AI-based passive radar multi-source electromagnetic positioning matching method and system

Abstract

The application provides an AI-based passive radar multi-source electromagnetic positioning matching method and system, relating to the technical field of artificial intelligence, first, a multi-source electromagnetic signal data set of a target area is acquired, the multi-source electromagnetic signal data set is an electromagnetic radiation signal synchronously captured by multiple signal collection devices in a preset time window, then, feature extraction is performed on the multi-source electromagnetic signal data set to obtain a signal feature set containing signal intensity distribution, frequency domain correlation and space-time correlation features, then, the signal feature set is fused based on a dynamic weight distribution strategy to generate a multi-dimensional positioning feature set, then, an AI positioning matching model is called to analyze positioning parameters to obtain the spatial position and signal propagation parameters of a target electromagnetic radiation source. Finally, positioning optimization instructions are generated according to the parameters and fed back to the signal collection device cluster for parameter calibration, so that the electromagnetic positioning accuracy and reliability can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, in particular to an AI-based passive radar multi-source electromagnetic positioning matching method and system. BACKGROUND

[0002] In today's electromagnetic environment monitoring and target positioning field, accurately and efficiently determining the position of electromagnetic radiation sources is of great significance and is widely used in many fields such as military reconnaissance, radio management, security monitoring, etc.

[0003] Traditional electromagnetic positioning methods mainly rely on a single signal feature or a small number of fixed features for positioning, such as only relying on signal strength or simple frequency characteristics. This approach, due to the single feature dimension, is difficult to fully reflect the complex characteristics of electromagnetic signals, and in complex electromagnetic environments, such as the presence of multipath effects, signal interference, etc., the positioning accuracy will be severely affected, and the position of the target electromagnetic radiation source cannot be accurately determined.

[0004] Moreover, most existing positioning methods lack effective feedback adjustment mechanisms, and after completing a positioning, they cannot real-time calibrate and optimize the parameters of the signal acquisition device according to the positioning results, cannot adapt to the dynamic changes of the electromagnetic environment, and are difficult to continuously maintain high positioning accuracy. SUMMARY

[0005] In view of the above-mentioned problems, in combination with the first aspect of the present application, the present application embodiment provides an AI-based passive radar multi-source electromagnetic positioning matching method, which comprises:

[0006] Obtaining a multi-source electromagnetic signal data set of a target area, the multi-source electromagnetic signal data set comprising electromagnetic radiation signals synchronously captured by a plurality of signal acquisition devices within a preset time window;

[0007] Performing signal feature extraction processing on the multi-source electromagnetic signal data set to obtain a signal feature set corresponding to each signal acquisition device, the signal feature set containing signal strength distribution characteristics, signal frequency domain correlation characteristics and signal space-time correlation characteristics;

[0008] Based on a preset dynamic weight allocation strategy, performing multi-source feature fusion processing on a plurality of signal feature sets to generate a multi-dimensional positioning feature set of the target electromagnetic radiation source;

[0009] Calling an AI positioning matching model to perform positioning parameter analysis processing on the multi-dimensional positioning feature set to generate spatial position parameters and signal propagation parameters of the target electromagnetic radiation source;

[0010] generate positioning optimization instructions according to the spatial position parameters and the signal propagation parameters, and feed back the positioning optimization instructions to the signal acquisition device cluster to trigger a parameter calibration operation.

[0011] In still another aspect, the embodiment of the present application also provides an AI-based passive radar multi-source electromagnetic positioning matching system, which comprises a processor and a machine readable storage medium, the machine readable storage medium is connected with the processor, the machine readable storage medium is used for storing programs, instructions or codes, and the processor is used for executing the programs, instructions or codes in the machine readable storage medium to realize the above-mentioned method.

[0012] Based on the above aspects, the present application can obtain a multi-source electromagnetic signal data set of a target area, collect electromagnetic radiation signals synchronously captured by multiple signal acquisition devices within a preset time window, perform signal feature extraction processing on the multi-source electromagnetic signal data set, obtain a signal feature set containing signal intensity distribution features, signal frequency domain correlation features and signal space-time correlation features, deeply mine the inherent characteristics of electromagnetic signals from multiple dimensions, and comprehensively reflect the essential features of signals. Based on a preset dynamic weight distribution strategy, multiple signal feature sets are subjected to multi-source feature fusion processing to generate a multi-dimensional positioning feature set of the target electromagnetic radiation source, the importance of different features is fully considered to dynamically change with actual conditions, the accuracy and comprehensiveness of positioning features are improved, and the limitations of single feature positioning are overcome. An AI positioning matching model is called to perform positioning parameter analysis processing on the multi-dimensional positioning feature set to generate spatial position parameters and signal propagation parameters of the target electromagnetic radiation source, the powerful learning and analysis capability of the AI positioning matching model is used to effectively improve the accuracy and efficiency of positioning parameter analysis, positioning optimization instructions are generated according to the spatial position parameters and the signal propagation parameters, and the positioning optimization instructions are fed back to the signal acquisition device cluster to trigger a parameter calibration operation, forming a closed-loop positioning optimization system, which can adjust the signal acquisition device according to the positioning result in real time, continuously improve the accuracy and reliability of subsequent positioning, and realize high-precision, intelligent and dynamic optimization of passive radar multi-source electromagnetic positioning. BRIEF DESCRIPTION OF DRAWINGS

[0013] Figure 1 is an execution flow schematic diagram of the AI-based passive radar multi-source electromagnetic positioning matching method provided by the embodiment of the present application.

[0014] Figure 2 is a schematic diagram of exemplary hardware and software components of the AI-based passive radar multi-source electromagnetic positioning matching system provided by the embodiment of the present application. DETAILED DESCRIPTION

[0015] The present application will be specifically described below in combination with the drawings of the specification, Figure 1is a flowchart of an AI-based passive radar multi-source electromagnetic positioning matching method provided by an embodiment of the present application. The AI-based passive radar multi-source electromagnetic positioning matching method will be described in detail below.

[0016] Step S110: Obtain a multi-source electromagnetic signal data set of a target area, which includes electromagnetic radiation signals synchronously captured by multiple signal acquisition devices within a preset time window.

[0017] In this embodiment, the target area can be a spatial area with a specific range, the boundaries of which can be defined by geographical coordinates or specific physical markers. In order to comprehensively obtain the electromagnetic signal conditions in the target area, multiple signal acquisition devices can be reasonably deployed in the target area. These signal acquisition devices have the ability to receive electromagnetic radiation signals, and their working principle is based on physical mechanisms such as electromagnetic induction, which can convert the received electromagnetic radiation into electrical signals for subsequent processing.

[0018] The distribution of signal acquisition devices needs to be planned according to factors such as the topography of the target area, the complexity of the electromagnetic environment, etc. For example, for areas with complex terrain and electromagnetic signals that are easily blocked, the number of signal acquisition devices may need to be increased, and a decentralized layout can be used to ensure that electromagnetic signals in every corner can be effectively captured. For areas with relatively flat terrain and relatively simple electromagnetic environment, the number of devices can be appropriately reduced and a relatively centralized layout can be used.

[0019] The preset time window is a predetermined time period, denoted by T, with a starting time t1 and an ending time t2, i.e. T = [t1, t2]. Within this time period, all signal acquisition devices will start the acquisition operation synchronously according to a unified clock signal. The purpose of synchronous acquisition is to ensure that the signals obtained by each device are consistent in time, so as to facilitate accurate signal analysis and processing later.

[0020] Each signal acquisition device will continuously receive electromagnetic radiation signals within the preset time window T and convert them into digital signals for storage. Let there be n signal acquisition devices, and the signal data collected by the i-th signal acquisition device within the preset time window T is denoted by Si(T), i = 1, 2, …, n. All signal data collected by the n signal acquisition devices are summarized together to form a multi-source electromagnetic signal data set S, i.e. S = {S1(T), S2(T), …, Sn(T)}.

[0021] Step S120: Perform signal feature extraction processing on the multi-source electromagnetic signal data set to obtain a signal feature set corresponding to each signal acquisition device, which includes signal intensity distribution features, signal frequency domain correlation features, and signal space-time correlation features.

[0022] After obtaining the multi-source electromagnetic signal data set S, in order to extract valuable information from these multi-source electromagnetic signal data, signal feature extraction processing is needed. Due to the different positions of different signal collection devices, the received electromagnetic signals will also be different, so it is necessary to perform feature extraction for each signal collection device respectively to obtain the corresponding signal feature set.

[0023] The signal intensity distribution feature is used to describe the intensity variation of electromagnetic signals at different times and different positions, which can reflect the intensity distribution law of electromagnetic radiation sources and the attenuation characteristics of signals in the propagation process. The signal frequency domain correlation feature mainly focuses on the characteristics of signals in the frequency domain, including the frequency components of signals, the relationship between frequency components, etc., which has important value for identifying different types of electromagnetic radiation sources and analyzing the modulation mode of signals. The signal space-time correlation feature combines the information of signals in time and space, and through the analysis of the propagation time difference and spatial position relationship between different signal collection devices, the position and propagation path of electromagnetic radiation sources can be more accurately determined.

[0024] Step S121: performing signal preprocessing operation on the electromagnetic radiation signal to obtain standardized electromagnetic signal data, the signal preprocessing operation including signal sampling rate conversion, background noise suppression and time-frequency alignment processing.

[0025] Before performing signal feature extraction, signal preprocessing operation needs to be performed on each signal Si(T) in the multi-source electromagnetic signal data set S to improve the quality and consistency of the signal, and to obtain standardized electromagnetic signal data.

[0026] Signal sampling rate conversion is the first step of preprocessing operation. Different signal collection devices may have different sampling rates, in order to facilitate subsequent processing, the sampling rates of all signals need to be unified. Let the original sampling rate of the i-th signal collection device be fi, and the target sampling rate be f0. The signal Si(T) can be converted by interpolation or decimation. For signals with a sampling rate higher than the target sampling rate, the method of decimation is used to reduce the number of sampling points according to a certain decimation factor; for signals with a sampling rate lower than the target sampling rate, the method of interpolation is used to insert appropriate sampling points in the signal to increase the sampling rate. After sampling rate conversion, the i-th signal becomes Si'(T), and its sampling rate is f0.

[0027] Background noise suppression is the second step of the preprocessing operation. In actual environments, the electromagnetic signals received by the signal acquisition device are often disturbed by various background noises, which can affect the subsequent feature extraction and analysis results. Therefore, a suitable filtering algorithm is needed to suppress the background noise of the signal Si'(T). For example, an adaptive filtering algorithm can be used to automatically adjust the parameters of the filter according to the statistical characteristics of the signal to achieve the best noise suppression effect. After background noise suppression, the ith signal becomes Si''(T), and the noise level is effectively reduced.

[0028] Time-frequency alignment processing is the third step of the preprocessing operation. Due to the time and frequency deviations that may exist during the installation and debugging of different signal acquisition devices, the signals collected may have inconsistencies in time and frequency. In order to eliminate such inconsistencies, the signal Si''(T) needs to be processed for time-frequency alignment. Correlation analysis and other methods can be used to determine the time and frequency deviations by finding the correlation between signals, and the signals are corrected accordingly. After time-frequency alignment processing, the standardized electromagnetic signal data Si'''(T) is obtained, i=1, 2, …, n.

[0029] Step S122: Perform time-frequency joint analysis processing on the standardized electromagnetic signal data, and extract signal intensity distribution features, including signal peak intensity sequence, average intensity fluctuation curve, and intensity attenuation trend parameters.

[0030] After obtaining the standardized electromagnetic signal data Si'''(T), perform time-frequency joint analysis processing to extract signal intensity distribution features. Time-frequency joint analysis can simultaneously show the changes of signals in time and frequency, which helps to better understand the characteristics of signals.

[0031] First, segment and window the standardized electromagnetic signal data Si'''(T) within a preset time interval Δt to generate multiple signal time-frequency analysis windows. Let the start time of the preset time interval Δt be tstart, and the end time be tend, and tstart and tend are within the preset time window T. Segment the standardized electromagnetic signal data Si'''(T) within the time interval [tstart, tend] according to Δt, and each segment of signal corresponds to a time-frequency analysis window. For the jth time-frequency analysis window, its time interval is [tj, tj+Δt], and the corresponding signal data is Sij(T).

[0032] Then, perform short-time Fourier transform processing on each signal time-frequency analysis window Sij(T) to obtain a time-frequency energy distribution map. Short-time Fourier transform is a commonly used time-frequency analysis method, which obtains the energy distribution of the signal at different times and frequencies by windowing the signal and then performing Fourier transform on the signal in each window. Let the time-frequency energy distribution map of the jth time-frequency analysis window after short-time Fourier transform be Eij(f, t), where f represents frequency and t represents time.

[0033] Extract the signal peak intensity sequence from the time-frequency energy distribution map Eij(f, t). The signal peak intensity sequence contains the maximum energy value in each analysis window and its corresponding timestamp. For the jth time-frequency analysis window, find the maximum energy value Emaxj and its corresponding timestamp tmaxj in the time-frequency energy distribution map Eij(f, t). Combine the maximum energy values and corresponding timestamps of all analysis windows to obtain the signal peak intensity sequence Pj = {(Emax1, tmax1), (Emax2, tmax2), …, (Emaxm, tmaxm)}, where m is the number of analysis windows.

[0034] Calculate the average energy value of all analysis windows and generate the average intensity fluctuation curve. The average intensity fluctuation curve contains the trend parameters of energy change over time. For the jth time-frequency analysis window, calculate its average energy value Eavgj, that is, integrate the time-frequency energy distribution map Eij(f, t) in frequency and time, and then divide by the area of the window. Arrange the average energy values of all analysis windows in chronological order to obtain the average intensity fluctuation curve Eavg(t).

[0035] Perform decay model fitting processing on the signal peak intensity sequence Pj to obtain the intensity decay trend parameters. The intensity decay trend parameters include the exponential decay coefficient and the linear decay slope. Least squares method or other methods can be used to fit the signal peak intensity sequence Pj, and select appropriate decay models such as exponential decay model or linear decay model. For the exponential decay model, its expression is E(t) = E0*exp(-αt), where E0 is the initial energy value and α is the exponential decay coefficient; for the linear decay model, its expression is E(t) = E0-βt, where β is the linear decay slope. The exponential decay coefficient α and the linear decay slope β obtained by fitting are the intensity decay trend parameters.

[0036] Finally, combine the signal peak intensity sequence Pj, the average intensity fluctuation curve Eavg(t), and the intensity decay trend parameters to splice the signal intensity distribution characteristics Ii1.

[0037] Step S123: performing multi-scale frequency domain decomposition processing on the standardized electromagnetic signal data to extract signal frequency domain correlation features, the signal frequency domain correlation features including a main frequency component amplitude spectrum, a harmonic component phase difference, and a frequency band energy proportion distribution.

[0038] The standardized electromagnetic signal data Si'''(T) is subjected to multi-scale frequency domain decomposition processing to extract signal frequency domain correlation features. Multi-scale frequency domain decomposition can decompose the signal into components of different frequency scales, which helps to analyze the frequency structure of the signal and the relationship between the frequency components.

[0039] A preset wavelet basis function is called to perform multi-level wavelet decomposition processing on the standardized electromagnetic signal data Si'''(T) to generate a plurality of frequency band component signals. Wavelet decomposition is a multi-resolution analysis method that decomposes a signal into wavelet coefficients of different scales and positions by selecting an appropriate wavelet basis function. Let the preset wavelet basis function be ψ(t), and after multi-level wavelet decomposition, a plurality of frequency band component signals Sik(T) are obtained, k=1, 2, …, K, where K is the number of decomposition levels.

[0040] Each frequency band component signal Sik(T) is subjected to energy normalization processing to calculate the frequency band energy proportion distribution. The frequency band energy proportion distribution includes the fundamental frequency component energy ratio, the second harmonic energy ratio, and the high-order harmonic energy ratio. For the kth frequency band component signal Sik(T), its energy Ek is calculated, i.e., the signal Sik(T) is squared and integrated. Then, the total energy of all frequency band component signals Etotal=∑Ek, k=1, 2, …, K is calculated. Finally, the energy proportion Rk=Ek / Etotal, k=1, 2, …, K of each frequency band component signal is calculated. The energy proportion corresponding to the fundamental frequency component is denoted as R1, the energy proportion corresponding to the second harmonic component is denoted as R2, and the energy proportions corresponding to the high-order harmonic components are denoted as R3, R4, …, RK, and thus the frequency band energy proportion distribution R={R1, R2, R3, …, RK} is obtained.

[0041] The fundamental frequency component signal Si1(T) is subjected to phase demodulation processing to extract the main frequency component amplitude spectrum. The main frequency component amplitude spectrum includes the spectral line distribution of the fundamental frequency amplitude change over time. Hilbert transform and other methods can be used to demodulate the phase of the fundamental frequency component signal Si1(T) to obtain its instantaneous amplitude and phase information. Then, the change of the instantaneous amplitude over time is plotted to obtain the main frequency component amplitude spectrum A1(t).

[0042] The phase difference calculation process is performed on the second harmonic component signal Si2(T) to obtain a harmonic component phase difference. The harmonic component phase difference contains a sequence of phase offsets of the second harmonic relative to the fundamental frequency. The phase offset of the second harmonic relative to the fundamental frequency can be calculated by comparing the phase information of the second harmonic component signal Si2(T) and the fundamental frequency component signal Si1(T). Let the phase offset of the second harmonic relative to the fundamental frequency at the t-th moment be φ(t), and the phase offsets at all moments are combined together to obtain the harmonic component phase difference Φ={φ(t1), φ(t2), …, φ(tn)}, where n is the number of sampling points of the signal.

[0043] Finally, the fundamental frequency component amplitude spectrum A1(t), the harmonic component phase difference Φ, and the frequency band energy proportion distribution R are combined as a signal frequency domain correlation feature Ii2.

[0044] Step S124: performing spatio-temporal correlation analysis processing on the standardized electromagnetic signal data captured by the plurality of signal acquisition devices to extract signal spatio-temporal correlation features, the signal spatio-temporal correlation features including a signal arrival time difference sequence, a propagation path correlation matrix, and a multipath effect interference coefficient.

[0045] The standardized electromagnetic signal data Si'''(T) captured by the plurality of signal acquisition devices i=1, 2, …, n is subjected to spatio-temporal correlation analysis processing to extract signal spatio-temporal correlation features. Spatio-temporal correlation analysis can combine the information of signals in time and space to more accurately determine the position and propagation path of the electromagnetic radiation source.

[0046] First, the signal propagation path difference parameters are calculated according to the spatial position coordinates of the signal acquisition devices. Let the spatial position coordinates of the i-th signal acquisition device be (xi, yi, zi), and the spatial position coordinates of the j-th signal acquisition device be (xj, yj, zj). Then the path length difference ΔLij=|Lij-Lji| of the signal propagating from the electromagnetic radiation source to the i-th signal acquisition device and the j-th signal acquisition device, where Lij is the path length of the signal propagating from the electromagnetic radiation source to the i-th signal acquisition device, and Lji is the path length of the signal propagating from the electromagnetic radiation source to the j-th signal acquisition device; the path angle difference Δθij can be obtained by calculating the included angle between the two signal acquisition devices and the electromagnetic radiation source. The path length difference and the path angle difference between all pairs of signal acquisition devices are combined together to obtain the propagation path difference parameter D={ΔLij, Δθij|i, j=1, 2, …, n; i≠j}.

[0047] The normalized electromagnetic signal data Si'''(T) captured by multiple signal collection devices are subjected to cross-correlation analysis to generate a signal arrival time difference sequence. Cross-correlation analysis is a commonly used signal processing method that determines the time delay between two signals by calculating their correlation. For the normalized electromagnetic signal data Si'''(T) and Sj'''(T) captured by the ith signal collection device and the jth signal collection device, the cross-correlation function Rij(τ) is calculated, where τ is the time delay. The time delay τij corresponding to the maximum value of the cross-correlation function Rij(τ) is found, which is the signal arrival time difference between the ith signal collection device and the jth signal collection device. Combining the signal arrival time differences between all pairs of signal collection devices together, the signal arrival time difference sequence ΔT = {ΔTij|i, j = 1, 2, …, n; i≠j} is obtained.

[0048] Based on the propagation path difference parameter D and the signal arrival time difference sequence ΔT, a propagation path correlation matrix is constructed. The propagation path correlation matrix contains the time delay consistency coefficients and path interference weights between each propagation path. Let the time delay consistency coefficient be Cij and the path interference weight be Wij. The time delay consistency coefficient and the path interference weight can be calculated according to the relationship between the propagation path difference parameter and the signal arrival time difference sequence. For example, the time delay consistency coefficient Cij can be calculated by comparing the consistency of the path length difference and the signal arrival time difference, and the path interference weight Wij can be determined according to factors such as the path angle difference and the signal strength. Combining the time delay consistency coefficients and the path interference weights between all pairs of signal collection devices together, the propagation path correlation matrix M = {Cij, Wij|i, j = 1, 2, …, n; i≠j} is obtained.

[0049] The received signals of each signal collection device are subjected to multipath effect detection processing to calculate a multipath effect interference coefficient. Multipath effect refers to the phenomenon that signals encounter obstacles during propagation and are reflected, refracted, etc., resulting in multiple signals arriving at the receiving end simultaneously, thus causing interference. Channel estimation and other methods can be used to detect the multipath effect of the received signals of each signal collection device, and the energy ratio of the direct wave to the reflected wave and the normalized time delay spread are calculated. Let the energy ratio of the direct wave to the reflected wave of the ith signal collection device be Ei and the normalized time delay spread be Di. Combining the energy ratio of the direct wave to the reflected wave and the normalized time delay spread of all signal collection devices together, the multipath effect interference coefficient N = {Ei, Di|i = 1, 2, …, n} is obtained.

[0050] Finally, the signal arrival time difference sequence ΔT, the propagation path correlation matrix M, and the multipath effect interference coefficient N are combined and spliced into the signal space correlation feature Ii3.

[0051] Step S125: combine the signal intensity distribution feature, the signal frequency domain correlation feature, and the signal space-time correlation feature to obtain the signal feature set.

[0052] The signal intensity distribution feature Ii1, the signal frequency domain correlation feature Ii2, and the signal space-time correlation feature Ii3 obtained in the previous step are combined and spliced together to obtain the signal feature set Ii corresponding to the i-th signal acquisition device, i.e., Ii={Ii1, Ii2, Ii3}, i=1, 2, …, n.

[0053] Step S130: based on a preset dynamic weight distribution strategy, performing multi-source feature fusion processing on the plurality of signal feature sets to generate a multi-dimensional positioning feature set of the target electromagnetic radiation source.

[0054] After obtaining the signal feature set Ii corresponding to each signal acquisition device, the signal feature sets need to be processed based on a preset dynamic weight distribution strategy to generate a multi-dimensional positioning feature set of the target electromagnetic radiation source. The dynamic weight distribution strategy can dynamically adjust the weight of each signal feature set according to factors such as signal quality and environmental conditions, thereby improving the accuracy and reliability of feature fusion.

[0055] Step S131: obtaining a signal quality evaluation parameter corresponding to each signal acquisition device, wherein the signal quality evaluation parameter includes a signal-to-noise ratio level, a signal stability index, and an environmental interference suppression coefficient.

[0056] In order to implement the dynamic weight distribution strategy, the signal quality evaluation parameter corresponding to each signal acquisition device needs to be obtained first. The signal-to-noise ratio level is used to measure the proportional relationship between the useful signal and the noise in the signal, which reflects the clarity of the signal. The signal-to-noise ratio level can be obtained by calculating the power spectral density of the signal and the power spectral density of the noise, and then taking their ratio. The signal stability index is used to measure the stability of the signal over time, which can be calculated by analyzing the changes in parameters such as amplitude and frequency of the signal. The environmental interference suppression coefficient is used to measure the suppression ability of the signal acquisition device to environmental interference, which can be determined according to factors such as the hardware characteristics of the signal acquisition device and the filtering algorithm used. Let the signal-to-noise ratio level of the i-th signal acquisition device be SNRi, the signal stability index be Sti, and the environmental interference suppression coefficient be EIi. Then the signal quality evaluation parameter Qi of the i-th signal acquisition device is {SNRi, Sti, EIi}, i=1, 2, …, n.

[0057] Step S132: calculating the initial fusion weight of each signal acquisition device according to the signal quality evaluation parameter, wherein the initial fusion weight is positively correlated with the signal-to-noise ratio level, the signal stability index, and the environmental interference suppression coefficient.

[0058] The initial fusion weight of each signal acquisition device is calculated according to the signal quality evaluation parameter Q i. Since the initial fusion weight is positively correlated with the signal-to-noise ratio level, the signal stability index, and the environmental interference suppression coefficient, a weighted summation method can be used to calculate the initial fusion weight. Let the weight of the signal-to-noise ratio level be w 1, the weight of the signal stability index be w 2, and the weight of the environmental interference suppression coefficient be w 3, and w 1 + w 2 + w 3 = 1. Then the initial fusion weight W i0 of the i th signal acquisition device can be obtained by the following calculation logic: W i0 = w 1 × S N R i + w 2 × S t i + w 3 × E I i. It should be noted here that the signal-to-noise ratio level, the signal stability index, and the environmental interference suppression coefficient need to be normalized before calculation to ensure that they are in the same dimension and value range, avoiding the addition of different dimension units. Normalization can use common linear normalization methods, for example, for a parameter x, the normalized parameter x' = (x - min(x)) / (max(x) - min(x)), where min(x) and max(x) are the minimum and maximum values of the parameter in all signal acquisition devices. Through such normalization, each parameter is mapped to the interval [0, 1], ensuring the rationality of the calculation of the initial fusion weight in dimension and value.

[0059] Step S133: performing dynamic adjustment processing on the initial fusion weight to generate a set of dynamic fusion weights, the dynamic adjustment processing including time decay compensation, spatial consistency constraint, and device state adaptation.

[0060] After obtaining the initial fusion weight W i0, it also needs to be dynamically adjusted to generate a set of dynamic fusion weights that are more in line with the actual situation. Dynamic adjustment processing mainly considers three factors, namely time decay compensation, spatial consistency constraint, and device state adaptation.

[0061] Step S1331: obtaining the historical weight adjustment record of the signal acquisition device, and extracting the time decay factor, which is calculated according to the device running time and the recent calibration time interval.

[0062] Firstly, time decay compensation is performed. The historical weight adjustment records of the signal acquisition device need to be obtained, and the time decay factor is extracted from these records. During the operation of the device, its performance may change over time, for example, the sensitivity of the sensor may decrease, resulting in poor quality of the acquired signal. Therefore, the time decay factor needs to be calculated according to the device running time and the recent calibration time interval. Let the device running time be T_run and the recent calibration time interval be T_calib, then the time decay factor γ can be calculated by a predefined function f(T_run, T_calib), that is, γ = f(T_run, T_calib), which can be designed according to the actual performance change law of the device, for example, it can be a linear function or an exponential function. Generally speaking, the longer the device running time, the longer the recent calibration time interval, and the smaller the value of the time decay factor, which means that the weight of the signal acquisition device needs to be attenuated to a greater extent.

[0063] Step S1332: Calculate the spatial consistency constraint coefficient according to the spatial distribution density between the signal acquisition devices, which is proportional to the distribution density.

[0064] Then consider the spatial consistency constraint. The spatial consistency constraint coefficient is calculated according to the spatial distribution density between the signal acquisition devices. The spatial distribution density reflects the tightness of the distribution of the signal acquisition devices in the target area. If the signal acquisition devices are distributed more densely, the correlation between the signals they acquire may be stronger, so these devices can be given relatively high weights; on the contrary, if the distribution is sparse, the weight can be appropriately reduced. Let the spatial distribution density of the signal acquisition devices be ρ, and the spatial consistency constraint coefficient β can be calculated by a linear function g(ρ), that is, β = g(ρ), and β is proportional to ρ. For example, when the spatial distribution density increases, the spatial consistency constraint coefficient also increases accordingly, which means that these signal acquisition devices will play a more important role in the feature fusion process.

[0065] Step S1333: Real-time monitoring of the working state parameters of the signal acquisition device, generating a device state adaptation coefficient, the working state parameters include power supply stability level and hardware aging index.

[0066] Then the device state adaptation is performed. The working state parameters of the signal acquisition device are monitored in real time, which include the power supply stability level and the hardware aging index. The power supply stability level reflects the stability of the power supply of the signal acquisition device. If the power supply is unstable, it may cause fluctuations in the collected signals, affecting the signal quality. The hardware aging index reflects the aging degree of the device hardware. Hardware aging may cause the performance of the device to decline. Let the power supply stability level be P_stab and the hardware aging index be H_aging. The device state adaptation coefficient a can be calculated by a comprehensive function h(P_stab, H_aging), that is, a = h(P_stab, H_aging). This function will comprehensively consider the influence of power supply stability level and hardware aging index on device performance. For example, when the power supply stability level is low or the hardware aging index is high, the value of the device state adaptation coefficient will decrease accordingly, thereby reducing the weight of the signal acquisition device.

[0067] Step S1334: Combine the time decay factor, the spatial consistency constraint coefficient, and the device state adaptation coefficient into a dynamic adjustment coefficient matrix.

[0068] The calculated time decay factor γ, spatial consistency constraint coefficient β, and device state adaptation coefficient a are combined into a dynamic adjustment coefficient matrix A. Matrix A is a diagonal matrix, and the elements on the diagonal are the products of γ, β, and a corresponding to each signal acquisition device, that is, Aii = γi × βi × ai, where i represents the i-th signal acquisition device. Such a matrix structure can conveniently adjust the initial fusion weight of each signal acquisition device.

[0069] Step S1335: Perform element-wise multiplication operation on the initial fusion weight using the dynamic adjustment coefficient matrix to generate a dynamic fusion weight set.

[0070] The initial fusion weight Wio is subjected to element-wise multiplication operation using the dynamic adjustment coefficient matrix A to generate a dynamic fusion weight set Wi. Specifically, for the i-th signal acquisition device, its dynamic fusion weight Wi = Aii × Wio. Through such dynamic adjustment, the weight of each signal acquisition device can be adjusted in real time according to factors such as time, space, and device state, thereby improving the accuracy and reliability of feature fusion.

[0071] Step S134: Based on the dynamic fusion weight set, the multiple signal feature sets are subjected to weighted fusion and splicing processing to obtain a multi-dimensional positioning feature set, which includes a fused signal strength feature, a fused frequency domain correlation feature, and a fused space-time correlation feature.

[0072] After obtaining the dynamic fusion weight set Wi, the multiple signal feature sets Ii are weighted and fused based on the set. Since the signal feature set Ii includes the signal intensity distribution feature Ii1, the signal frequency domain correlation feature Ii2 and the signal space-time correlation feature Ii3, the features in these three aspects need to be weighted and fused respectively.

[0073] For the signal intensity distribution feature, let the signal intensity distribution feature of the i-th signal acquisition device be Ii1, and the dynamic fusion weight be Wi. The fused signal intensity feature I1_fused can be obtained by weighting and splicing the signal intensity distribution features of all signal acquisition devices. The specific calculation logic is: first, multiply the signal intensity distribution feature Ii1 of each signal acquisition device by its corresponding dynamic fusion weight Wi to obtain the weighted signal intensity distribution feature Wi×Ii1. Then, splice all the weighted signal intensity distribution features in a set order to form the fused signal intensity feature I1_fused.

[0074] For the signal frequency domain correlation feature, similarly, let the signal frequency domain correlation feature of the i-th signal acquisition device be Ii2, and the dynamic fusion weight be Wi. The fused frequency domain correlation feature I2_fused can be obtained by multiplying the signal frequency domain correlation feature Ii2 of each signal acquisition device by its corresponding dynamic fusion weight Wi to obtain the weighted signal frequency domain correlation feature Wi×Ii2, and then splicing all the weighted signal frequency domain correlation features.

[0075] For the signal space-time correlation feature, let the signal space-time correlation feature of the i-th signal acquisition device be Ii3, and the dynamic fusion weight be Wi. The fused space-time correlation feature I3_fused can be obtained by multiplying the signal space-time correlation feature Ii3 of each signal acquisition device by its corresponding dynamic fusion weight Wi to obtain the weighted signal space-time correlation feature Wi×Ii3, and then splicing all the weighted signal space-time correlation features.

[0076] Combining the fused signal intensity feature I1_fused, the fused frequency domain correlation feature I2_fused and the fused space-time correlation feature I3_fused together, a multi-dimensional positioning feature set I_fused={I1_fused, I2_fused, I3_fused} is obtained.

[0077] Step S135: performing dimensionality reduction and redundancy removal processing on the multi-dimensional positioning feature set to generate an optimized multi-dimensional positioning feature set.

[0078] After obtaining the multi-dimensional positioning feature set I_fused, in order to reduce the dimension of the features and improve the efficiency of subsequent processing, it is necessary to perform dimension reduction and redundancy removal processing. Dimension reduction and redundancy removal processing can use methods such as principal component analysis (PCA). The basic idea of principal component analysis is to convert the original high-dimensional features into a set of new, mutually independent low-dimensional features through linear transformation, which can preserve as much information as possible of the original features.

[0079] Specifically, first, the multi-dimensional positioning feature set I_fused is regarded as a matrix, each row of the matrix represents a sample, and each column represents a feature. Then, the covariance matrix of the matrix is calculated, which reflects the correlation between the features. Next, the covariance matrix is decomposed to obtain the eigenvalues and corresponding eigenvectors. The eigenvalues represent the variance of each principal component, and the larger the variance, the more information the principal component contains. The eigenvectors corresponding to the first k eigenvalues with larger variances are selected, and the original feature matrix is multiplied by the matrix composed of the k eigenvectors to obtain the dimension-reduced feature matrix.

[0080] In the process of dimension reduction, some redundant features can also be removed. Redundant features refer to those that contribute little to the final positioning result or are highly correlated with other features. By calculating the correlation coefficients between the features, features with high correlation coefficients can be merged or removed, thereby further reducing the number of features.

[0081] After dimension reduction and redundancy removal processing, an optimized multi-dimensional positioning feature set I_fused_optimized is generated. This multi-dimensional positioning feature set not only has lower dimension, but also removes redundant information, and can be more effectively used for subsequent positioning parameter analysis processing.

[0082] Step S140: calling an AI positioning matching model to perform positioning parameter analysis processing on the multi-dimensional positioning feature set to generate the spatial position parameters and signal propagation parameters of the target electromagnetic radiation source.

[0083] After obtaining the optimized multi-dimensional positioning feature set I_fused_optimized, an AI positioning matching model is called to perform positioning parameter analysis processing to generate the spatial position parameters and signal propagation parameters of the target electromagnetic radiation source. The AI positioning matching model is a trained intelligent model that can learn the mapping relationship between the multi-dimensional positioning feature set and the spatial position parameters and signal propagation parameters of the target electromagnetic radiation source.

[0084] Step S141: inputting the multi-dimensional positioning feature set into a pre-trained AI positioning matching model for feature mapping processing to generate a high-dimensional abstract feature vector.

[0085] First, the optimized multi-dimensional positioning feature set I_fused_optimized is input into the pre-trained AI positioning matching model for feature mapping processing. The AI positioning matching model internally includes multiple neural network layers, such as the input layer, the hidden layer, and the output layer. During the feature mapping process, the multi-dimensional positioning feature set I_fused_optimized enters the model from the input layer, undergoes a series of nonlinear transformations in the hidden layer, and finally generates a high-dimensional abstract feature vector V in the output layer. The high-dimensional abstract feature vector V is an abstract representation of the multi-dimensional positioning feature set I_fused_optimized, which contains more potential information and can more accurately reflect the characteristics of the target electromagnetic radiation source.

[0086] Step S142: Perform regional grid division processing on the high-dimensional abstract feature vector to generate a candidate positioning grid set, each candidate positioning grid containing a spatial coordinate range and a signal propagation parameter range.

[0087] After obtaining the high-dimensional abstract feature vector V, perform regional grid division processing on it. Regional grid division is to divide the target region into multiple small grids, each grid corresponding to a candidate positioning region. Specifically, according to the boundary and resolution requirements of the target region, the target region is divided into multiple small cubic grids in space, each grid having a specific spatial coordinate range. At the same time, combined with the characteristics of signal propagation and historical data, set the corresponding signal propagation parameter range for each grid, such as signal propagation speed, path loss coefficient, etc.

[0088] Let the coordinate range of the target region in the three-dimensional space be [x_min, x_max], [y_min, y_max], and [z_min, z_max], and the resolution of the grid be Δx, Δy, and Δz. Then the target region can be divided into Nx=(x_max-x_min) / Δx, Ny=(y_max-y_min) / Δy, and Nz=(z_max-z_min) / Δz grids, and the spatial coordinate range of each grid can be represented as [x_i, x_i+Δx], [y_j, y_j+Δy], and [z_k, z_k+Δz], where i=0, 1, …, Nx-1, j=0, 1, …, Ny-1, k=0, 1, …, Nz-1. For each grid, according to historical data and signal propagation model, set its signal propagation parameter range, such as signal propagation speed range [v_min, v_max] and path loss coefficient range [L_min, L_max]. Combining all the grids and their corresponding spatial coordinate ranges and signal propagation parameter ranges, we get the candidate positioning grid set G={G1, G2, …, GN}, where N=Nx×Ny×Nz.

[0089] Step S143: Calculate the matching score between each candidate positioning grid and the high-dimensional abstract feature vector, which is generated based on the feature similarity algorithm and the propagation model constraint condition.

[0090] Next, the matching score between each candidate positioning grid and the high-dimensional abstract feature vector V is calculated. The matching score is used to measure the matching degree of each candidate positioning grid with the target electromagnetic radiation source. The higher the score, the more likely the grid is the location of the target electromagnetic radiation source. The matching score is generated based on the feature similarity algorithm and the propagation model constraint condition.

[0091] For example, step S1431: perform block feature extraction processing on the high-dimensional abstract feature vector to generate signal strength feature blocks, frequency domain correlation feature blocks, and space-time correlation feature blocks.

[0092] First, perform block feature extraction processing on the high-dimensional abstract feature vector V. According to the properties and roles of different features in the high-dimensional abstract feature vector V, it is divided into signal strength feature blocks V1, frequency domain correlation feature blocks V2, and space-time correlation feature blocks V3. For example, according to the dimensions of the feature vector and the correlation between the features, the features in the first part of the dimensions are taken as the signal strength feature blocks V1, the features in the middle part of the dimensions are taken as the frequency domain correlation feature blocks V2, and the features in the last part of the dimensions are taken as the space-time correlation feature blocks V3.

[0093] Step S1432: Perform similarity calculation between the preset feature template of each candidate positioning grid and the signal strength feature blocks to generate signal strength matching degrees.

[0094] For each candidate positioning grid Gi, it has a preset feature template Ti. Perform similarity calculation between the signal strength feature part in the preset feature template Ti of the candidate positioning grid Gi and the signal strength feature blocks V1 to generate the signal strength matching degree Si1. The similarity calculation can use common similarity measurement methods, such as Euclidean distance, cosine similarity, etc. Taking cosine similarity as an example, the signal strength matching degree Si1 can be obtained by calculating the cosine value of the vector angle between the signal strength feature blocks V1 and the preset signal strength feature part of the candidate positioning grid Gi. The closer the cosine value is to 1, the higher the similarity between the two, and the higher the signal strength matching degree Si1.

[0095] Step S1433: Perform compliance comparison between the frequency domain constraint condition of each candidate positioning grid and the frequency domain correlation feature blocks to generate a frequency domain compliance score.

[0096] The frequency domain constraint condition Fi of each candidate positioning grid Gi is compared with the frequency domain correlation feature block V2 for compliance, to generate a frequency domain compliance score Si2. The frequency domain constraint condition Fi is a series of constraints on frequency ranges and frequency relationships set according to the location of the candidate positioning grid and the electromagnetic environment. For example, a certain candidate positioning grid can stipulate that the main frequency of the signal must be within a certain specific frequency range, and the phase difference between harmonic components must satisfy a certain relationship. The frequency information in the frequency domain correlation feature block V2 is compared with the frequency domain constraint condition Fi, and a higher score is given if the constraint condition is met, or a lower score is given if the constraint condition is partially met or not met.

[0097] Step S1434: The spatio-temporal constraint condition of each candidate positioning grid is logically checked with the spatio-temporal correlation feature block, to generate a spatio-temporal checking score.

[0098] The spatio-temporal constraint condition Si of each candidate positioning grid Gi is logically checked with the spatio-temporal correlation feature block V3, to generate a spatio-temporal checking score Si3. The spatio-temporal constraint condition Si is a series of constraints on time and space relationships set according to the spatial location of the candidate positioning grid and the signal propagation characteristics. For example, a certain candidate positioning grid can stipulate that the time difference of signal arrival at different signal collection devices must be within a certain specific range, and the propagation path of the signal must comply with a certain geometric relationship. The time and space information in the spatio-temporal correlation feature block V3 is logically checked with the spatio-temporal constraint condition Si, and a higher score is given if the constraint condition is met, or a lower score is given if the constraint condition is partially met or not met.

[0099] Step S1435: The signal strength matching degree, the frequency domain compliance score, and the spatio-temporal checking score are weighted and summed, to generate the matching degree score.

[0100] The signal strength matching degree Si1, the frequency domain compliance score Si2, and the spatio-temporal checking score Si3 are weighted and summed, to generate the matching degree score Si. Let the weight of the signal strength matching degree be w_s1, the weight of the frequency domain compliance score be w_s2, and the weight of the spatio-temporal checking score be w_s3, and w_s1+w_s2+w_s3=1. Then the matching degree score Si=w_s1×Si1+w_s2×Si2+w_s3×Si3.

[0101] Step S1436: The matching degree score is error compensated according to historical positioning error data, to generate an optimized matching degree score.

[0102] The matching degree score Si is error-compensated according to historical positioning error data to generate an optimized matching degree score Si_optimized. The historical positioning error data records the actual positioning error of each candidate positioning grid in the past positioning process. The matching degree score of each candidate positioning grid can be adjusted according to the error data. For example, if a candidate positioning grid often has a large error in historical positioning, its score in the current matching degree score can be appropriately reduced; conversely, if the positioning error of a candidate positioning grid is small, its score can be appropriately increased. The error compensation process can be implemented using a compensation function f_error(Si, E_history), where Si is the un-compensated matching degree score, and E_history is the historical positioning error data. After error compensation, the optimized matching degree score Si_optimized is obtained.

[0103] Step S144: Select the candidate positioning grid with the highest matching degree score as the target positioning grid, and extract the weighted average coordinates within the spatial coordinate range of the target positioning grid as the spatial position parameter, which is calculated according to the signal propagation parameter and the historical positioning data within the target positioning grid.

[0104] After obtaining the optimized matching degree score Si_optimized of each candidate positioning grid, the candidate positioning grid with the highest matching degree score is selected as the target positioning grid G_target. The target positioning grid G_target has the highest matching degree, indicating that it is most likely the position of the target electromagnetic radiation source.

[0105] The weighted average coordinates within the spatial coordinate range of the target positioning grid G_target are extracted as the spatial position parameter. The calculation of the weighted average coordinates needs to consider the signal propagation parameter and the historical positioning data within the target positioning grid. Let the spatial coordinate range of the target positioning grid G_target be [x_target_min, x_target_max], [y_target_min, y_target_max] and [z_target_min, z_target_max], and the coordinates of each positioning point in the historical positioning data be (x_j, y_j, z_j), with the corresponding signal propagation parameter being p_j and the weight being w_j. First, the relationship between the weight w_j and the signal propagation parameter p_j needs to be determined. Generally, the signal propagation parameter can reflect the reliability or accuracy of the signal at the positioning point, for example, the reliability of the positioning point is higher when the signal strength is larger and the propagation path is clearer, and the weight should also be larger. The weight w_j can be calculated through a pre-defined function, such as w_j = g(p_j), which maps the signal propagation parameter p_j to the corresponding weight value.

[0106] Next, the weighted average coordinate is calculated. For the x coordinate, the weighted average coordinate x_avg can be obtained by multiplying the x coordinate value xj of each positioning point by its corresponding weight wj, then adding all the products, and dividing by the sum of all weights. That is, first calculate the sum of all xj*wj sum_x = ∑(xj*wj) (the summation is performed on all historical positioning points within the target positioning grid), and the sum of all weights sum_w = ∑wj, then x_avg = sum_x / sum_w.

[0107] Similarly, for the y coordinate, first calculate the sum of all yj*wj sum_y = ∑(yj*wj), y_avg = sum_y / sum_w. For the z coordinate, first calculate the sum of all zj*wj sum_z = ∑(zj*wj), z_avg = sum_z / sum_w.

[0108] In this way, the weighted average coordinate (x_avg, y_avg, z_avg) within the spatial coordinate range of the target positioning grid is obtained, which is taken as the spatial position parameter of the target electromagnetic radiation source.

[0109] Step S145: Extract the median value of the signal propagation parameter range of the target positioning grid as the signal propagation parameter.

[0110] The target positioning grid G_target has a pre-set signal propagation parameter range, such as a signal propagation speed range [v_min, v_max], a path loss coefficient range [L_min, L_max], etc. For each signal propagation parameter, the median value of its range is extracted as the final signal propagation parameter.

[0111] For the signal propagation speed, the median value v = (v_min + v_max) / 2. For the path loss coefficient, the median value L = (L_min + L_max) / 2. Similarly, for other signal propagation parameters involved in the target positioning grid, the median value is calculated in this way, and these median values are combined to obtain the signal propagation parameters of the target electromagnetic radiation source.

[0112] Step S150: Generate positioning optimization instructions according to the spatial position parameter and the signal propagation parameter, and feed back the positioning optimization instructions to the signal acquisition device cluster to trigger parameter calibration operations.

[0113] After obtaining the spatial position parameter and the signal propagation parameter of the target electromagnetic radiation source, positioning optimization instructions need to be generated according to these information, and then the instructions are fed back to the signal acquisition device cluster to trigger the parameter calibration operation of the signal acquisition device, so as to improve the accuracy of subsequent positioning.

[0114] Step S151: Comparing the spatial position parameter with the geographical position database of the signal acquisition device to generate a positioning deviation vector.

[0115] First, the obtained spatial position parameter (x_avg, y_avg, z_avg) of the target electromagnetic radiation source is compared with the geographical position database of the signal acquisition device. The geographical position database of the signal acquisition device records the accurate geographical position information of each signal acquisition device, and the geographical position coordinates of the ith signal acquisition device are (xi, yi, zi).

[0116] The deviation between the spatial position parameter of the target electromagnetic radiation source and the geographical position coordinates of each signal acquisition device is calculated. For the x coordinate, the deviation Δxi=x_avg-xi; for the y coordinate, the deviation Δyi=y_avg-yi; and for the z coordinate, the deviation Δzi=z_avg-zi. Combining these deviations together, the positioning deviation vector ΔVi=(Δxi, Δyi, Δzi) corresponding to the ith signal acquisition device is obtained. Such calculation is performed for all signal acquisition devices to obtain the positioning deviation vector set {ΔV1, ΔV2, …, ΔVn} of all signal acquisition devices.

[0117] Step S152: Calculating device calibration parameters including antenna pointing adjustment angle, receiving sensitivity compensation value, and sampling rate optimization coefficient according to the positioning deviation vector.

[0118] The device calibration parameters of each signal acquisition device are calculated according to the positioning deviation vector set {ΔV1, ΔV2, …, ΔVn}. The device calibration parameters include antenna pointing adjustment angle, receiving sensitivity compensation value, and sampling rate optimization coefficient.

[0119] For the antenna pointing adjustment angle, it is closely related to the positioning deviation vector. Through the positioning deviation vector, the azimuth and angle information of the target electromagnetic radiation source relative to the signal acquisition device can be determined. For example, the horizontal angle deviation can be calculated according to the x and y coordinate deviations in the positioning deviation vector, and the vertical angle deviation can be calculated according to the z coordinate deviation and the horizontal direction information. Let the horizontal antenna pointing adjustment angle of the ith signal acquisition device be θi_hori, and the vertical antenna pointing adjustment angle be θi_vert, which can be calculated according to the positioning deviation vector ΔVi through trigonometric functions and other methods.

[0120] The reception sensitivity compensation value is related to the size of the positioning deviation vector. Generally speaking, a larger positioning deviation may mean that the signal quality received by the signal collection device is poor, and the reception sensitivity needs to be improved. The reception sensitivity compensation value Ci_sens can be calculated by a predefined function h(||ΔVi||), where ||ΔVi|| represents the length of the positioning deviation vector ΔVi, and the function maps the length of the positioning deviation vector to the corresponding reception sensitivity compensation value.

[0121] The sampling rate optimization coefficient can also be adjusted according to the positioning deviation vector. If the positioning deviation is large, it may be necessary to increase the sampling rate to obtain more detailed signal information. The sampling rate optimization coefficient Ci_rate can be calculated by a function k(||ΔVi||), which maps the length of the positioning deviation vector to the corresponding sampling rate optimization coefficient.

[0122] The antenna pointing adjustment angle θi_hori, θi_vert, the reception sensitivity compensation value Ci_sens, and the sampling rate optimization coefficient Ci_rate of each signal collection device are combined together to obtain the device calibration parameter Pi of the i-th signal collection device.

[0123] Step S153: Match the signal propagation parameters with the preset propagation model library to generate a model update instruction, which includes a path loss coefficient correction amount and a multipath effect suppression strategy.

[0124] The signal propagation parameters of the target electromagnetic radiation source are matched with the preset propagation model library. The preset propagation model library contains signal propagation models in various scenarios, and each model has corresponding parameters such as path loss coefficient and multipath effect processing method.

[0125] For the path loss coefficient, the path loss coefficient L of the target electromagnetic radiation source is compared with the path loss coefficients of each model in the propagation model library. The closest propagation model to L is found, and then the path loss coefficient correction amount ΔL is calculated according to the difference between them. For example, if the path loss coefficient in the propagation model is L_model, then the path loss coefficient correction amount ΔL = L - L_model.

[0126] For the multipath effect suppression strategy, the most suitable multipath effect suppression strategy is selected according to the signal propagation parameters of the target electromagnetic radiation source and the information in the propagation model library. Different propagation models may correspond to different multipath effect suppression methods, such as adaptive filtering, beamforming, etc. According to the matching result, the specific multipath effect suppression strategy is determined and used as part of the model update instruction.

[0127] The path loss coefficient correction amount AL and the multipath effect suppression strategy are combined together to obtain a model update instruction P_model.

[0128] Step S154: Combine the device calibration parameter and the model update instruction into a positioning optimization instruction.

[0129] The device calibration parameter Pi of each signal acquisition device and the model update instruction P_model are combined together to form a positioning optimization instruction P_optimize. For the i-th signal acquisition device, its positioning optimization instruction P_optimize_i = {Pi, P_model}. The positioning optimization instructions of all signal acquisition devices are summarized to obtain a complete set of positioning optimization instructions {P_optimize_1, P_optimize_2, …, P_optimize_n}.

[0130] Step S155: Perform executability verification processing on the positioning optimization instruction according to the hardware constraint conditions of the signal acquisition device, to generate a verified positioning optimization instruction.

[0131] Before feeding back the positioning optimization instruction to the signal acquisition device cluster, it is necessary to perform executability verification processing on the positioning optimization instruction according to the hardware constraint conditions of the signal acquisition device. The hardware constraint conditions of the signal acquisition device include the maximum adjustment angle of the antenna, the adjustable range of the receiving sensitivity, the maximum and minimum limits of the sampling rate, etc.

[0132] For the antenna pointing adjustment angle, check whether θi_hori and θi_vert are within the maximum adjustment angle range of the antenna. If it is out of range, adjust it to the maximum adjustment angle value. For the receiving sensitivity compensation value Ci_sens, check whether it is within the adjustable range of the receiving sensitivity. If it is out of range, adjust it to the boundary value of the adjustable range. For the sampling rate optimization coefficient Ci_rate, check whether it is within the maximum and minimum limit range of the sampling rate. If it is out of range, adjust it to the value within the limit range.

[0133] For the path loss coefficient correction amount AL and the multipath effect suppression strategy in the model update instruction, it is also necessary to check whether it matches the hardware capability of the signal acquisition device. For example, some multipath effect suppression strategies may require specific hardware support, if the signal acquisition device does not have this hardware, it is necessary to select other appropriate strategies.

[0134] After the executability verification process, the positioning optimization instructions are adjusted accordingly to generate a set of verified positioning optimization instructions {P_optimize_1_verified, P_optimize_2_verified, …, P_optimize_n_verified}. These verified positioning optimization instructions are fed back to the signal acquisition device cluster, which will perform parameter calibration operations according to these instructions, thereby improving the accuracy and reliability of subsequent positioning of the target electromagnetic radiation source.

[0135] Figure 2 An exemplary hardware and software components of the AI-based passive radar multi-source electromagnetic positioning matching system 100 that can implement the idea of the present application are shown. For example, the processor 120 can be used in the AI-based passive radar multi-source electromagnetic positioning matching system 100 and used to perform the functions in the present application.

[0136] The AI-based passive radar multi-source electromagnetic positioning matching system 100 can be a general-purpose server or a special-purpose server, both of which can be used to implement the AI-based passive radar multi-source electromagnetic positioning matching method of the present application. Although only one server is shown in the present application, for the sake of convenience, the functions described in the present application can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0137] For example, the AI-based passive radar multi-source electromagnetic positioning matching system 100 can include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and different forms of storage media 140, such as a disk, a ROM, or a RAM, or any combination thereof. Exemplarily, the AI-based passive radar multi-source electromagnetic positioning matching system 100 can also include program instructions stored in a ROM, a RAM, or other types of non-transitory storage media, or any combination thereof. The method of the present application can be implemented according to these program instructions. The AI-based passive radar multi-source electromagnetic positioning matching system 100 also includes an I / O interface 150 between the computer and other input / output devices.

[0138] For the convenience of description, only one processor is described in the AI-based passive radar multi-source electromagnetic positioning matching system 100. However, it should be noted that the AI-based passive radar multi-source electromagnetic positioning matching system 100 in the present application can also include multiple processors, so the steps performed by one processor described in the present application can also be jointly performed or separately performed by multiple processors. For example, if the processor of the AI-based passive radar multi-source electromagnetic positioning matching system 100 performs steps A and B, it should be understood that steps A and B can also be jointly performed by two different processors or separately performed in one processor. For example, a first processor performs step A, a second processor performs step B, or a first processor and a second processor jointly perform steps A and B.

[0139] In addition, the embodiment of the present application also provides a readable storage medium, wherein computer executable instructions are preset, and when a processor executes the computer executable instructions, the AI-based passive radar multi-source electromagnetic positioning matching method is realized.

[0140] It should be noted that, in order to simplify the description of the present application and to help the understanding of one or more embodiments of the present application, in the foregoing description of the embodiments of the present application, various features are sometimes combined into one embodiment, figure or description thereof.

Claims

1. A passive radar multi-source electromagnetic positioning matching method based on AI, characterized in that: The method comprises: Acquire a multi-source electromagnetic signal data set of a target area, wherein the multi-source electromagnetic signal data set includes electromagnetic radiation signals synchronously captured by multiple signal acquisition devices within a preset time window; Performing signal feature extraction processing on the multi-source electromagnetic signal data set to obtain a signal feature set corresponding to each signal acquisition device, wherein the signal feature set includes a signal intensity distribution feature, a signal frequency domain correlation feature, and a signal space-time correlation feature; Based on a preset dynamic weight allocation strategy, a multi-source feature fusion process is performed on the plurality of signal feature sets to generate a multi-dimensional positioning feature set of the target electromagnetic radiation source; Calling the AI ​​positioning matching model to perform positioning parameter analysis on the multi-dimensional positioning feature set to generate the spatial position parameters and signal propagation parameters of the target electromagnetic radiation source; generating a positioning optimization instruction according to the spatial position parameter and the signal propagation parameter, and feeding the positioning optimization instruction back to the signal acquisition device cluster to trigger a parameter calibration operation; Generating a positioning optimization instruction according to the spatial position parameter and the signal propagation parameter includes: Comparing the spatial position parameter with the geographical location database of the signal acquisition device to generate a positioning deviation vector; Calculating device calibration parameters based on the positioning deviation vector, wherein the device calibration parameters include antenna pointing adjustment angle, receiving sensitivity compensation value and sampling rate optimization coefficient; Matching the signal propagation parameters with a preset propagation model library to generate a model update instruction, wherein the model update instruction includes a path loss coefficient correction amount and a multipath effect suppression strategy; combining the device calibration parameters and the model update instructions into positioning optimization instructions; The positioning optimization instruction is subjected to executable verification processing according to the hardware constraint conditions of the signal acquisition device to generate a verified positioning optimization instruction.

2. The AI-based passive radar multi-source electromagnetic positioning matching method according to claim 1 is characterized in that: The performing signal feature extraction processing on the multi-source electromagnetic signal data set to obtain a signal feature set corresponding to each signal acquisition device includes: Performing a signal preprocessing operation on the electromagnetic radiation signal to obtain standardized electromagnetic signal data, wherein the signal preprocessing operation includes signal sampling rate conversion, background noise suppression, and time-frequency alignment processing; Performing time-frequency joint analysis on the standardized electromagnetic signal data to extract signal strength distribution characteristics, wherein the signal strength distribution characteristics include a signal peak strength sequence, an average strength fluctuation curve, and an strength attenuation trend parameter; Performing multi-scale frequency domain decomposition processing on the standardized electromagnetic signal data to extract signal frequency domain correlation features, wherein the signal frequency domain correlation features include the amplitude spectrum of the main frequency component, the phase difference of the harmonic component, and the frequency band energy ratio distribution; Performing spatiotemporal correlation analysis on the standardized electromagnetic signal data captured by multiple signal acquisition devices to extract spatiotemporal correlation features of the signals, wherein the spatiotemporal correlation features of the signals include a signal arrival time difference sequence, a propagation path correlation matrix, and a multipath effect interference coefficient; The signal strength distribution feature, the signal frequency domain correlation feature and the signal space-time correlation feature are combined and spliced ​​into the signal feature set.

3. The AI-based passive radar multi-source electromagnetic positioning matching method according to claim 2, characterized in that: The performing time-frequency joint analysis on the standardized electromagnetic signal data to extract signal strength distribution characteristics includes: Performing segmented windowing processing on the standardized electromagnetic signal data within a preset time interval to generate multiple signal time-frequency analysis windows; Perform short-time Fourier transform processing on each signal time-frequency analysis window to obtain a time-frequency energy distribution map; Extracting a signal peak intensity sequence from the time-frequency energy distribution diagram, wherein the signal peak intensity sequence includes a maximum energy value in each analysis window and a corresponding timestamp; Calculating average energy values ​​of all analysis windows and generating an average intensity fluctuation curve, wherein the average intensity fluctuation curve includes a trend parameter of energy variation over time; Performing attenuation model fitting processing on the signal peak intensity sequence to obtain intensity attenuation trend parameters, wherein the intensity attenuation trend parameters include an exponential attenuation coefficient and a linear attenuation slope; The signal peak intensity sequence, the average intensity fluctuation curve and the intensity attenuation trend parameter are combined and spliced ​​into the signal intensity distribution feature.

4. The AI-based passive radar multi-source electromagnetic positioning matching method according to claim 2, characterized in that: The performing multi-scale frequency domain decomposition processing on the standardized electromagnetic signal data to extract signal frequency domain correlation features includes: Calling a preset wavelet basis function to perform multi-level wavelet decomposition processing on the standardized electromagnetic signal data to generate multiple frequency band component signals; Performing energy normalization processing on each frequency band component signal and calculating the energy proportion distribution of each frequency band, wherein the frequency band energy proportion distribution includes the fundamental frequency component energy ratio, the second harmonic energy ratio and the higher harmonic energy ratio; Performing phase demodulation processing on the fundamental frequency component signal to extract the amplitude spectrum of the main frequency component, wherein the amplitude spectrum of the main frequency component includes the spectrum line distribution of the fundamental frequency amplitude changing with time; Performing phase difference calculation processing on the second harmonic component signal to obtain a harmonic component phase difference, wherein the harmonic component phase difference includes a phase offset sequence of the second harmonic relative to the fundamental frequency; The main frequency component amplitude spectrum, the harmonic component phase difference and the frequency band energy proportion distribution are combined into the signal frequency domain correlation feature.

5. The AI-based passive radar multi-source electromagnetic positioning matching method according to claim 2, characterized in that: The performing of spatiotemporal correlation analysis on the standardized electromagnetic signal data captured by the plurality of signal acquisition devices to extract spatiotemporal correlation features of the signals includes: Calculating signal propagation path difference parameters according to the spatial position coordinates of the signal acquisition device, wherein the propagation path difference parameters include path length difference and path angle difference; Performing cross-correlation analysis on the standardized electromagnetic signal data captured by multiple signal acquisition devices to generate a signal arrival time difference sequence, wherein the signal arrival time difference sequence includes time delays between any two signal acquisition devices; Constructing a propagation path correlation matrix based on the propagation path difference parameter and the signal arrival time difference sequence, wherein the propagation path correlation matrix includes a delay consistency coefficient and a path interference weight between each propagation path; Performing multipath effect detection processing on the received signal of each signal acquisition device and calculating the multipath effect interference coefficient, wherein the multipath effect interference coefficient includes the energy ratio of the direct wave and the reflected wave and the normalized delay spread; The signal arrival time difference sequence, the propagation path correlation matrix and the multipath effect interference coefficient are combined and spliced ​​into the signal spatiotemporal correlation feature.

6. The AI-based passive radar multi-source electromagnetic positioning matching method according to claim 1, characterized in that: The method of performing multi-source feature fusion processing on the plurality of signal feature sets based on a preset dynamic weight allocation strategy to generate a multi-dimensional positioning feature set of the target electromagnetic radiation source includes: Acquire signal quality evaluation parameters corresponding to each signal acquisition device, wherein the signal quality evaluation parameters include a signal-to-noise ratio level, a signal stability index, and an environmental interference suppression coefficient; Calculating an initial fusion weight of each signal acquisition device according to the signal quality evaluation parameter, wherein the initial fusion weight is positively correlated with the signal-to-noise ratio level, the signal stability index, and the environmental interference suppression coefficient; Dynamically adjusting the initial fusion weights to generate a dynamic fusion weight set, wherein the dynamic adjustment includes time decay compensation, spatial consistency constraint, and device state adaptation; Performing weighted fusion and splicing processing on the plurality of signal feature sets based on the dynamic fusion weight set to obtain a multi-dimensional positioning feature set, wherein the multi-dimensional positioning feature set includes a fused signal strength feature, a fused frequency domain correlation feature, and a fused spatiotemporal correlation feature; Performing dimensionality reduction and redundancy removal processing on the multi-dimensional positioning feature set to generate an optimized multi-dimensional positioning feature set.

7. The AI-based passive radar multi-source electromagnetic positioning matching method according to claim 6, characterized in that: The dynamically adjusting the initial fusion weights to generate a dynamic fusion weight set includes: Obtaining historical weight adjustment records of the signal acquisition device and extracting a time decay factor, the time decay factor being calculated based on the device's operating time and the time interval between the most recent calibration; Calculating a spatial consistency constraint coefficient according to the spatial distribution density between the signal acquisition devices, wherein the spatial consistency constraint coefficient is proportional to the distribution density; Real-time monitoring of the operating status parameters of the signal acquisition device to generate a device status adaptation coefficient, wherein the operating status parameters include the power supply stability level and the hardware aging index; Combining the time attenuation factor, the spatial consistency constraint coefficient, and the device state adaptation coefficient into a dynamic adjustment coefficient matrix; The dynamic adjustment coefficient matrix is ​​used to perform an element-by-element product operation on the initial fusion weights to generate a dynamic fusion weight set.

8. The AI-based passive radar multi-source electromagnetic positioning matching method according to claim 1, characterized in that: The calling of the AI ​​positioning matching model to perform positioning parameter parsing processing on the multi-dimensional positioning feature set to generate spatial position parameters and signal propagation parameters of the target electromagnetic radiation source includes: Inputting the multi-dimensional positioning feature set into a pre-trained AI positioning matching model for feature mapping processing to generate a high-dimensional abstract feature vector; Performing regional grid division processing on the high-dimensional abstract feature vector to generate a set of candidate positioning grids, each candidate positioning grid including a spatial coordinate range and a signal propagation parameter range; Calculating a matching score between each candidate positioning grid and the high-dimensional abstract feature vector, wherein the matching score is generated based on a feature similarity algorithm and propagation model constraints; Select the candidate positioning grid with the highest matching score as the target positioning grid, extract the weighted average coordinates within the spatial coordinate range of the target positioning grid as the spatial position parameter, and calculate the weighted average coordinates based on the signal propagation parameters and the historical positioning data in the target positioning grid; The median value of the signal propagation parameter range of the target positioning grid is extracted as the signal propagation parameter.

9. An AI-based passive radar multi-source electromagnetic positioning matching system, characterized in that: It includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the AI-based passive radar multi-source electromagnetic positioning matching method described in any one of claims 1 to 8.

Images