the clusters that meet the requirements as the final clustering results according to the characteristics such as the density threshold and fitting curve parameters of each cluster.

[0050] The beneficial effects of the present invention are as follows: The present invention provides a method for sorting time difference of arrival (TDOA) parameters of a low-earth orbit satellite passive positioning system. The TDOA parameters of each target radiation source are obtained by low-earth orbit satellites. The TDOA parameters are preprocessed by a data processing center and modeled to fully extract the multi-complex features of the target data set. Using the basic principles of grid clustering and density clustering methods, and combining the idea of data hierarchical clustering processing, effective clustering and sorting of the TDOA parameters of each target are carried out in the case where the target data distribution is approximately linear, there is a certain proportion of noise in the data set, there are high and low precision curves in terms of parameter accuracy, and there are high and low density curves in terms of density.

[0051] Other advantages, objectives and features of the present invention will be described to some extent in the subsequent specification, and to some extent, will be obvious to those skilled in the art based on the study of the following text, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following specification. Brief Description of the Drawings

[0052] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be described in detail preferably with reference to the accompanying drawings, where:

[0053] Figure 1 is a flowchart of the method for sorting TDOA parameters of a low-earth orbit satellite passive positioning system;

[0054] Figure 2 is a schematic diagram of data clusters formed based on network, hierarchical and density clustering;

[0055] Figure 3 is a schematic diagram of clusters formed based on the principle of position connection;

[0056] Figure 4 is a schematic diagram before clustering of the target TDOA parameters;

[0057] Figure 5 is a schematic diagram after clustering of the target TDOA parameters. Detailed Embodiments

[0058] The following specific examples illustrate the embodiments of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the drawings provided in the following embodiments only illustrate the basic concept of the present invention schematically. Without conflict, the following embodiments and the features in the embodiments can be combined with each other.

[0059] Among them, the attached drawings are only for illustrative purposes, showing only schematic diagrams rather than physical diagrams, and should not be construed as a limitation on the present invention; in order to better illustrate the embodiments of the present invention, some components in the attached drawings will be omitted, enlarged or reduced, which do not represent the dimensions of the actual product; for those skilled in the art, it is understandable that some well-known structures and their descriptions in the attached drawings may be omitted.

[0060] In the attached drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components; in the description of the present invention, it should be understood that if there are terms such as "upper", "lower", "left", "right", "front", "rear", etc. indicating the orientation or positional relationship, they are based on the orientation or positional relationship shown in the attached drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, the terms describing the positional relationship in the attached drawings are only for illustrative purposes and should not be construed as a limitation on the present invention. For those of ordinary skill in the art, the specific meanings of the above terms can be understood according to specific circumstances.

[0061] First, some nouns or terms appearing in the embodiments of the present invention are explained as follows:

[0062] Clustering: In the present invention, it is an abbreviation of the "cluster analysis algorithm", which refers to an algorithm that converges the same or similar data into one category.

[0063] Cluster: In the present invention, it refers to a data set in which grid cells are connected to each other through a clustering algorithm. The data within the same cluster has the greatest similarity, and the similarity between data in different clusters is the smallest.

[0064] As Figure 1 shown, this embodiment provides a method for sorting time difference of arrival parameters of a low-earth orbit satellite passive positioning system, and the method specifically includes the following steps:

[0065] S1. The data processing center performs data preprocessing based on the multi-target time difference of arrival parameters, and then performs parameter modeling to fully extract the characteristics of the target radiation source data set, including the following steps:

[0066] S11. According to the data set obtained by the satellite, which includes the observation time series and the corresponding time difference series; at the discrete time index m, let X m represent the relative time point of the m-th observation parameter, in seconds (s), and let Y m represent the corresponding time difference value, in microseconds (μs);

[0067] S12. On the two-dimensional plane, with X m as the abscissa and Y m as the ordinate, draw a plane graph composed of the time axis and the time difference axis;

[0068] S13. Through simulation experiments for comparison and analysis, the basic dataset features of multi-target time difference of arrival parameters in the satellite positioning scenario are fully extracted.

[0069] In this example, the target dataset features include:

[0070] One curve corresponds to a single target; high-precision and low-precision curves may exist simultaneously, where the amplitude jitter of the high-precision time difference of arrival is small, while that of the low-precision time difference of arrival is large;

[0071] It fluctuates by several microseconds per unit time; the target data density varies; within the same time period, different curves are separated in terms of the time difference amplitude;

[0072] Two adjacent curves are separated by obvious blank or sparse areas, and the larger the area of the blank or sparse area, the more distinct the boundary between different curves can be distinguished.

[0073] S2. As shown in Figure 2 and Figure 3 , considering the target dataset features in this positioning scenario, the data is clustered and grouped by combining the basic principles of hierarchical, grid, and density clustering algorithms to achieve the sorting of time difference of arrival parameters;

[0074] S21. According to the value ranges of discrete time data and time difference data, grid division is performed: the time length of each grid is set to 1 second, and the time difference width is set to 1 microsecond;

[0075] S22. For the grid cells in unit time, according to the positions of the grids without data, the sorting results of several data clusters in unit time are obtained;

[0076] S23. For all clusters SC lk in each unit time, the root mean square error is calculated in turn to detect whether all the data in this data cluster can be fitted into a straight line with a small error. If the root mean square error is lower than the set threshold, this cluster can be fitted into a straight line with a small error and no further processing is required; if the root mean square error is higher than the set threshold, it is determined that this data cluster cannot be fitted into a straight line with a small error, and this cluster needs to be further split. Then, the density of each grid cell in this cluster is detected to determine the peak and valley positions of its density, and this cluster is split into two new clusters at the valley position;

[0077] S24. The data clusters in each unit time are stored to obtain the sorting results of each target in unit time.

[0078] S3. According to the principle of position connection, the data clusters in step S24 are matched and connected between clusters to further merge each cluster;

[0079] In the cluster set Cluster, according to the density threshold of each cluster and the fitting curve parameters and other characteristics, the cluster that meets the requirements is determined as the final clustering result.

[0080] Step S3 of this example specifically includes:

[0081] S31, for each data cluster per unit time generated in step S24, the sorting result within the first second is stored in a cluster set Cluster, and J=2 is set;

[0082] S32, connect the clusters in the Jth second with the clusters in the Cluster; the positions of the clusters are connected, which means that there is at least one edge or one grid vertex connecting two grids between the grid units they contain; if there is a convergence, that is, multiple clusters in the Cluster are connected to the same cluster in the Jth second, or if there is a bifurcation, that is, a cluster in the Cluster is connected to multiple clusters in the Jth second, it is necessary to select the cluster that best matches both parties for inter-cluster connection based on the linear fitting parameters of the data in their respective clusters;

[0083] The cluster that fails to connect to any cluster in the Cluster in the Jth second is stored in the Cluster as a new cluster; the connected clusters are merged into the corresponding clusters in the Cluster;

[0084] S33, after the Jth second is processed, J=J+1, and step S32 is repeated until the last second;

[0085] Among them, the characteristic of all clusters in the cluster set Cluster is that their clusters last to the most recent time unit. For example, when performing the J-th second operation, each cluster in Cluster contains a cluster at the J-1th second.

[0086] S34. In the cluster set Cluster, according to the density threshold and fitting curve parameters of each cluster and other characteristics, the cluster that meets the requirements is determined as the final clustering result.

[0087] Experimental verification: The present invention uses the Matlab simulation platform to implement the method of the present invention. This embodiment uses a relatively typical scenario for simulation verification, that is, it consists of two clock-synchronized low-orbit satellites, three targets to be measured, a radar, and a data processing center. The altitude of the low-orbit satellite is 750 km from the ground, the starting longitude and latitude of satellite A is (112, 16.7), and the ending longitude and latitude are (112.2, 14.9), the starting longitude and latitude of satellite B is (112.1, 16.1), and the ending longitude and latitude are (112.3, 14.3), the observation time is 30 seconds, the height of the target to be measured is set to 10m, the longitude distribution range is 110.2-110.3, the latitude distribution range is 10.1-10.5, and the noise ratio is 10% of the original data volume.

[0088] More specifically, two low-earth orbit satellites move at a uniform speed, collect the data T1 and T2 radiated by the target to be measured, perform a difference operation on them to obtain the TDOA, and transmit it to the data processing server. The multi-target time difference parameter clustering and sorting model based on hierarchical, grid, and density clustering fusion is used to cluster the TDOA data.

[0089] Figure 4 The TDOA data characteristics (data volume, density, parameter accuracy) of each target in this application scenario are shown, and the data distribution of each target is approximately linear.

[0090] Figure 5 The sorting diagram of the time difference parameters of each target output by the multi-target TDOA clustering and sorting model constructed by the present invention is shown. It can be seen that within the 30S observation time range, the time difference parameters of the three targets to be measured are successfully clustered and sorted, and the external evaluation indicators of the clustering results are shown in Table 1.

[0091] Table 1 Evaluation of Clustering Results

[0092]

[0093]

[0094] Among them, the larger the parameter accuracy value, the greater the oscillation amplitude of the target data and the greater the impact on TDOA clustering, as shown by the date1 target in Figure 2 The parameter measurement rate determines the total amount of data obtained by the low-earth orbit satellite for each target within the observation time range, that is, the original data volume = observation time * parameter measurement rate.

[0095] The present invention studies the basic principles of combining grid-based and density clustering algorithms in a low-earth orbit satellite passive positioning system to accurately cluster the arrival time difference parameters of each target in the presence of multiple target numbers, time-frequency aliasing, and multiple complex characteristics of target TDOA parameters. The present invention constructs a multi-target arrival time difference parameter clustering and sorting model based on hierarchical, grid, and density clustering fusion to specifically solve the situation where the target data in the low-earth orbit satellite passive positioning scenario has multiple complex characteristics, such as the shape of the target data curve has a near-linear characteristic on the plane composed of time and time difference; there are a certain proportion of random outliers in the data; in terms of parameter accuracy, there are high- and low-precision curves; in terms of density, there are high- and low-density curves. The simulation results confirm that through the constructed multi-target arrival time difference parameter clustering and sorting model, the time difference parameters of each target can be effectively sorted, and the recall rate and precision rate of the clustering results are good.

[0096] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the present technical solution, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A method for sorting time difference of arrival parameters of a low-earth orbit satellite passive positioning system, characterized in that: It includes the following steps: S1. The data processing center performs data preprocessing according to the multi-target time difference of arrival parameters, and then conducts parameter modeling to fully extract the characteristics of the target radiation source dataset. S2. The observed data is divided into grids, and based on the basic principle of hierarchical clustering, for the grid cells in a unit time, data cluster sorting is performed according to density characteristics. Step S2 includes: S21. According to the value range of the discrete-time data X in the observation sequence m and the time difference data Y m in the two-dimensional plane formed by the time axis and the time difference axis is divided into grid cell structures of equal length and width, and the time length of each grid is set to 1 second, the time difference width is set to 1 microsecond, and the observation parameter data points corresponding to each grid are recorded; wherein, the time axis is the horizontal axis of the two-dimensional plane, and the time difference axis is the vertical axis of the two-dimensional plane; S22. Initialize the clustering iteration times l = 1 and the number of clusters k = 1, k ∈ [1, K], where K is the total number of clusters in this unit time, and l = 1 represents the unit time from 0 to 1 s, that is, within the l-th second. S23. Obtain all grid cell sets \(G\) divided by the grid within the \(l\)-th second l =\(\{g\) l1 , g l2 , \(\cdots\), g li \}, i\in N\), where \(g\) li represents the grid cell with the sequence number \(i\) in the \(l\)-th second, \(i\) represents the grid number sequence number, and \(N\) represents a positive integer; S24. Divide each grid cell within this second into several clusters SC according to the positions of the grid cells with no data lk , where each cluster contains one or more grid cells and the time difference amplitudes of the multiple grid cells are continuous, that is, there are no grid cells with no data among them; where k = 1, 2,... S25. According to all clusters SC obtained in the first second lk , successively determine whether each cluster needs to be further split; S26. After the l-th second is processed, let l = l + 1, k = 1 and repeat steps S22 - S25 until l = L, and the clustering ends, where L represents the last second, that is, the maximum observation time. The said step S25 includes: S251. Obtain the number N of grid cells included in each cluster at the l-th second lk ; S252. If N lk ≥ 3 in a certain cluster, calculate the root mean square error RMSE(x,h) of the data in this cluster. If the root mean square error is lower than the set root mean square threshold RMSE_thr, do not split this cluster. Otherwise, further splitting processing is required. The methods for splitting clusters include: A. Denote the density of each grid cell in the cluster as D lk (g li ); B. By finding the positions of the density wave peaks and valleys within the cluster, at the density valley positions, i.e., satisfying D lk (g li ) < D lk (g l(i-1) ) and D lk (g li ) < D lk (g l(i+1) ), further split it and generate a new data cluster within the l-th second; S3. Through the principle of position connection, inter-cluster connection is performed on the clusters in different unit times, and according to the density threshold and fitting parameters in the data clusters, the clusters that meet the requirements are selected as the final clustering results.

2. The method for sorting time difference of arrival parameters of a low-earth orbit satellite passive positioning system according to claim 1, characterized in that: Step S1 specifically includes the following steps: S11. Based on the data set obtained from the low-earth orbit satellite, which includes the observation time series and the corresponding time difference series; at discrete time index m, let X m represent the relative time point of the m-th observation parameter, in seconds, and let Y m represent the corresponding time difference value, in microseconds; S12. On a two-dimensional plane, with X m as the abscissa and Y m as the ordinate, draw a planar graph composed of a time axis and a time difference axis; S13. Through simulation experiment comparison and analysis, the basic dataset characteristics of the multi-target time difference of arrival parameters in the satellite positioning scenario are fully extracted.

3. The method for sorting time difference of arrival parameters of a low-earth orbit satellite passive positioning system according to claim 1, characterized in that: In step S1, the characteristics of the target radiation source dataset include: One curve corresponds to a single target; there are high-precision curves or low-precision curves, or both high-precision curves and low-precision curves exist at the same time, where the amplitude jitter of the high-precision time difference of arrival is not large, and the amplitude jitter of the low-precision time difference of arrival is relatively large. The fluctuation in a unit time is on the order of several microseconds; the target data density is different; within the same time period, different curves show a separated state in terms of the time difference amplitude. Two adjacent curves are separated by obvious blank or sparse regions, and the larger the blank or sparse region, the more distinct the boundary between different curves can be distinguished.

4. The method for sorting time difference of arrival parameters of a low-earth orbit satellite passive positioning system according to claim 1, characterized in that: Step S3 specifically includes the following steps: S31. According to the clusters generated in each unit time, the sorting results in the first second are stored in the cluster set Cluster, and set J = 2. S32. Match and connect all clusters SC within the J-th second Jk with the clusters C in Cluster (J-1)n in an inter-cluster matching connection; where the positions of the clusters are connected, which means that there is at least one edge or a grid vertex adjacent to the grid cells g Ji contained in each of them, that is, there are two grids where there is at least one edge or a grid vertex adjacent to each other, that is, there exists Y(C (J-1)n ) ≤ Y max (SC Jk ) + 1 and Y(C (J-1)n ) ≥ Y min (SC Jk ) - 1; where Y(·) represents the arrival time difference amplitude, C (J-1)n represents the cluster existing in the unit time on the left side of the J-th second, n = 1, 2,..., k = 1, 2,..., where n represents the sub-cluster number index in C (J-1)n ​ When multiple clusters in Cluster are connected to the same cluster in the J-th second, that is, the confluence situation, calculate the root mean square error of each cluster in Cluster and its corresponding fitting function, and calculate the fitting residuals of each cluster in Cluster with the time mean and the time difference of arrival mean of the cluster in the J-th second, and thus select the cluster in Cluster that best matches the cluster in the J-th second for inter-cluster connection. When one cluster in Cluster is connected to multiple clusters in the J-th second, that is, the bifurcation situation, also select the clusters that best match each other according to the linear fitting parameters for inter-cluster connection. The clusters that fail to be connected to any cluster in Cluster in the J-th second are stored as new clusters in the cluster set Cluster, and the connected ones are merged into the corresponding clusters in Cluster. S33. After the J-th second is processed, let J = J + 1, and repeat step S32 until J = L, where L represents the last second, that is, the maximum observation time. S34. In the cluster set Cluster, judge the clusters that meet the requirements as the final clustering results according to the density threshold and the fitting curve parameter characteristics of each cluster.

Images