A radar working mode recognition method, device and electronic equipment based on multi-station asynchronous PDW time slot trusted fusion and quality-aware timing modeling

Abstract

This invention relates to a radar operating mode recognition method, apparatus, and electronic equipment based on multi-station asynchronous PDW time slot reliable fusion and quality-aware timing modeling. The method acquires PDW observation data from multiple receiving stations, corrects the Time of Arrival (TOA) using a time reference station as a benchmark, and adaptively constructs time slots based on local PRI (Primary Rank). It calculates slot-level reliability by combining cross-station support, parameter consistency, local morphology, station reliability, and reference time constraints, filters reliable time slots, and fuses TOA, RF (Primary Rank), and PW (Primary Rank). It performs missing pulse detection and candidate compensation on the fused sequence, inputting slot-level reliability, cross-station support, missing pulse markers, and PDW features into the operating mode recognition network. The network predicts mode switching boundaries, observation quality, and missing risk, selectively updates the timing state, and outputs the operating mode using information geometric prototype classification. This invention improves the recognition stability under conditions of multi-station asynchronous operation, missing pulses, false pulses, and parameter errors.

Description

Technical Field

[0001] This invention relates to the field of radar signal processing technology, specifically to a radar operating mode identification method, device, and electronic equipment based on multi-station asynchronous PDW time slot reliable fusion and quality-aware timing modeling. It is applicable to scenarios with missing pulses, false pulses, parameter measurement errors, and mode switching boundary disturbances under asynchronous observation conditions of multiple receiving stations. Background Technology

[0002] Multifunctional radars typically perform various tasks such as search, tracking, target confirmation, and track-while-search by changing the pulse repetition interval (PRI), radio frequency (RF), pulse width (PW), and their timing scheduling. Operating mode recognition, which aims to determine the current or historical operating status of a radar source based on its pulse description word (PDW) sequence, is a key technology in electronic reconnaissance, threat assessment, and situational awareness.

[0003] In real-world distributed reconnaissance scenarios, individual receiving stations are often affected by obstruction, noise, clock asynchrony, changes in propagation conditions, and local interference, making them prone to missed pulses, false pulses, and parameter measurement deviations. Multi-station observations can provide complementary information spatially, but inconsistencies in the number of pulses, local sequence numbers, and time bases among different receiving stations, coupled with the potential for independent errors and related interference at each station, make it difficult to directly match the same real pulse line by line across multiple stations.

[0004] Existing working pattern recognition methods mostly use single-station PDW sequences as input, typically assuming that the input sequence already has good integrity and purity. When the input comes from asynchronous observations from multiple stations, simple splicing, averaging, or substitution based on one station can easily lead to problems such as false associations, spurious pulse spread, and missing real pulses, which in turn affect the subsequent recognition network's learning of the timing patterns of PRI, RF, and PW.

[0005] For example, existing patent CN115575902A discloses a multi-station fusion sorting scheme for radar signals based on primary and secondary observation station PDW data, time difference of arrival matching, and clustering result verification; existing patent CN116203508A further discloses a self-attention homogeneous matching method for asynchronous cross-system or asynchronous same-system PDW data. These schemes demonstrate that the temporal correlation, parameter matching, and fusion processing of multi-station PDW data are already established research areas.

[0006] Furthermore, Notaro et al. disclosed a radar radiation source classification method in 2019 that utilizes recurrent neural networks to extract the time-series dependence of pulse flow attributes; Du et al. disclosed a Bayesian attention network for radar operating mode recognition in 2023. Therefore, simply inputting PDW sequences into a recurrent network, attention network, or prototype classifier is insufficient to constitute the technical essence of this invention.

[0007] On the other hand, traditional neural networks rely heavily on general structures such as convolutional neural networks, recurrent neural networks, or Transformers for the recognition of PDW sequences, lacking specific modeling for parameter errors, missing pulses, spurious pulses, and mode switching boundaries unique to fused PDW sequences. When local interval anomalies or boundary abrupt changes occur, ordinary time series models are prone to writing anomalous observations into the hidden state, causing state contamination and misclassification. Summary of the Invention

[0008] In view of this, the present invention proposes a radar operating mode identification method, device and electronic equipment based on reliable fusion of time slots of multi-station asynchronous PDW and quality-aware timing modeling, so as to at least solve the problems faced by distributed receiving station PDW observation conditions, such as inter-station time inconsistency, pulse miscorrelation, coexistence of missing pulses and false pulses, parameter error interference and easy confusion of operating mode switching boundaries.

[0009] Unlike the existing technologies mentioned above, the main technical line of this invention is not a simple parallel of multi-station fusion, temporal networks, and classifiers, but rather: using locally PRI-driven adaptive time slots to reconstruct asynchronous multi-station observations into unified pulse occupancy; enabling cross-station support, parameter consistency, local temporal morphology, station historical reliability, and reference time constraints to jointly form slot-level credibility; and then explicitly passing this slot-level credibility, supporting station information, and missing markers to subsequent temporal state updates, so that the front-end fusion quality can constrain the back-end recognition process.

[0010] Specifically, the present invention is achieved through the following technical solution:

[0011] According to a first aspect of the embodiments of this specification, a radar operating mode recognition method based on multi-station asynchronous PDW time slot reliable fusion and quality-aware timing modeling is provided, comprising the following steps:

[0012] Step S1: Acquire PDW observation data from multiple receiving stations observing the same radar radiation source. The PDW observation data includes the observation values ​​of the set PDW parameters, which include TOA parameters, RF parameters, and PW parameters.

[0013] Step S2: Based on the TOA parameter observation values ​​of all observation pulses from all receiving stations, divide the time into several time slots along the time axis, and determine the local PRI estimate value corresponding to each time slot based on the statistical results of the difference between the center positions of adjacent time slots within the preset neighborhood time window of each time slot; for any time slot, if the TOA parameter observation value of at least one observation pulse from a certain receiving station falls within the time boundary of that time slot, then that receiving station is recorded as the support station of that time slot.

[0014] Step S3: Calculate the credibility score of each time slot based on the number of support stations corresponding to each time slot, the consistency of the observations of each support station on each PDW parameter, and the matching degree of each time slot with the observations of each time slot within its preset neighborhood time window on each PDW parameter, and determine the credible time slot based on the credibility score.

[0015] Step S4: For each reliable time slot, the observed values ​​of each PDW parameter of the pulse obtained from each support station are weighted and fused to obtain the PDW parameter values ​​of the fused pulse. The PDW fusion sequence is obtained by arranging the TOA parameter values ​​of all fused pulses in ascending order.

[0016] Step S5: Perform pulse missing detection on the PDW fusion sequence, and if there is a pulse missing between adjacent fusion pulses, insert a candidate pulse between the adjacent fusion pulses to obtain the final PDW fusion sequence.

[0017] Step S6: Input the final PDW fusion sequence into the working mode recognition network to obtain the working mode recognition result of the radar radiation source.

[0018] According to a second aspect of the embodiments of this specification, a radar operating mode identification device based on multi-station asynchronous PDW time slot reliable fusion and quality-aware timing modeling is provided, the device comprising:

[0019] The observation data acquisition unit is used to acquire PDW observation data from multiple receiving stations observing the same radar radiation source. The PDW observation data includes the observation values ​​of set PDW parameters, which include TOA parameters, RF parameters, and PW parameters.

[0020] The time slot construction unit is used to divide the time slot along the time axis according to the TOA parameter observation values ​​of all observation pulses of all receiving stations, and to determine the local PRI estimate value corresponding to each time slot according to the statistical results of the difference between the center positions of adjacent time slots within the preset neighborhood time window of each time slot; for any time slot, if the TOA parameter observation value of at least one observation pulse of a certain receiving station falls within the time boundary of the time slot, then the receiving station is recorded as the support station of the time slot.

[0021] The time slot evaluation unit is used to calculate the credibility score of each time slot based on the number of support stations corresponding to each time slot, the consistency of the observations of each support station on each PDW parameter, and the matching degree of each time slot with the observations of each time slot within its preset neighborhood time window on each PDW parameter, and to determine the credible time slot based on the credibility score.

[0022] The multi-station pulse fusion unit is used to perform weighted fusion of the observed values ​​of each PDW parameter of the pulse obtained from each support station for each reliable time slot, so as to obtain the PDW parameter values ​​of the fused pulse, and sort them in ascending order according to the TOA parameter values ​​of all fused pulses to obtain the PDW fusion sequence.

[0023] A pulse missing detection unit is used to detect pulse missing in the PDW fusion sequence, and insert a candidate pulse between adjacent fusion pulses when there is a pulse missing between adjacent fusion pulses to obtain the final PDW fusion sequence.

[0024] The working mode recognition unit is used to input the final PDW fusion sequence into the working mode recognition network to obtain the working mode recognition result of the radar radiation source.

[0025] According to a third aspect of the embodiments of this specification, an electronic device is provided, comprising: a processor; and a computer-readable storage medium storing computer program instructions that, when executed by the processor, cause the processor to perform the method as described in the first aspect.

[0026] The embodiments of the present invention have at least the following technical effects:

[0027] (1) By dividing time slots along the time axis, the present invention transforms the cross-station matching problem of multi-station observation pulses into the problem of determining the affiliation between each observation pulse and the time slot, thus avoiding the misassociation caused by direct matching station by station and pulse by pulse under the conditions of inconsistent number of observation pulses, inconsistent local sequence numbers and inconsistent time references. Furthermore, the width of the time slot in the present invention is adaptively determined according to the local PRI estimate, which can adapt to the variation range of radar PRI in different working modes, and takes into account the slot's ability to capture real pulses and suppress false pulses.

[0028] (2) This invention calculates the credibility score of each time slot by considering multiple factors such as the number of support stations, the consistency of observations of each support station on each PDW parameter, and the matching degree of observations with the neighboring time slots on each PDW parameter. Based on this, the credibility time slots are selected, avoiding the simple judgment that multiple support stations are real and single support stations are fake, which is based solely on the number of support stations. In particular, the introduction of the matching degree of observations of the neighboring time slots enables the rescue of real pulses with weak support based on the continuity of the neighboring parameter morphology near the mode switching boundary or in areas with strong local interference, while suppressing false pulses related to multiple stations and abnormal observation points near the boundary.

[0029] (3) The present invention performs weighted fusion of PDW observations of support stations in each trusted time slot. Since the fusion object has been screened for credibility, the pollution of spurious pulses and obvious abnormal observations on the fusion result is eliminated, which significantly improves the purity and parameter accuracy of the fused PDW sequence and provides high-quality input for subsequent working mode recognition.

[0030] (4) The present invention performs pulse missing detection on the fused PDW sequence and inserts candidate pulses when a missing pulse is confirmed, thereby making up for the time series gaps caused by the missing pulses of each station and enhancing the temporal integrity of the fused PDW sequence.

[0031] In summary, this invention, through the combined use of adaptive time slot partitioning, multi-factor reliability assessment, weighted fusion, and missing pulse interpolation, can automatically fuse and clean up multi-station signals based solely on multi-station PDW observation data, without requiring any prior information about the operating mode or knowledge of the mode switching time. Furthermore, by inputting the fused, high-quality PDW sequence into the operating mode recognition network, stable identification of the radar source's operating mode can be achieved. This invention demonstrates excellent suppression capabilities against non-ideal factors commonly found in distributed electronic reconnaissance scenarios, such as missing pulses, false pulses, parameter measurement errors, and inconsistencies in inter-station time bases, significantly improving the engineering usability and accuracy of radar source operating mode recognition in practical reconnaissance environments. Attached Figure Description

[0032] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Some specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings in an exemplary and non-limiting manner. The same reference numerals in the drawings indicate the same or similar parts or components. Those skilled in the art should understand that these drawings are not necessarily drawn to scale. In the drawings:

[0033] Figure 1This is a schematic diagram illustrating the process of radar system performing radar radiation source operating mode identification in an exemplary embodiment of the present invention;

[0034] Figure 2 This is a flowchart illustrating an exemplary embodiment of the present invention of a radar operating mode identification method based on multi-station asynchronous PDW time slot trusted fusion and quality-aware timing modeling;

[0035] Figure 3 This is a schematic diagram of a time slot shown in an exemplary embodiment of the present invention;

[0036] Figure 4 This is a block diagram illustrating an electronic device according to an exemplary embodiment of the present invention;

[0037] Figure 5 This is a block diagram of a radar operating mode recognition device based on multi-station asynchronous PDW time slot trusted fusion and quality-aware timing modeling, as illustrated in an exemplary embodiment of the present invention. Detailed Implementation

[0038] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the invention as detailed in the appended claims.

[0039] The terminology used in this invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The singular forms “a,” “the,” and “the” used in this invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0040] To facilitate understanding of the technical solution of this invention, a mathematical model is first performed on the problem of radar radiation source working mode identification in a distributed observation scenario with multiple receiving stations.

[0041] Assuming there is a total The receiving station, the first When multiple receiving stations observe the same radar radiation source within the same observation time period, the resulting PDW observation sequence is as follows:

[0042] (1)

[0043] in, For the first The number of pulses observed by each receiving station during the observation period. For the first The receiving station for the first The PDW observation vector of an observation pulse can be expressed as:

[0044] (2)

[0045] in, These represent the observed values ​​of the TOA, RF, and PW parameters, respectively.

[0046] For the For each receiving station, the observed pulse and the actual pulse emitted by the radar source are usually not in one-to-one correspondence. Let... A set of indices for real pulses. For the first The set of indices of the actual observed pulses at each receiving station. For the first A set of indices for leaky pulses at each receiving station. For the first The set of indices of receiving stations regarding spurious pulses, where the first... The observation sequence of each receiving station can be abstractly represented as:

[0047] (3)

[0048] in, Indicates the first The observation mapping corresponding to each receiving station is used to describe the effects of inter-station clock deviation, propagation path differences, and slow drift. Indicates parameter measurement error; This operator represents sorting operations based on the TOA parameter value from smallest to largest.

[0049] Specifically, no. Each receiving station receives the real pulse. The observed values ​​of the TOA parameter can be expressed as:

[0050] (4)

[0051] in, For the first The clock ratio error of each receiving station relative to the time reference station For the first Global time deviation of each receiving station relative to the time reference station For slow drift term, This represents the random measurement error of the TOA parameter.

[0052] For time reference stations ,make , and Then the observed values ​​of the RF parameters and PW parameters can be expressed as:

[0053] (5)

[0054] in, and These represent the measurement errors of the RF parameters and the PW parameters, respectively. Represents the observed values ​​of the RF parameters. This represents the observed value of the PW parameter.

[0055] when When, it indicates the first The receiving station has a missing pulse. Missing pulses increase the interval between TOA parameter values ​​of adjacent observation pulses, causing the observed PRI to appear as an accumulation of multiple true PRIs.

[0056] (6)

[0057] in, Indicates the first The first receiving station The observed pulse and the first PRI between observation pulses; Represents the first pulse in the real pulse sequence. The first real pulse and the first The real PRI between real pulses; and These represent the start and end indices of the actual pulse intervals corresponding to two adjacent observed pulses, respectively.

[0058] When there is a missing pulse between two adjacent observation pulses, the observed PRI is approximately equal to the sum of multiple true PRI within that interval.

[0059] when When, it indicates the first The receiving station exhibits spurious pulses. Spurious pulses do not correspond to genuine pulses, and they will alter the interval structure of local TOA parameter values, potentially causing abnormally short intervals or parameter abrupt changes in the observed PRI sequence.

[0060] when When this occurs, it indicates the presence of parameter measurement errors in the PDW observation sequence. Parameter measurement errors will cause the observed values ​​of TOA parameters, RF parameters, and PW parameters to deviate from their true values, thereby affecting PRI calculation, cross-site matching, and subsequent working mode identification.

[0061] When leaked pulses, spurious pulses, and parameter measurement errors coexist, the first... The mixed nonideal observations from multiple receiving stations can be expressed as:

[0062] (7)

[0063] Therefore, identifying the operating mode of radar radiation sources in a distributed observation scenario with multiple receiving stations requires solving at least three problems: first, the time references between different receiving stations are inconsistent; second, the observation sequence is incomplete, impure, or inaccurate due to missing pulses, false pulses, and parameter measurement errors; and third, the ranges of PRI, RF, and PW parameter values ​​may overlap between different operating modes, making it difficult to reliably identify them using only single-pulse parameters.

[0064] To address the aforementioned issues, the radar system of this invention generates a PDW fusion sequence through multi-station observation unification and timing consistency correction, time slot construction, multi-factor quality assessment of the time slot, and robust sequence optimization. The PDW fusion sequence is then input into the working mode recognition network, which outputs pulse-level, window-level, or sample-level working mode recognition results.

[0065] In various embodiments of the present invention, the operating modes of the radar radiation source include, but are not limited to, Velocity Search (VS) mode, High Pulse Repetition Frequency Range While Search (HRWS) mode, Medium Pulse Repetition Frequency Range While Search (MRWS) mode, Track and Search-Search (TAS-S) mode in the Track and Search combined mode, Track and Search-Track (TAS-T) mode in the Track and Search combined mode, and Single Target Tracking (STT) mode. Since different operating modes of the radar radiation source typically exhibit different temporal variation patterns in the PDW observation sequence, such as fixed, jittery, agile, grouped, staggered, and slip-varying patterns, different operating modes have different temporal variation patterns in PDW parameters such as PRI, RF, and PW, and some operating modes may have overlapping parameter ranges. This radar system does not rely on the instantaneous parameters of a single pulse when identifying the operating mode. Instead, it uses the temporal variation characteristics of the PDW sequence of continuous pulses and the complementary information between multi-station observations to determine the operating mode of the radar radiation source.

[0066] like Figure 1As described above, when performing radar radiation source operating mode identification services, this radar system receives PDW observation data streams obtained from observations of the same radar radiation source by multiple receiving stations. Each receiving station's output PDW observation data stream includes at least the observed values ​​of PDW parameters such as TOA, RF, and PW. Due to different observation conditions at each receiving station, the PDW observation data streams of each receiving station may contain missing pulses, spurious pulses, parameter measurement errors, and time reference deviations. Therefore, the number of pulses, local pulse numbers, and TOA start points are usually inconsistent between different receiving stations.

[0067] After receiving PDW observation data streams from multiple receiving stations, this radar system first performs field unification processing, mapping the TOA, RF, and PW of each receiving station to a unified data format and removing obviously invalid observations. Subsequently, the PRI sequence of each station is obtained based on the difference between adjacent TOA, and the RF, PW, and PRI are scaled or normalized for subsequent fusion and identification network use.

[0068] After unifying the fields, this radar system selects a time reference station from multiple receiving stations. The time reference station is determined comprehensively based on the number of pulses, local PRI stability, anomalous pulse ratio, and parameter measurement stability. For non-time reference stations, this radar system first estimates their global time deviation relative to the time reference station based on the difference between similar TOA values, and then further estimates the affine clock error and local drift error, thereby correcting the TOA sequences of different receiving stations to a unified time scale.

[0069] After time correction, this radar system adaptively constructs time slots based on the corrected TOA sequence and local PRI estimates. In this embodiment, a time slot refers to a discrete time interval unit divided along the corrected time axis to accommodate the observed TOA parameters of the observed pulses from all receiving stations. Each time slot represents a possible actual pulse transmission moment, and its width is adaptively adjusted according to the local PRI estimate. In other words, each time slot represents a possible actual pulse location. If observed pulses from different receiving stations fall into the same time slot, these observed pulses are considered to likely correspond to the same actual pulse, and the observed pulse closest to the center of the time slot from each receiving station is preferentially retained as the representative observation.

[0070] For each time slot, this radar system calculates its reliability score based on cross-station support, intra-slot parameter consistency, local morphological consistency, receiver reliability, and reference time constraint scores. The reliability score of each time slot is compared with a corresponding threshold to determine its reliability. Generally, if a time slot has good parameter consistency with multiple stations, it is considered a reliable time slot; if a time slot is supported by only a single station but has consistent local timing patterns with the preceding and following stations, it can also be considered a reliable time slot; if a time slot exhibits abnormally short intervals, parameter spikes, or significant inconsistencies with local morphology, it is considered a suspected spurious pulse time slot and is deleted.

[0071] For a reliable time slot, this radar system robustly fuses the observations of TOA, RF, and PW parameters from each station within the time slot. The fusion can employ weighted median, in-point weighted average, or reference skeleton constraint fusion methods, where the station weights are determined based on the station's parameter residual scale and historical reliability. After pulse missing interpolation of the PDW fused sequence, this radar system obtains the final PDW fused sequence, which includes the TOA, RF, PW, PRI of each fused pulse, and the corresponding reliability score.

[0072] The final PDW fusion sequence is used as input to the working mode recognition network. The working mode recognition network first constructs the difference, second-order difference, relative change and local statistical features based on RF, PW and PRI; then estimates the mode switching probability and pulse quality probability; then updates the temporal state and suppresses the contamination of the hidden state by missing pulses, false pulses and low-quality observations; finally, it outputs the working mode category through information geometric prototype classification.

[0073] The output of this radar system includes the operating mode category, category confidence level, mode switching boundary probability, pulse quality probability, and missing pulse risk warning. If the operating mode identification results are consistent across multiple consecutive windows, a stable operating mode is output; if the boundary probability continues to increase, a possible operating mode switch is indicated; if the quality probability remains low or the missing pulse risk is high, the current identification result is significantly affected by missing pulses, false pulses, or parameter errors.

[0074] The following detailed explanation of each step is provided in conjunction with the accompanying drawings.

[0075] Figure 2 This is a flowchart illustrating an exemplary embodiment of the present invention of a radar operating mode recognition method based on multi-station asynchronous PDW time slot reliable fusion and quality-aware timing modeling, as shown below. Figure 2 As shown, the radar radiation source operating mode identification method includes the following steps:

[0076] Step S1: Acquire PDW observation data from multiple receiving stations observing the same radar radiation source. The PDW observation data includes the observation values ​​of the set PDW parameters, which include TOA parameters, RF parameters, and PW parameters.

[0077] Step S2: Based on the TOA parameter observation values ​​of all observation pulses from all receiving stations, divide the time axis into several time slots, and determine the local PRI estimate value corresponding to each time slot based on the statistical results of the difference between the center positions of adjacent time slots within the preset neighborhood time window of each time slot; for any time slot, if the TOA parameter observation value of at least one observation pulse from a certain receiving station falls within the time boundary of that time slot, then that receiving station is recorded as a support station for that time slot.

[0078] Step S3: Calculate the credibility score of each time slot based on the number of support stations corresponding to each time slot, the consistency of the observations of each support station on each PDW parameter, and the matching degree of each time slot with the observations of each time slot within its preset neighborhood time window on each PDW parameter, and determine the credible time slot based on the credibility score.

[0079] Step S4: For each reliable time slot, the observed values ​​of each PDW parameter of the pulse obtained from each support station are weighted and fused to obtain the PDW parameter values ​​of the fused pulse. The fused pulses are then sorted in ascending order according to the TOA parameter values ​​of all fused pulses to obtain the PDW fusion sequence.

[0080] Step S5: Perform pulse missing detection on the PDW fusion sequence, and if there is a pulse missing between adjacent fusion pulses, insert a candidate pulse between the adjacent fusion pulses to obtain the final PDW fusion sequence.

[0081] like Figure 2 As illustrated in the radar radiation source operating mode identification method, this embodiment of the invention, through the combination of adaptive time slot partitioning, multi-factor reliability assessment, weighted fusion, and missing pulse interpolation, can automatically complete the fusion and purification of multi-station signals based solely on multi-station PDW observation data, without requiring any prior operating mode information or knowledge of the mode switching time. Based on this, inputting the fused high-quality PDW sequence into the operating mode identification network enables stable identification of the radar radiation source's operating mode.

[0082] This invention has a good ability to suppress non-ideal factors that are common in distributed electronic reconnaissance scenarios, such as missing pulses, false pulses, parameter measurement errors, and inconsistencies in time references between stations, and significantly improves the engineering usability and recognition accuracy of radar radiation source working mode identification in actual reconnaissance environments.

[0083] In practical applications, PDW observation data contains observation errors, which can be uniformly described as observation biases formed by the combined effects of inter-station clock deviations, slow drift, propagation path differences, and random measurement errors.

[0084] Based on this, in some embodiments of the present invention, step S2 includes:

[0085] One receiving station is selected from the plurality of receiving stations as a time reference station, and the time reference station is used as a reference to perform time deviation correction on the observed values ​​of the TOA parameters of the other receiving stations; the time deviation correction includes at least one of global time deviation correction, clock ratio error correction and local drift compensation.

[0086] Based on the corrected TOA parameter observations, the center position of each time slot is determined along the time axis, and the local PRI estimate value corresponding to each time slot is determined based on the statistical results of the difference between the center positions of adjacent time slots within the preset neighborhood time window of each time slot.

[0087] The slot width of each time slot is determined based on the local PRI estimate corresponding to each time slot.

[0088] Specifically, a time reference station is first adaptively selected from multiple receiving stations. Let the time reference station be... Then, coarse registration is performed based on the PRI differential sequence or local pulse interval pattern of each station, and a robust set of matched pulse pairs between receiving station s and time reference station r is obtained using random sample consensus estimation. Then, estimate the global time deviation of the receiving station s relative to the time reference station r using the following formula. :

[0089] (8)

[0090] in, Indicates receiving station Observation pulse index, , For receiving station The number of observed pulses; superscript " " indicates the observed value; Indicates receiving station The Observed values ​​of the TOA parameter for each observation pulse; This represents the set of cross-station matched pulse pairs obtained through coarse registration using the PRI differential mode and random sampling consistency test. This indicates that the i-th observation pulse of receiving station s and the k-th observation pulse of time reference station r are determined to be candidate co-originating pulses.

[0091] This embodiment uses a pulse interval mode to establish the matching relationship, which can avoid the ambiguity of modulo PRI caused by direct nearest neighbor matching when the fixed deviation exceeds one PRI or when the PRI is periodic.

[0092] Based on the global time deviation For the receiving station Observation pulse Correct the TOA parameter value:

[0093] (9)

[0094] in, These are the TOA parameter values ​​after global time deviation correction.

[0095] Then, the receiving station is used according to the following formula. Observation pulse Clock ratio error correction and local drift compensation are performed using the TOA parameter values:

[0096] (10)

[0097] The superscript "cal" indicates the value after time correction; Indicates receiving station The The TOA parameter values ​​after global time deviation, clock ratio error and local drift compensation are completed by each observation pulse; The clock proportional error in formula (4) The estimated value; This represents the local drift term in formula (4). The estimated function, and The estimation method can be found in the relevant technical solutions by those skilled in the art, and will not be described in detail in this embodiment.

[0098] We can first use robust matched pulse pairs to form an initial time skeleton, and then perform coarse clustering on the corrected TOA observations to obtain the location centers of the time slots arranged in ascending time order. To avoid near-zero adjacency differences in the merged sequence after the same real pulse is received simultaneously by multiple receiving stations, the local PRI estimate for the j-th time slot is obtained by statistically analyzing the differences between the centers of adjacent initial slots:

[0099] (11)

[0100] in, This represents the median operator. Indicates the first The observed values ​​of the TOA parameters of each observation pulse after time deviation correction. , , Indicates the total number of time slots. Indicates the first The preset domain time window corresponding to each time slot.

[0101] Next, according to the first Local PRI estimates for each time slot Calculate the first The half width of the slot corresponding to each time slot :

[0102] (12)

[0103] Where clip(·) represents clipping operation; This is the ratio coefficient for the slot width; , These represent the preset lower limit and preset upper limit of the slot half-width, respectively; This refers to the constraint coefficient for the spacing between adjacent slots, as exemplified. .

[0104] This embodiment uses the constraint η<0.5 to ensure that the time boundaries of adjacent time slots do not overlap in principle; when an individual observation simultaneously meets the candidate conditions of two slots, it is uniquely assigned to the nearest slot center.

[0105] In constructing such Figure 3 After the time slot shown, the following formula is used to obtain the first... The set of support stations corresponding to each time slot :

[0106] (13)

[0107] It is worth noting that for multiple observation pulses falling into the same time slot from the same receiving station, priority is given to retaining those closest to the center of the time slot. The most recent observation pulse is used as the representative observation.

[0108] In some scenarios, the observation periods of multiple receiving stations span two different operating modes. Mode switching typically does not directly cause abrupt changes in the local clock ratio error or propagation path delay at the receiving station, but it does alter the PRI scheduling pattern, pulse correlation, TOA coarse-match residual distribution, and the receiver's local operating point. Therefore, using the same correction parameter to correct the TOA parameter values ​​of the observed pulses may lead to local timing deviations in the TOA parameter values ​​near the mode switching boundary.

[0109] Based on this, in some embodiments, step S2 further includes:

[0110] Based on the PDW observation data of pulses observed by each support station in each time slot, fingerprint information of each initial time slot is calculated. The fingerprint information includes: the difference between the current time slot and its previous time slot regarding the center position of the time slot; the mean RF parameter corresponding to the current time slot, which refers to the mean RF parameter values ​​of all observed pulses in the support station set of the current time slot; the mean PW parameter corresponding to the current time slot, which refers to the mean PW parameter values ​​of all observed pulses in the support station set of the current time slot; the difference in PRI parameter values ​​between the current time slot and its previous time slot; and at least one of the following: the PRI standard deviation within a preset neighborhood time window of the current time slot.

[0111] Based on the fingerprint information of each initial time slot and the fingerprint information within the preset neighboring time window of each initial time slot, mode switching detection is performed.

[0112] When a mode switch is detected in the radar radiation source during the observation phase, the mode switch boundary is determined and the observation pulse subsequence outside the mode switch boundary region is obtained. A new time slot is constructed based on the observation pulse subsequence.

[0113] The new time slot and the original time slot within the mode switching boundary area are used together as the final time slot.

[0114] Specifically, in this embodiment, the first step is calculated using the following formula. Fingerprint information corresponding to each time slot :

[0115] (14)

[0116] in, , ,

[0117] The horizontal line indicates that the corresponding parameter has been normalized. ; Indicates the first The standard deviation of PRI within the preset domain time window of each time slot.

[0118] Then based on fingerprint information Calculate the first Boundary mutation score for each time slot :

[0119] (15)

[0120] in, and They represent the first The left side of each time slot is continuous Each time slot and the right continuum The collection of fingerprint information within each time slot The preset neighborhood time window length; and These are preset weighting coefficients.

[0121] Finally, the boundary discrimination formula is used to determine whether the radar radiation source switches modes during the observation phase:

[0122] (16)

[0123] in, Represents the boundary score sequence The median absolute deviation, Indicates all valid time slots The boundary score sequence is composed of effective time slots, which are those with continuous time slots on both its left and right sides. Each time slot.

[0124] If the above boundary discrimination formula holds, then the first... The position center of the first time slot is determined as the mode switching boundary center, and the first time slot is determined as the mode switching boundary center. Each time slot and its left and right sides The time region consisting of several time slots is marked as the mode switching boundary region, where... The preset boundary region half-width, .

[0125] In some embodiments, the observation sequences on both sides of the mode switching boundary region are subjected to segmented secondary correction. Specifically, using the detected mode switching boundary as the segmentation position, the observation sequences of each receiving station are divided into observation segments on the left side of the boundary and observation segments on the right side of the boundary. For each observation segment, the global time deviation, clock ratio error, and local drift of each receiving station relative to the time reference station are re-estimated using the TOA observation values ​​within the observation segment according to the aforementioned time consistency correction method, and the TOA parameter values ​​within the observation segment are then subjected to secondary correction based on the re-estimated parameters. The observation segments on both sides of the mode switching boundary region each use their own re-estimated time correction parameters, thereby reducing the correction error caused by changes in local timing characteristics near the mode switching. After completing the segmented secondary correction, the TOA parameter values ​​after segmented correction are re-divided into new time slots according to the time axis. The newly divided time slots and the original time slots within the mode switching boundary region are used as the final time slots, and subsequent reliability assessments and parameter fusion are performed on the final time slots.

[0126] In this embodiment, mode switching boundary detection is performed. When a mode switching boundary is detected, a new time slot is re-estimated on the left side of the boundary using the observation data on the left side, and another new time slot is re-estimated on the right side of the boundary using the observation data on the right side. Each is calibrated independently, so that the TOA parameter values ​​on both sides of the boundary are closer to the true values ​​after calibration.

[0127] In some embodiments, the formula for calculating the credibility score of each time slot in step S3 above is as follows:

[0128] (17)

[0129] in, Indicates the first Credibility score for each time slot , This represents the preset weighting coefficient. The value can be determined in advance based on the verification data. For example, .

[0130] Indicates the first Cross-site support corresponding to each time slot Indicates the first Number of stations supported per time slot Indicates the total number of receiving stations;

[0131] Indicates the first The parameters within each time slot are consistent. Indicates the first PDW parameters within each time slot The degree of cross-site dispersion, Indicates the first The receiving station is at the first PDW parameters within each time slot The observed values, Indicates the first The set of support stations corresponding to each time slot , Both represent the index of the receiving station. Indicates PDW parameters The corresponding preset allowable deviation scale, A positive number set to prevent the denominator from being zero;

[0132] Indicates the first The consistency of local morphology corresponding to each time slot This represents the mean operator;

[0133] Indicates the first The reliability of the receiving station corresponding to each time slot Indicates the first Each receiving station PDW parameters On reliability,

[0134] Indicates the first Each receiving station PDW parameters The residual scale relative to other support stations;

[0135] Indicates the first The reference time constraint score corresponding to each time slot Indicates the first The center position of each time slot Indicates the first Half the width of a time slot, Indicates the first The ideal center position corresponding to each time slot is the time point calculated by the time reference station based on its own clock information.

[0136] Ideal central location Determined recursively based on the time skeleton: When the time reference station has a representative observation in the current time slot, the recursive result is robustly corrected using the TOA parameter value corrected by that representative observation; when the time reference station has a missing pulse, let... That is, compensation is made by using the previous reliable time slot center and the current local PRI estimate.

[0137] After calculating the credibility score for each time slot, the credible time slots can be determined through the following steps:

[0138] Determine whether each time slot is within the mode switching boundary area;

[0139] If it is within the mode switching boundary area, the credibility score of the time slot is compared with the preset first retention threshold. If the credibility score is not less than the first retention threshold, the time slot is determined to be credible.

[0140] If the time slot is outside the mode switching boundary region, the credibility score of the time slot is compared with the preset second retention threshold. If the credibility score is not less than the second retention threshold, the time slot is determined to be credible. The first retention threshold is not greater than the second retention threshold.

[0141] Specifically, different retention thresholds are used depending on whether the candidate slot is in the mode switching boundary region:

[0142] (18)

[0143] like Then determine the first If a time slot is deemed reliable, it will be retained; if Then determine the first The time slots were deemed unreliable and were deleted. To reduce the probability of real pulses being accidentally deleted near mode switching, The first retention threshold, This is a second retention threshold, the value of which can be predetermined based on the verification data. For example, , .

[0144] After obtaining a reliable time slot, the observed values ​​of each PDW parameter of the pulse obtained from each support station are weighted and fused to obtain the PDW parameter values ​​of the fused pulse.

[0145] Specifically, for each reliable time slot, parameter fusion can be achieved by combining weighted median and inlier weighted average:

[0146] (19)

[0147] in, Indicates the PDW parameter type; Indicates receiving station In the Parameters obtained within each time slot The observed value, when When taking TOA, RF and PW respectively, These represent the TOA, RF, and PW observation values ​​for the corresponding observation pulses, respectively. Indicates the first Parameters within each time slot The weighted median; Indicates the first Within the first time slot Each receiving station sets parameters Normalized reliability.

[0148] Next, determine the parameters based on the weighted median. The corresponding set of inward points:

[0149] (20)

[0150] in, Indicates the first Parameters within each time slot The corresponding set of in-point stations; Indicates parameters The preset interior point determination threshold is used to limit the allowable deviation of each station's observation value relative to the weighted median.

[0151] When there are no abnormal observations in a certain time slot, the set of inliers includes all supporting stations of that time slot.

[0152] When parameter Observations with weighted median The absolute deviation between them is no greater than At that time, the receiving station In parameters The observed pulses are determined to be interior points; otherwise, they are determined to be abnormal observations and will not be included in the subsequent weighted average of this parameter.

[0153] Next, a weighted average is performed on the parameter observations in the set of inlier stations:

[0154] (twenty one)

[0155] in, Indicates the first Parameters within each time slot The final fusion value.

[0156] In some embodiments, step S5 above performs pulse missing detection on the PDW fusion sequence, including:

[0157] For adjacent fusion pulses in the PDW fusion sequence, calculate the time interval between the adjacent fusion pulses;

[0158] The number of pulse intervals contained in the time interval is determined based on the local PRI estimate value corresponding to the time slot where the adjacent fused pulse is located.

[0159] When the number of pulse intervals is equal to 1, it is determined that there is no pulse missing between adjacent fusion pulses;

[0160] If the number of pulse intervals is greater than 1, it is determined whether the deviation between the time interval and the theoretical time interval is less than a preset pulse missing determination error threshold. If it is not greater than the pulse missing determination error threshold, it is determined that there is a pulse missing between the adjacent fused pulses, and the number of missing pulses is the number of pulse intervals minus 1. If it is greater than the pulse missing determination error threshold, it is determined that there is no pulse missing between the adjacent fused pulses.

[0161] Specifically, for adjacent fusion pulses and First, calculate the time interval. :

[0162] (twenty two)

[0163] Correspondingly, the number of pulse intervals between these two adjacent fusion pulses It can be calculated using the following formula:

[0164] (twenty three)

[0165] in, This indicates the rounding operation.

[0166] when When, determine that there is no missing pulse between adjacent fusion pulses; when At that time, the estimated number of missing pulses is At this point, further judgment needs to be made based on formula (24):

[0167] (twenty four)

[0168] in, A preset pulse missing detection error threshold is used to limit the time interval. and The theoretical time interval corresponding to each local PRI Allowable deviation between Setting the value too low may miss true missing pulses due to measurement errors, while setting it too high may incorrectly identify normal PRI changes as missing pulses. Those skilled in the art can set the value based on experience.

[0169] When the deviation is not greater than If the time interval is determined to be an integer multiple of the local PRI, it can be determined that there is a missing pulse between the adjacent fusion pulses.

[0170] If a pulse is found to be missing between adjacent fusion pulses, a candidate pulse is inserted between the adjacent fusion pulses, and the TOA parameter value of the candidate pulse is... as follows:

[0171] (25)

[0172] The RF and PW parameter values ​​for the candidate pulses are as follows:

[0173] (26)

[0174] In some embodiments, a missing compensation flag can be set for the inserted candidate pulse. Setting the fusion pulse directly obtained from a trusted time slot .

[0175] The fused pulses after parameter fusion and finite missing interpolation are sorted in ascending order of TOA parameter values, and PRI is recalculated.

[0176] (27)

[0177] Then the first The network input vector of a fused pulse is represented as:

[0178] (28)

[0179] in, Let represent the PDW eigenvector of the t-th fusion pulse; This indicates the credibility score of the time slot where the fused pulse is located. This indicates the cross-site support of the time slot containing the fused pulse. This indicates a missing compensation marker.

[0180] Therefore, the quality, support, and missing information generated by front-end integration can directly participate in back-end time-series state updates.

[0181] In some embodiments, the working mode recognition network in step S6 is configured as follows:

[0182] Receive the final PDW fusion sequence and use it as the input sequence;

[0183] For each fusion pulse in the input sequence, dynamic geometric features are generated based on the RF parameter value, PW parameter value and PRI parameter value of the fusion pulse, and interval anomaly indication features are generated based on the PRI difference information, PRI relative change information, local PRI discreteness information and missing compensation mark corresponding to the fusion pulse.

[0184] Based on the geometric features, predict the boundary probability of each fusion pulse belonging to the working mode switching boundary, and based on the geometric features and the interval anomaly indication features, predict the quality probability and missing threshold of each fusion pulse.

[0185] Based on the boundary probability, the quality probability, and the missing threshold of each fused pulse, the hidden state of the fused pulse is selectively updated.

[0186] The hidden states updated by each fusion pulse are aggregated at the sequence level to obtain the aggregated state representation of the input sequence, and the working mode category corresponding to the input sequence is determined based on the aggregated state representation.

[0187] In the working mode recognition stage, the final PDW fusion sequence is input into the working mode recognition network to obtain the working mode recognition result of the radar radiation source.

[0188] Continue to refer to Figure 1 The network's encoding module extracts the dynamic geometric features of the input sequence. :

[0189] (29)

[0190] in, Indicates the first The first-order difference between the TOA parameter vector of the fusion pulse and the TOA parameter vector of the previous fusion pulse; express The second difference; and They respectively represent the first The mean and standard deviation of the TOA parameter vector within a preset time window centered on a fusion pulse; Indicates the first The intensity of local parameter changes at each fusion pulse.

[0191] The network's encoding module encodes dynamic geometric features:

[0192] (30)

[0193] in, The feature encoding function can be composed of one or more fully connected layers and nonlinear activation functions, and its parameters are determined during network training. Representing dynamic geometric features The encoded features obtained after processing by the aforementioned feature encoding function.

[0194] The encoding module also constructs interval anomaly alert features based on PRI difference, relative PRI change, local PRI standard deviation, and missing compensation marker. :

[0195] (31)

[0196] in, This represents the first difference of PRI. Indicates the first The standard deviation of PRI within a preset time window centered on a single pulse. Indicates the first Missing compensation markers for each fusion pulse.

[0197] Subsequently, the prediction module of the working pattern recognition network calculates the boundary probability based on the encoded features and the interval anomaly cue features. Quality probability and missing threshold :

[0198] (32)

[0199] in, and These are the boundary prediction weight matrix and the boundary prediction bias term, respectively; and These are the quality prediction weight matrix and the quality prediction bias term, respectively. Indicates will and Perform feature splicing; and These are the missing gate prediction weight matrix and the missing gate prediction bias term, respectively. This is the sigmoid activation function. The parameters of the various weight matrices and bias terms are determined during network training.

[0200] The network's state memory module selectively updates the hidden state of the fused pulse based on the boundary probability, the quality probability, and the missing threshold of each fused pulse:

[0201] (33)

[0202] in, This is the final hidden state after quality probability adjustment. This represents the intermediate hidden state after joint adjustment by boundary probability and missing gate. This is a state of missing compensation. This is a normal candidate state. To reset the candidate state; and It consists of two GRU units used to process sequence data.

[0203] Therefore, near the mode switching boundary, the higher Reduce the continuation ratio of old mode states and increase the weight of reset candidate states; at long intervals or suspected leaky pulses, a higher weight is applied. Suppressing unreliable current observation writes makes the state rely more on historical memory or missing state compensation; at low-reliability time slots, lower This further weakens the influence of the current candidate state on the final hidden state. This gating mechanism couples slot-level confidence with temporal state updates, rather than simply linking them together between fusion and recognition.

[0204] During the classification phase, the information geometric prototype classifier maps state representations to a probabilistic simplex and calculates the Fisher-Rao distance between the sample distribution and the class prototype distribution.

[0205] The information geometric prototype classifier in this embodiment is for lengths of... State aggregation is performed on the robust state sequence:

[0206] (34)

[0207] in, This indicates the number of time steps contained in the input sequence. Indicates the first The robust state vector corresponding to each time step This represents the window-level state representation obtained by averaging and aggregating the robust state vectors over all time steps.

[0208] The information geometry prototype classifier maps aggregate states to a probabilistic simplex, obtaining the sample probability distribution:

[0209] (35)

[0210] in, For probability embedding mapping matrix, These are the corresponding bias vectors, both of which are network parameters determined during model training; Used to represent window-level state The mapping is to an unnormalized score vector in the probability embedding space; This represents the normalization exponential function, used to convert the unnormalized score into a sample probability distribution where each component is non-negative and sums to 1. .

[0211] No. Prototype probability distribution corresponding to class working mode for:

[0212] (36)

[0213] prototype parameters It can be set as a trainable parameter, and initialized with the average embedding of the k-th class samples in the training set; in another embodiment, it can also be updated after each round of training based on the robust mean of the embeddings of the k-th class samples. .

[0214] The Bhattacharyya coefficient between the sample probability distribution and the class prototype probability distribution is:

[0215] (37)

[0216] Corresponding Fisher-Rao distance for:

[0217] (38)

[0218] No. Category score of work mode for:

[0219] (39)

[0220] in, The preset temperature adjustment parameters, These are trainable network parameters used to perform an overall offset adjustment on the score of the k-th class.

[0221] Applying amplitude limiting to the Bhattacharyya coefficients before calculating the inverse cosine avoids floating-point errors causing the input to exceed the effective domain; Equation (39) retains only one scale parameter. This is to avoid the scaling factor repeating the temperature parameter.

[0222] Final working mode probability for:

[0223] (40)

[0224] The final identified category is:

[0225] (41)

[0226] Based on the above embodiments of the present invention, it can be seen that the radar radiation source operating mode identification method of the present invention has at least the following advantages:

[0227] (1) This invention can establish relative time consistency under multi-station asynchronous observation conditions by selecting time reference stations, global time deviation correction, clock ratio error correction, local drift compensation and boundary segmentation secondary correction, thereby reducing cross-station miscorrelation caused by clock inconsistency.

[0228] (2) This invention transforms the complex cross-station point-by-point matching problem into the problem of determining the attribution between each observation pulse and the time slot by adaptively constructing time slots. It can naturally accommodate different numbers of support stations such as single-station support, dual-station support and multi-station support, and is more suitable for distributed observation scenarios where missing pulses and false pulses coexist.

[0229] (3) By performing multi-factor credibility scoring on the time slot, this invention avoids the simple judgment that three-station support is real and single-station support is false, and can rescue weak support real pulses, while suppressing three-station related false pulses and abnormal points near the boundary.

[0230] (4) This invention improves the integrity of fusion while controlling the purity of output by robust parameter fusion, local morphological compensation and output sequence purification, so as to provide stable input for subsequent working mode recognition.

[0231] (5) The working mode recognition network of the present invention sets up dynamic geometric coding, missing perception memory, quality anomaly gating and information geometric prototype classification mechanisms for parameter errors, missing pulses, false pulses and mode boundaries in PDW fusion sequences, respectively, thereby improving the recognition robustness under non-ideal conditions.

[0232] (6) The present invention adopts the same field unification, PRI calculation, time sequence consistency correction, time slot construction, credibility assessment, parameter fusion and network inference process in the training configuration stage and the actual recognition stage, so that the working pattern recognition network can maintain stable recognition capability under the conditions of missing pulses, false pulses, parameter measurement errors and inter-station time inconsistency.

[0233] Figure 4 This is a schematic diagram of an electronic device illustrated in this specification according to an exemplary embodiment. Please refer to... Figure 4 At the hardware level, the device includes a processor 410, an internal bus 420, a network interface 430, memory 440, a hardware acceleration device 450, and non-volatile memory 460, and may also include other hardware required for its functions. One or more embodiments of the present invention can be implemented in software, for example, the processor 410 reads the corresponding computer program from the non-volatile memory 460 into the memory 440 and then runs it. Of course, in addition to software implementation, one or more embodiments of the present invention do not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. That is to say, the execution subject of the above processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0234] Figure 5 This is a structural block diagram of a radar operating mode recognition device based on multi-station asynchronous PDW time slot trusted fusion and quality-aware timing modeling, as illustrated in an exemplary embodiment of the present invention. The radar operating mode recognition device can be applied to applications such as... Figure 4 The electronic device shown implements the technical solution of the present invention. The radar operating mode recognition device includes: an observation data acquisition unit 510, a time slot construction unit 520, a time slot evaluation unit 530, a multi-station pulse fusion unit 540, a pulse missing detection unit 550, and an operating mode recognition unit 560, wherein:

[0235] The observation data acquisition unit 510 is used to acquire PDW observation data from multiple receiving stations observing the same radar radiation source. The PDW observation data includes the observation values ​​of set PDW parameters, which include TOA parameters, RF parameters, and PW parameters.

[0236] The time slot construction unit 520 is used to divide the time slots along the time axis according to the TOA parameter observation values ​​of all observation pulses of all receiving stations, and to determine the local PRI estimate value corresponding to each time slot according to the statistical results of the difference between the center positions of adjacent time slots within the preset neighborhood time window of each time slot; for any time slot, if the TOA parameter observation value of at least one observation pulse of a certain receiving station falls within the time boundary of the time slot, then the receiving station is recorded as the support station of the time slot.

[0237] The time slot evaluation unit 530 is used to calculate the credibility score of each time slot based on the number of support stations corresponding to each time slot, the consistency of the observation values ​​of each support station on each PDW parameter, and the matching degree of each time slot with the observation values ​​of each time slot within its preset neighborhood time window on each PDW parameter, and to determine the credible time slot based on the credibility score.

[0238] The multi-station pulse fusion unit 540 is used to perform weighted fusion of the observed values ​​of each PDW parameter of the pulse obtained from each support station for each reliable time slot, to obtain the PDW parameter values ​​of the fused pulse, and to sort them in ascending order according to the TOA parameter values ​​of all fused pulses to obtain the PDW fusion sequence.

[0239] The pulse missing detection unit 550 is used to perform pulse missing detection on the PDW fusion sequence, and insert a candidate pulse between the adjacent fusion pulses when there is a pulse missing between adjacent fusion pulses to obtain the final PDW fusion sequence.

[0240] The working mode recognition unit 560 is used to input the final PDW fusion sequence into the working mode recognition network to obtain the working mode recognition result of the radar radiation source.

[0241] In some embodiments, the time slot construction unit 520 is configured to select one receiving station from the plurality of receiving stations as a time reference station, and use the time reference station as a reference to correct the time deviation of the observed values ​​of the TOA parameters of the other receiving stations; determine the slot center position of each time slot along the time axis based on the corrected TOA parameter observed values, and determine the local PRI estimate value corresponding to each time slot based on the statistical results of the difference between the center positions of adjacent time slots within a preset neighborhood time window of each time slot; and determine the slot width of each time slot based on the local PRI estimate value corresponding to each time slot.

[0242] In some embodiments, the time slot construction unit 520 is further configured to: calculate fingerprint information of each initial time slot based on the PDW observation data of the pulses observed by each support station; perform mode switching detection based on the fingerprint information of each initial time slot and the fingerprint information within a preset neighborhood time window of each initial time slot; when a mode switching is detected in the radar radiation source during the observation phase, determine the mode switching boundary and obtain the observation pulse subsequence outside the mode switching boundary region, construct a new time slot based on the observation pulse subsequence; and use the new time slot and the original time slot within the mode switching boundary region together as the final time slot.

[0243] In some embodiments, the time slot evaluation unit 530 is used to determine whether each time slot is within a mode switching boundary region; if it is within a mode switching boundary region, the reliability score of the time slot is compared with a preset first retention threshold, and if the reliability score is not less than the first retention threshold, the time slot is determined to be reliable; if it is outside a mode switching boundary region, the reliability score of the time slot is compared with a preset second retention threshold, and if the reliability score is not less than the second retention threshold, the time slot is determined to be reliable; wherein, the first retention threshold is not greater than the second retention threshold.

[0244] In some embodiments, the pulse missing detection unit 550 is configured to: calculate the time interval between adjacent fused pulses in the PDW fusion sequence; determine the number of pulse intervals included in the time interval based on the local PRI estimate corresponding to the time slot where the adjacent fused pulses are located; determine that there is no pulse missing between the adjacent fused pulses if the number of pulse intervals is equal to 1; determine whether the deviation between the time interval and the theoretical time interval is less than a preset pulse missing determination error threshold if the deviation is not greater than the pulse missing determination error threshold; determine that there is a pulse missing between the adjacent fused pulses if the deviation is less than the pulse missing determination error threshold, and the number of missing pulses is the number of pulse intervals minus 1; determine that there is no pulse missing between the adjacent fused pulses if the deviation is greater than the pulse missing determination error threshold; wherein, the theoretical time interval is obtained by multiplying the number of pulse intervals by the local PRI estimate.

[0245] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of the present invention according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0246] Accordingly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in any of the above embodiments.

[0247] Accordingly, embodiments of the present invention also provide a computer program product configured to perform the methods described in any of the above embodiments.

[0248] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer, which can take the form of a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email sending and receiving device, game console, tablet computer, wearable device, or any combination of these devices.

[0249] In a typical configuration, a computer includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0250] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0251] Computer-readable media, including both permanent and non-permanent, removable and non-removable media, can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random-access memory (SRAM), dynamic random-access memory (DRAM), other types of random-access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, disk storage, quantum memory, graphene-based storage media or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0252] While this specification contains numerous specific implementation details, these should not be construed as limiting the scope of any invention or the scope of the claims, but rather are primarily intended to describe features of specific embodiments of a particular invention. Certain features described in the various embodiments herein may also be implemented in combination in a single embodiment. Conversely, various features described in a single embodiment may also be implemented separately in various embodiments or in any suitable sub-combination. Furthermore, while features may function in certain combinations as described above and even initially claimed in this way, one or more features from a claimed combination may be removed from that combination in some cases, and a claimed combination may refer to a sub-combination or a variation thereof.

[0253] Similarly, although the operations are depicted in a specific order in the accompanying drawings, this should not be construed as requiring these operations to be performed in the specific order shown or sequentially, or requiring all illustrated operations to be performed to achieve the desired result. In some cases, multitasking and parallel processing may be advantageous. Furthermore, the separation of various system modules and components in the above embodiments should not be construed as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.

[0254] Thus, specific embodiments of the subject matter have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions recited in the claims may be performed in a different order and still achieve the desired result. Furthermore, the processes depicted in the drawings are not necessarily shown in a specific order or sequence to achieve the desired result. In some implementations, multitasking and parallel processing may be advantageous.

[0255] It should be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0256] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A radar operating mode recognition method based on multi-station asynchronous PDW time slot reliable fusion and quality-aware temporal modeling, characterized in that, Includes the following steps: Step S1: Acquire PDW observation data from multiple receiving stations observing the same radar radiation source. The PDW observation data includes the observed values ​​of TOA parameters, RF parameters, and PW parameters. Step S2: Based on the TOA parameter observations of all receiving stations, divide the time axis into several time slots, and determine the local PRI estimate and support station corresponding to each time slot based on the statistical results of the difference between the center positions of adjacent time slots within the preset neighborhood time window of each time slot. Step S3: Calculate the credibility score of each time slot based on the number of support stations corresponding to each time slot, the consistency of the observations of each support station on each PDW parameter, and the matching degree of each time slot with the observations of each time slot within its preset neighborhood time window on each PDW parameter, and determine the credible time slot based on the credibility score. Step S4: For each reliable time slot, the observed values ​​of each PDW parameter of the observation pulse under its supporting station are weighted and fused to obtain the PDW parameter values ​​of the fused pulse, and then sorted in ascending order according to the TOA parameter values ​​of all fused pulses to obtain the PDW fusion sequence. Step S5: Perform pulse missing detection on the PDW fusion sequence, and if there is a pulse missing between adjacent fusion pulses, insert a candidate pulse between the adjacent fusion pulses to obtain the final PDW fusion sequence. Step S6: Input the final PDW fusion sequence into the working mode recognition network to obtain the working mode recognition result of the radar radiation source.

2. The method according to claim 1, characterized in that, Step S2 includes: Select one receiving station from the plurality of receiving stations as a time reference station, and use the time reference station as a reference to correct the time deviation of the observed values ​​of the TOA parameters of the other receiving stations. Based on the corrected TOA parameter observations, the center position of each time slot is determined along the time axis, and the local PRI estimate value corresponding to each time slot is determined based on the statistical results of the difference between the center positions of adjacent time slots within the preset neighborhood time window of each time slot. The slot width of each time slot is determined based on the local PRI estimate corresponding to each time slot.

3. The method according to claim 2, characterized in that: The time deviation correction includes at least one of global time deviation correction, clock ratio error correction, and local drift compensation; The formula for calculating the local PRI estimate is as follows: ; The formula for calculating the slot width is: ; in, Indicates the first Local PRI estimates for each time slot This represents the median operator. Indicates the first The observed values ​​of the TOA parameters of each observation pulse after time deviation correction. , , Indicates the total number of time slots. Indicates the first The preset neighborhood time window corresponding to each time slot Indicates the first The half width of the slot corresponding to each time slot This indicates the amplitude limiting operator. This is the preset slot width ratio coefficient. , These represent the preset upper and lower limits of the slot width, respectively.

4. The method according to claim 1, characterized in that, Step S2 further includes: Based on the PDW observation data of the observation pulses under its support station for each time slot, calculate the fingerprint information of each initial time slot; Based on the fingerprint information of each initial time slot and the fingerprint information within the neighboring preset time window of each initial time slot, mode switching detection is performed. When a mode switch is detected in the radar radiation source during the observation phase, the mode switch boundary is determined and the observation pulse sub-sequence outside the mode switch boundary is obtained. A new time slot is constructed based on the observation pulse sub-sequence. The new time slot and the time slot inside the mode switching boundary are used together as the final time slot.

5. The method according to claim 4, characterized in that, The fingerprint information for each initial time slot includes at least one of the following: The difference between the current time slot and its previous time slot regarding the center position of the time slot; The mean RF parameter value corresponding to the current time slot, wherein the mean RF parameter value refers to the mean RF parameter value of all observed pulses in the support station set of the current time slot; The mean PW parameter value corresponding to the current time slot, wherein the mean PW parameter value refers to the average value of the PW parameter values ​​of all observed pulses in the support station set of the current time slot; The difference in PRI parameter values ​​between the current time slot and its previous time slot, and the standard deviation of PRI within the preset neighborhood time window of the current time slot.

6. The method according to claim 1, characterized in that, The formula for calculating the credibility score of each time slot in step S3 is as follows: ； in, Indicates the first Credibility score for each time slot , This indicates the preset weighting coefficients; Indicates the first Cross-site support corresponding to each time slot; Indicates the first The parameters within each time slot are consistent. Indicates the first Local morphological consistency corresponding to each time slot; Indicates the first The reliability of the receiving station corresponding to each time slot; Indicates the first The reference time constraint score corresponding to each time slot.

7. The method according to claim 4, characterized in that, Step S3 determines the reliable time slot based on the reliability score: Determine whether each time slot is within the mode switching boundary area; If it is within the mode switching boundary area, the credibility score of the time slot is compared with the preset first retention threshold. If the credibility score is not less than the first retention threshold, the time slot is determined to be credible. If the time slot is outside the mode switching boundary area, the credibility score of the time slot is compared with the preset second retention threshold. If the credibility score is not less than the second retention threshold, the time slot is determined to be credible. The first retention threshold is no greater than the second retention threshold.

8. The method according to claim 1, characterized in that, Step S5 performs pulse loss detection on the PDW fusion sequence, including: For adjacent fusion pulses in the PDW fusion sequence, calculate the time interval between the adjacent fusion pulses; The number of pulse intervals contained in the time interval is determined based on the local PRI estimate value corresponding to the time slot where the adjacent fused pulse is located. When the number of pulse intervals is equal to 1, it is determined that there is no pulse missing between adjacent fusion pulses; If the number of pulse intervals is greater than 1, it is determined whether the deviation between the time interval and the theoretical time interval is less than a preset pulse missing determination error threshold. If it is not greater than the pulse missing determination error threshold, it is determined that there is a pulse missing between the adjacent fused pulses, and the number of missing pulses is the number of pulse intervals minus 1. If it is greater than the pulse missing determination error threshold, it is determined that there is no pulse missing between the adjacent fused pulses. The theoretical time interval is obtained by multiplying the number of pulse intervals by the local PRI estimate.

9. The method according to claim 1, characterized in that, The working mode recognition network in step S6 is configured as follows: Receive the final PDW fusion sequence and use it as the input sequence; For each fusion pulse in the input sequence, dynamic geometric features are generated based on the RF parameter value, PW parameter value and PRI parameter value of the fusion pulse, and interval anomaly indication features are generated based on the PRI difference information, PRI relative change information, local PRI discreteness information and missing compensation mark corresponding to the fusion pulse. Based on the geometric features, predict the boundary probability of each fusion pulse belonging to the working mode switching boundary, and based on the geometric features and the interval anomaly indication features, predict the quality probability and missing threshold of each fusion pulse. Based on the boundary probability, the quality probability, and the missing threshold of each fused pulse, the hidden state of the fused pulse is selectively updated. The hidden states updated by each fusion pulse are aggregated at the sequence level to obtain the aggregated state representation of the input sequence, and the working mode category corresponding to the input sequence is determined based on the aggregated state representation.

10. A radar radiation source operating mode identification device based on time slot quality assessment, characterized in that, include: The observation data acquisition unit is used to acquire PDW observation data from multiple receiving stations observing the same radar radiation source. The PDW observation data includes the observed values ​​of TOA parameters, RF parameters, and PW parameters. The time slot construction unit is used to divide the time slots along the time axis according to the TOA parameter observations of all receiving stations, and to determine the local PRI estimate and support station corresponding to each time slot based on the statistical results of the difference between the center positions of adjacent time slots within the preset neighborhood time window of each time slot. The time slot evaluation unit is used to calculate the credibility score of each time slot based on the number of support stations corresponding to each time slot, the consistency of the observations of each support station on each PDW parameter, and the matching degree of each time slot with the observations of each time slot within its preset neighborhood time window on each PDW parameter, and to determine the credible time slot based on the credibility score. The multi-station pulse fusion unit is used to perform weighted fusion of the observed values ​​of each PDW parameter of the observed pulse under each reliable time slot, to obtain the PDW parameter values ​​of the fused pulse, and to sort them in ascending order according to the TOA parameter values ​​of all fused pulses to obtain the PDW fusion sequence. A pulse missing detection unit is used to detect pulse missing in the PDW fusion sequence, and insert a candidate pulse between adjacent fusion pulses when there is a pulse missing between adjacent fusion pulses to obtain the final PDW fusion sequence. The working mode recognition unit is used to input the final PDW fusion sequence into the working mode recognition network to obtain the working mode recognition result of the radar radiation source.

11. An electronic device, characterized in that, include: processor; as well as A computer-readable storage medium storing computer program instructions that, when executed by the processor, cause the processor to perform the method as described in any one of claims 1 to 9.

Images