A multi-platform cooperative instantaneous signal passive positioning method and system

Abstract

The application discloses a kind of multi-platform coordination instantaneous signal passive positioning method and system, it is related to data processing technical field.The method includes: by multiple signal receiving platforms distributed in different spatial positions, multiple electromagnetic signals are synchronously intercepted and respectively corresponding to extract pulse description word;Based on pulse description word, construct the pulse matching model based on geometric constraint, output obtains the matching degree evaluation value between any two pulses;For any pulse intercepted by any signal receiving platform, according to the matching degree evaluation value, search and aggregate in the pulse list of all other platforms to form candidate pulse group;Based on the signal arrival time of all pulses in candidate pulse group and the spatial position information of corresponding platform, the spatial coordinates of radiation source are solved through cooperative positioning, and the positioning result is output.The application effectively improves the accuracy and stability of radiation source positioning.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a multi-platform collaborative instantaneous signal passive positioning method and system. Background Technology

[0002] With the increasing complexity of electronic warfare environments, target radiation sources often employ characteristics such as low intercept and frequency agility, making it difficult to continuously and stably capture the radio frequency signals they intentionally or unintentionally emit using traditional methods. In existing technologies, multi-platform cooperative passive localization mainly relies on accurately correlating the pulses intercepted by each platform, i.e., determining whether the pulses received by different platforms originate from the same radiation source.

[0003] However, in dense and interwoven electromagnetic environments, pulse loss, parameter measurement errors, and the asynchronicity of pulse lists on different platforms make traditional pulse matching methods prone to generating erroneous pulse associations. This leads to a large number of false targets or association failures in subsequent positioning calculations, making it difficult to guarantee the accuracy and stability of radiation source positioning. Summary of the Invention

[0004] This invention provides a multi-platform collaborative instantaneous signal passive positioning method and system, aiming to solve the technical problem that existing technologies cannot guarantee the accuracy and stability of radiation source positioning.

[0005] In view of the above problems, the present invention provides a multi-platform collaborative instantaneous signal passive positioning method and system.

[0006] In a first aspect, the present invention provides a multi-platform collaborative instantaneous signal passive positioning method, comprising: Multiple signal receiving platforms distributed in different spatial locations synchronously intercept multiple electromagnetic signals and extract pulse descriptors from each signal. Based on the pulse descriptor, a pulse matching model based on geometric constraints is constructed, and the matching degree evaluation value between any two pulses is obtained. For any pulse intercepted by any of the signal receiving platforms, a candidate pulse group is formed by searching and aggregating the pulse lists of all other platforms based on the matching degree evaluation value. Based on the signal arrival time of all pulses in the candidate pulse group and the spatial location information of the corresponding platform, the spatial coordinates of the radiation source are calculated through collaborative positioning, and the positioning result is output.

[0007] Secondly, the present invention provides a multi-platform collaborative instantaneous signal passive positioning system, comprising: The signal interception module is used to simultaneously intercept multiple electromagnetic signals from multiple signal receiving platforms distributed in different spatial locations and extract pulse descriptors for each signal. The pulse matching modeling module is used to construct a pulse matching model based on geometric constraints based on the pulse descriptor and output the matching degree evaluation value between any two pulses. The candidate pulse aggregation module is used to search and aggregate any pulse intercepted by any of the signal receiving platforms to form a candidate pulse group based on the matching degree evaluation value in the pulse lists of all other platforms. The positioning calculation output module is used to calculate the spatial coordinates of the radiation source based on the signal arrival time of all pulses in the candidate pulse group and the spatial location information of the corresponding platform through collaborative positioning, and output the positioning result.

[0008] One or more technical solutions provided in this invention have at least the following technical effects or advantages: This invention provides a multi-platform collaborative passive localization method and system for instantaneous signals, achieving high-precision and high-reliability passive localization of instantaneous radio frequency signals. First, multi-source synchronous interception and PDW extraction provide a unified and accurate signal feature basis for subsequent processing. Second, by constructing a geometric constraint matching model, platform spatial relationships are integrated into pulse similarity evaluation, fundamentally improving the ability to identify signals from the same source in dense, interleaved pulse streams and effectively suppressing false associations. Furthermore, through cross-platform dynamic search and pulse aggregation, candidate pulse sets corresponding to the same potential radiation source can be flexibly and robustly formed, overcoming the challenges posed by pulse loss and asynchronicity. Finally, collaborative localization calculation based on a high-quality set of signals from the same source improves the convergence and accuracy of algorithms such as time-difference localization. This invention forms a complete technical closed loop from signal processing and intelligent association to precise calculation, enhancing the ability to rapidly locate and generate situational awareness for low-interception, instantaneous signals in complex electromagnetic environments, and improving the reliability and accuracy of passive localization of instantaneous signals in electronic warfare scenarios. Attached Figure Description

[0009] Figure 1 A flowchart illustrating a multi-platform collaborative instantaneous signal passive positioning method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a multi-platform collaborative instantaneous signal passive positioning system provided in an embodiment of the present invention; The components represented by each number in the attached diagram are explained below: Signal interception module 11, pulse matching modeling module 12, candidate pulse aggregation module 13, and positioning solution output module 14. Detailed Implementation

[0010] This invention provides a multi-platform collaborative instantaneous signal passive positioning method and system, which addresses the technical problem that existing technologies cannot guarantee the accuracy and stability of radiation source positioning.

[0011] Example 1, as Figure 1 As shown, the present invention provides a multi-platform collaborative instantaneous signal passive positioning method, the method comprising: S100: Multiple signal receiving platforms distributed in different spatial locations synchronously intercept multiple electromagnetic signals and extract pulse descriptors for each signal.

[0012] In this embodiment of the invention, multiple signal receiving platforms distributed in different spatial locations simultaneously intercept multiple electromagnetic signals and extract pulse descriptors for each signal. In the dense electromagnetic environment of electronic warfare, the transient radio frequency signals emitted by non-cooperative enemy targets are characterized by short duration and susceptibility to interference. A single signal receiving platform cannot completely capture the signal characteristics, and the original electromagnetic signals are mixed with noise and clutter, which will lead to distortion of signal parameter measurements if used directly. To provide accurate and reliable basic data for subsequent cross-platform pulse matching, homogeneous aggregation, and radiation source localization, it is necessary to simultaneously intercept electromagnetic signals through multiple spatially distributed signal receiving platforms, extract standardized pulse descriptors that have undergone rationality verification, and eliminate invalid and erroneous measurement parameters to ensure the accuracy of subsequent localization processes.

[0013] Step S100 in the method provided in this embodiment of the invention includes: The step of extracting the pulse descriptor includes: The raw electromagnetic signals collected by each signal receiving platform are instantaneously detected based on a preset amplitude threshold to determine the start and end times of the pulses and record the signal arrival time. For the intercepted pulse, the direction of arrival of the signal is calculated by using the fixed phase difference between the signals received by the multi-channel receiving array of the corresponding platform and the phase interferometry method. For the same pulse, an instantaneous frequency measurement circuit is used to measure the corresponding carrier frequency; The extracted signal arrival direction, signal arrival time, and carrier frequency are each checked for reasonableness based on a preset physical possibility range threshold. The parameter combinations that pass the check are then encapsulated into a descriptor for the current pulse.

[0014] First, the raw electromagnetic signals acquired by each signal receiving platform are instantaneously detected based on a preset amplitude threshold to determine the start and end times of the pulse and record the signal arrival time. Instantaneous detection is a signal acquisition method that determines the start and end times of the pulse in real time based on changes in signal amplitude. The preset amplitude threshold refers to a pre-set voltage threshold used to distinguish valid pulse signals from environmental noise. The signal arrival time is the timestamp corresponding to the pulse start time and is a time-domain parameter for subsequent collaborative positioning. The signal receiving platform continuously compares the amplitude of the acquired raw electromagnetic signals with the preset amplitude threshold. When the signal amplitude continuously exceeds the preset amplitude threshold, it is determined as the pulse start and a timestamp is marked; when the signal amplitude falls back below the preset amplitude threshold, it is determined as the pulse end, completing the acquisition of a single complete pulse, and the pulse start time is recorded as the signal arrival time.

[0015] For example, the preset amplitude threshold of the signal receiving platform is set to 0.5V. The platform collects the original electromagnetic signal in real time. When the signal amplitude rises to 0.6V and continues to exceed the threshold at a certain moment, the pulse is determined to start and the signal arrival time is recorded as 10:00:00.123. When the signal amplitude falls back to 0.4V, the pulse is determined to end, thus completing the capture and time recording of the instantaneous pulse.

[0016] Secondly, for the intercepted pulse, the signal arrival direction is calculated using the fixed phase difference between the signals received by the multi-channel receiving array of the corresponding platform, employing the phase interferometry method. A multi-channel receiving array refers to a receiving device composed of multiple array elements mounted on the signal receiving platform, capable of synchronously receiving the same pulse signal. The phase interferometry method calculates the signal incident angle using the fixed phase difference between the signals received by multiple array elements. The formula is: Signal arrival direction = arcsin(wavelength × phase difference / (2π × array element spacing)). The signal arrival direction refers to the angle at which the pulse signal is incident on the receiving platform, and is a spatial parameter of the pulse descriptor. For the intercepted pulse, the fixed spacing between the array elements in the platform's multi-channel receiving array and the phase difference between the signals received by different array elements are obtained. These are then combined with the signal wavelength and substituted into the phase interferometry formula to calculate the signal arrival direction.

[0017] For example, the array element spacing of the multi-channel signal receiving platform is 0.15m, the wavelength of the current pulse signal is 0.3m, and the measured phase difference between the array elements is π / 2. Substituting into the formula, the signal arrival direction is calculated as: signal arrival direction = arcsin(0.3×π / 2 / (2π×0.15)) = arcsin(0.5) = 30°, thus the arrival direction of the pulse signal is 30°.

[0018] Secondly, for the same pulse, an instantaneous frequency measurement circuit is used to measure the corresponding carrier frequency. The instantaneous frequency measurement circuit refers to the hardware circuit on the signal receiving platform that can measure the carrier frequency of a pulse signal in real time. The carrier frequency refers to the carrier frequency parameter of the pulse electromagnetic signal, which is a frequency domain parameter of the pulse descriptor. For a single intercepted pulse, the instantaneous frequency measurement circuit of the signal receiving platform measures the carrier frequency corresponding to the pulse in real time, thus extracting the frequency domain parameter. For example, for the intercepted pulse mentioned above, the instantaneous frequency measurement circuit determines that the carrier frequency of the pulse is 5 GHz.

[0019] Furthermore, the extracted signal arrival direction, signal arrival time, and carrier frequency are each subjected to a validity check based on a preset physical possibility range threshold. The parameter combinations that pass the check are then encapsulated into the current pulse descriptor. The physical possibility range threshold is a pre-defined range of valid parameters used to filter out obviously erroneous or invalid pulse measurement parameters. The pulse descriptor is a standardized data unit formed by encapsulating the validated parameters such as signal arrival direction, signal arrival time, and carrier frequency. The extracted signal arrival direction, signal arrival time, and carrier frequency are each subjected to a validity check based on the preset physical possibility range threshold. Only parameters that pass all threshold checks are retained, and the qualified parameter combinations are encapsulated into the current pulse descriptor.

[0020] For example, the signal arrival direction threshold is set to [-60°, 60°], the carrier frequency threshold is set to [1GHz, 18GHz], and the signal arrival time threshold is limited to the current detection time window and the pulse interval is non-negative. The signal arrival direction of 30°, the signal arrival time of 10:00:00.123, and the carrier frequency of 5GHz extracted in this case all meet the corresponding threshold requirements. After verification, they are encapsulated as the description word of the pulse.

[0021] In this embodiment of the invention, by simultaneously intercepting instantaneous electromagnetic signals across multiple platforms, the precise extraction and rationality verification of pulse time-domain, spatial-domain, and frequency-domain parameters are completed. Erroneous measurement parameters caused by noise interference are eliminated, and standardized pulse descriptors are formed. This provides clean and reliable basic data for subsequent cross-platform pulse matching and homogeneous pulse aggregation, avoids pulse correlation errors caused by parameter distortion, and improves the accuracy of basic data for subsequent radiation source positioning.

[0022] S200: Based on the pulse description word, construct a pulse matching model based on geometric constraints, and output the matching degree evaluation value between any two pulses.

[0023] In this embodiment of the invention, a pulse matching model based on geometric constraints is constructed based on the pulse descriptor, outputting a matching degree evaluation value between any two pulses. In the dense electromagnetic environment of electronic warfare, the number of pulses intercepted by multiple platforms is enormous, with overlapping and parameter measurement errors. Relying solely on single pulse parameters cannot accurately determine whether pulses from different platforms originate from the same radiation source. Furthermore, traditional matching methods lack geometric constraints such as azimuth, time difference, and spatial consistency, making them susceptible to noise interference and mismatches. To achieve accurate same-source determination of cross-platform pulses, a dedicated pulse matching model needs to be constructed based on the pulse descriptor extracted by S100, combined with azimuth constraints, time difference constraints, and spatial consistency constraints. Stable discrimination capability is obtained through supervised training, outputting a quantified pulse matching degree evaluation value, providing a reliable basis for subsequent same-source pulse aggregation.

[0024] Step S200 in the method provided in this embodiment of the invention includes: The geometric constraints include at least orientation constraints, time difference constraints, and spatial consistency constraints.

[0025] The steps for constructing the pulse matching model include: Historical electromagnetic signal data synchronously intercepted by the multiple signal receiving platforms under the same detection scenario are collected, and the corresponding set of historical pulse descriptors is extracted. Based on prior information about the true location of radiation sources, pulses in the historical pulse descriptor set are manually sorted and labeled to identify pulse pairs originating from the same radiation source, forming positive sample pairs, and pulses from different sources are randomly combined to form negative sample pairs, thus forming a training sample set. A pulse matching model based on a Siamese network architecture is constructed, wherein the Siamese network architecture includes two feature extraction sub-networks with identical structures and shared weights. Each sub-network is used to process the pulse descriptor of a pulse and extract high-order features. The matching degree evaluation value between the two pulses is calculated and output through a fusion discriminant layer based on the consistency between azimuth cosine similarity and theoretical time difference. Using the training sample set as input and the homologous labels of the sample pairs as supervision signals, the impulse matching model is subjected to end-to-end supervised training. When the matching accuracy of the model's output on the independent validation set for distinguishing positive and negative sample pairs is higher than a preset accuracy threshold, training stops, and the construction of the pulse matching model is completed.

[0026] First, historical electromagnetic signal data synchronously intercepted by multiple signal receiving platforms under the same detection scenario is collected, and a historical pulse descriptor set is extracted accordingly. Historical electromagnetic signal data refers to the raw electromagnetic signal data synchronously intercepted and stored by multiple signal receiving platforms under the same detection scenario. The same detection scenario refers to a detection condition where conditions such as radiation source distribution, platform location, and electromagnetic environment remain consistent. The historical pulse descriptor set is a collection of pulse descriptors extracted from historical electromagnetic signal data through step S100. A detection scenario consistent with the actual positioning task is selected, and multiple spatially distributed signal receiving platforms are invoked to synchronously intercept electromagnetic signals. Following the pulse extraction and verification process in S100, pulse descriptors for all historical signals are extracted and aggregated to form a unified historical pulse descriptor set.

[0027] For example, an electronic warfare scenario identical to the actual mission is selected, and three spatially distributed signal receiving platforms are used to simultaneously intercept historical electromagnetic signals. Following step S100, a set of historical pulse descriptors containing parameters such as signal arrival direction 30°, signal arrival time 10:00:00.123, and carrier frequency 5GHz is extracted. The set contains thousands of valid pulse data intercepted by each platform.

[0028] Secondly, based on the prior true location information of the radiation source, pulses in the historical pulse descriptor set are manually sorted and labeled. Pulses originating from the same radiation source are identified as positive sample pairs, and pulses from different sources are randomly combined to form negative sample pairs, thus forming a training sample set. The prior true location information of the radiation source refers to the precise spatial coordinates of the radiation source obtained through calibration equipment or calibration methods in historical detections. A positive sample pair is a sample pair in the historical pulse descriptor set consisting of two pulses originating from the same true radiation source. A negative sample pair is a sample pair in the historical pulse descriptor set randomly formed from two pulses originating from different true radiation sources. The training sample set is a dataset used for model training, formed by combining positive and negative sample pairs in a balanced ratio. Based on the prior true location information of the radiation source, pulses in the historical pulse descriptor set are manually sorted and labeled. Pulses from the same source are paired to form positive sample pairs, and pulses from different sources are randomly paired to form negative sample pairs. These positive and negative sample pairs are then integrated in a balanced ratio to form a pulse matching training sample set.

[0029] For example, given that the true coordinates of the prior radiation source are (1000, 500, 100), the radiation source pulses intercepted by platform A and platform B are paired as positive sample pairs; the radiation source pulse is randomly paired with another radiation source pulse at coordinates (2000, 800, 200) as a negative sample pair; finally, a training sample set containing 10,000 positive sample pairs and 10,000 negative sample pairs is formed.

[0030] Furthermore, a pulse matching model based on a Siamese network architecture is constructed. This Siamese network architecture comprises two structurally identical feature extraction sub-networks with shared weights. Each sub-network processes the pulse descriptor of a pulse and extracts high-order features. A fusion discriminant layer based on azimuth cosine similarity and theoretical time difference consistency calculates and outputs a matching degree evaluation value between the two pulses. The Siamese network architecture consists of two identical neural networks with shared parameter weights, used to process two sets of homologous data to be discriminated. Weight sharing means that the two feature extraction sub-networks use identical network parameters, ensuring consistency in feature extraction rules. The feature extraction sub-network is a network unit used to perform feature mapping on the pulse descriptor and extract high-order features in the time, frequency, and spatial domains. Geometric constraints, including azimuth constraints, time difference constraints, and spatial consistency constraints, are the physical prior conditions for pulse homology determination. The fusion discriminant layer combines azimuth cosine similarity and theoretical time difference consistency to complete feature fusion and output the pulse matching degree. The matching degree evaluation value is a quantitative value representing the probability of homology between two pulses; the higher the value, the greater the probability of homology.

[0031] Specifically, a pulse matching model based on a twin network architecture is constructed, configuring two feature extraction subnetworks with identical structures and shared weights to process the descriptor of a pulse and extract high-order features respectively; azimuth constraints, time difference constraints, and spatial consistency constraints are introduced, and feature fusion is completed in the fusion discriminant layer by combining azimuth cosine similarity and theoretical time difference consistency. The model finally outputs the matching degree evaluation value of the two pulses.

[0032] For example, a twin network pulse matching model is built. The two sub-networks receive the pulse descriptor extracted by S100 and the pulse descriptor of another platform, respectively. After extracting high-order features, the orientation constraints, time difference constraints, and spatial consistency constraints are fused. The matching degree evaluation value is calculated through the fusion discriminant layer.

[0033] Furthermore, using the training sample set as input and the homology labels of sample pairs as supervision signals, the impulse matching model undergoes end-to-end supervised training. End-to-end training refers to directly inputting the original impulse descriptors into the model, with automatic optimization from input to output, eliminating the need for manual feature processing. Supervised training refers to a training method that uses the homology labels of the training sample set as supervision signals to guide the model in adjusting its parameters. Homology labels are used to indicate whether sample pairs are homologous; homologous is labeled as 1, and dissimilar is labeled as 0. By inputting the training sample set into the impulse matching model and using the homology labels of sample pairs as supervision signals, the network parameters are iteratively optimized using an end-to-end training method, reducing the error between the model's output matching degree and the true labels. For example, positive and negative sample pairs are input into the model, with positive sample pairs labeled as 1 and negative sample pairs labeled as 0; the Siamese network parameters are iteratively optimized through backpropagation, allowing the model to gradually learn the inherent rules of impulse homology matching.

[0034] Finally, when the matching accuracy of the model's output on the independent validation set for distinguishing positive and negative sample pairs exceeds a preset accuracy threshold, training stops, and the construction of the pulse matching model is complete. The independent validation set refers to an independent set of pulse samples that has no overlap with the training sample set and is used to test the model's generalization ability. The preset accuracy threshold is a pre-set minimum discrimination accuracy standard that the model can be used in practice. The model performance is tested using the independent validation set, and the accuracy of the model's output matching accuracy for distinguishing positive and negative sample pairs is statistically analyzed. When this accuracy exceeds the preset accuracy threshold, training stops, and model construction is complete. The two pulse descriptors to be matched are input into the trained model, and the matching accuracy is directly output.

[0035] For example, with a preset model discrimination accuracy threshold of 95%, and using an independent validation set for testing, the pulse matching model achieves a discrimination accuracy of 96% for positive and negative sample pairs, thus meeting the requirements and stopping training; when two pulse descriptors to be matched are input into the pulse matching model, the pulse matching model outputs a matching degree evaluation value of 0.92.

[0036] In this embodiment of the invention, a twin network pulse matching model is constructed by combining three types of geometric constraints: orientation, time difference, and spatial consistency. Through supervised training, a precise quantitative assessment of pulse homology across platforms is achieved, effectively solving the matching failure problem caused by pulse interleaving and parameter errors in dense electromagnetic environments. The model outputs a stable and reliable matching degree evaluation value, providing a precise basis for subsequent candidate pulse group aggregation and reducing the probability of pulse misassociation from the source.

[0037] S300: For any pulse intercepted by any of the signal receiving platforms, a candidate pulse group is formed by searching and aggregating the pulse lists of all other platforms based on the matching degree evaluation value.

[0038] In this embodiment of the invention, for any pulse intercepted by any signal receiving platform, a candidate pulse group is formed by searching and aggregating the pulse lists of all other platforms based on the matching degree evaluation value. In dense electromagnetic countermeasures environments, the number of pulses intercepted by different signal receiving platforms is enormous and measurement errors exist. If homogeneous aggregation is performed directly based on pulse matching degree, a large number of heterogeneous interference pulses are easily mixed in. At the same time, inconsistent pulse homogeneity within the candidate pulse group and insufficient number of platforms participating in positioning will directly lead to large errors or even false positioning in subsequent collaborative positioning calculations. To ensure that the pulse set entering the positioning stage has high homogeneity purity and sufficient platform support, an initial matching degree threshold needs to be scientifically set through validation set data. Then, abnormal pulses are eliminated through intra-group consistency verification. The minimum number of platforms is determined by combining offline simulation, and the threshold is iteratively optimized by adaptively lowering it to finally form a candidate pulse group that meets the requirements of homogeneity, consistency, and number of platforms, providing a reliable data foundation for subsequent accurate positioning.

[0039] Step S300 in the method provided in this embodiment of the invention includes: Based on the current pulse, set the initial matching degree threshold; Traverse the pulse lists of all other platforms, filter out pulses whose matching degree evaluation value is higher than the initial matching degree threshold, and add them to the initial candidate pulse group; Obtain the matching degree evaluation value among all pulse pairs in the initial candidate pulse group; If any two pulses have matching evaluation values ​​lower than the intra-group consistency threshold, then the pulse with the lowest matching evaluation value is removed from the initial candidate pulse group. After removing the pulses, the consistency judgment within the group is re-performed based on the remaining pulses; If any two pulses still have matching evaluation values ​​lower than the intra-group consistency threshold, the elimination and re-judgment operation continues until the matching evaluation values ​​of all pulse pairs in the group are not lower than the intra-group consistency threshold, thus forming an intermediate candidate pulse group. If the minimum number of platforms is not met, the initial matching threshold is lowered, and the traversal screening and intra-group consistency judgment are re-executed until the intermediate candidate pulse group meets the minimum number of platforms requirement. If the minimum number of platforms is met, the intermediate candidate pulse group will be determined as the final output candidate pulse group.

[0040] First, based on the current pulse, set an initial matching threshold.

[0041] Among them, the initial matching degree threshold is set based on the current pulse, including: The pulse matching model is invoked, and the validation set pulse descriptors intercepted and labeled by multiple signal receiving platforms in historical detection are input to obtain the statistical distribution of the validation set pulse pair matching degree evaluation values. Based on the homologous labels of the validation set pulse pairs, extract the set of positive sample pair matching evaluation values ​​and the set of negative sample pair matching evaluation values ​​from the statistical distribution, respectively. Calculate the upper quartile of the set of matching evaluation values ​​for the negative sample pairs, and use the upper quartile as the recommended setting value for the initial matching threshold; Before the association filtering begins, the initial matching threshold is set to be equal to or higher than the recommended setting value.

[0042] First, the pulse matching model is invoked, taking into account the validation set pulse descriptors captured and labeled by multiple signal receiving platforms during historical detection, to obtain the statistical distribution of the matching degree evaluation values ​​of the validation set pulse pairs. The validation set pulse descriptors refer to the independent pulse descriptor dataset captured by multiple signal receiving platforms in historical detection scenarios, labeled with source-like tags, and not used in model training. The statistical distribution of the matching degree evaluation values ​​of the validation set pulse pairs refers to the overall probability distribution of the matching degree evaluation values ​​output by the model after pairing all pulses in the validation set, reflecting the distribution pattern of the matching degree values, such as the concentration interval and dispersion. The pulse matching model trained in S200 is then invoked, inputting all pulse descriptors in the validation set in pairs into the model to obtain the matching degree evaluation value for each pulse pair; based on all matching degree evaluation values, the corresponding statistical distribution, such as histograms and probability density curves, is generated.

[0043] For example, the trained pulse matching model is called, and the input contains 1,000 validation set pulse descriptors covering 5 signal receiving platforms and 10 different radiation sources. All pulses are paired up to obtain 499,500 pulse pair matching degree values. The model is then input to obtain the matching degree evaluation value for each pair. The generated statistical distribution shows that the matching degree is mainly concentrated in the range of 0.1 to 0.8, with the range of 0.3 to 0.5 accounting for the highest proportion.

[0044] Secondly, based on the homology labels of the pulse pairs in the validation set, sets of positive and negative sample pair matching evaluation values ​​are extracted from the statistical distribution. Homology labels are pre-labeled tags on the pulse pairs in the validation set, used to identify whether the pulse pairs originate from the same radiation source; homology is labeled 1, and heterology is labeled 0. The set of positive sample pair matching evaluation values ​​is the sum of matching evaluation values ​​for pulse pairs in the validation set with a homology label of 1. The set of negative sample pair matching evaluation values ​​is the sum of matching evaluation values ​​for pulse pairs in the validation set with a homology label of 0. Based on the homology labels of the pulse pairs in the validation set, all pulse pair matching evaluation values ​​are classified and filtered: pulse pairs with a label of 1 are assigned to the positive sample pair set, and those with a label of 0 are assigned to the negative sample pair set, thus completing the extraction of the two sets.

[0045] For example, the matching degree values ​​of the above 499,500 pulse pairs are filtered based on the homologous labels: there are 50,000 positive sample pairs with label 1, and the matching degree values ​​are summarized into a positive sample pair set {0.62, 0.75, 0.81, ..., 0.93}; there are 449,500 negative sample pairs with label 0, and the matching degree values ​​are summarized into a negative sample pair set {0.11, 0.23, 0.35, ..., 0.72}.

[0046] Next, the upper quartile of the negative sample pair matching evaluation value set is calculated, and this upper quartile is used as the recommended setting value for the initial matching threshold. The upper quartile refers to the value at the 75th percentile after the negative sample pair matching evaluation value set is sorted in ascending order; that is, 75% of the negative sample matching values ​​are ≤ this value, and 25% of the values ​​are > this value. The recommended setting value for the initial matching threshold is a reference value using the upper quartile of the negative sample pair matching degree, used to initially define the matching degree threshold for high-probability homologous pulses. The values ​​in the negative sample pair matching evaluation value set are sorted in ascending order; the value corresponding to the 75th percentile is determined using the quartile calculation method, and this value is the recommended setting value for the initial matching threshold. For example, after sorting the negative sample pair set {0.11, 0.23, 0.35, ..., 0.72} in ascending order, the value of the 75th percentile is calculated to be 0.68, therefore the recommended setting value for the initial matching threshold is 0.68.

[0047] Therefore, before the correlation screening begins, the initial matching degree threshold is set to be equal to or higher than the recommended value. The initial matching degree threshold is a critical value used to screen cross-platform pulses. Setting it to be equal to or higher than the recommended value ensures high homology purity of the initial candidate pulse group and reduces the risk of irrelevant pulses entering subsequent calculations. The lower edge of the matching degree distribution of the positive sample pair matching degree evaluation set refers to the boundary of the interval with lower values ​​in the positive sample pair matching degree set, used to determine the specific magnitude of threshold adjustment. Before conducting cross-platform pulse correlation screening, the initial matching degree threshold is set to be equal to or higher than the recommended value; if the threshold needs to be adjusted upwards, the specific value can be determined by referring to the lower edge of the positive sample pair matching degree distribution, such as increasing the recommended value by 0.05.

[0048] For example, the lower edge of the positive sample pair matching set is 0.70. To ensure the purity of the initial candidate pulse group, the initial matching threshold is finally set to 0.73 by increasing the value by 0.05 based on the recommended value of 0.68. If the screening efficiency is desired, the threshold can also be set directly to the recommended value of 0.68.

[0049] Subsequently, the pulse lists of all other platforms are traversed, and pulses with matching degree evaluation values ​​higher than the initial matching degree threshold are selected and added to the initial candidate pulse group. The pulse list refers to the set of all valid pulse descriptors intercepted by a single signal receiving platform and extracted and validated in step S100. The initial candidate pulse group refers to the set formed by selecting pulses from other platforms with matching degrees higher than the initial threshold, using a reference pulse from a certain platform as the core. Any pulse intercepted by any signal receiving platform is selected as the reference pulse, and the pulse lists of all other signal receiving platforms are traversed; the reference pulse and each traversed pulse are input into the trained pulse matching model to obtain their matching degree evaluation values; pulses with matching degree evaluation values ​​higher than the initial matching degree threshold are selected and added to the initial candidate pulse group.

[0050] For example, the reference pulse P0 intercepted by platform A is selected: signal arrival direction 30°, arrival time 10:00:00.123, carrier frequency 5GHz. The pulse lists of platforms B, C, D, and E are traversed, and each platform has 200, 180, 220, and 190 valid pulses respectively. P0 and other pulses are input into the matching model one by one, and 8 pulses with a matching degree > 0.73 are selected, including 3 from platform B, 2 from platform C, 2 from platform D, and 1 from platform E. The 8 pulses together with P0 form the initial candidate pulse group.

[0051] Next, the matching degree evaluation values ​​of all pulse pairs in the initial candidate pulse group are obtained. A pulse pair within the initial candidate group refers to any pair of pulses in the initial candidate pulse group. The matching degree evaluation value of each pulse pair within the initial candidate group is the matching degree value output by the pulse matching model, used to verify the homogeneity of pulses within the group. All pulses in the initial candidate pulse group are paired up, and each pair is input into the trained pulse matching model. The matching degree evaluation value of each pair is obtained one by one, and the results are summarized to form a set of matching degrees for pulse pairs within the group.

[0052] For example, the initial candidate pulse group includes a baseline pulse P0 and 8 selected pulses, for a total of 9 pulses. Pairing them together yields 36 pulse pairs. Each of the 36 pulse pairs is then input into the matching model to obtain the corresponding set of matching degree evaluation values: {P0-P1: 0.85, P0-P2: 0.81, ...}.

[0053] Furthermore, if the matching degree evaluation value of any two pulses is lower than the intra-group consistency threshold, the pulse with the lowest matching degree evaluation value is removed from the initial candidate pulse group. The intra-group consistency threshold is a critical value determined through backtracking analysis of historically successfully located task data, used to determine whether pulses within a group possess source consistency. The pulse with the lowest matching degree evaluation value refers to the pulse with the lowest average matching degree after pairing with other pulses within the group, or at least one pair of pulses with a matching degree lower than the threshold and the smallest value. The matching degree evaluation values ​​of all pulse pairs within the group are compared with the intra-group consistency threshold; if the matching degree value of any pair of pulses is lower than this threshold, the pulse with the lowest matching degree evaluation value is removed from the initial candidate pulse group.

[0054] For example, the consistency threshold within the group is set to 0.70. Comparing the matching degree values ​​of 36 pulse pairs, it is found that the matching degree of P1-P2 is 0.67, which is lower than the threshold. The average matching degree of each pulse is calculated: the average matching degree of P2 with the other 8 pulses is 0.69, which is the lowest among all pulses. Therefore, pulse P2 is removed from the initial candidate pulse group.

[0055] Furthermore, after removing the pulses, the consistency judgment within the group is re-performed based on the remaining pulses: If any two pulses still have a matching score lower than the intra-group consistency threshold, the elimination and re-judgment operation continues until the matching score of all pulse pairs in the group is not lower than the intra-group consistency threshold, forming an intermediate candidate pulse group. The remaining pulses refer to the set of pulses remaining in the initial candidate pulse group after eliminating low-matching pulses. The intermediate candidate pulse group refers to the set of pulses where, after multiple iterations of elimination, the matching score of all pulse pairs in the group is not lower than the intra-group consistency threshold. After eliminating pulses, the remaining pulses are re-paired, and the matching model is used to obtain new intra-group pulse pair matching scores. It is then checked again whether any pair of pulses has a matching score lower than the intra-group consistency threshold. If so, the low-matching pulse elimination and re-pairing verification operation is repeated until the matching score of all pulse pairs in the group is ≥ the intra-group consistency threshold. The pulse set at this point is the intermediate candidate pulse group.

[0056] For example, after removing P2, the initial candidate group has 8 pulses remaining. These are then re-paired to obtain 28 pulse pairs, which are input into the model to obtain a new set of matching degrees. The verification shows that the matching degree of P4-P5 (0.68) is still lower than 0.70, and the average matching degree of P5 is 0.71, which is the lowest among the remaining pulses. Therefore, P5 is removed. The matching degree values ​​of the remaining 7 pulses (21 pairs) are verified again. All values ​​are ≥0.70, such as P0-P1: 0.85, P1-P3: 0.72, P6-P7: 0.78, etc. The iteration stops, and the 7 pulses form an intermediate candidate pulse group.

[0057] The steps for obtaining the minimum number of platforms required include: Offline simulation analysis is performed based on the historical pulse descriptor set and prior radiation source location information; The offline simulation analysis includes at least randomly selecting pulse subsets from different number of platforms from the historical pulse descriptor set and performing cooperative localization calculations on each subset. The error distribution between the positioning solution results corresponding to pulse subsets of different platform numbers and the prior true location information of the radiation source is calculated. From the error distribution, determine the minimum number of platforms corresponding to when the positioning error is stably lower than the preset positioning accuracy threshold, and set the minimum number of platforms as the basic platform number requirement. Based on the fault tolerance redundancy required in the actual task, the minimum number of platforms is increased on top of the basic platform number requirement to obtain the final minimum number of platforms required.

[0058] First, offline simulation analysis is performed based on a historical pulse descriptor set and prior true location information of the radiation source. The historical pulse descriptor set refers to the aggregated dataset of pulse descriptors intercepted by multiple signal receiving platforms under the same electronic countermeasures detection scenarios in the past, extracted and validated through the S100 step. The prior true location information of the radiation source refers to the precise three-dimensional spatial coordinates of the radiation source in the historical detection scenarios, obtained in advance through professional methods such as calibration and verification. Offline simulation analysis refers to a data analysis method that simulates the positioning process under different numbers of platforms based on historical data in a non-real-time, non-site simulation environment, without the participation of actual signal receiving platforms.

[0059] Specifically, we compile a set of historical pulse descriptors and their corresponding real-time radiation source locations, and build an offline simulation environment consistent with the parameters of historical detection scenarios. We define the simulation variable as the number of platforms participating in the positioning, fix invariants such as radiation source locations and electromagnetic environment parameters, and initiate offline simulation analysis with the impact of the number of platforms on positioning accuracy as the objective.

[0060] For example, a set of historical pulse descriptors containing 10 enemy radiation sources (real locations such as (1000m, 500m, 100m), (2000m, 800m, 200m), etc.) and 8 signal receiving platforms is obtained; an offline simulation environment is built to replicate the electromagnetic interference intensity, platform location and other parameters of the historical scene, and offline simulation analysis is started.

[0061] Secondly, the offline simulation analysis includes at least the following: randomly selecting pulse subsets from the historical pulse descriptor set, with varying numbers of platforms, and performing cooperative localization calculations on each subset. A pulse subset refers to a combination of pulse descriptors originating from the same radiation source, randomly selected from the historical pulse descriptor set according to different platform numbers. Cooperative localization calculation refers to the process of calculating the spatial coordinates of the radiation source using a weighted least squares algorithm, combining the pulse's signal arrival time with the corresponding platform's spatial location information. In the offline simulation environment, for each radiation source's historical pulse descriptor, pulse subsets of 2, 3, 4... up to the total number of platforms are randomly selected; for each number of pulse subsets, a weighted least squares algorithm is used to perform cooperative localization calculations, and the radiation source coordinate results obtained in each calculation are recorded.

[0062] For example, for radiation source 1, at its actual location (1000m, 500m, 100m), a subset of pulses from 2 platforms, a subset of pulses from 3 platforms, and so on, a subset of pulses from 8 platforms are randomly selected from the historical pulse set. The weighted least squares algorithm is used to solve each subset to obtain the solved coordinates of platform 2 (1050m, 540m, 110m), platform 3 (1020m, 510m, 105m), platform 4 (1005m, 502m, 101m), etc.

[0063] Next, the error distribution between the positioning results corresponding to pulse subsets with different platform numbers and the prior true location information of the radiation source is statistically analyzed. Positioning error refers to the spatial distance error between the calculated radiation source coordinates and the prior true location. Error distribution refers to the statistical distribution of positioning errors obtained from multiple random selections of pulse subsets for different platform numbers, including mean, variance, and 95% confidence interval. For the pulse subset calculation results of each platform number, the positioning error is calculated according to the formula; for 2, 3, 4… platform numbers, the positioning errors of at least 100 random selections of subsets are summarized, and the mean, variance, maximum value, and upper limit of the 95% confidence interval are statistically analyzed for each platform number to form the error distribution corresponding to different platform numbers.

[0064] For example, the positioning error of the solution results for each platform is calculated: the mean error for 2 platforms is 85m, and the variance is 25m. 2 The mean error of the three platforms is 35m, and the variance is 8m. 2 The mean error of the 4 platform solutions is 15m, and the variance is 3m. 2 The average solution error for 5 platforms and above is stable at around 12m. These data are compiled into an error distribution table and visualization curve of the number of platforms - mean / variance of positioning error.

[0065] Furthermore, from the error distribution, the minimum number of platforms required for the positioning error to be stably lower than a preset positioning accuracy threshold is determined, and this minimum number of platforms is set as the basic platform number requirement. The preset positioning accuracy threshold refers to the upper limit of positioning error set according to the actual electronic countermeasures mission requirements; the calculated error must be stably lower than this threshold to meet the basic positioning accuracy requirements of the mission. A stable positioning error below the threshold means that, with a certain number of platforms, the mean of the positioning error and the upper limit of the 95% confidence interval are both lower than the preset threshold, and the error variance is less than the preset fluctuation threshold. The basic platform number requirement refers to the minimum number of platforms for which the positioning error is stably lower than the preset accuracy threshold; it is the minimum number of platforms required to meet the basic positioning accuracy of the mission. By comparing the error distribution of different platform numbers with the preset positioning accuracy threshold, all platform numbers with stable positioning errors below the threshold are selected; the minimum value among these numbers is then set as the basic platform number requirement.

[0066] For example, the preset positioning accuracy threshold is 50m and the error fluctuation threshold is 10m. 2 2. The mean error of the platform is 85m, which is greater than 50m, so the condition is not met; 3. The mean error of the platform is 35m, and the upper limit of the 95% confidence interval is 48m, both of which are less than 50m, and the variance is 8m. 2 <10m 2 The requirement is met; therefore, the basic platform number requirement is determined to be 3.

[0067] Finally, based on the fault tolerance redundancy in the actual task, the basic platform number requirement is increased to obtain the final minimum platform number requirement. The fault tolerance redundancy refers to the additional number of platforms set according to the actual task scenario to improve the anti-interference capability and fault tolerance of the positioning. The final minimum platform number requirement is the sum of the basic platform number requirement and the fault tolerance redundancy, which is the minimum number of platforms that subsequent candidate pulse groups must cover. The fault tolerance redundancy is determined based on the actual electronic warfare task scenario, such as setting it to 1 for normal scenarios and 2 for strong interference / high fault tolerance scenarios; the basic platform number requirement and the fault tolerance redundancy are added together to obtain the final minimum platform number requirement.

[0068] For example, this task is in a strong interference environment, and the fault tolerance redundancy is set to 1; the basic platform number requirement is 3, so the final minimum platform number requirement = 3 + 1 = 4, that is, the subsequent candidate pulse group needs to cover at least 4 signal receiving platforms.

[0069] Furthermore, if the minimum number of platforms is not met, the initial matching degree threshold is lowered, and the traversal screening and intra-group consistency judgment are re-executed until the resulting intermediate candidate pulse group meets the minimum number of platforms requirement. Threshold reduction rule: The current initial matching degree threshold is reduced by a fixed proportion, and the reduced value is not lower than the median of the set of negative sample pair matching degree evaluation values, avoiding the introduction of a large number of heterogeneous pulses. The number of signal receiving platforms to which the pulses in the intermediate candidate pulse group belong is counted. If the minimum number of platforms is not met, the initial matching degree threshold is lowered according to the threshold reduction rule. Using the new threshold as the standard, the pulse lists of all other platforms are traversed again, and pulses with matching degrees higher than the new threshold are selected to generate a new initial candidate group. Intra-group consistency judgment is performed on the new initial candidate group, calculating the matching degree of all pulse pairs, removing low-matching degree pulses, and iteratively verifying to form a new intermediate candidate group. The number of platforms is verified again; if it still does not meet the standard, the above operations are repeated until the number of platforms covered by the intermediate candidate group meets the minimum number of platforms requirement.

[0070] For example, the final minimum number of platforms required is 4. The current intermediate candidate pulse group only covers 3 platforms, A, B, and C, which does not meet the requirement. The current initial matching degree threshold is 0.73, and the median matching degree of negative sample pairs is 0.52. After adjusting by 95%, the threshold is 0.73 × 95% = 0.6935, which is higher than 0.52 and meets the rule. Using 0.6935 as the new threshold, the pulse lists of platforms D and E are re-traversed, and 2 pulses with a matching degree > 0.6935 are selected to generate a new initial candidate group. The intra-group consistency check is performed on the new initial candidate group, and 1 pulse with a matching degree of 0.68 is removed. The resulting new intermediate candidate group covers 4 platforms, A, B, C, and D, which meets the minimum number of platforms requirement.

[0071] Furthermore, if the minimum platform number requirement is met, the intermediate candidate pulse group is determined as the final output candidate pulse group. The final candidate pulse group refers to the pulse set that simultaneously satisfies the requirements that the matching degree of all pulse pairs within the group is ≥ the intra-group consistency threshold and the number of covered platforms is ≥ the minimum platform number. It is the final pulse data used for subsequent collaborative positioning calculations. The number of signal receiving platforms covered by the intermediate candidate pulse group is verified. If it reaches or exceeds the minimum platform number requirement, and the matching degree of all pulse pairs within the group is not lower than the intra-group consistency threshold, then no further iteration is needed, and the intermediate candidate pulse group is directly determined as the final output candidate pulse group.

[0072] For example, after threshold reduction and iterative screening, the new intermediate candidate pulse group contains 7 pulses, covering 4 signal receiving platforms, meeting the minimum platform requirement of 4, and the intra-group consistency threshold of matching degree of all pulse pairs in the group is ≥0.70; therefore, the intermediate candidate pulse group is determined as the final candidate pulse group, and the cross-platform pulse aggregation process is completed.

[0073] In this embodiment of the invention, the purity of the initial pulse screening is ensured by scientifically setting the initial matching degree threshold, and the interference pulses from different sources are effectively eliminated through the consistency check within the group. The minimum number of platforms is determined by offline simulation, and the complete aggregation of pulses from the same source is achieved by adaptively lowering the threshold. The final generated candidate pulse group has high consistency and sufficient platform support, which fundamentally avoids false positioning and solution failure caused by pulse association errors and insufficient number of platforms, and provides high-quality and high-reliability pulse data for subsequent collaborative positioning and solution.

[0074] S400: Based on the signal arrival time of all pulses in the candidate pulse group and the spatial location information of the corresponding platform, the spatial coordinates of the radiation source are calculated through collaborative positioning, and the positioning result is output.

[0075] In this embodiment of the invention, based on the signal arrival time of all pulses in the candidate pulse group and the spatial location information of the corresponding platform, the spatial coordinates of the radiation source are calculated through collaborative positioning, and the positioning result is output. In electronic warfare scenarios, the accuracy of collaborative positioning is directly affected by the platform's spatial geometry and the rationality of pulse weight allocation. If ordinary least squares calculation is directly used without considering the differences in platform geometry and pulse matching degree, it is easy to lead to large calculation errors. At the same time, the coordinates obtained from the candidate pulse group may exceed the platform's physical detection range, resulting in false positioning results. To improve the positioning accuracy and validity of the radiation source, it is necessary to first calculate the geometric accuracy factor of the platform cluster to quantify the quality of the configuration, allocate pulse weights by combining the matching degree and the geometric accuracy factor, use weighted least squares to calculate the initial coordinate values, and verify and eliminate invalid calculation results through spatial domain and detection distance, iteratively optimizing the weights until a precise positioning result that conforms to physical constraints is obtained.

[0076] Step S400 in the method provided in this embodiment of the invention includes: Based on the spatial location information of each pulse source platform in the candidate pulse group, calculate the geometric precision factor of the platform cluster corresponding to the current candidate pulse group; The reciprocal of the matching degree evaluation value corresponding to each pulse in the candidate pulse group is used as the initial weight, and normalization is performed in combination with the geometric precision factor to determine the final fusion weight of each pulse in the solution. Using the signal arrival time of all pulses in the candidate pulse group and the spatial location information of the corresponding platform, a weighted least squares algorithm is used to perform cooperative localization calculation to obtain the initial value of the spatial coordinates of the radiation source. Based on the initial spatial coordinates and the spatial location information of all platforms, verify whether the current coordinates are within the visible geographic airspace and the maximum theoretical detection distance of all platforms. If the verification is successful, the initial spatial coordinates will be output as the final positioning result. If the verification fails, the allocation strategy of the final fusion weights is readjusted, and the weighted least squares solution and spatial verification operation are iteratively executed until the spatial coordinates that pass the verification are obtained and output.

[0077] First, based on the spatial location information of each pulse source platform in the candidate pulse group, the geometric precision factor of the platform cluster corresponding to the current candidate pulse group is calculated.

[0078] Specifically, based on the spatial location information of the source platform for each pulse in the candidate pulse group, the geometric precision factor of the platform cluster corresponding to the current candidate pulse group is calculated, including: Select the spatial position of any one platform in the candidate pulse group as the reference origin, and calculate the spatial position vector of all other platforms relative to the reference origin. Normalize the spatial position vectors of all platforms to obtain the corresponding unit direction vectors; Based on the unit direction vectors of all platforms, calculate the volume characteristic value of the spatial geometric configuration composed of multiple unit direction vectors; Set a very small positive number as the lower limit threshold of volume. If the volume feature value is less than the lower limit threshold of volume, then the current volume feature value is directly set as the lower limit threshold of volume. The reciprocal of the volume feature value after processing the volume lower limit threshold is mapped to the geometric precision factor of the platform cluster, wherein the smaller the value of the geometric precision factor, the better the spatial geometry of the platform cluster.

[0079] First, the spatial location of any platform in the candidate pulse group is selected as the reference origin, and the spatial position vectors of all other platforms relative to the reference origin are calculated. The reference origin is any spatial location selected from the platforms covered by the candidate pulse group as the reference point for calculating relative positions. The spatial position vector is a local coordinate system established with the reference origin as the center, converting the three-dimensional geographic coordinates of other platforms into three-dimensional rectangular coordinate vectors originating from the origin, directly reflecting the relative spatial relationship between platforms. From all signal receiving platforms corresponding to the candidate pulse group, the three-dimensional geographic coordinates (X0, Y0, Z0) of one platform are randomly selected as the reference origin; the three-dimensional geographic coordinates (X0, Y0, Z0) of all other platforms are extracted. i ,Y i Z i ), calculate the spatial position vector of each platform relative to the reference origin: v i =(X i -X0,Y i -Y0,Z i −Z0).

[0080] For example, the candidate pulse group covers platforms A (0m, 0m, 100m), B (500m, 0m, 100m), C (250m, 433m, 100m), and D (250m, 144m, 200m); platform A is selected as the reference origin, and the spatial position vector of each platform is calculated: Platform B: v B =(500−0,0−0,100−100)=(500,0,0); Platform C: v C =(250−0,433−0,100−100)=(250,433,0); Platform D: v D =(250−0,144−0,200−100)=(250,144,100).

[0081] Secondly, the spatial position vectors of all platforms are normalized to obtain the corresponding unit direction vectors. Normalization involves dividing the spatial position vector by its magnitude to convert it into a unit direction vector with a magnitude of 1, thus eliminating the influence of vector length on subsequent geometric configuration calculations. The magnitude |v| is calculated for each spatial position vector. i |=(X i -X0) 2 +(Y i -Y0) 2 +(Z i −Z0) 2 Dividing each component of the vector by its magnitude yields the unit direction vector u. i =v i / ∣v i |. For example, platform B: module length |v B |=500, unit direction vector u B =(1,0,0); Platform C: Module length |v C |=250 2 +433 2 ≈500, unit direction vector u C =(0.5,0.866,0); Platform D: Module length |v D |=250 2 +144 2 +100 2 ≈300, unit direction vector u D ≈(0.833,0.48,0.333).

[0082] Furthermore, based on the unit direction vectors of all platforms, the volume characteristic value of the spatial geometric configuration composed of multiple unit direction vectors is calculated. The volume characteristic value of the spatial geometric configuration refers to the quantified volume of the spatial geometry composed of the unit direction vectors of all platforms, reflecting the spatial distribution quality of the platform cluster; a larger volume indicates a better configuration. The unit direction vectors of all platforms are constructed into a matrix, with each row corresponding to a unit direction vector. The volume characteristic value of the spatial geometric configuration corresponding to this matrix is ​​calculated using the determinant or the spatial geometry volume formula. For example, based on the unit direction vector u... B =(1,0,0), u C =(0.5,0.866,0), u D Construct a matrix with ≈(0.833,0.48,0.333) and calculate the determinant to obtain the volume eigenvalue ≈0.288.

[0083] Furthermore, a very small positive number is set as the lower limit threshold for volume. If the volume feature value is less than the lower limit threshold, the current volume feature value is directly set as the lower limit threshold. The lower limit threshold is a preset very small positive number, such as 1 × 10⁻⁶. −10 This is used to prevent a volume feature value of 0 from causing a denominator of 0 in subsequent geometric precision factor calculations. The calculated volume feature value is compared with a volume lower limit threshold. If the volume feature value is less than the threshold, it is replaced with the volume lower limit threshold; if it is greater than or equal to the threshold, the original value is retained. For example, the preset volume lower limit threshold is 1 × 10⁻⁶. −10 0.288 > 1 × 10 −10 Therefore, the retained volume characteristic value is 0.288.

[0084] Finally, the reciprocal of the volume feature value after processing with the lower volume threshold is mapped to the geometric precision factor (GDOP) of the platform cluster. A smaller GDOP indicates a better spatial geometry of the platform cluster. The GDOP is an indicator that quantifies the impact of the platform cluster's spatial geometry on positioning accuracy; a smaller value indicates a better configuration and higher positioning accuracy. The calculation formula is: GDOP = 1 / volume feature value. Taking the reciprocal of the volume feature value after lower limit processing yields the GDOP of the platform cluster; if the volume feature value has been replaced with the lower limit threshold, then GDOP = 1 / lower volume threshold.

[0085] For example, the volume characteristic value after lower limit processing is 0.288, and the geometric precision factor GDOP = 1 / 0.288 ≈ 3.472; if the volume characteristic value is 5 × 10 −11 Less than 1×10 −10 Then replace with 1×10 −10 Geometric precision factor GDOP = 1 / (1×10) −10 ) = 1 × 1010 .

[0086] Secondly, the reciprocal of the matching degree evaluation value corresponding to each pulse in the candidate pulse group is used as the initial weight. This initial weight is then normalized using the geometric accuracy factor to determine the final fusion weight for each pulse in the solution. The initial weight refers to using the reciprocal of the matching degree evaluation value of each pulse in the candidate pulse group as the base weight; the lower the matching degree, the smaller the initial weight. The normalization process involves adjusting the initial weights in conjunction with the geometric accuracy factor and scaling them proportionally until the total weight sums to 1, ensuring the rationality of the weight allocation. The final fusion weight is the final weight for each pulse participating in the positioning solution, comprehensively considering both the pulse matching degree and the advantages and disadvantages of the platform's geometric configuration.

[0087] Specifically, the matching degree evaluation value corresponding to each pulse in the candidate pulse group is extracted, and its reciprocal is calculated as the initial weight. ,in Let be the matching degree evaluation value of the i-th pulse; multiply each initial weight by the geometric precision factor to obtain the adjusted initial weight. GDOP; normalizes all adjusted weights: , where n is the number of pulses in the candidate pulse group, ensuring that the sum of all final fusion weights is 1.

[0088] For example, the candidate pulse group contains 4 pulses: P0 (match 0.95), P1 (match 0.90), P2 (match 0.85), and P3 (match 0.80); the geometric precision factor GDOP ≈ 3.472. Initial weights: P0: 1 / 0.95 ≈ 1.0526; P1: 1 / 0.90 ≈ 1.1111; P2: 1 / 0.85 ≈ 1.1765; P3: 1 / 0.80 = 1.25. Adjusted weights: P0: 1.0526 × 3.472 ≈ 3.654; P1: 1.1111 × 3.472 ≈ 3.858; P2: 1.1765 × 3.472 ≈ 4.085; P3: 1.25 × 3.472 ≈ 4.34. Normalized: Total = 3.654 + 3.858 + 4.085 + 4.34 ≈ 15.937; P0: 3.654 / 15.937 ≈ 0.229; P1: 3.858 / 15.937 ≈ 0.242; P2: 4.085 / 15.937 ≈ 0.256; P3: 4.34 / 15.937 ≈ 0.273; Final fusion weight sum = 0.229 + 0.242 + 0.256 + 0.273 = 1, which meets the requirements.

[0089] Next, using the signal arrival times of all pulses in the candidate pulse group and the spatial location information of the corresponding platform, a weighted least squares algorithm is used for collaborative localization to obtain the initial spatial coordinates of the radiation source. The weighted least squares algorithm introduces weights into the least squares solution, allowing pulses with higher confidence levels to have a greater impact on the solution results, thus improving the accuracy. The initial spatial coordinates of the radiation source refer to the three-dimensional coordinates (X', Y', Z') obtained through the weighted least squares solution; these are preliminary localization results and their effectiveness needs further verification.

[0090] Specifically, a positioning equation based on the time of arrival (TOA) is constructed: |r−p i |=c×(t) i -t0), where r is the coordinate of the radiation source, p i Let be the spatial coordinates of the i-th platform, c be the electromagnetic wave propagation speed, and t be the coordinates of the i-th platform. i Let ti be the arrival time of the i-th pulse signal, and t0 be the emission time of the radiation source; the final fusion weights are introduced into the equation to construct a weighted least squares objective function: Solve the objective function to obtain the initial values ​​of the spatial coordinates (X', Y', Z') of the radiation source.

[0091] For example, the platform coordinates of the candidate pulse group are: A(0,0,100), B(500,0,100), C(250,433,100), D(250,144,200); pulse arrival times are: P0 (Platform A, t=10:00:00.001), P1 (Platform B, t=10:00:00.00267), P2 (Platform C, t=10:00:00.00233), P3 (Platform D, t=10:00:00.002); electromagnetic wave velocity c=3×10 8 m / s; Substituting into the weighted least squares equation and combining the final fusion weights (0.229, 0.242, 0.256, 0.273), the initial coordinate values ​​are obtained: X'=1000m, Y'=800m, Z'=200m.

[0092] Furthermore, based on the initial spatial coordinates and the spatial location information of all platforms, it is verified whether the current coordinates are within the visible geographic airspace and maximum theoretical detection range of all platforms. The visible geographic airspace refers to the physical line-of-sight range determined by geometric calculations based on the platform's location, altitude, and high-precision terrain data; this can be obtained in advance based on the platform's own status before processing. The maximum theoretical detection range refers to the maximum detectable distance conservatively estimated based on the radar equations combined with signal parameters and propagation models, assuming the worst-case scenario. The visible geographic airspace range and maximum theoretical detection range of each platform are obtained in advance; the spatial distance from the initial coordinates (X', Y', Z') to each platform is calculated. Verify that the initial coordinates of the two points are within the visible geographic spatial domain on all platforms, and that all d i Check if the distance is less than or equal to the maximum theoretical detection distance of the corresponding platform; if both conditions are met, the verification passes, otherwise the verification fails.

[0093] If the verification passes, the initial spatial coordinates will be output as the final positioning result. For example, the maximum theoretical detection distance of platforms A, B, C, and D is 2000m, and their visible geographic spatial domains cover X∈[0,1500], Y∈[0,1000], and Z∈[0,500]. The initial coordinates (1000,800,200) are approximately 1284.5m to platform A (≤2000m) and approximately 953.9m to platform B (≤2000m). The coordinates are located within the visible geographic spatial domains of all platforms; therefore, the verification passes. This will be output as the final positioning result, which is the spatial coordinates of the radiation source: X=1000m, Y=800m, Z=200m.

[0094] If the verification fails, the final fusion weight allocation strategy is readjusted, and weighted least squares calculation and spatial verification operations are iteratively executed until the verified spatial coordinates are obtained and output. If the verification fails, the final fusion weight allocation strategy is readjusted; new final fusion weights are calculated based on the new weight allocation strategy, weighted least squares calculation is re-executed to obtain new initial coordinate values, and spatial and detection distance verification is performed again; the above iterative process is repeated until the verified spatial coordinates are obtained and output.

[0095] For example, if the initial coordinates are (3000, 2500, 600), the distance to platform A is approximately 3956m > 2000m, and Z = 600 exceeds the visible space Z ≤ 500, the verification fails. Adjust the weight allocation strategy, changing the initial weights to... Recalculate the final fusion weights: P0: 1 / (0.95) 2 ≈1.108; P1: 1 / (0.90) 2 ≈1.235; P2: 1 / (0.85) 2 ≈1.384; P3: 1 / (0.80)²≈1.562; After normalization based on GDOP≈3.472, we obtain the new weights: P0≈0.215, P1≈0.234, P2≈0.259, P3≈0.292; We recalculate to obtain the new initial coordinate values ​​(1100, 850, 220), verify that they are located in the visible space and the distance to each platform is ≤2000m, and output the coordinates after verification.

[0096] In this embodiment of the invention, the configuration quality of the computing platform cluster is quantified by the geometric precision factor, and differentiated fusion weights are assigned by the pulse matching degree. The weighted least squares algorithm is used to improve the positioning accuracy. By verifying the visible geographic spatial domain and the maximum theoretical detection distance, false positioning results that exceed physical constraints are effectively eliminated. For cases where the verification fails, the weight strategy is iteratively optimized, and finally, accurate radiation source positioning results that conform to physical constraints are output. This solves the problem that traditional positioning solutions do not consider platform configuration and pulse reliability and are prone to false positioning, thus improving the accuracy and effectiveness of passive positioning of instantaneous signal radiation sources in electronic warfare scenarios.

[0097] Through the specific implementation methods described above, the embodiments of the present invention achieve the following technical effects: This invention provides a multi-platform collaborative instantaneous signal passive positioning method and system, achieving accurate passive positioning of instantaneous radio frequency signal radiation sources from non-cooperative adversaries in electronic warfare scenarios. By synchronously intercepting electromagnetic signals across multiple platforms, the method extracts and verifies the rationality of pulse time-domain, spatial-domain, and frequency-domain parameters, eliminating invalid interference parameters and forming standardized pulse descriptors. This provides clean and reliable basic data for subsequent processes, avoiding cascading errors caused by parameter distortion. A twin network pulse matching model is constructed by combining three types of geometric constraints: azimuth, time difference, and spatial consistency. Supervised training enables accurate quantitative evaluation of cross-platform pulse homology, solving the matching failure problem caused by pulse interleaving and parameter errors in dense electromagnetic environments. By scientifically setting matching degree thresholds, intra-group consistency verification, and minimum platform number constraints, iterative optimization aggregates candidate pulse groups with high homology purity that meet positioning requirements, suppressing heterogeneous pulse interference and false associations at the source. Combining the advantages and disadvantages of platform cluster geometric configurations with pulse credibility allocation weights, a weighted least squares algorithm is used to calculate positioning coordinates. Iterative optimization is verified through visible spatial domain and maximum detection distance verification, effectively eliminating false positioning results and improving positioning accuracy. Ultimately, this improved the accuracy, reliability, and efficiency of passive positioning of instantaneous signal radiation sources in complex and dense electromagnetic environments.

[0098] Example 2, as Figure 2 As shown, the present invention provides a multi-platform collaborative instantaneous signal passive positioning system, the system comprising: The signal interception module 11 is used to simultaneously intercept multiple electromagnetic signals from multiple signal receiving platforms distributed in different spatial locations and extract pulse descriptors for each signal. The pulse matching modeling module 12 is used to construct a pulse matching model based on geometric constraints based on the pulse descriptor and output the matching degree evaluation value between any two pulses. The candidate pulse aggregation module 13 is used to search and aggregate any pulse intercepted by any of the signal receiving platforms to form a candidate pulse group based on the matching degree evaluation value in the pulse list of all other platforms. The positioning calculation output module 14 is used to calculate the spatial coordinates of the radiation source through collaborative positioning based on the signal arrival time of all pulses in the candidate pulse group and the spatial position information of the corresponding platform, and output the positioning result.

[0099] In one embodiment, the signal interception module 11 is further configured to: The step of extracting the pulse descriptor includes: The raw electromagnetic signals collected by each signal receiving platform are instantaneously detected based on a preset amplitude threshold to determine the start and end times of the pulses and record the signal arrival time. For the intercepted pulse, the direction of arrival of the signal is calculated by using the fixed phase difference between the signals received by the multi-channel receiving array of the corresponding platform and the phase interferometry method. For the same pulse, an instantaneous frequency measurement circuit is used to measure the corresponding carrier frequency; The extracted signal arrival direction, signal arrival time, and carrier frequency are each checked for reasonableness based on a preset physical possibility range threshold. The parameter combinations that pass the check are then encapsulated into a descriptor for the current pulse.

[0100] In one embodiment, the pulse matching modeling module 12 is further configured to: The geometric constraints include at least orientation constraints, time difference constraints, and spatial consistency constraints.

[0101] The steps for constructing the pulse matching model include: Historical electromagnetic signal data synchronously intercepted by the multiple signal receiving platforms under the same detection scenario are collected, and the corresponding set of historical pulse descriptors is extracted. Based on prior information about the true location of radiation sources, pulses in the historical pulse descriptor set are manually sorted and labeled to identify pulse pairs originating from the same radiation source, forming positive sample pairs, and pulses from different sources are randomly combined to form negative sample pairs, thus forming a training sample set. A pulse matching model based on a Siamese network architecture is constructed, wherein the Siamese network architecture includes two feature extraction sub-networks with identical structures and shared weights. Each sub-network is used to process the pulse descriptor of a pulse and extract high-order features. The matching degree evaluation value between the two pulses is calculated and output through a fusion discriminant layer based on the consistency between azimuth cosine similarity and theoretical time difference. Using the training sample set as input and the homologous labels of the sample pairs as supervision signals, the impulse matching model is subjected to end-to-end supervised training. When the matching accuracy of the model's output on the independent validation set for distinguishing positive and negative sample pairs is higher than a preset accuracy threshold, training stops, and the construction of the pulse matching model is completed.

[0102] In one embodiment, the candidate pulse aggregation module 13 is further configured to: Based on the current pulse, set the initial matching degree threshold; Traverse the pulse lists of all other platforms, filter out pulses whose matching degree evaluation value is higher than the initial matching degree threshold, and add them to the initial candidate pulse group; Obtain the matching degree evaluation value among all pulse pairs in the initial candidate pulse group; If any two pulses have matching evaluation values ​​lower than the intra-group consistency threshold, then the pulse with the lowest matching evaluation value is removed from the initial candidate pulse group. After removing the pulses, the consistency judgment within the group is re-performed based on the remaining pulses; If any two pulses still have matching evaluation values ​​lower than the intra-group consistency threshold, the elimination and re-judgment operation continues until the matching evaluation values ​​of all pulse pairs in the group are not lower than the intra-group consistency threshold, thus forming an intermediate candidate pulse group. If the minimum number of platforms is not met, the initial matching threshold is lowered, and the traversal screening and intra-group consistency judgment are re-executed until the intermediate candidate pulse group meets the minimum number of platforms requirement. If the minimum number of platforms is met, the intermediate candidate pulse group will be determined as the final output candidate pulse group.

[0103] Among them, the initial matching degree threshold is set based on the current pulse, including: The pulse matching model is invoked, and the validation set pulse descriptors intercepted and labeled by multiple signal receiving platforms in historical detection are input to obtain the statistical distribution of the validation set pulse pair matching degree evaluation values. Based on the homologous labels of the validation set pulse pairs, extract the set of positive sample pair matching evaluation values ​​and the set of negative sample pair matching evaluation values ​​from the statistical distribution, respectively. Calculate the upper quartile of the set of matching evaluation values ​​for the negative sample pairs, and use the upper quartile as the recommended setting value for the initial matching threshold; Before the association filtering begins, the initial matching threshold is set to be equal to or higher than the recommended setting value.

[0104] The steps for obtaining the minimum number of platforms required include: Offline simulation analysis is performed based on the historical pulse descriptor set and prior radiation source location information; The offline simulation analysis includes at least randomly selecting pulse subsets from different number of platforms from the historical pulse descriptor set and performing cooperative localization calculations on each subset. The error distribution between the positioning solution results corresponding to pulse subsets of different platform numbers and the prior true location information of the radiation source is calculated. From the error distribution, determine the minimum number of platforms corresponding to when the positioning error is stably lower than the preset positioning accuracy threshold, and set the minimum number of platforms as the basic platform number requirement. Based on the fault tolerance redundancy required in the actual task, the minimum number of platforms is increased on top of the basic platform number requirement to obtain the final minimum number of platforms required.

[0105] In one embodiment, the positioning solution output module 14 is further configured to: Based on the spatial location information of each pulse source platform in the candidate pulse group, calculate the geometric precision factor of the platform cluster corresponding to the current candidate pulse group; The reciprocal of the matching degree evaluation value corresponding to each pulse in the candidate pulse group is used as the initial weight, and normalization is performed in combination with the geometric precision factor to determine the final fusion weight of each pulse in the solution. Using the signal arrival time of all pulses in the candidate pulse group and the spatial location information of the corresponding platform, a weighted least squares algorithm is used to perform cooperative localization calculation to obtain the initial value of the spatial coordinates of the radiation source. Based on the initial spatial coordinates and the spatial location information of all platforms, verify whether the current coordinates are within the visible geographic airspace and the maximum theoretical detection distance of all platforms. If the verification is successful, the initial spatial coordinates will be output as the final positioning result. If the verification fails, the allocation strategy of the final fusion weights is readjusted, and the weighted least squares solution and spatial verification operation are iteratively executed until the spatial coordinates that pass the verification are obtained and output.

[0106] Specifically, based on the spatial location information of the source platform for each pulse in the candidate pulse group, the geometric precision factor of the platform cluster corresponding to the current candidate pulse group is calculated, including: Select the spatial position of any one platform in the candidate pulse group as the reference origin, and calculate the spatial position vector of all other platforms relative to the reference origin. Normalize the spatial position vectors of all platforms to obtain the corresponding unit direction vectors; Based on the unit direction vectors of all platforms, calculate the volume characteristic value of the spatial geometric configuration composed of multiple unit direction vectors; Set a very small positive number as the lower limit threshold of volume. If the volume feature value is less than the lower limit threshold of volume, then the current volume feature value is directly set as the lower limit threshold of volume. The reciprocal of the volume feature value after processing the volume lower limit threshold is mapped to the geometric precision factor of the platform cluster, wherein the smaller the value of the geometric precision factor, the better the spatial geometry of the platform cluster.

[0107] It should be noted that 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 protection scope of the present invention.

Claims

1. A multi-platform collaborative instantaneous signal passive positioning method, characterized in that, The method includes: Multiple signal receiving platforms distributed in different spatial locations synchronously intercept multiple electromagnetic signals and extract pulse descriptors from each signal. Based on the pulse descriptor, a pulse matching model based on geometric constraints is constructed, and the matching degree evaluation value between any two pulses is obtained. For any pulse intercepted by any of the signal receiving platforms, a candidate pulse group is formed by searching and aggregating the pulse lists of all other platforms based on the matching degree evaluation value. Based on the signal arrival time of all pulses in the candidate pulse group and the spatial location information of the corresponding platform, the spatial coordinates of the radiation source are calculated through collaborative positioning, and the positioning result is output.

2. The multi-platform collaborative instantaneous signal passive positioning method according to claim 1, characterized in that, The pulse descriptor extraction steps include: The raw electromagnetic signals collected by each signal receiving platform are instantaneously detected based on a preset amplitude threshold to determine the start and end times of the pulses and record the signal arrival time. For the intercepted pulse, the direction of arrival of the signal is calculated by using the fixed phase difference between the signals received by the multi-channel receiving array of the corresponding platform and the phase interferometry method. For the same pulse, an instantaneous frequency measurement circuit is used to measure the corresponding carrier frequency; The extracted signal arrival direction, signal arrival time, and carrier frequency are each checked for reasonableness based on a preset physical possibility range threshold. The parameter combinations that pass the check are then encapsulated into a descriptor for the current pulse.

3. The multi-platform collaborative instantaneous signal passive positioning method according to claim 1, characterized in that, The geometric constraints include at least orientation constraints, time difference constraints, and spatial consistency constraints.

4. The multi-platform collaborative instantaneous signal passive positioning method according to claim 1, characterized in that, The steps for constructing the pulse matching model include: Historical electromagnetic signal data synchronously intercepted by the multiple signal receiving platforms under the same detection scenario are collected, and the corresponding set of historical pulse descriptors is extracted. Based on prior information about the true location of radiation sources, pulses in the historical pulse descriptor set are manually sorted and labeled to identify pulse pairs originating from the same radiation source, forming positive sample pairs, and pulses from different sources are randomly combined to form negative sample pairs, thus forming a training sample set. A pulse matching model based on a Siamese network architecture is constructed, wherein the Siamese network architecture includes two feature extraction sub-networks with identical structures and shared weights. Each sub-network is used to process the pulse descriptor of a pulse and extract high-order features. The matching degree evaluation value between the two pulses is calculated and output through a fusion discriminant layer based on the consistency between azimuth cosine similarity and theoretical time difference. Using the training sample set as input and the homologous labels of the sample pairs as supervision signals, the impulse matching model is subjected to end-to-end supervised training. When the matching accuracy of the model's output on the independent validation set for distinguishing positive and negative sample pairs is higher than a preset accuracy threshold, training stops, and the construction of the pulse matching model is completed.

5. The multi-platform collaborative instantaneous signal passive positioning method according to claim 1, characterized in that, For any pulse intercepted by any of the aforementioned signal receiving platforms, based on the matching degree evaluation value, a candidate pulse group is searched and aggregated in the pulse lists of all other platforms, including: Based on the current pulse, set the initial matching degree threshold; Traverse the pulse lists of all other platforms, filter out pulses whose matching degree evaluation value is higher than the initial matching degree threshold, and add them to the initial candidate pulse group; Obtain the matching degree evaluation value among all pulse pairs in the initial candidate pulse group; If any two pulses have matching evaluation values ​​lower than the intra-group consistency threshold, then the pulse with the lowest matching evaluation value is removed from the initial candidate pulse group. After removing the pulses, the consistency judgment within the group is re-performed based on the remaining pulses; If any two pulses still have matching evaluation values ​​lower than the intra-group consistency threshold, the elimination and re-judgment operation continues until the matching evaluation values ​​of all pulse pairs in the group are not lower than the intra-group consistency threshold, thus forming an intermediate candidate pulse group. If the minimum number of platforms is not met, the initial matching threshold is lowered, and the traversal screening and intra-group consistency judgment are re-executed until the intermediate candidate pulse group meets the minimum number of platforms requirement. If the minimum number of platforms is reached, the intermediate candidate pulse group will be determined as the final output candidate pulse group.

6. The multi-platform collaborative instantaneous signal passive positioning method according to claim 5, characterized in that, Based on the current pulse, set an initial matching threshold, including: The pulse matching model is invoked, and the validation set pulse descriptors intercepted and labeled by multiple signal receiving platforms in historical detection are input to obtain the statistical distribution of the validation set pulse pair matching degree evaluation values. Based on the homologous labels of the validation set pulse pairs, extract the set of positive sample pair matching evaluation values ​​and the set of negative sample pair matching evaluation values ​​from the statistical distribution, respectively. Calculate the upper quartile of the set of matching evaluation values ​​for the negative sample pairs, and use the upper quartile as the recommended setting value for the initial matching threshold; Before the association filtering begins, the initial matching threshold is set to be equal to or higher than the recommended setting value.

7. The multi-platform collaborative instantaneous signal passive positioning method according to claim 5, characterized in that, The steps for obtaining the minimum number of platforms required include: Offline simulation analysis is performed based on the historical pulse descriptor set and prior radiation source location information; The offline simulation analysis includes at least randomly selecting pulse subsets from different number of platforms from the historical pulse descriptor set and performing cooperative localization calculations on each subset. The error distribution between the positioning solution results corresponding to pulse subsets of different platform numbers and the prior true location information of the radiation source is calculated. From the error distribution, determine the minimum number of platforms corresponding to when the positioning error is stably lower than the preset positioning accuracy threshold, and set the minimum number of platforms as the basic platform number requirement. Based on the fault tolerance redundancy required in the actual task, the minimum number of platforms is increased on top of the basic platform number requirement to obtain the final minimum number of platforms required.

8. The multi-platform collaborative instantaneous signal passive positioning method according to claim 1, characterized in that, Based on the signal arrival time of all pulses in the candidate pulse group and the spatial location information of the corresponding platform, the spatial coordinates of the radiation source are calculated through collaborative positioning, and the positioning results are output, including: Based on the spatial location information of each pulse source platform in the candidate pulse group, calculate the geometric precision factor of the platform cluster corresponding to the current candidate pulse group; The reciprocal of the matching degree evaluation value corresponding to each pulse in the candidate pulse group is used as the initial weight, and normalization is performed in combination with the geometric precision factor to determine the final fusion weight of each pulse in the solution. Using the signal arrival time of all pulses in the candidate pulse group and the spatial location information of the corresponding platform, a weighted least squares algorithm is used to perform cooperative localization calculation to obtain the initial value of the spatial coordinates of the radiation source. Based on the initial spatial coordinates and the spatial location information of all platforms, verify whether the current coordinates are within the visible geographic airspace and the maximum theoretical detection distance of all platforms. If the verification is successful, the initial spatial coordinates will be output as the final positioning result. If the verification fails, the allocation strategy of the final fusion weights is readjusted, and the weighted least squares solution and spatial verification operation are iteratively executed until the spatial coordinates that pass the verification are obtained and output.

9. The multi-platform collaborative instantaneous signal passive positioning method according to claim 8, characterized in that, Based on the spatial location information of the source platform for each pulse in the candidate pulse group, the geometric precision factor of the platform cluster corresponding to the current candidate pulse group is calculated, including: Select the spatial position of any one platform in the candidate pulse group as the reference origin, and calculate the spatial position vector of all other platforms relative to the reference origin. Normalize the spatial position vectors of all platforms to obtain the corresponding unit direction vectors; Based on the unit direction vectors of all platforms, calculate the volume characteristic value of the spatial geometric configuration composed of multiple unit direction vectors; Set a very small positive number as the lower limit threshold of volume. If the volume feature value is less than the lower limit threshold of volume, then the current volume feature value is directly set as the lower limit threshold of volume. The reciprocal of the volume feature value after processing the volume lower limit threshold is mapped to the geometric precision factor of the platform cluster, wherein the smaller the value of the geometric precision factor, the better the spatial geometry of the platform cluster.

10. A multi-platform collaborative instantaneous signal passive positioning system, characterized in that, For implementing the multi-platform collaborative instantaneous signal passive positioning method according to any one of claims 1-9, the system comprises: The signal interception module is used to simultaneously intercept multiple electromagnetic signals from multiple signal receiving platforms distributed in different spatial locations and extract pulse descriptors for each signal. The pulse matching modeling module is used to construct a pulse matching model based on geometric constraints based on the pulse descriptor and output the matching degree evaluation value between any two pulses. The candidate pulse aggregation module is used to search and aggregate any pulse intercepted by any of the signal receiving platforms to form a candidate pulse group based on the matching degree evaluation value in the pulse lists of all other platforms. The positioning calculation output module is used to calculate the spatial coordinates of the radiation source based on the signal arrival time of all pulses in the candidate pulse group and the spatial location information of the corresponding platform through collaborative positioning, and output the positioning result.

Images