Radar signal multi-station fusion sorting method

Abstract

A multi-station fusion sorting method for radar signals belongs to the field of electronic countermeasures. It solves the problems that in the case of high-density pulse streams, the time difference matching accuracy of the existing radar signal sorting method based on multiple time differences under multi-station cluster collaboration is low, and it is difficult to ensure the accuracy of sorting only relying on the time difference parameter TDOA. The present invention realizes the sorting of radar signals based on the PCN set pair potential and the optimized plane change main sorting. The present invention clusters the pulses with close characteristic parameters in each observation station, and then uses the multi-station time difference matching TDOA parameter as the basis for fusion verification, weakening the influence of the time difference matching error on the sorting result, thereby increasing the accuracy of the sorting algorithm under the condition of high-density pulse streams. It also proposes to introduce the SDIF algorithm to determine the potential transformation width W of each clustering heap i , and then determine the optimal transformation width W, so as to realize the recognition of the repetition period and the type of PRF modulation of the radar to be measured. The present invention is mainly used for radar signal sorting.

Description

Technical Field

[0001] The invention belongs to the field of electronic countermeasures, and in particular relates to a radar signal fusion and sorting method in the direction of electronic reconnaissance. Background Art

[0002] Radar radiation source sorting technology is also called radiation source pulse deinterleaving. Its main purpose is to remove cooperative pulses from randomly interleaved pulse streams and classify non-cooperative pulses according to their radars. Radar radiation source sorting is a key link in the reconnaissance system and plays a decisive role in the subsequent radar behavior identification and situation assessment. With the increasing complexity of the electromagnetic environment, various anti-interference technologies such as low interception and low signal-to-noise ratio environments have caused the radiation source characteristic parameters extracted by the reconnaissance receiver to be lost or inaccurate; the increasing number of radiation sources has caused the pulse density to continue to increase to millions or even tens of millions; the radiation source modulation method has expanded from a simple single frequency to various complex intra-pulse modulation methods including frequency modulation and coded phase modulation.

[0003] The development of sorting algorithms has gone through a process from the inter-pulse modulation PRI change characteristics, intra-pulse modulation characteristics, to the introduction of machine learning methods. At present, the unsupervised clustering pre-sorting based on machine learning, combined with the main sorting method algorithm model based on the inter-pulse modulation PRI change characteristics, conforms to the actual battlefield environment that lacks prior information, and is widely used in engineering. Among them, typical unsupervised clustering algorithms mainly include five categories: partition clustering method, hierarchical clustering method, grid clustering method, density clustering method and fuzzy clustering method. Commonly used main sorting algorithms for PRI mainly include extended association method, sequence difference histogram method, cumulative difference histogram method, PRI transformation method and plane transformation method.

[0004] Xidian University has proposed a radar signal sorting method based on multiple time differences under multi-station cluster coordination in its patent application, which is titled "Radar signal sorting method based on multiple time differences under multi-station cluster coordination" (application number 202210074764.X, application publication number CN 114488028 A). This method uses TDOA / FDOA (i.e., arrival time difference / arrival frequency difference) extracted from multi-station coordinated paired pulses to eliminate false time differences and achieve accurate radar signal sorting. Although this invention solves the problem that the single-station radar signal sorting method cannot sort multi-function scheduling re-frequency radar signals and the false time difference problem of the dual-station cluster coordinated signal sorting method, it still has some shortcomings. Judging from the simulation parameter settings of this method, the pulse flow density is relatively low, and the matching tolerance depends on prior information. In an electromagnetic environment with high pulse flow density, it is difficult to avoid pulse matching errors between the primary and secondary stations, which increases the calculation error of the arrival time difference of the pulses between the primary and secondary stations. This sorting method, which relies too much on the arrival time difference (TDOA) parameters and ignores the characteristic parameters of the signal itself, such as the carrier frequency and pulse width, is greatly affected by the TDOA parameters. The accuracy of radar signal sorting is difficult to guarantee in a complex electromagnetic environment. Summary of the invention

[0005] The purpose of the present invention is to solve the problem that the time difference matching accuracy of the existing radar signal sorting method based on multiple groups of time differences under the coordination of multiple station clusters is low in the case of high-density pulse flow, and the accuracy of TDOA sorting relying solely on the time difference parameter is difficult to guarantee. The present invention provides a radar signal multi-station fusion sorting method based on partial connection number (PCN) set pair potential and improved plane transformation method.

[0006] A radar signal multi-station fusion sorting method is implemented based on a main observation station and N sub-observation stations. All observation stations are used to detect radar signals output by the radar radiation source to be measured, and the radar signal detected by the receiver of each observation station within a preset time period is a series of pulse sequences, and the pulse sequence is used as a PDW data sample set, and each pulse signal in the pulse sequence is used as a sample; the characteristic parameters of the pulse signal include carrier frequency CF, pulse width PW, arrival angle AOA and pulse arrival time TOA;

[0007] The multi-station fusion sorting method includes the following steps:

[0008] S1. All secondary observation stations send the PDW data sample sets they detect to the main observation station;

[0009] S2, in the main observation station, preprocess the PDW data sample sets of all observation stations;

[0010] First, determine the pulse signals as cluster centers in the PDW data sample set of each observation station, and mark the pulse signals as cluster centers as having cluster centers, and mark the remaining pulse signals not as cluster centers as having no cluster centers. Then, use the pulse signals in the PDW data sample set of the main observation station as the reference to match the arrival time difference TDOA with the corresponding pulse signals in the PDW data sample set of each secondary observation station, and mark the arrival time difference TDOA that successfully matches in the corresponding samples of the PDW data sample set of all observation stations. At the same time, remove the samples that fail to match from the PDW data sample set of the observation station where they are located, and obtain the preprocessed PDW data sample set of each observation station.

[0011] S3, using the PCN set pair potential clustering method, characteristic parameter clustering is performed on the pulse signals in the preprocessed PDW data sample set of each observation station, and multiple cluster piles are formed in the preprocessed PDW data sample set of each observation station;

[0012] S4, multi-station arrival time difference TDOA clustering fusion verification, to obtain the updated PDW data sample set;

[0013] Determine the average arrival time difference TDOA of each cluster pile in the preprocessed PDW data sample set of each observation station, and according to the relationship between the average arrival time difference TDOA in each preprocessed PDW data sample set, perform decision-level fusion between the cluster piles that meet the clustering conditions to obtain the updated PDW data sample set;

[0014] Among them, each cluster pile in the updated PDW data sample set in each observation station contains one or more sets of characteristic parameters, and each set of characteristic parameters of the cluster pile includes average carrier frequency, average pulse width and average arrival angle, and each set of characteristic parameters of each cluster pile serves as the characteristic parameters of the radar radiation source to be measured corresponding to the cluster pile in the corresponding working mode;

[0015] S5, optimized plane change main sorting, output sorting results;

[0016] By introducing the SDIF algorithm, the TOA of all pulse signals in each cluster of the updated PDW data sample set in each observation station is calculated at m levels, and the m potential transformation widths W of each cluster are obtained. i ; i = 1, 2, 3 ... m, m is an integer, W i is the i-th potential transformation width;

[0017] Then for each potential transformation width W i Performing plane transformation on the pulse arrival time TOA of all pulse signals in the corresponding cluster pile to obtain a plane transformation characteristic curve;

[0018] Finally, the optimal transformation width W of each cluster stack is determined according to the m plane transformation characteristic curves corresponding to each cluster stack, and the repetition rate modulation type of the cluster stack is determined according to the change law of the plane transformation characteristic curve corresponding to the optimal transformation width W, wherein,

[0019] The optimal transformation width W of the cluster stack is used as the repetition period PRI of a radar radiation source to be measured corresponding to the cluster stack;

[0020] The repetition frequency modulation type of the cluster stack is used as the repetition frequency modulation type of a radar radiation source to be measured corresponding to the cluster stack.

[0021] Preferably, in step S2, the implementation method of first determining each pulse signal in the PDW data sample set of each observation station as the cluster center is:

[0022] S21, using LOF method to remove outlier samples in the PDW data sample set of each observation station;

[0023] S22, using the maximum distance product method to determine the pulse signals in each PDW data sample set as the cluster center after removing the outlier samples.

[0024] Preferably, in step S2, the implementation method of using each pulse signal in the PDW data sample set of the primary observation station as a reference to perform time difference of arrival TDOA matching with the corresponding pulse signal in the PDW data sample set of each secondary observation station includes:

[0025] The pulse matching time difference window is determined according to the spatial positions of the main and secondary observation stations. Within the same pulse matching time difference window, the carrier frequency CF, pulse width PW, and arrival angle AOA of the pulse signal of the main observation station are matched with the carrier frequency CF, pulse width PW, and arrival angle AOA of the corresponding pulse signal in each secondary observation station. If the matching results of each characteristic parameter are within the preset threshold, the match is determined to be successful, and the pulse arrival time TOA of the pulse signal of the main observation station is subtracted from the pulse arrival time TOA of the corresponding pulse signal in the corresponding secondary observation station to obtain the arrival time difference TDOA.

[0026] Preferably, in step S3, the PCN set pair potential clustering method is used to cluster the characteristic parameters of the pulse signals in the preprocessed PDW data sample set of each observation station, and a plurality of cluster piles are formed in the preprocessed PDW data sample set of each observation station in the following implementation manner:

[0027] S31, treating each pulse signal with a cluster center mark in the preprocessed PDW data sample set of each observation station as a cluster pile;

[0028] S32, using the PCN set pair potential clustering method, each pulse signal without cluster center mark in the preprocessed PDW data sample set of each observation station is clustered and analyzed with multiple pulse signals with cluster center mark to obtain multiple set pair potential values; the pulse signal without cluster center mark corresponding to the maximum set pair potential value is classified into the cluster pile where the corresponding pulse signal with cluster center mark is located;

[0029] At the same time, the characteristic parameters of the cluster pile to which each pulse signal without a cluster center mark belongs are updated according to each classification result.

[0030] Preferably, in step S32, the characteristic parameters of the cluster heap to which each pulse signal without a cluster center mark belongs are updated according to each classification result in the following manner:

[0031] Taking the average of the characteristic parameters of all pulse signals in the cluster pile after each classification, and taking the average of all characteristic parameters as a set of characteristic parameters of the cluster pile;

[0032] Among them, the carrier frequency CF of all pulse signals in the cluster stack is averaged as the average carrier frequency CF of the cluster stack; the pulse width PW of all pulse signals in the cluster stack is averaged as the average pulse width of the cluster stack; the arrival angle AOA of all pulse signals in the cluster stack is averaged as the average arrival angle of the cluster stack.

[0033] Preferably, the average time difference of arrival TDOA of each cluster pile in step S4 is calculated as follows:

[0034] The arrival time difference TDOA of all pulse signals in each clustering pile are clustered, and the mean of the arrival time difference TDOA of all pulse signals in the clustering result with the largest number of pulse signals in the clustering pile is selected as the average arrival time difference TDOA of the clustering pile.

[0035] Preferably, the implementation method of clustering the arrival time differences TDOA of all pulse signals in each clustering stack is:

[0036] The arrival time difference TDOA of each pulse signal in each cluster stack is compared with the arrival time difference TDOA of the remaining pulse signals in the stack, and the difference is compared with the preset arrival time difference TDOA difference range. When the difference is within the preset arrival time difference TDOA difference range, the two pulse signals corresponding to the difference are clustered.

[0037] Preferably, in step S4, the clustering condition for decision-level fusion is: subtract the average arrival time difference TDOA of any cluster pile in each preprocessed PDW data sample set from the average arrival time difference TDOA of the remaining cluster piles, and compare the difference with the preset average arrival time difference difference range; when the difference is within the preset average arrival time difference difference range, it is deemed that the clustering condition for decision-level fusion is met.

[0038] Preferably, in step S5, the optimal transformation width W of each cluster stack is determined according to the m plane transformation characteristic curves corresponding to the cluster stack in the following manner:

[0039] The plane entropy value of each plane transformation feature curve is calculated, and the potential transformation width corresponding to the plane transformation feature curve with the smallest entropy value among the m plane transformation feature curves corresponding to each cluster pile is taken as the optimal transformation width W of the cluster pile.

[0040] Principle analysis: Based on the PCN set-pair potential single-station clustering sorting, the present invention performs decision-level fusion of each clustering result based on the pulse multi-station arrival time difference TDOA, and obtains the number of radar radiation sources to be measured and the identification of the characteristic parameters of each radar radiation source to be measured. Finally, the arrival time TOA parameters of the fused clustering results are extracted. Through the plane transformation main sorting, the pulse repetition frequency modulation type and repetition period PRI of the radar radiation source to be measured can also be recognized.

[0041] The technical idea of ​​the present invention is: in view of the problem that the performance of traditional clustering algorithms is greatly affected by parameters such as prior thresholds, the partial connection coefficient theory in the mathematical field is introduced to achieve parameter-free clustering by characterizing the development trend of samples and cluster centers. At the same time, the multi-station receiving model of the main and secondary stations is established by utilizing the fact that the arrival time difference of signals emitted by the same radiation source at different observation stations is basically stable. The TDOA arrival time difference parameters of the pulses to each station are obtained by matching the time difference window of the received pulses at each station, and the single-station sorting results are subjected to decision-level fusion verification to solve the problem of over-segmentation caused by large-bandwidth signals across channels, distinguish rabbit ear pulses from narrow pulses, and adapt to multi-function radar systems. The main sorting part introduces the idea of ​​the arrival time difference histogram of the SDIF algorithm to optimize the potential transformation width W of the plane transformation. i The search process, compared with the traditional plane transformation method, steps through the transformation width within a given range, which greatly reduces the algorithm calculation amount.

[0042] The beneficial effects brought by the present invention are:

[0043] The patent mentioned in the background technology realizes the radar signal sorting process by extracting TDOA / FDOA from multi-station coordinated paired pulses, eliminating false time differences, and realizing accurate radar signal sorting. It overly relies on the arrival time difference (TDOA) parameter for sorting, while ignoring the influence of the signal's own characteristic parameters such as carrier frequency and pulse width on the sorting results. The performance of this sorting method is greatly affected by the TDOA parameters.

[0044] The radar signal sorting process implemented by the present invention is: first clustering pulses with similar characteristic parameters, and then using multi-station time difference matching TDOA parameters as the basis for fusion verification, weakening the influence of time difference matching error on the sorting result, thereby increasing the accuracy of the sorting algorithm under high-density pulse flow conditions; at the same time, compared with the patent mentioned in the technical background, the present invention can also identify the target repetition period PRI and the repetition frequency modulation type, adding more reliable basis for subsequent target behavior recognition and other processing.

[0045] The present invention also has the following advantages:

[0046] First, the present invention introduces the partial connection coefficient theory in the field of mathematics, and realizes parameter-free clustering by comparing the size of the set potential between samples and cluster centers, thus getting rid of the limitation that the performance of previous clustering algorithms is greatly affected by the prior threshold.

[0047] Second, the introduction of multi-station fusion theory can utilize the position relationship between the primary and secondary stations to obtain one-dimensional stable arrival time difference parameter information compared to the traditional single-station sorting. The clustering results within each station can be fused at the decision level, which can effectively distinguish the rabbit ear effect from narrow pulses, solve the problem of over-segmentation of channelized receivers when facing large-bandwidth signals across multiple sub-channels, and enhance the system's adaptability to multi-functional radar systems.

[0048] Third, the histogram statistics idea in the SDIF algorithm is introduced to optimize the potential transformation width W of the plane transformation i Compared with the traditional plane transformation method, which traverses the transformation width step by step within a given range, the algorithm calculation amount is greatly reduced and the operation speed of the sorting method of the present invention is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a schematic diagram of the principle of the radar signal multi-station fusion sorting method of the present invention;

[0050] Figure 2 It is a schematic diagram of the principle of preprocessing the PDW data sample set of the observation station;

[0051] Figure 3 It is a simulated three-dimensional distribution diagram of the characteristic parameters of the pulse signal in the pulse sequence detected by the observation station;

[0052] Figure 4The PCN set pair potential clustering method is used to Figure 3 The characteristic parameters of each pulse signal are clustered to obtain the pre-sorting result map of each observation station. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0055] See also Figure 1 The present embodiment is described. The radar signal multi-station fusion sorting method described in the present embodiment is implemented based on a main observation station and N sub-observation stations. All observation stations are used to detect the radar signal output by the radar radiation source to be measured, and the radar signal detected by the receiver of each observation station within a preset time period is a series of pulse sequences, and the pulse sequence is used as a PDW data sample set, and each pulse signal in the pulse sequence is used as a sample; the characteristic parameters of the pulse signal include carrier frequency CF, pulse width PW, arrival angle AOA and pulse arrival time TOA;

[0056] The multi-station fusion sorting method includes the following steps:

[0057] S1. All secondary observation stations send the PDW data sample sets they detect to the main observation station;

[0058] S2, in the main observation station, preprocess the PDW data sample sets of all observation stations;

[0059] First, determine the pulse signals as cluster centers in the PDW data sample set of each observation station, and mark the pulse signals as cluster centers as having cluster centers, and mark the remaining pulse signals not as cluster centers as having no cluster centers. Then, use the pulse signals in the PDW data sample set of the main observation station as the reference to match the arrival time difference TDOA with the corresponding pulse signals in the PDW data sample set of each secondary observation station, and mark the arrival time difference TDOA that successfully matches in the corresponding samples of the PDW data sample set of all observation stations. At the same time, remove the samples that fail to match from the PDW data sample set of the observation station where they are located, and obtain the preprocessed PDW data sample set of each observation station.

[0060] S3, using the PCN set pair potential clustering method, characteristic parameter clustering is performed on the pulse signals in the preprocessed PDW data sample set of each observation station, and multiple cluster piles are formed in the preprocessed PDW data sample set of each observation station;

[0061] S4, multi-station arrival time difference TDOA clustering fusion verification, to obtain the updated PDW data sample set;

[0062] Determine the average arrival time difference TDOA of each cluster pile in the preprocessed PDW data sample set of each observation station, and according to the relationship between the average arrival time difference TDOA in each preprocessed PDW data sample set, perform decision-level fusion between the cluster piles that meet the clustering conditions to obtain the updated PDW data sample set;

[0063] Among them, each cluster pile in the updated PDW data sample set in each observation station contains one or more sets of characteristic parameters, and each set of characteristic parameters of the cluster pile includes average carrier frequency, average pulse width and average arrival angle, and each set of characteristic parameters of each cluster pile serves as the characteristic parameters of the radar radiation source to be measured corresponding to the cluster pile in the corresponding working mode;

[0064] S5, optimized plane change main sorting, output sorting results;

[0065] By introducing the SDIF algorithm, the TOA of all pulse signals in each cluster of the updated PDW data sample set in each observation station is calculated at m levels, and the m potential transformation widths W of each cluster are obtained. i ; i = 1, 2, 3 ... m, m is an integer, W i is the i-th potential transformation width;

[0066] Then for each potential transformation width W i Performing plane transformation on the pulse arrival time TOA of all pulse signals in the corresponding cluster pile to obtain a plane transformation characteristic curve;

[0067] Finally, the optimal transformation width W of each cluster stack is determined according to the m plane transformation characteristic curves corresponding to each cluster stack, and the repetition rate modulation type of the cluster stack is determined according to the change law of the plane transformation characteristic curve corresponding to the optimal transformation width W, wherein,

[0068] The optimal transformation width W of the cluster stack is used as the repetition period PRI of a radar radiation source to be measured corresponding to the cluster stack;

[0069] The repetition frequency modulation type of the cluster stack is used as the repetition frequency modulation type of a radar radiation source to be measured corresponding to the cluster stack.

[0070] The radar signal multi-station fusion sorting method described in the present invention is mainly used to realize the sorting of radar signals, that is, to distinguish and determine the radar radiation sources to be tested; the present invention first clusters pulses with similar characteristic parameters, and then uses multi-station time difference matching TDOA parameters as the basis for fusion verification, weakening the influence of time difference matching error on the sorting result, thereby increasing the accuracy of the sorting algorithm under high-density pulse flow conditions; at the same time, compared with the patent mentioned in the technical background, the present invention can also identify the target repetition period PRI and the repetition rate modulation type, adding more reliable basis for subsequent target behavior recognition and other processing.

[0071] The proposed multi-station arrival time difference TDOA clustering fusion verification theory:

[0072] Traditional pulse repetition period estimation algorithms cannot distinguish the parameters of various multi-function radars with complex intra-pulse and inter-pulse modulation methods, and the de-interleaving effect is limited.

[0073] The present invention uses multiple observation stations to establish a multi-sensor time difference fusion system, and uses the arrival time difference (TDOA) of the same pulse at different observation stations as a new one-dimensional parameter basis to perform fusion verification on the clustering results of conventional single stations based on the principle of similar characteristic parameters, thereby making up for its limitations such as difficulty in adapting to multi-function radars. At the same time, TDOA parameters can also be used as a basis for target positioning.

[0074] The key to time difference matching is to find the serial number corresponding to the same pulse in each observation station, establish a corresponding relationship and extract parameters to calculate the arrival time difference TDOA. This part involves the selection strategy of the time difference window. The selection of time difference window parameters has nothing to do with the parameters of the radiation source itself, but only with the distance between the observation stations and the spatial layout, thus meeting the requirements for signal sorting of non-cooperative radiation sources in actual environments.

[0075] Taking a single secondary observation station as an example, the pulses received by the main observation station and the corresponding secondary observation station are paired with time difference to obtain the pulse arrival time difference TDOA parameters. When a false pulse is matched, the carrier frequency, pulse width and other pulse description word parameters can be used for matching to determine the matching pulse pair and the real time difference.

[0076] After processing the pulses in the main observation station and the secondary observation station, several groups of clustering results are obtained. The pulse parameters in each clustering result are extracted for clustering verification based on the arrival time difference TDOA. Several groups of clustering results that meet the TDOA pulse clustering verification conditions are fused at the decision level to solve the problem of excessive segmentation of large bandwidth signals across multiple channels and adapt to the different signal characteristic parameters of multi-function radars in different working states. Rabbit ear pulses will be removed when the pulse TDOA time difference between the main and secondary stations is matched because they do not meet the time difference matching conditions.

[0077] The time difference matching method essentially uses the characteristic that the pulses emitted by the same radiation source have the same time difference when arriving at the observation station and the main station to perform pulse fusion verification. It can adapt to the characteristics of different characteristic parameters of multi-function radar in different working states, and can more scientifically distinguish between narrow pulse width signals and rabbit ear signals, as well as perform feature-level fusion on the phenomenon of excessive segmentation caused by large bandwidth signals across channels. Compared with conventional fusion algorithms, it has lower time complexity.

[0078] For further details, see Figure 2 In step S2, the implementation method of first determining each pulse signal as the cluster center in the PDW data sample set of each observation station is:

[0079] S21, using LOF method to remove outlier samples in the PDW data sample set of each observation station;

[0080] S22, using the maximum distance product method to determine the pulse signals in each PDW data sample set as the cluster center after removing the outlier samples.

[0081] In this preferred implementation, the outliers in the PDW data sample set of the observation station mainly come from data measurement errors and various interferences in space. Removing the outliers is beneficial to reducing the subsequent calculation amount of the algorithm and improving the algorithm accuracy.

[0082] The maximum distance product method is used to further improve the selection of cluster centers and better thin the initial cluster centers.

[0083] For further details, see Figure 2 In step S2, the implementation method of using each pulse signal in the PDW data sample set of the main observation station as a reference to match the arrival time difference TDOA with the corresponding pulse signal in the PDW data sample set of each secondary observation station includes:

[0084] The pulse matching time difference window is determined according to the spatial positions of the main and secondary observation stations. Within the same pulse matching time difference window, the carrier frequency CF, pulse width PW, and arrival angle AOA of the pulse signal of the main observation station are matched with the carrier frequency CF, pulse width PW, and arrival angle AOA of the corresponding pulse signal in each secondary observation station. If the matching results of each characteristic parameter are within the preset threshold, the match is determined to be successful, and the pulse arrival time TOA of the pulse signal of the main observation station is subtracted from the pulse arrival time TOA of the corresponding pulse signal in the corresponding secondary observation station to obtain the arrival time difference TDOA.

[0085] In this preferred embodiment, the matching method essentially utilizes the characteristics that the receiving characteristic parameters of the same pulse at different observation stations are close, without considering the different characteristic parameters of different working modes of the multi-function radar. The selection of the time difference window depends only on the spatial position distribution of the observation station, so the multi-station matching model conforms to the characteristics of the actual electronic countermeasure environment without prior information.

[0086] Furthermore, in step S3, the PCN set pair potential clustering method is used to cluster the characteristic parameters of the pulse signals in the preprocessed PDW data sample set of each observation station, and a plurality of cluster piles are formed in the preprocessed PDW data sample set of each observation station.

[0087] S31, treating each pulse signal with a cluster center mark in the preprocessed PDW data sample set of each observation station as a cluster pile;

[0088] S32, using the PCN set pair potential clustering method, each pulse signal without cluster center mark in the preprocessed PDW data sample set of each observation station is clustered and analyzed with multiple pulse signals with cluster center mark to obtain multiple set pair potential values; the pulse signal without cluster center mark corresponding to the maximum set pair potential value is classified into the cluster pile where the corresponding pulse signal with cluster center mark is located;

[0089] At the same time, the characteristic parameters of the cluster pile to which each pulse signal without a cluster center mark belongs are updated according to each classification result.

[0090] In this preferred embodiment, the PCN set-pair potential characteristic parameter clustering method introduces the partial connection coefficient decision theory, and the sample classification is achieved by calculating the set-pair potential size of the sample set and the cluster center set to characterize the development trend of the sample toward the cluster center. This clustering method belongs to parameter-free clustering and does not require any prior information. The cluster center is dynamically corrected as the sample classification is continuously updated, which reduces the receiver measurement error and makes the clustering result closer to the actual signal output of the target.

[0091] Furthermore, in step S32, the characteristic parameters of the cluster pile to which each pulse signal without a cluster center mark belongs are updated according to each classification result as follows:

[0092] Taking the average of the characteristic parameters of all pulse signals in the cluster pile after each classification, and taking the average of all characteristic parameters as a set of characteristic parameters of the cluster pile;

[0093] Among them, the carrier frequency CF of all pulse signals in the cluster stack is averaged as the average carrier frequency CF of the cluster stack; the pulse width PW of all pulse signals in the cluster stack is averaged as the average pulse width of the cluster stack; the arrival angle AOA of all pulse signals in the cluster stack is averaged as the average arrival angle of the cluster stack.

[0094] Furthermore, the average time difference of arrival TDOA of each cluster in step S4 is calculated as follows:

[0095] The arrival time difference TDOA of all pulse signals in each clustering pile are clustered, and the mean of the arrival time difference TDOA of all pulse signals in the clustering result with the largest number of pulse signals in the clustering pile is selected as the average arrival time difference TDOA of the clustering pile.

[0096] Furthermore, the implementation method of clustering the arrival time difference TDOA of all pulse signals in each cluster pile is as follows:

[0097] The arrival time difference TDOA of each pulse signal in each cluster stack is compared with the arrival time difference TDOA of the remaining pulse signals in the stack, and the difference is compared with the preset arrival time difference TDOA difference range. When the difference is within the preset arrival time difference TDOA difference range, the two pulse signals corresponding to the difference are clustered.

[0098] Furthermore, in step S4, the clustering condition for decision-level fusion is: subtract the average arrival time difference TDOA of any cluster pile in each preprocessed PDW data sample set from the average arrival time difference TDOA of the remaining cluster piles, and compare the difference with the preset average arrival time difference difference range. When the difference is within the preset average arrival time difference difference range, it is deemed that the clustering condition for decision-level fusion is met.

[0099] In this preferred embodiment, the initial clustering stack is fused at the decision level based on the arrival time difference TDOA parameters, which can effectively adapt to the different characteristics of the characteristic parameters corresponding to various working modes of the multi-function radar system and solve the "increase batch" problem. At the same time, in the face of the problem of excessive segmentation of large bandwidth signals across channels, redundant pulse clustering stacks can be fused based on TDOA parameters.

[0100] Furthermore, in step S5, the optimal transformation width W of each cluster stack is determined according to the m plane transformation characteristic curves corresponding to the cluster stack as follows:

[0101] The plane entropy value of each plane transformation feature curve is calculated, and the potential transformation width corresponding to the plane transformation feature curve with the smallest entropy value among the m plane transformation feature curves corresponding to each cluster pile is taken as the optimal transformation width W of the cluster pile.

[0102] When the present invention is specifically applied:

[0103] (1) A main observation station and two secondary observation stations can be selected to establish a multi-station reception model, transmit the pulse stream information received by the secondary station to the main station, and use the LOF method to mark the outliers in each station;

[0104] (2 The maximum distance product method is used to iteratively determine the cluster center of the potential effective pulse signal of the received pulse stream;

[0105] (3) Determine the receiving pulse time difference window based on the distance and position relationship between the main station and each sub-observation station;

[0106] (4) Perform time difference window matching and arrival time difference TDOA calculation on the pulses received by each station. When the time difference window matches multiple pulses at the same time, auxiliary matching is performed using parameters such as carrier frequency and pulse width.

[0107] (5) Use the sample carrier frequency, pulse width, arrival angle and center to perform single-station sorting based on PCN set pair potential, and classify the samples into the cluster center with the largest corresponding set pair potential;

[0108] (6) Based on the arrival time difference parameters of the pulses of each sorting result, a fusion check is performed, and a decision-level fusion is performed on several groups of clustering results with similar arrival time difference parameters and meeting the threshold conditions to complete the pre-sorting part;

[0109] (7) Extract the TOA parameters of the pre-sorted clustered pulse arrival time and find the potential transformation width W by making a level difference histogram i

[0110] (8) With W i A plane transformation is performed to transform the width, and the signal repetition rate modulation type is obtained by observing the characteristic curve.

[0111] In specific application, the technical effect of the present invention is verified by the following data:

[0112] A total of nine radar radiation sources with different modulation types were set up, including four single-frequency, one three-difference, one repetition frequency jitter, one agile frequency, one multi-function radar and one narrow pulse agile frequency radar. A mixed signal with characteristic parameters shown in Table 1 was generated to verify the actual sorting effect of the method of the present invention.

[0113] Table 1 Radar signal simulation parameter settings

[0114]

[0115]

[0116] The pre-sorting clustering process is shown in the attached Figure 3 and Figure 4As shown, the present invention obtains the attached Figure 4 The potential valid data set in the PDW data set is determined by the maximum distance product method, and then the PDW data is clustered by single-station PCN set pair potential to obtain several groups of cluster piles. Then, the above clustering results are fused and verified based on the multi-station arrival time difference TDOA parameters. The recognition of frequency agility, multi-function radar and narrow pulse signals can be realized with high sorting accuracy. The final sorting results are shown in Table 2.

[0117] Table 2 Radar signal simulation parameter settings and sorting results

[0118]

[0119] Although the present invention is described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the present invention. It should therefore be understood that many modifications may be made to the exemplary embodiments and that other arrangements may be devised without departing from the spirit and scope of the present invention as defined by the appended claims. It should be understood that the various dependent claims and features described herein may be combined in a manner different from that described in the original claims. It should also be understood that features described in conjunction with individual embodiments may be used in other described embodiments.

Claims

1. A radar signal multi-station fusion sorting method, which is implemented based on a main observation station and N sub-observation stations. All observation stations are used to detect the radar signal output by the radar radiation source to be measured, and the radar signal detected by the receiver of each observation station within a preset time period is a series of pulse sequences, and the pulse sequence is used as a PDW data sample set, and each pulse signal in the pulse sequence is used as a sample; the characteristic parameters of the pulse signal include carrier frequency CF, pulse width PW, arrival angle AOA and pulse arrival time TOA; It is characterized in that The multi-station fusion sorting method includes the following steps: S1. All secondary observation stations send the PDW data sample sets they detect to the main observation station; S2, in the main observation station, preprocess the PDW data sample sets of all observation stations; First, determine the pulse signals as cluster centers in the PDW data sample set of each observation station, and mark the pulse signals as cluster centers as having cluster centers, and mark the remaining pulse signals not as cluster centers as having no cluster centers. Then, use the pulse signals in the PDW data sample set of the main observation station as the reference to match the arrival time difference TDOA with the corresponding pulse signals in the PDW data sample set of each secondary observation station, and mark the arrival time difference TDOA that successfully matches in the corresponding samples of the PDW data sample set of all observation stations. At the same time, remove the samples that fail to match from the PDW data sample set of the observation station where they are located, and obtain the preprocessed PDW data sample set of each observation station. S3, using the PCN set pair potential clustering method, characteristic parameter clustering is performed on the pulse signals in the preprocessed PDW data sample set of each observation station, and multiple cluster piles are formed in the preprocessed PDW data sample set of each observation station; S4, multi-station arrival time difference TDOA clustering fusion verification, to obtain the updated PDW data sample set; Determine the average arrival time difference TDOA of each cluster pile in the preprocessed PDW data sample set of each observation station, and according to the relationship between the average arrival time difference TDOA in each preprocessed PDW data sample set, perform decision-level fusion between the cluster piles that meet the clustering conditions to obtain the updated PDW data sample set; Among them, each cluster pile in the updated PDW data sample set in each observation station contains one or more sets of characteristic parameters, and each set of characteristic parameters of the cluster pile includes average carrier frequency, average pulse width and average arrival angle, and each set of characteristic parameters of each cluster pile serves as the characteristic parameters of the radar radiation source to be measured corresponding to the cluster pile in the corresponding working mode; S5, optimized plane change main sorting, output sorting results; By introducing the SDIF algorithm, the TOA of all pulse signals in each cluster of the updated PDW data sample set in each observation station is calculated at m levels, and the m potential transformation widths W of each cluster are obtained. i ; i = 1, 2, 3 ... m, m is an integer, W i is the i-th potential transformation width; Then for each potential transformation width W i Performing plane transformation on the pulse arrival time TOA of all pulse signals in the corresponding cluster pile to obtain a plane transformation characteristic curve; Finally, the optimal transformation width W of each cluster stack is determined according to the m plane transformation characteristic curves corresponding to each cluster stack, and the repetition rate modulation type of the cluster stack is determined according to the change law of the plane transformation characteristic curve corresponding to the optimal transformation width W, wherein, The optimal transformation width W of the cluster stack is used as the repetition period PRI of a radar radiation source to be measured corresponding to the cluster stack; The repetition frequency modulation type of the cluster stack is used as the repetition frequency modulation type of a radar radiation source to be measured corresponding to the cluster stack.

2. The radar signal multi-station fusion sorting method according to claim 1 is characterized in that: In step S2, the implementation method of first determining each pulse signal as the cluster center in the PDW data sample set of each observation station is: S21, using LOF method to remove outlier samples in the PDW data sample set of each observation station; S22, using the maximum distance product method to determine the pulse signals in each PDW data sample set as the cluster center after removing the outlier samples.

3. The radar signal multi-station fusion sorting method according to claim 1 or 2, characterized in that: In step S2, the implementation method of using each pulse signal in the PDW data sample set of the main observation station as a reference to match the arrival time difference TDOA with the corresponding pulse signal in the PDW data sample set of each secondary observation station includes: The pulse matching time difference window is determined according to the spatial positions of the main and secondary observation stations. Within the same pulse matching time difference window, the carrier frequency CF, pulse width PW, and arrival angle AOA of the pulse signal of the main observation station are matched with the carrier frequency CF, pulse width PW, and arrival angle AOA of the corresponding pulse signal in each secondary observation station. If the matching results of each characteristic parameter are within the preset threshold, the match is determined to be successful, and the pulse arrival time TOA of the pulse signal of the main observation station is subtracted from the pulse arrival time TOA of the corresponding pulse signal in the corresponding secondary observation station to obtain the arrival time difference TDOA.

4. The radar signal multi-station fusion sorting method according to claim 1 is characterized in that: In step S3, the PCN set pair potential clustering method is used to cluster the characteristic parameters of the pulse signals in the preprocessed PDW data sample set of each observation station, and a plurality of cluster piles are formed in the preprocessed PDW data sample set of each observation station. The implementation method is as follows: S31, treating each pulse signal with a cluster center mark in the preprocessed PDW data sample set of each observation station as a cluster pile; S32, using the PCN set pair potential clustering method, each pulse signal without cluster center mark in the preprocessed PDW data sample set of each observation station is clustered and analyzed with multiple pulse signals with cluster center mark to obtain multiple set pair potential values; the pulse signal without cluster center mark corresponding to the maximum set pair potential value is classified into the cluster pile where the corresponding pulse signal with cluster center mark is located; At the same time, the characteristic parameters of the cluster pile to which each pulse signal without a cluster center mark belongs are updated according to each classification result.

5. The radar signal multi-station fusion sorting method according to claim 4 is characterized in that: In step S32, the implementation method of updating the characteristic parameters of the cluster heap to which each pulse signal without a cluster center mark belongs according to each classification result is as follows: Taking the average of the characteristic parameters of all pulse signals in the cluster pile after each classification, and taking the average of all characteristic parameters as a set of characteristic parameters of the cluster pile; Among them, the carrier frequency CF of all pulse signals in the cluster stack is averaged as the average carrier frequency CF of the cluster stack; the pulse width PW of all pulse signals in the cluster stack is averaged as the average pulse width of the cluster stack; the arrival angle AOA of all pulse signals in the cluster stack is averaged as the average arrival angle of the cluster stack.

6. The radar signal multi-station fusion sorting method according to claim 1 is characterized in that: The calculation method of the average arrival time difference TDOA of each cluster pile in step S4 is: The arrival time difference TDOA of all pulse signals in each clustering pile are clustered, and the mean of the arrival time difference TDOA of all pulse signals in the clustering result with the largest number of pulse signals in the clustering pile is selected as the average arrival time difference TDOA of the clustering pile.

7. The radar signal multi-station fusion sorting method according to claim 6 is characterized in that: The implementation method of clustering the arrival time difference TDOA of all pulse signals in each cluster pile is: The arrival time difference TDOA of each pulse signal in each cluster stack is compared with the arrival time difference TDOA of the remaining pulse signals in the stack, and the difference is compared with the preset arrival time difference TDOA difference range. When the difference is within the preset arrival time difference TDOA difference range, the two pulse signals corresponding to the difference are clustered.

8. The radar signal multi-station fusion sorting method according to claim 1 is characterized in that: In step S4, the clustering condition for decision-level fusion is as follows: subtract the average arrival time difference TDOA of any cluster pile in each preprocessed PDW data sample set from the average arrival time difference TDOA of the remaining cluster piles, and compare the difference with the preset average arrival time difference difference range. When the difference is within the preset average arrival time difference difference range, it is deemed that the clustering condition for decision-level fusion is met.

9. The radar signal multi-station fusion sorting method according to claim 1, characterized in that: In step S5, the optimal transformation width W of each cluster pile is determined according to the m plane transformation characteristic curves corresponding to the cluster pile as follows: The plane entropy value of each plane transformation feature curve is calculated, and the potential transformation width corresponding to the plane transformation feature curve with the smallest entropy value among the m plane transformation feature curves corresponding to each cluster pile is taken as the optimal transformation width W of the cluster pile.

Images