Intelligent electromagnetic situation awareness system based on multi-sensor data fusion

Abstract

The invention relates to the technical field of radio monitoring and electromagnetic situation awareness, in particular to a multi-sensor data fusion intelligent electromagnetic situation awareness system, which comprises an environment reconstruction module for calculating and generating a pure ideal state reference vector which is not interfered at the current moment; the disturbance deduction module is configured to call to-be-detected target characteristic parameters from a preset signal characteristic parameter library and generate a simulation state vector with disturbance; the difference calculation module is used for calculating a numerical value difference between the real-time observation vector and the pure ideal state reference vector to obtain a real difference residual error, and calculating a numerical value difference between the disturbance simulation state vector and the pure ideal state reference vector to obtain a theoretical difference residual error; the coupling judgment module is used for calculating the geometric direction consistency of the real difference residual error and the theoretical difference residual error in the feature space and outputting a spectrum situation sensing result; according to the method, the false alarm rate of the system is remarkably reduced, and the judgment stability in a low signal-to-noise ratio environment is ensured.

Description

Technical Field

[0001] This invention relates to the field of radio monitoring and electromagnetic situational awareness technology, specifically to an intelligent electromagnetic situational awareness system that integrates multi-sensor data. Background Technology

[0002] In the current complex electromagnetic situational awareness environment, distributed sensor arrays need to achieve real-time identification and analysis of abnormal interference signals in monitoring areas with a large number of legitimate background signals, such as high-power radio stations, base station signals, and random environmental noise.

[0003] To achieve spectrum monitoring, existing solutions generally rely on energy detection methods or statistical analysis methods based on historical data. While these methods are effective in scenarios with clear signal characteristics, the highly dynamic and complex electromagnetic environment presents several limitations: severe masking effects; high-power legitimate background signals often mask weak anomalous interference, making it difficult for monitoring systems to set uniform detection thresholds and limiting weak signal detection capabilities; significant environmental noise interference; traditional energy detection methods are highly susceptible to random noise fluctuations and environmental attenuation, such as rain attenuation and building obstruction, resulting in high false alarm rates and poor decision robustness in low signal-to-noise ratio environments; and a lack of proactive extrapolation capabilities. Existing technologies are mostly passive monitoring, heavily reliant on known interference samples for training, lacking the ability to proactively simulate and match complex interference that has never appeared before but conforms to physical laws, making it difficult to support high-precision, semantic-level situational awareness. Therefore, how to accurately eliminate legitimate background signals and improve the sensitivity and accuracy of weak signal detection in complex background noise and strong interference environments through physical modeling and structured feature matching has become an urgent technical problem to be solved. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides an intelligent electromagnetic situational awareness system based on multi-sensor data fusion. Specifically, the technical solution of this invention includes:

[0005] The environment reconstruction module is configured to call the static geographic environment data and propagation model parameters of the monitoring area, and calculate and generate the pure ideal state reference vector that is undisturbed at the current moment based on the deterministic physical modeling method.

[0006] The disturbance inference module is configured to retrieve the target feature parameters to be detected from a preset signal feature parameter library, and superimpose the target feature parameters onto the pure ideal state reference vector to generate a disturbance-triggered simulation state vector.

[0007] The differential calculation module is configured to receive real-time observation vectors collected by a distributed sensor array, calculate the numerical difference between the real-time observation vectors and the pure ideal state reference vectors to obtain the real difference residuals, and calculate the numerical difference between the perturbed simulation state vectors and the pure ideal state reference vectors to obtain the theoretical difference residuals.

[0008] The coupled decision module is configured to calculate the geometrical orientation consistency between the actual difference residual and the theoretical difference residual in the feature space based on the vector similarity matching algorithm, and output the spectrum situational awareness result based on the consistency calculation result.

[0009] Preferably, the environment reconstruction module is configured to construct a clean, ideal-state baseline vector that is undisturbed at the current moment by performing the following operations:

[0010] Acquire topographic elevation data, building obstruction data, and transmission parameters of legitimate stations within the monitoring area;

[0011] Using the ray tracing algorithm, the theoretical power spectral density and phase data at each node in the distributed sensor array are calculated in a forward manner without random noise or external interference.

[0012] The theoretical power spectral density and phase data are aggregated to generate the pure ideal state reference vector.

[0013] Preferably, the perturbation inference module is configured to generate a perturbed simulation state vector by superimposing the target feature parameters onto the pure ideal state reference vector by performing the following operations:

[0014] In response to the current monitoring task requirements, select the corresponding interference type or anomaly mode from the signal feature parameter library;

[0015] The selected interference type or anomaly mode is converted into mathematical parameter factors, which include the nominal center frequency, center frequency drift, bandwidth change and modulation mode characteristic data of the target to be detected.

[0016] Numerical perturbation components are generated based on the mathematical parameter factors, and the numerical perturbation components are applied to the corresponding frequency index position of the pure ideal state reference vector in a mathematical superposition manner to generate the perturbation simulation state vector containing structured anomaly features.

[0017] Preferably, the difference calculation module is configured to calculate the difference between the real-time observation vector and the pure ideal state reference vector to obtain the reality difference residual by performing the following operations:

[0018] Perform time-frequency synchronization alignment operation between the real-time observation vector and the pure ideal state reference vector;

[0019] Perform vector subtraction operation to remove the legitimate background signal reference component contained in the pure ideal state reference vector by subtracting the pure ideal state reference vector from the real-time observation vector;

[0020] The residual vector component obtained after the subtraction operation, which does not contain the legitimate background signal reference component but retains the real abnormal signal, environmental random noise, and sensor thermal noise, is determined as the real difference residual.

[0021] Preferably, the coupling decision module is configured to calculate the geometrical orientation consistency between the actual difference residual and the theoretical difference residual in the feature space by performing the following operation based on a vector similarity matching algorithm:

[0022] Calculate the vector dot product of the actual difference residual and the theoretical difference residual to obtain the inner product scalar;

[0023] The product of the Euclidean norm of the actual difference residual and the Euclidean norm of the theoretical difference residual is calculated to obtain the normalized scalar.

[0024] Dividing the inner product scalar by the normalized scalar yields the cosine similarity score, which represents the angle between the two residual vectors in the feature space.

[0025] Preferably, the coupling decision module is configured to output spectrum situational awareness results based on the consistency calculation results by performing the following operations:

[0026] Call the preset decision threshold;

[0027] If the similarity score is greater than the decision threshold, it is determined that there is an abnormal signal in the real-time observation vector that matches the target feature parameters, and an abnormal alarm and target type are output.

[0028] If the similarity score is less than or equal to the decision threshold, it is determined that the difference in the real-time observation vector mainly comes from environmental random noise, and a normal situation report is output.

[0029] Preferably, the system further includes a feedback correction module, which is configured to:

[0030] When the coupling decision module outputs a normal situation report, the propagation model parameters in the environment reconstruction module are corrected for error minimization based on the statistical distribution characteristics of the actual difference residuals.

[0031] When the coupling decision module outputs an abnormal alarm, the corresponding real difference residual is digitally marked and stored in the signal feature parameter library to update the target feature parameter.

[0032] Preferably, the real-time observation vector includes: complex baseband IQ data or power spectral density data of multiple monitoring nodes collected simultaneously by the distributed sensor array and time-synchronized.

[0033] Preferably, the signal characteristic parameter library includes: FM broadcast sideband characteristic parameters, navigation signal clock drift pattern parameters, and radar signal pulse repetition interval pattern parameters.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] 1. This invention uses digital twin reconstruction and ray tracing algorithms to generate a high-precision pure ideal state reference through physical modeling. Compared with the traditional statistical averaging method, this system can accurately model and remove the background signal of high-power legal stations, effectively solving the problem of strong signals masking weak interference. This differential extraction mechanism significantly increases the proportion of anomalous signals in the residual, enabling the system to maintain extremely high detection sensitivity to weak anomalous signals even in complex electromagnetic environments.

[0036] 2. This invention enhances the robustness of decision-making in complex environments by feature space matching; it adopts a coupled decision-making mechanism based on vector similarity, and outputs the result by calculating the geometrical consistency between the actual residual and the theoretical residual in the feature space; the algorithm utilizes the orthogonality between structured signals and random noise in the feature space, focusing on signal structure features rather than absolute amplitude matching, effectively eliminating signal amplitude fluctuation interference caused by weather attenuation, building obstruction, etc., significantly reducing the false alarm rate of the system, and ensuring decision-making stability in low signal-to-noise ratio environments;

[0037] 3. This invention possesses proactive inference capabilities, enabling forward-looking situational awareness; it transforms expert experience into a computable mathematical model, and generates perturbation-injected simulation vectors by proactively injecting interference features; this changes the limitations of traditional monitoring systems that passively wait for samples and heavily rely on historical data, enabling the system to perform forward-looking simulation and matching of complex interferences that conform to physical laws but have never appeared before; combined with the digital labeling function of the feedback correction module, the system can achieve incremental learning of the feature library, continuously improving its ability to identify and classify new interference signals;

[0038] 4. This invention employs frequency domain synchronization alignment technology in differential calculation to ensure the real-time performance of multi-node data processing. Simultaneously, the online calibration mechanism established through the feedback correction module can correct the propagation model parameters in real time based on the statistical characteristics of the actual residuals. This adaptive evolution capability enables the system to automatically sense changes in long-term propagation characteristics within the monitoring area caused by seasonal changes or temporary buildings, ensuring that the digital twin model always maintains a high degree of consistency with the real physical environment and sustains long-term monitoring accuracy. Attached Figure Description

[0039] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0040] Figure 1 This is a structural diagram of the system of the present invention. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0042] Example 1:

[0043] Please see Figure 1 An intelligent electromagnetic situational awareness system that integrates multi-sensor data fusion, comprising:

[0044] The environment reconstruction module is configured to call the static geographic environment data and propagation model parameters of the monitoring area, and calculate and generate the pure ideal state reference vector that is undisturbed at the current moment based on the deterministic physical modeling method.

[0045] The disturbance inference module is configured to retrieve the target feature parameters to be detected from a preset signal feature parameter library, and superimpose the target feature parameters onto the pure ideal state reference vector to generate a disturbance-triggered simulation state vector.

[0046] The differential calculation module is configured to receive real-time observation vectors collected by a distributed sensor array, calculate the numerical difference between the real-time observation vectors and the pure ideal state reference vectors to obtain the real difference residuals, and calculate the numerical difference between the perturbed simulation state vectors and the pure ideal state reference vectors to obtain the theoretical difference residuals.

[0047] The coupled decision module is configured to calculate the geometrical orientation consistency between the actual difference residual and the theoretical difference residual in the feature space based on the vector similarity matching algorithm, and output the spectrum situational awareness result based on the consistency calculation result.

[0048] This embodiment provides an intelligent electromagnetic situational awareness system based on multi-sensor data fusion. This system aims to solve the problems of difficult weak signal detection and high false alarm rate in traditional spectrum monitoring due to complex background environments and high levels of random noise interference. The system adopts a digital twin reconstruction + differential coupling verification architecture and includes the following core modules:

[0049] The purpose of the environment reconstruction module is to construct a physically pure reference benchmark. In this embodiment, the module is configured to call static geographic environment data of the monitoring area, such as digital elevation model (DEM) and propagation model parameters, and calculate and generate a pure ideal benchmark vector that is undisturbed at the current moment based on deterministic physical modeling methods rather than historical statistical methods. This vector represents the theoretical electromagnetic data that the sensor should receive under completely ideal conditions, without external malicious interference and random noise.

[0050] The purpose of the perturbation inference module is to actively generate the theoretical feature fingerprint of the target to be detected. In this embodiment, the module is configured to retrieve the target feature parameters to be detected from the preset signal feature parameter library, such as known interference source features, and superimpose the target feature parameters onto the pure ideal state reference vector to generate a perturbation simulation state vector. This process is equivalent to simulating the interference scenario in the digital twin space.

[0051] The differential calculation module aims to extract the deviation components between reality and theory. In this embodiment, the module is configured to receive real-time observation vectors acquired by a distributed sensor array. The module performs dual-track differential calculation: calculating the numerical difference between the real-time observation vector and the pure ideal state reference vector to obtain the reality difference residual; and calculating the numerical difference between the perturbed simulation state vector and the pure ideal state reference vector to obtain the theoretical difference residual.

[0052] The coupling decision module aims to filter environmental noise and confirm the existence of the target. In this embodiment, the module is configured to calculate the geometrical orientation consistency between the actual difference residual and the theoretical difference residual in the feature space based on the vector similarity matching algorithm, and output the spectrum situational awareness result based on the consistency calculation result.

[0053] By constructing a pure ideal state reference vector, this system can accurately remove legitimate background signals, such as high-power radio stations, using a physical model, thus solving the problem of strong signals masking weak interference. At the same time, by calculating the consistency between the real difference residual and the theoretical difference residual, and by utilizing the orthogonality between structured signals and random noise in the feature space, the system can maintain extremely high detection sensitivity for weak anomalous signals even under extremely noisy environments.

[0054] Example 2:

[0055] The environment reconstruction module is configured to construct a clean, ideal baseline vector that is undisturbed at the current moment by performing the following operations:

[0056] Acquire topographic elevation data, building obstruction data, and transmission parameters of legitimate stations within the monitoring area;

[0057] Using the ray tracing algorithm, the theoretical power spectral density and phase data at each node in the distributed sensor array are calculated in a forward manner without random noise or external interference.

[0058] The theoretical power spectral density and phase data are aggregated to generate the pure ideal state reference vector.

[0059] This embodiment provides a detailed description of the specific implementation of the environment reconstruction module; the process by which the environment reconstruction module constructs the undisturbed pure ideal state reference vector at the current moment is as follows:

[0060] The module acquires terrain elevation data within the monitoring area, such as SRTM data with a precision of 30 meters, building occlusion data, 3D city models, and transmission parameters of legitimate stations; wherein, the transmission parameters of legitimate stations include, but are not limited to, determined physical quantities such as center frequency, transmission power, antenna gain, antenna height, and beam pointing.

[0061] The module uses a ray tracing algorithm to calculate the theoretical power spectral density and phase data at each node in the distributed sensor array in a forward manner, without random noise or external interference. The ray tracing algorithm used here refers to the simulation of the direct, reflected, diffracted and scattered paths of electromagnetic waves in complex physical environments based on electromagnetic wave propagation theory, such as the Longley-Rice model or ITU-R P.1546 recommendation, thereby obtaining a deterministic field strength distribution.

[0062] The theoretical power spectral density and phase data are aggregated according to frequency index and spatial nodes to generate a pure ideal state reference vector; the specific aggregation method is as follows: The first... Each sensing node at frequency Theoretical power spectral density at With phase data Convert to complex form And according to node sequence number and frequency index The size of the construction is Complex characteristic matrix or concatenated vector ; The imaginary unit;

[0063] By introducing high-precision GIS data and ray tracing algorithms, the reference vector generated in this embodiment is based on the physical calculation results of first principles, rather than the statistical average of historical data. This means that the reference vector does not contain interference or anomalies that have not been discovered in history, and can truly reflect the current electromagnetic propagation environment, such as the shadow effect caused by building obstruction, thereby greatly improving the accuracy of subsequent differential calculations.

[0064] Example 3:

[0065] The perturbation inference module is configured to superimpose the target feature parameters onto the pure ideal state reference vector to generate a perturbed simulation state vector by performing the following operations:

[0066] In response to the current monitoring task requirements, select the corresponding interference type or anomaly mode from the signal feature parameter library;

[0067] The selected interference type or anomaly mode is converted into mathematical parameter factors, which include the nominal center frequency, center frequency drift, bandwidth change and modulation mode characteristic data of the target to be detected.

[0068] Numerical perturbation components are generated based on the mathematical parameter factors, and the numerical perturbation components are applied to the corresponding frequency index position of the pure ideal state reference vector in a mathematical superposition manner to generate the perturbation simulation state vector containing structured anomaly features.

[0069] This embodiment provides a detailed description of the specific implementation of the perturbation inference module; the operation of the perturbation inference module to superimpose the target feature parameters onto the pure ideal state reference vector to generate the perturbation-induced simulation state vector is as follows:

[0070] In response to current monitoring task requirements, such as when the system receives an instruction to investigate illegal radio broadcasts, it selects the corresponding interference type or abnormal mode from the signal characteristic parameter library.

[0071] The selected interference type or anomaly pattern is converted into mathematical parameter factors; these mathematical parameter factors are not vague descriptions, but specific, calculable variables, including: the nominal center frequency of the target to be detected. Center frequency drift Bandwidth variation Modulation mode characteristic data, such as IQ constellation diagram mapping rules;

[0072] Numerical perturbation components are generated based on mathematical parameter factors, and these components are applied to the corresponding frequency index positions of the pure ideal state reference vector through mathematical superposition. For example, if the complex value of the pure reference at a certain frequency point is... The numerical perturbation components are The sum of the values ​​is 1, 2, 3, 4, 5, 6, 7, 8, 9, 1 ... The amplitude of the disturbance component With phase Generated by mapping preset modulation mode feature data; amplitude A power dimension consistent with the pure ideal state reference vector, such as dBm or linear power density, is adopted to ensure physical alignment during numerical superposition; this generates the perturbed simulation state vector containing structured anomaly features; numerical perturbation components. The generation logic is as follows: based on the center frequency drift. and bandwidth change Define a complex perturbation function within the corresponding frequency band, and index other frequencies as follows: ;

[0073] This embodiment uses a knowledge parameterization injection method to transform expert experience into a computable mathematical model; the system does not need to passively wait for abnormal signals to appear to train the AI ​​model, but can actively deduce various possible abnormal situations; this comprehensive analysis method enables the system to detect complex interferences that have never been seen before but conform to physical laws.

[0074] Example 4:

[0075] The difference calculation module is configured to calculate the difference between the real-time observation vector and the pure ideal state reference vector to obtain the reality difference residual by performing the following operations:

[0076] Perform time-frequency synchronization alignment operation between the real-time observation vector and the pure ideal state reference vector;

[0077] Perform vector subtraction operation to remove the legitimate background signal reference component contained in the pure ideal state reference vector by subtracting the pure ideal state reference vector from the real-time observation vector;

[0078] The residual vector component obtained after the subtraction operation, which does not contain the legitimate background signal reference component but retains the real abnormal signal, environmental random noise, and sensor thermal noise, is determined as the real difference residual.

[0079] This embodiment provides a detailed description of the specific implementation of the difference calculation module; the operation of the difference calculation module to calculate the difference between the real-time observation vector and the pure ideal state reference vector to obtain the actual difference residual is as follows:

[0080] Perform time-frequency synchronization alignment on the real-time observation vector and the pure ideal state reference vector to ensure that they correspond on the same time window and frequency grid; the specific alignment operation includes: calculating the real-time observation vector. With reference vector cross-correlation function ;

[0081] in, For the time delay search factor, superscript This indicates that a conjugate operation is performed on a complex sequence to extract the phase difference features of the signal;

[0082] Through search The point of maximum delay As the synchronization starting point, cyclic shift compensation is performed on the real-time observation vector to eliminate phase deviation caused by spatial propagation delay. To meet the real-time requirements of millisecond-level broadband spectrum situational awareness, the above cross-correlation function calculation process is preferably implemented in the frequency domain using Fast Fourier Transform (FFT). That is, after performing FFT on the two sets of vectors, a dot product is performed, and then the inverse FFT is used to transform back to the time domain, thereby reducing the computational complexity from... Reduce to ;

[0083] Perform vector subtraction; specifically, subtract the pure ideal state reference vector from the real-time observation vector; since the pure ideal state reference vector has accurately modeled the legitimate signals in the region, such as TV tower and base station signals, this step can effectively remove the legitimate background signal reference components contained in the pure ideal state reference vector;

[0084] The remaining vector components obtained after the subtraction operation are determined as the real difference residual. It should be emphasized that this residual does not contain the legitimate background signal reference component in a physical sense, but retains the following three parts: the real abnormal signal to be detected; random noise in the environment, such as atmospheric noise; and the thermal noise of the sensor itself.

[0085] Traditional energy detection methods struggle to set thresholds under strong background signals. This embodiment transforms the signal detection problem into a residual analysis problem through differential calculation. By eliminating high-power legitimate background signals, it significantly increases the proportion of anomalous signals, even those with low power, in the remaining residuals, laying a high signal-to-noise ratio (SNR) foundation for subsequent coupling decisions.

[0086] Example 5:

[0087] The coupling decision module is configured to calculate the geometrical orientation consistency between the actual difference residual and the theoretical difference residual in the feature space by performing the following operation based on a vector similarity matching algorithm:

[0088] Calculate the vector dot product of the actual difference residual and the theoretical difference residual to obtain the inner product scalar;

[0089] The product of the Euclidean norm of the actual difference residual and the Euclidean norm of the theoretical difference residual is calculated to obtain the normalized scalar.

[0090] Dividing the inner product scalar by the normalized scalar yields the cosine similarity score, which represents the angle between the two residual vectors in the feature space.

[0091] This embodiment provides a detailed explanation of the core algorithm of the coupling decision module. The coupling decision module calculates the geometrical orientation consistency between the actual difference residual and the theoretical difference residual in the feature space based on the vector similarity matching algorithm, specifically by calculating the cosine similarity score.

[0092] Define the reality difference residual as a vector. Define the theoretical difference residual as a vector. Both of these vectors are preferably multidimensional complex vectors or high-dimensional spectral feature vectors.

[0093] Calculate the vector dot product of the actual difference residual and the theoretical difference residual to obtain the inner product scalar. The calculation formula is as follows: ;

[0094] in, This indicates the conjugate operation of the theoretical difference residual; Represents the dimension index of a vector, such as the frequency point index;

[0095] The normalized scalar is obtained by multiplying the Euclidean norm of the actual difference residuals with the Euclidean norm of the theoretical difference residuals. The calculation formula is as follows: ;

[0096] in, The total number of dimensional indices of the feature vector; the similarity score The range of values ​​is normalized to . Within the interval, the closer the value is to 1, the better the observed actual residual matches the preset theoretical interference characteristics geometrically; dividing the inner product scalar by the normalized scalar yields the cosine similarity score representing the angle between the two residual vectors in the feature space. : ;

[0097] in, Let be the angle between two vectors in the feature space;

[0098] The cosine similarity score Physically, it characterizes the proportion of projected energy of the actual residual in the direction of theoretical anomalies. Because normalization is performed through norm multiplication during calculation, the decision depends only on the structural characteristics of the signal rather than its absolute amplitude. Therefore, even when the electromagnetic environment experiences non-frequency-selective attenuation due to weather conditions such as rain or obstruction, the system can still accurately identify weak anomalies through directional consistency. The core advantage of this similarity algorithm lies in its matched filter characteristics. Since environmental noise and thermal noise are statistically random and disordered, they differ from structured... The inner product between theoretical difference residuals approaches zero; while if It contains real abnormal signals, and its structure will be similar to... High matching makes The value increases significantly; by calculating directional consistency rather than absolute error, this invention effectively eliminates the impact of overall signal amplitude fluctuations, such as overall attenuation caused by weather, on the decision.

[0099] Example 6:

[0100] The coupling decision module is configured to output spectrum situational awareness results based on the consistency calculation results by performing the following operations:

[0101] Call the preset decision threshold;

[0102] If the similarity score is greater than the decision threshold, it is determined that there is an abnormal signal in the real-time observation vector that matches the target feature parameters, and an abnormal alarm and target type are output.

[0103] If the similarity score is less than or equal to the decision threshold, it is determined that the difference in the real-time observation vector mainly comes from environmental random noise, and a normal situation report is output.

[0104] This embodiment provides a detailed explanation of the decision logic of the coupled decision module; the operation of the coupled decision module in outputting the spectrum situational awareness result based on the consistency calculation result is as follows:

[0105] Call the preset decision threshold This threshold is usually preset according to the CFAR (False Alarm Rate) criterion, for example, a value of 0.7.

[0106] If the calculated similarity score Greater than the decision threshold If the system detects an abnormal signal in the real-time observation vector that matches the target feature parameters, it will determine that such an abnormal signal exists. The system will then output an abnormal alarm and the identified target type, which corresponds to the interference mode selected in Example 3.

[0107] If the similarity score Less than or equal to the decision threshold If the difference in the real-time observation vector is determined to be mainly due to random environmental noise, a normal situation report will be output.

[0108] This dual-threshold logic provides a clear decision boundary; compared to traditional energy-based thresholds which are susceptible to background level fluctuations, similarity-based thresholds are more robust; it can clearly distinguish between meaningless noise fluctuations and meaningful signal anomalies, thereby significantly reducing false alarm interference faced by maintenance personnel.

[0109] Example 7:

[0110] The system also includes a feedback correction module, which is configured to:

[0111] When the coupling decision module outputs a normal situation report, the propagation model parameters in the environment reconstruction module are corrected for error minimization based on the statistical distribution characteristics of the actual difference residuals.

[0112] When the coupling decision module outputs an abnormal alarm, the corresponding real difference residual is digitally marked and stored in the signal feature parameter library to update the target feature parameter.

[0113] This embodiment also includes a feedback correction module for the adaptive evolution of the system. The feedback correction module performs the following operations:

[0114] When the coupling decision module outputs a normal situation report, it means that the current discrepancy is mainly caused by model error or random noise. At this time, the system initiates an online calibration procedure: the module, based on the statistical distribution characteristics of the actual discrepancy residuals, such as mean deviation, uses the average historical residuals from multiple sets of normal states within a sliding window to adjust the propagation model parameters in the environment reconstruction module, including but not limited to the road loss index. Standard deviation of shadow fading Error minimization correction is performed; the correction uses the stochastic gradient descent (SGD) algorithm, and a loss function is defined. The propagation model parameters are updated using the L2 mean of the residuals of the actual differences, according to the following formula. : ;

[0115] in This is the preset step size coefficient; This represents the iteration number index for propagation model parameter correction. This timed sliding window update mechanism effectively filters out instantaneous multipath fluctuations, enabling the digital twin system to adapt to long-term propagation characteristics changes within the monitoring area. When the coupling decision module outputs an abnormal alarm, it means a real threat signal has been captured. At this time, the module digitizes and stores the corresponding real-world difference residual in the signal feature parameter library to update the target feature parameters, achieving incremental learning for the system. For novel interference signals, the system can record their feature fingerprints, enriching the feature library and enabling faster and more accurate detection of similar signals in the future.

[0116] Example 8:

[0117] The real-time observation vector includes: complex baseband IQ data or power spectral density data of multiple monitoring nodes collected simultaneously by the distributed sensor array and time-synchronized.

[0118] This embodiment provides a detailed description of the data format for real-time observation vectors. Specifically, real-time observation vectors include: complex baseband IQ data or power spectral density (PSD) data from multiple monitoring nodes, acquired simultaneously by a distributed sensor array and time-synchronized. Time synchronization calibration is achieved through a timing system, such as GNSS or Precision Time Protocol (PTP), requiring that the synchronization error between monitoring nodes not exceed [a certain value]. This is to ensure that the data from each node can meet the phase alignment requirements for spatial coherence processing when monitoring in the millimeter wave or ultra-high frequency band.

[0119] Complex baseband IQ data: contains amplitude and phase information of the signal, suitable for scenarios requiring high-precision phase alignment and modulation identification;

[0120] Power spectral density data: contains only amplitude spectrum information, suitable for distributed node scenarios with limited data transmission bandwidth;

[0121] The data format is clearly defined as time-synchronized and calibrated, which ensures the spatial coherence of multi-node data. This is a prerequisite for effective differential calculation and spatial feature matching.

[0122] Example 9:

[0123] The signal characteristic parameter library includes: FM broadcast sideband characteristic parameters, navigation signal clock drift pattern parameters, and radar signal pulse repetition interval pattern parameters.

[0124] This embodiment provides a detailed description of the contents of the signal feature parameter library; the signal feature parameter library stores prior knowledge in a specific field, specifically including: FM broadcast sideband feature parameters: used to identify illegal black broadcasts, specific pilot frequency deviations and sideband modulation indices; navigation signal clock drift pattern parameters: used to identify GNSS spoofing signals, clock drift rates inconsistent with real satellite signals; radar signal pulse repetition interval (PRI) mode parameters: used to identify unauthorized radar occupancy, specific pulse sequence timing structures;

[0125] By pre-setting these high-dimensional feature parameters, the system is not just detecting energy, but identifying the genes of the signal; this enables the system to effectively distinguish between legitimate signals in the same frequency band, such as normal broadcasts, and illegal signals, such as black broadcasts with specific sideband characteristics, thus achieving semantic-level spectrum situational awareness.

[0126] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An intelligent electromagnetic situational awareness system that integrates multi-sensor data, characterized in that, include: The environment reconstruction module is configured to call the static geographic environment data and propagation model parameters of the monitoring area, and calculate and generate the pure ideal state reference vector that is undisturbed at the current moment based on the deterministic physical modeling method. The disturbance inference module is configured to retrieve the target feature parameters to be detected from a preset signal feature parameter library, and superimpose the target feature parameters onto the pure ideal state reference vector to generate a disturbance-triggered simulation state vector. The differential calculation module is configured to receive real-time observation vectors collected by a distributed sensor array, calculate the numerical difference between the real-time observation vectors and the pure ideal state reference vectors to obtain the real difference residuals, and calculate the numerical difference between the perturbed simulation state vectors and the pure ideal state reference vectors to obtain the theoretical difference residuals. The coupled decision module is configured to calculate the geometrical orientation consistency between the actual difference residual and the theoretical difference residual in the feature space based on the vector similarity matching algorithm, and output the spectrum situational awareness result based on the consistency calculation result.

2. The intelligent electromagnetic situational awareness system based on multi-sensor data fusion according to claim 1, characterized in that, The environment reconstruction module is configured to construct a clean, ideal-state baseline vector that is undisturbed at the current moment by performing the following operations: Acquire topographic elevation data, building obstruction data, and transmission parameters of legitimate stations within the monitoring area; Using the ray tracing algorithm, the theoretical power spectral density and phase data at each node in the distributed sensor array are calculated in a forward manner without random noise or external interference. The theoretical power spectral density and phase data are aggregated to generate the pure ideal state reference vector.

3. The intelligent electromagnetic situational awareness system based on multi-sensor data fusion according to claim 1, characterized in that, The perturbation inference module is configured to superimpose the target feature parameters onto the pure ideal state reference vector by performing the following operations to generate a perturbated simulation state vector: In response to the current monitoring task requirements, select the corresponding interference type or anomaly mode from the signal feature parameter library; The selected interference type or anomaly mode is converted into mathematical parameter factors, which include the nominal center frequency, center frequency drift, bandwidth change and modulation mode characteristic data of the target to be detected. Numerical perturbation components are generated based on the mathematical parameter factors, and the numerical perturbation components are applied to the corresponding frequency index position of the pure ideal state reference vector in a mathematical superposition manner to generate the perturbation simulation state vector containing structured anomaly features.

4. The intelligent electromagnetic situational awareness system based on multi-sensor data fusion according to claim 1, characterized in that, The difference calculation module is configured to calculate the difference between the real-time observation vector and the pure ideal state reference vector to obtain the real-world difference residual by performing the following operations: Perform time-frequency synchronization alignment operation between the real-time observation vector and the pure ideal state reference vector; Perform vector subtraction operation to remove the legitimate background signal reference component contained in the pure ideal state reference vector by subtracting the pure ideal state reference vector from the real-time observation vector; The residual vector component obtained after the subtraction operation, which does not contain the legitimate background signal reference component but retains the real abnormal signal, environmental random noise, and sensor thermal noise, is determined as the real difference residual.

5. The intelligent electromagnetic situational awareness system based on multi-sensor data fusion according to claim 1, characterized in that, The coupling decision module is configured to calculate the geometrical orientation consistency between the actual difference residual and the theoretical difference residual in the feature space by performing the following operation based on a vector similarity matching algorithm: Calculate the vector dot product of the actual difference residual and the theoretical difference residual to obtain the inner product scalar; The product of the Euclidean norm of the actual difference residual and the Euclidean norm of the theoretical difference residual is calculated to obtain the normalized scalar. Dividing the inner product scalar by the normalized scalar yields the cosine similarity score, which represents the angle between the two residual vectors in the feature space.

6. The intelligent electromagnetic situational awareness system based on multi-sensor data fusion according to claim 5, characterized in that, The coupling decision module is configured to output spectrum situational awareness results based on the consistency calculation results by performing the following operations: Call the preset decision threshold; If the similarity score is greater than the decision threshold, it is determined that there is an abnormal signal in the real-time observation vector that matches the target feature parameters, and an abnormal alarm and target type are output. If the similarity score is less than or equal to the decision threshold, it is determined that the difference in the real-time observation vector mainly comes from environmental random noise, and a normal situation report is output.

7. The intelligent electromagnetic situational awareness system based on multi-sensor data fusion according to claim 1, characterized in that, The system further includes a feedback correction module, which is configured to: When the coupling decision module outputs a normal situation report, the propagation model parameters in the environment reconstruction module are corrected for error minimization based on the statistical distribution characteristics of the actual difference residuals. When the coupling decision module outputs an abnormal alarm, the corresponding real difference residual is digitally marked and stored in the signal feature parameter library to update the target feature parameter.

8. The intelligent electromagnetic situational awareness system based on multi-sensor data fusion according to claim 1, characterized in that, The real-time observation vector includes: complex baseband IQ data or power spectral density data of multiple monitoring nodes collected simultaneously by the distributed sensor array and time-synchronized.

9. The intelligent electromagnetic situational awareness system based on multi-sensor data fusion according to claim 3, characterized in that, The signal characteristic parameter library includes: FM broadcast sideband characteristic parameters, navigation signal clock drift pattern parameters, and radar signal pulse repetition interval pattern parameters.

Images