A GNSS interference analysis method and system based on big data

Abstract

This invention relates to the field of GNSS interference analysis technology, and particularly to a GNSS interference analysis method and system based on big data. The method includes the following steps: deploying multi-source signal acquisition sensors to collect GNSS signals and performing standardization processing to generate a multimodal standard dataset; performing non-uniform sampling alignment on the multimodal standard dataset to obtain a GNSS synchronized spatiotemporal data stream; mapping the GNSS synchronized spatiotemporal data stream to a three-dimensional spatial grid to construct a multidimensional tensor, obtaining a GNSS interference detection feature tensor. Therefore, this invention solves the problems of data asynchrony, low spatial resolution, and single alarm strategy in traditional GNSS interference detection by multi-source multimodal data fusion and three-dimensional spatial propagation modeling, thus improving the accuracy of interference detection and the level of intelligent response.

Description

Technical Field

[0001] This invention relates to the field of GNSS interference analysis technology, and in particular to a GNSS interference analysis method and system based on big data. Background Technology

[0002] Traditional methods often rely on data from a single GNSS receiver, making it difficult to simultaneously acquire multiple types of observations such as carrier phase, Doppler shift, pseudorange residuals, and carrier-to-noise ratio. This results in insufficient interference feature dimensions, hindering accurate characterization of complex interference patterns. At the data fusion level, existing technologies often employ simple averaging and fixed-weighting rules, failing to consider the spatiotemporal heterogeneity between observations and lacking dynamic fusion strategies. This can easily amplify noise interference and reduce detection reliability. In interference identification, mainstream methods are mostly based on single-dimensional signal analysis, such as frequency domain anomaly detection or time-series mutation analysis, neglecting the joint feature representation of signals in the time, space, and frequency dimensions. They lack high-order tensor feature extraction and fusion mechanisms for multimodal signals. Traditional interference detection algorithms mostly employ static logic such as threshold methods and static model methods, which are difficult to adapt to the complex and ever-changing electromagnetic environment of cities. Especially in densely built-up areas, signal-blocked areas, and areas with strong reflection interference, there is a high risk of missed detections and false alarms. In addition, most current interference analysis systems use two-dimensional display methods, which cannot build a complete interference propagation model based on the city's three-dimensional BIM structure or geographic elevation information. They also lack the ability to dynamically compensate for spatial shielding effects and have not achieved correlation modeling with physical environments such as buildings and critical infrastructure. More importantly, in terms of early warning response, existing systems generally lack a graded response mechanism based on interference intensity and impact range, making it difficult to quickly output refined alarm strategies for specific areas or facilities. Summary of the Invention

[0003] Therefore, it is necessary to provide a GNSS interference analysis method based on big data to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, a GNSS interference analysis method based on big data is proposed, the method comprising the following steps: Step S1: Deploy multi-source signal acquisition sensors to collect GNSS signals, and perform standardization processing to generate a multimodal standard dataset; Step S2: Perform non-uniform sampling alignment on the multimodal standard dataset to obtain the GNSS synchronized spatiotemporal data stream; map the GNSS synchronized spatiotemporal data stream to a three-dimensional spatial grid to construct a multidimensional tensor, and obtain the GNSS interference detection feature tensor; Step S3: Construct a baseline network based on the GNSS interference detection feature tensor, and output optimized coordinate DEM to generate anti-interference high-precision DEM data; construct a three-dimensional propagation base from the anti-interference high-precision DEM data to generate a three-dimensional electromagnetic propagation base for the city; perform path simulation on the three-dimensional electromagnetic propagation base for the city, calculate the loss, and obtain a priority response list. Step S4: Design a multi-level alarm strategy for the priority response list and optimize the reward function threshold to obtain the GNSS interference detection model; construct a holographic three-dimensional interference situation map based on the GNSS interference detection model and generate a big data GNSS interference analysis report.

[0005] The beneficial effects of this invention lie in the fact that it achieves synchronization of data with different sampling frequencies and timestamps through non-uniform sampling alignment technology, enabling the original data to be accurately mapped onto a unified spatiotemporal grid, forming a multidimensional tensor, and thus effectively extracting interference features from GNSS signals. The baseline network and optimized coordinate digital elevation model (DEM) output constructed based on this feature tensor not only improve the spatial resolution of geographic information but also enhance the spatial accuracy and anti-interference capability of interference detection. The construction of a three-dimensional propagation substrate, combined with building BIM models, enables accurate simulation and loss calculation of electromagnetic wave propagation paths in complex urban environments, thereby generating a priority response list and improving the targeting and timeliness of interference response. Based on this, a multi-level alarm strategy is designed and the reward function threshold is optimized, forming a GNSS interference detection model with adaptive adjustment capabilities, enabling interference events to be identified and responded to in stages under different impact ranges and intensity levels. Finally, by constructing a holographic three-dimensional interference situation map, complex interference information is presented in an intuitive three-dimensional spatial form. Combined with big data analysis technology, a systematic and structured GNSS interference analysis report is generated, realizing a complete closed loop from data acquisition, feature extraction, spatial modeling, path simulation to alarm strategy formulation and situation visualization. The entire process enhances the detection accuracy and spatiotemporal resolution of GNSS interference at the data level, while simultaneously enabling comprehensive monitoring and response management of electromagnetic interference in complex urban environments. Therefore, this invention, through multi-source, multi-modal data fusion and three-dimensional spatial propagation modeling, solves the problems of data asynchrony, low spatial resolution, and limited alarm strategies in traditional GNSS interference detection, thereby improving the accuracy of interference detection and the level of intelligent response.

[0006] Preferably, step S1 includes the following steps: Step S11: Deploy lightweight GNSS signal acquisition software on the mobile terminal, configure a sampling rate of 10Hz, capture signal strength, carrier-to-noise ratio and pseudorange residual in real time, and generate raw data from the mobile terminal. Step S12: Deploy an INS / GNSS integrated navigation system with an attitude angle accuracy of 0.05° on the vehicle terminal to collect Doppler frequency shift, vehicle acceleration, and GNSS data, and generate raw vehicle observation data; Step S13: Deploy a multi-frequency multi-mode GNSS receiver at the fixed monitoring station to collect reference data of phase center deviation at a sampling rate of 100Hz, and generate high-precision data for the fixed monitoring station; Step S14: Adaptive wavelet packet transform is used to jointly reduce noise from the original data of the mobile terminal, the original observation data of the vehicle, and the high-precision data of the fixed monitoring station. The wavelet basis function of Daubechies-8 is selected according to the signal frequency bands L1: 1575.42MHz and B1: 1561.098MHz to generate the GNSS noise reduction dataset. Step S15: Unify the geographic location benchmark of the GNSS noise reduction dataset to generate a multimodal standard dataset.

[0007] This invention achieves multi-dimensional, high-frequency data acquisition of GNSS signals by deploying multi-layered, multi-band GNSS and auxiliary sensors on different platforms. This comprehensively captures multimodal raw data from moving terminals, vehicle dynamics, and fixed monitoring stations, covering key indicators such as signal strength, carrier-to-noise ratio, pseudorange residuals, Doppler shift, and phase center deviation. Data sampling frequencies are set to 10Hz for the moving terminal and 100Hz for the fixed monitoring station, ensuring the timeliness and accuracy of signals in different scenarios. By introducing an INS / GNSS integrated navigation system with an attitude angle accuracy of 0.05°, vehicle acceleration and dynamic attitude information are further supplemented, enriching the spatiotemporal physical characteristics of the observed data. In the data fusion stage, adaptive wavelet packet transform is used for joint noise reduction of multi-source heterogeneous signals. Daubechies-8 wavelet basis functions are selected for the L1 band (1575.42MHz) and B1 band (1561.098MHz) to effectively separate noise and useful signal components, improving the signal-to-noise ratio and feature integrity of the data. This noise reduction process not only lowers random noise and systematic errors in the acquired data but also preserves the multi-scale characteristics of the signal, providing high-quality input for subsequent interference analysis. Finally, by unifying the geographic location benchmark, the noise-reduced data collected from different devices and platforms are spatially calibrated and temporally aligned to form a unified multimodal standard dataset, achieving cross-device and cross-temporal consistency and data fusion. The entire data processing chain ensures the spatiotemporal synchronization and feature consistency of the multi-source raw data, significantly enhancing the usability and analytical depth of GNSS signals, and laying a solid data foundation for subsequent interference detection and localization.

[0008] Preferably, step S2 includes the following steps: Step S21: Perform non-uniform sampling alignment on the multimodal standard dataset to obtain the GNSS synchronized spatiotemporal data stream; Step S22: Map the GNSS synchronized spatiotemporal data stream to a 10m×10m three-dimensional spatial grid, and construct a spatial topology matrix by combining the spatial location of high-precision data from fixed monitoring stations to obtain the GNSS spatial topology matrix; Step S23: Perform three-level extraction processing on the GNSS spatial topology relation matrix to obtain the three-level extracted signal feature tensor.

[0009] This invention constructs a unified GNSS synchronized spatiotemporal data stream through time interpolation and frequency normalization, ensuring the consistency and comparability of various signal observation data on the same time axis, thereby improving the basic accuracy of spatiotemporal analysis. Based on the synchronized spatiotemporal data stream, the data is further mapped to a regular grid with a resolution of 10 meters × 10 meters. Combined with high-precision spatial positioning information from fixed monitoring stations, a GNSS spatial topology matrix with physical constraints is constructed. This spatial matrix not only reflects the relative spatial relationships between different observation points but also preserves the propagation distribution characteristics and correlation structure of various GNSS signals in geographic space. Subsequently, a three-level signal feature extraction process is performed on this spatial topology matrix, extracting interference-sensitive indicators at different levels from three dimensions: time-domain features (such as pseudorange change rate and phase jitter), frequency-domain features (such as subband energy entropy), and spatial-domain features (such as signal strength gradient and spatial correlation coefficient). These are then encapsulated in tensor form as a three-level extracted signal feature tensor, maintaining the unity and hierarchy of spatiotemporal signal distribution and multimodal characteristics. This signal feature tensor provides standardized and structured data input for subsequent interference identification and feature fusion analysis, significantly improving the adaptability, stability, and modeling depth of the interference detection algorithm.

[0010] Preferably, step S23 includes the following steps: Step S231: Calculate the pseudorange change rate within a window of 1s for the GNSS spatial topology matrix, and perform time-domain carrier phase jitter calculation to generate time-domain signal feature data; Step S232: Perform S-transform on the GNSS spatial topology matrix to obtain the S-transform decomposed signal; extract the energy entropy of the 1-20MHz sub-band from the S-transform decomposed signal to obtain frequency domain signal feature data; Step S233: Calculate the signal strength gradient of the GNSS spatial topology matrix to obtain signal strength gradient data; calculate the Pearson spatial correlation coefficient based on the signal strength gradient data to generate spatial-level signal feature data; Step S234: Construct a three-dimensional tensor from the time-domain signal feature data, frequency-domain signal feature data, and spatial-domain signal feature data according to time × space × feature, and generate the GNSS interference detection feature tensor.

[0011] This invention quantifies the stability of GNSS signals in the short term by calculating the pseudorange change rate within a 1-second sliding window in the time domain. Simultaneously, it performs perturbation analysis on the carrier phase to extract phase jitter indices, reflecting the transient characteristics of the signal under rapid interference. Secondly, in the frequency domain, it performs joint time-frequency analysis of the GNSS signal using the S-transform. While preserving temporal locality characteristics, it quantifies and extracts the sub-band energy entropy within the 1–20 MHz range, significantly enhancing the ability to identify spectral contamination features and helping to characterize the frequency domain distribution complexity and energy concentration of interfering signals. Thirdly, in the spatial domain, it first calculates the intensity gradient distribution of the GNSS signal in a spatial grid, quantifying the spatial rate of change of the signal field strength. Then, based on the signal gradient distribution between spatial grid points, it constructs a spatial signal correlation matrix using the Pearson correlation coefficient to identify the spatial propagation consistency and distribution structure of the interfering wave. Finally, the three types of dimensional features mentioned above are fused and encapsulated according to the three axes of time, space and feature to construct a three-dimensional structured GNSS interference detection feature tensor. This not only realizes the time-frequency-space integration of multimodal signal features, but also provides a standard input data structure with high resolution and high dimensionality for downstream interference pattern recognition and network modeling.

[0012] Preferably, step S3, which involves constructing a baseline network based on the GNSS interference detection feature tensor and optimizing the coordinate DEM output, includes: A baseline network for interference detection features is constructed based on the GNSS interference detection feature tensor to generate an interference detection feature baseline network. The interference detection feature baseline network is used to calculate the ratio of signal arrival time difference to arrival angle to generate TDOA-AOA ratio data. Obtain historical electromagnetic environment maps; perform prior distribution analysis on the TDOA-AOA ratio data and historical electromagnetic environment maps, and update the distribution by resampling to generate interference-resistant high-precision DEM data.

[0013] This invention constructs a baseline network based on the GNSS interference detection feature tensor, systematically integrating the response patterns of multi-source interference signals in the spatial and temporal domains. Furthermore, it extracts the Time Difference of Arrival (TDOA) and Angle of Arrival (AOA) information based on collaborative observation data among network nodes, using their ratio to construct high-precision spatiotemporal indicators. This enables a multi-angle, fine-grained characterization of the interference source location and propagation path. Specifically, through collaborative computation among nodes in the baseline network, the time delay difference of GNSS signals received at different locations is quantified. Combined with AOA estimation results, the ratio of TDOA to AOA is calculated, reflecting the directional and distance structure characteristics of the interference wave along different propagation paths. Subsequently, this ratio data is introduced into a historical electromagnetic environment map and combined with prior spatial data such as terrain, buildings, and electromagnetic reflection characteristics to perform probability distribution modeling, forming a spatial distribution model of interference propagation with geographic electromagnetic priors. Based on this, by dynamically updating the distribution weights through particle filtering or other adaptive resampling mechanisms, the propagation paths of high-probability interference and their impact range are reconstructed, thereby generating a high-precision digital elevation model (DEM) that integrates spatiotemporal indicator data and electromagnetic space prior knowledge. This DEM not only reflects the three-dimensional structural features affected by terrain and building obstruction during electromagnetic wave propagation, but also possesses the high-resolution electromagnetic inversion capability required for interference identification, significantly improving the modeling accuracy of signal propagation laws and the ability to restore the anti-interference situation in complex interference environments.

[0014] Preferably, step S3, which involves simulating the path of the three-dimensional electromagnetic propagation substrate in the city and calculating the loss, includes: Obtain the building BIM model; construct a three-dimensional propagation base by combining the anti-interference high-precision DEM data with the building BIM model to generate a three-dimensional electromagnetic propagation base for the city; The path of the three-dimensional electromagnetic propagation substrate in the city is determined and the path loss is simulated to obtain three-dimensional simulated path data; Based on the three-dimensional simulated path data, the interference intensity coverage area is statistically analyzed to generate an interference probability density distribution map. Critical infrastructure is identified by analyzing the interference probability density distribution map, resulting in a priority response list.

[0015] This invention effectively establishes a three-dimensional electromagnetic propagation substrate encompassing urban terrain features and building spatial structures by fusing building BIM models with high-precision anti-interference DEM data. This allows for accurate simulation of the propagation path and attenuation process of GNSS interference signals in complex urban environments. Specifically, the BIM model provides high-resolution data on building geometry, material properties, and spatial distribution, while the high-precision anti-interference DEM data reflects terrain undulations, surface reflection characteristics, and geographical features related to electromagnetic wave propagation. The combination of these two data forms a three-dimensional propagation simulation substrate with realistic physical shielding relationships and a ground electromagnetic environment. Based on this substrate, the system introduces a path simulation algorithm to generate three-dimensional path loss data covering urban space by calculating the propagation path, reflection, diffraction, and attenuation process of electromagnetic waves point by point. Subsequently, based on the path simulation results and according to the spatial attenuation law of interference signal power, the interference intensity coverage of each spatial unit is statistically analyzed, and an interference probability density distribution map is constructed, achieving an accurate representation of the spatial distribution of interference risk. Furthermore, the system spatially cross-matches the interference probability density map with the locations of critical infrastructure (such as airports, railway stations, and communication hubs) to identify important targets within high-risk interference areas. Based on the degree of interference and the importance of the facilities, it generates a priority response list, providing data support for subsequent emergency management, early warning triggering, and resource allocation. This process achieves closed-loop processing of electromagnetic propagation simulation, interference risk assessment, and target prioritization at the data level, significantly enhancing the spatial perception accuracy and intelligent decision-making capabilities of GNSS anti-interference analysis.

[0016] Preferably, step S4 includes the following steps: Step S41: Design a multi-level alarm strategy from the priority response list to obtain the multi-level alarm strategy; Step S42: Construct the state space based on the multi-level alarm strategy, and construct a model to optimize the reward function threshold to obtain the GNSS interference detection model; Step S43: Construct a holographic three-dimensional interference situation map based on the GNSS interference detection model, and generate a big data GNSS interference analysis report.

[0017] This invention constructs a multi-level alarm strategy with a hierarchical interference level response mechanism by using a priority response list as input. Based on this, it formally models the state space of GNSS interference monitoring behavior, thereby realizing graded response capabilities and dynamic adaptation to multi-source interference events. At the data level, the system first classifies the interference-affected area into levels based on the joint weights of interference intensity, spatial coverage, and the importance of target facilities, and generates three-level alarm strategies including local, regional, and wide-area, enabling the alarm system to have flexibility and priority orientation when handling interference events of different scales and intensities. Subsequently, the system maps the multi-level alarm strategy to a discrete set of states, constructing an interference state space, where the state transition probability is generated driven by historical interference events and path simulation data; and adaptively optimizes the reward function threshold with the objective function of improving detection accuracy and response timeliness, ultimately completing the training and construction of a GNSS interference detection model for interference level identification and dynamic early warning discrimination. In the model output stage, the system maps the identification results to a three-dimensional spatial visualization environment, combining geographic information system data and building BIM models to achieve gradient coloring of interference intensity in three-dimensional coordinates, forming a three-dimensional holographic three-dimensional interference situation map. Meanwhile, the system structurally summarizes the data output, anomaly distribution, alarm records and target correlation during the interference detection process, and generates a big data GNSS interference analysis report with traceability and statistical analysis functions. This realizes a closed-loop process from raw signal data to situational awareness and strategy feedback, thereby significantly enhancing the ability to quickly identify and respond to GNSS interference in complex urban environments.

[0018] Preferably, step S41 includes the following steps: Based on the judgment of interference intensity and influence radius of the priority response list, a multi-level alarm strategy is generated, which includes a first-level local alarm strategy, a second-level regional alarm strategy, and a third-level wide-area alarm strategy. When the interference intensity is greater than 10dB·Hz and the influence radius is less than 1km, it is judged as a level one local alarm strategy, a level one local command is generated, and it is transmitted to the motion terminal through the spare channel. When the interference intensity is greater than 15dB·Hz and the influence radius is less than 5km, it is judged as a level 2 area alarm strategy, a level 2 area command is generated, and it is transmitted to the motion terminal and vehicle terminal through the normal cloud channel. When the interference intensity is greater than 20 dB·Hz and the influence radius is greater than 5 km, it is judged as a level three wide-area alarm strategy, and a level three wide-area command is generated and transmitted to the motion terminal, vehicle terminal and fixed monitoring station through the fiber optic fast channel for rapid early warning.

[0019] This invention establishes a hierarchical GNSS interference alarm strategy system by performing structured analysis on interference events in a priority response list and combining two key indicators: interference intensity (in dB·Hz) and interference impact radius (in km). This significantly improves the precision of the alarm mechanism in terms of spatial resolution and intensity discrimination. Specifically, the scheme normalizes the raw interference data and calculates the signal attenuation gradient from the interference center to the boundary region based on a spatial interpolation algorithm, thereby estimating the actual coverage radius of the interference. Simultaneously, the system weighted and fused parameters such as carrier-to-noise ratio, signal strength jitter rate, and pseudorange residual variation amplitude of different monitoring nodes to deduce the effective interference power of the central interference source and calculate its interference intensity index. After completing the parameter analysis, the system sets threshold ranges and, according to predefined judgment rules, events with interference intensity greater than 10 dB·Hz and an impact radius less than 1 km are classified as Level 1 local alarm strategies, applicable to sudden interference in high-intensity but locally enclosed areas; interference intensity greater than 15 dB·Hz and an impact radius less than 5 km are classified as Level 2 regional alarm strategies, applicable to regional interference events simultaneously detected by multiple terminals; and interference intensity greater than 20 dB·Hz and an impact radius exceeding 5 km are classified as Level 3 wide-area alarm strategies, used for identifying large-area, continuous, or suspected source-moving interference behaviors. This hierarchical mechanism not only ensures a clear priority order for data response mechanisms when facing different levels of interference, but also forms a hierarchical early warning linkage structure in the GNSS monitoring network, enhancing the system's response efficiency and discrimination accuracy to spatially heterogeneous interference signals. It is particularly suitable for highly sensitive scenarios in complex urban environments or key areas, effectively supporting the precise implementation of subsequent early warning issuance, resource scheduling, and risk assessment linkage mechanisms.

[0020] Preferably, step S43 includes the following steps: Step S431: Perform interference detection based on the GNSS interference detection model and output GNSS interference detection output data; Step S432: Map the GNSS interference detection output data and multi-level alarm strategy to the 3D engine for visualization rendering to obtain a three-dimensional interference detection scene; perform interference intensity gradient coloring on the three-dimensional interference detection scene to obtain a holographic three-dimensional interference situation map; Step S433: Generate a report from the holographic 3D interference situation map to obtain a big data GNSS interference analysis report.

[0021] This invention achieves high-precision visualization and intelligent presentation of interference situations by integrating data results output from a GNSS interference detection model with multi-level alarm strategies, significantly improving the identification efficiency and completeness of GNSS interference events at both the spatial dimension and level of response. Specifically, the interference detection model first performs deep analysis on the input GNSS interference detection feature tensor. Through joint modeling of the time-frequency characteristics, spatial gradient, and statistical perturbation of the interference signal, it outputs structured GNSS interference detection data, including core parameters such as the location, intensity, impact range, and corresponding confidence level of the interference source. Subsequently, the system fuses this structured interference data with the established multi-level alarm strategies and maps it to a 3D visualization engine to construct a 3D interference detection scene map with multi-scale response capabilities. In this scene, not only is the dynamic distribution of the interference source in 3D space presented in real time, but also the intensity gradient of each interference region is rendered in layers, encoded using color gradients or voxel transparency, forming a holographic 3D interference situation map with strong body sense and clear hierarchy. Building upon this foundation, the system further extracts and statistically summarizes data from the 3D situation map, including spatial information structure, temporal evolution trajectory, interference level distribution ratio, and regional alarm frequency density. Combined with historical records and multi-source comparison mechanisms, it automatically generates a structured big data GNSS interference analysis report. The report covers interference characteristic indicator evolution trends, regional sensitivity distribution, spatiotemporal abrupt change boundary identification results, and potential anomaly source prediction, possessing comprehensive, continuous, and reproducible analytical capabilities. Therefore, by constructing a complete chain of data flow to a visualized path driven by a detection model, this solution addresses the problems of information fragmentation, weak spatial representation capabilities, and high reliance on manual intervention in traditional GNSS interference event interpretation, thereby improving the systematic and intelligent level of GNSS interference identification, early warning, and decision support.

[0022] This specification provides a big data-based GNSS interference analysis system for performing the aforementioned big data-based GNSS interference analysis method. The big data-based GNSS interference analysis system includes: The signal acquisition and standardization module is used to deploy multi-source signal acquisition sensors to acquire GNSS signals, perform standardization processing, and generate a multimodal standard dataset. The spatiotemporal alignment and feature construction module is used to perform non-uniform sampling alignment on the multimodal standard dataset to obtain the GNSS synchronized spatiotemporal data stream; the GNSS synchronized spatiotemporal data stream is mapped to a three-dimensional spatial grid to construct a multidimensional tensor to obtain the GNSS interference detection feature tensor; The high-precision DEM construction and propagation modeling module is used to construct a baseline network based on the GNSS interference detection feature tensor, optimize the coordinate DEM output, and generate anti-interference high-precision DEM data; construct a three-dimensional propagation base from the anti-interference high-precision DEM data to generate a three-dimensional electromagnetic propagation base for the city; perform path simulation on the three-dimensional electromagnetic propagation base for the city, calculate the loss, and obtain a priority response list. The interference detection and situation assessment module is used to design multi-level alarm strategies from the priority response list and optimize the reward function threshold to obtain a GNSS interference detection model; based on the GNSS interference detection model, a holographic three-dimensional interference situation map is constructed, and a big data GNSS interference analysis report is generated.

[0023] The beneficial effects of this invention lie in the fact that it achieves synchronization of data with different sampling frequencies and timestamps through non-uniform sampling alignment technology, enabling the original data to be accurately mapped onto a unified spatiotemporal grid, forming a multidimensional tensor, and thus effectively extracting interference features from GNSS signals. The baseline network and optimized coordinate digital elevation model (DEM) output constructed based on this feature tensor not only improve the spatial resolution of geographic information but also enhance the spatial accuracy and anti-interference capability of interference detection. The construction of a three-dimensional propagation substrate, combined with building BIM models, enables accurate simulation and loss calculation of electromagnetic wave propagation paths in complex urban environments, thereby generating a priority response list and improving the targeting and timeliness of interference response. Based on this, a multi-level alarm strategy is designed and the reward function threshold is optimized, forming a GNSS interference detection model with adaptive adjustment capabilities, enabling interference events to be identified and responded to in stages under different impact ranges and intensity levels. Finally, by constructing a holographic three-dimensional interference situation map, complex interference information is presented in an intuitive three-dimensional spatial form. Combined with big data analysis technology, a systematic and structured GNSS interference analysis report is generated, realizing a complete closed loop from data acquisition, feature extraction, spatial modeling, path simulation to alarm strategy formulation and situation visualization. The entire process enhances the detection accuracy and spatiotemporal resolution of GNSS interference from a data perspective, while also enabling comprehensive monitoring and response management of electromagnetic interference in complex urban environments. Attached Figure Description

[0024] Figure 1 This is a flowchart illustrating the steps of a GNSS interference analysis method based on big data. Figure 2 for Figure 1 A detailed flowchart illustrating the implementation steps of step S4. The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0025] The technical method of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this invention.

[0026] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0027] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0028] To achieve the above objectives, please refer to Figures 1 to 2 A GNSS interference analysis method based on big data, the method comprising the following steps: Step S1: Deploy multi-source signal acquisition sensors to collect GNSS signals, and perform standardization processing to generate a multimodal standard dataset; Step S2: Perform non-uniform sampling alignment on the multimodal standard dataset to obtain the GNSS synchronized spatiotemporal data stream; map the GNSS synchronized spatiotemporal data stream to a three-dimensional spatial grid to construct a multidimensional tensor, and obtain the GNSS interference detection feature tensor; Step S3: Construct a baseline network based on the GNSS interference detection feature tensor, and output optimized coordinate DEM to generate anti-interference high-precision DEM data; construct a three-dimensional propagation base from the anti-interference high-precision DEM data to generate a three-dimensional electromagnetic propagation base for the city; perform path simulation on the three-dimensional electromagnetic propagation base for the city, calculate the loss, and obtain a priority response list. Step S4: Design a multi-level alarm strategy for the priority response list and optimize the reward function threshold to obtain the GNSS interference detection model; construct a holographic three-dimensional interference situation map based on the GNSS interference detection model and generate a big data GNSS interference analysis report.

[0029] In this embodiment of the invention, reference is made to Figure 1 The diagram shown illustrates the steps of a GNSS interference analysis method based on big data according to the present invention. In this example, the GNSS interference analysis method based on big data includes the following steps: Step S1: Deploy multi-source signal acquisition sensors to collect GNSS signals, and perform standardization processing to generate a multimodal standard dataset; In this embodiment of the invention, a heterogeneous signal acquisition sensor array is deployed to achieve real-time acquisition of GNSS signals from different sources and frequency bands, and a unified data representation format is constructed to support subsequent data fusion processing. At the data level, GNSS receivers with high sampling accuracy and low noise characteristics are first selected, including multi-mode, multi-frequency receivers supporting GPS, GLONASS, Galileo, and BeiDou systems. This is supplemented by various types of auxiliary sensors, such as inertial navigation units (IMUs), reference clock modules, and spectrum analysis components, to achieve structural completion and noise characteristic constraints of the signals. During the acquisition process, each sensor synchronously records multi-dimensional physical parameters such as signal strength, code pseudorange, multipath characteristics, and carrier phase according to a set sampling period, and binds precise spatiotemporal tags to the raw sampling results at each moment to ensure data traceability. After initial data aggregation, cross-device standardized conversion rules are constructed, including frequency domain resampling, waveform normalization, amplitude equalization, and timestamp alignment, to unify all signals into a unified multimodal structure tensor representation. This structural form uses triples {timestamp, frequency band identifier, physical quantity vector} as basic units to form a standardized dataset that can be decoded by the system. This ensures that the data has temporal consistency, structural compatibility, and semantic clarity, providing a multi-dimensional coupled input source for subsequent interference detection and feature construction. In addition, to adapt to the inherent limitations of different sampling devices in terms of resolution, signal-to-noise ratio, etc., time-frequency domain compensation functions and noise removal operators are introduced to interpolate and reconstruct missing segments and perform trend smoothing on outlier segments. Ultimately, this completes the task of constructing a highly consistent and high-fidelity GNSS multimodal standard dataset.

[0030] Step S2: Perform non-uniform sampling alignment on the multimodal standard dataset to obtain the GNSS synchronized spatiotemporal data stream; map the GNSS synchronized spatiotemporal data stream to a three-dimensional spatial grid to construct a multidimensional tensor, and obtain the GNSS interference detection feature tensor; In this embodiment of the invention, a non-uniform sampling alignment algorithm is used to achieve synchronous representation of multi-source signals in the temporal and spatial domains, and a gridded mapping method is used to complete the standardized representation of high-dimensional data structures. At the data level, the multimodal standard dataset generated in step S1 is first subjected to time-series analysis to identify differences in sampling frequency, start time, and data missing rate among different source signals. Then, a non-uniform sampling alignment mechanism based on Dynamic Time Warping (DTW) or its improved algorithms is used to normalize the multi-source signals on the time axis, ensuring consistent alignment attributes at any given time point. For spatial alignment, a geographic reference system (such as WGS-84) is used to uniformly transform the three-dimensional spatial coordinates of the signal sampling points, and a weighted interpolation method is used to spatially complete sparse areas based on station deployment density, thereby obtaining a continuous and spatially consistent GNSS synchronized spatiotemporal data stream. In the aligned data stream, each sampling point records multi-dimensional features including timestamp, spatial location, signal strength, multipath estimation, carrier phase perturbation, and spectral energy density. These data are further embedded into a gridded coordinate system and tensor-mapped according to geographical region and time window division rules. The mapping process employs a fixed-time sliding window combined with spatial partitioning to divide the original spatiotemporal data stream into fourth-order tensor blocks of the form (time, lat, lon, feature), where each tensor unit contains a set of synchronously sampled GNSS signal physical quantities. During tensor construction, the dimensional sparsity problem caused by data density variations must be considered. Therefore, low-rank tensor completion algorithms (such as CP decomposition or Tucker decomposition) are introduced to densify the tensors, ensuring that the tensor structure possesses strong resolvability and subsequent modeling capabilities while maintaining the original physical information. The final constructed GNSS interference detection feature tensor possesses data characteristics such as multi-dimensional attribute fusion, spatiotemporal consistency assurance, and enhanced noise robustness, and can serve as the core input foundation for interference identification networks.

[0031] Step S3: Construct a baseline network based on the GNSS interference detection feature tensor, and output optimized coordinate DEM to generate anti-interference high-precision DEM data; construct a three-dimensional propagation base from the anti-interference high-precision DEM data to generate a three-dimensional electromagnetic propagation base for the city; perform path simulation on the three-dimensional electromagnetic propagation base for the city, calculate the loss, and obtain a priority response list. In this embodiment of the invention, based on multi-dimensional GNSS disturbance indices (such as signal phase shift, intensity attenuation, and multipath reflection anomalies) contained in the feature tensor, a signal disturbance baseline network is established using a sparse graph neural network or dynamic adjacency graph construction method. In this network, nodes represent spatial location units, and edge weights reflect the interference coupling relationship or propagation correlation between different regions. To extract an accurate terrain model, an interference-weighted geographic inversion strategy is further introduced, projecting the GNSS disturbance features back into the spatial coordinate dimension. The node location information is corrected and fine-tuned using least squares optimization or regularized reconstruction methods, outputting a set of optimized coordinates constructed under interference constraints. This optimized coordinate set is reconstructed in geospace into a high-precision digital elevation model (DEM), where each grid cell not only records terrain elevation information but also incorporates the statistical distribution of the local GNSS signal disturbance field, forming a terrain representation under interference-resistant constraints. Building upon this foundation, to support the analysis of electromagnetic propagation characteristics, a three-dimensional propagation substrate model needs to be constructed from the anti-interference DEM data. This involves embedding the terrain surface structure, building boundary information, material properties of ground features (reflectivity, absorptivity), and environmental electromagnetic parameters into a three-dimensional spatial scene to build a stereo propagation scenario suitable for GNSS signal propagation path simulation. After the three-dimensional propagation substrate is constructed, path tracing algorithms (such as ray tracing or finite difference time-domain method, FDTD) are used to numerically simulate the reflection, diffraction, occlusion, and loss behavior of GNSS signals in this space. Path loss scores are calculated by combining the propagation stability, signal strength attenuation, and noise coupling level of each path. Finally, all path loss results are mapped back to the original tensor structure and aggregated into a response list based on spatiotemporal priority. Each priority path corresponds to a propagation channel with the lowest interference level and optimal propagation stability, providing clear data support and path selection criteria for subsequent interference response strategies and model training.

[0032] Step S4: Design a multi-level alarm strategy for the priority response list and optimize the reward function threshold to obtain the GNSS interference detection model; construct a holographic three-dimensional interference situation map based on the GNSS interference detection model and generate a big data GNSS interference analysis report.

[0033] In this embodiment of the invention, the GNSS priority response list obtained in the previous step is transformed into a multi-level early warning signal system with triggerable characteristics, and the interference situation is spatialized and visualized through an algorithm model. At the data level, the spatiotemporal labels, loss values, signal strength fluctuation indicators, and environmental complexity parameters of each path unit in the priority response list are first deconstructed, and the disturbance intensity scoring matrix is ​​generated by aggregation and statistics according to time windows and spatial locations. This matrix is ​​used to construct a multi-level alarm strategy framework, in which response units are divided into light, moderate, and severe interference levels according to a preset disturbance level classification standard (e.g., quantile segmentation based on distribution functions or empirical thresholds), serving as the input driver for the alarm system. On this basis, a reward function optimization mechanism is introduced, using the policy gradient method or deep Q-network (DQN) commonly used in reinforcement learning to simulate and train the system response strategy under different alarm levels. The dynamic threshold setting of the reward function is optimized with the interference situation change trend and false alarm rate and false alarm rate as the objective function, thereby completing the training and iterative convergence of the GNSS interference detection model. This model possesses the ability to map input feature spatiotemporal sequences to output alarm levels, classifying input datasets to specific interference levels in real time and outputting matching spatial response areas. After model construction, voxelization technology is further introduced to project the prediction results onto a 3D geographic space. By integrating terrain elevation, Building Information Modeling (BIM) structure, and propagation path scenarios, a 3D interference situation map is constructed. The holographic 3D situation map uses color coding, transparency mapping, and dynamic playback to represent the temporal and spatial variations of the interference level, and provides interfaces for full-domain data expansion, local scaling, and cross-sectional slicing. Finally, based on this interference situation map, the system outputs a comprehensive big data analysis report, including statistics on interference source distribution, time-series evolution curves, response efficiency comparison analysis, and regional interference heatmaps, providing quantitative support for subsequent strategy evaluation and early warning system deployment.

[0034] Preferably, step S1 includes the following steps: Step S11: Deploy lightweight GNSS signal acquisition software on the mobile terminal, configure a sampling rate of 10Hz, capture signal strength, carrier-to-noise ratio and pseudorange residual in real time, and generate raw data from the mobile terminal. Step S12: Deploy an INS / GNSS integrated navigation system with an attitude angle accuracy of 0.05° on the vehicle terminal to collect Doppler frequency shift, vehicle acceleration, and GNSS data, and generate raw vehicle observation data; Step S13: Deploy a multi-frequency multi-mode GNSS receiver at the fixed monitoring station to collect reference data of phase center deviation at a sampling rate of 100Hz, and generate high-precision data for the fixed monitoring station; Step S14: Adaptive wavelet packet transform is used to jointly reduce noise from the original data of the mobile terminal, the original observation data of the vehicle, and the high-precision data of the fixed monitoring station. The wavelet basis function of Daubechies-8 is selected according to the signal frequency bands L1: 1575.42MHz and B1: 1561.098MHz to generate the GNSS noise reduction dataset. Step S15: Unify the geographic location benchmark of the GNSS noise reduction dataset to generate a multimodal standard dataset.

[0035] In this embodiment of the invention, the three types of raw observation data are uniformly structured and organized: signal strength, carrier-to-noise ratio (C / N0), and pseudorange residuals collected by the motion terminal; Doppler frequency shift, acceleration information, and GNSS observations collected by the vehicle terminal; and 100Hz high-frequency multimode phase data collected by the fixed monitoring station are aligned by timestamps and encoded in a unified vector form to form a multi-source GNSS observation data matrix. Subsequently, considering the frequency characteristics differences of different physical quantities in this matrix, an adaptive wavelet packet transform algorithm is used to project the signals of each channel onto the frequency subspace for multi-scale analysis. This algorithm selects the optimal wavelet packet decomposition path based on the entropy criterion. Specifically, within the L1 (1575.42 MHz) and B1 (1561.098 MHz) signal frequency bands, the Daubechies-8 wavelet basis function is selected for recursive decomposition at three or more levels. After each level of decomposition, the noise-dominant frequency band is identified, and high-frequency noise components are suppressed through soft thresholding or median filtering strategies. During wavelet packet reconstruction, low-frequency components and several mid-frequency components with characteristic distributions in the main information passband are retained and recombined into the reconstructed GNSS signal channels. While performing joint denoising, the differences in sampling frequencies of multi-source data must also be considered. Therefore, before denoising, interpolation resampling or weighted sliding window methods are used to normalize data at different sampling rates such as 10Hz and 100Hz to avoid the superposition of errors caused by time-domain misalignment. Finally, through this wavelet packet denoising process, a multi-channel GNSS denoised dataset that is compatible in both time and frequency domains, enhances characteristic signals, and suppresses noise components is formed, providing a stable and continuous input data source for subsequent positioning calculations, interference detection, and spatial modeling.

[0036] Preferably, step S2 includes the following steps: Step S21: Perform non-uniform sampling alignment on the multimodal standard dataset to obtain the GNSS synchronized spatiotemporal data stream; Step S22: Map the GNSS synchronized spatiotemporal data stream to a 10m×10m three-dimensional spatial grid, and construct a spatial topology matrix by combining the spatial location of high-precision data from fixed monitoring stations to obtain the GNSS spatial topology matrix; Step S23: Perform three-level extraction processing on the GNSS spatial topology relation matrix to obtain the three-level extracted signal feature tensor.

[0037] In this embodiment of the invention, during the non-uniform sampling alignment stage, it is necessary to address issues such as sampling frequency, temporal resolution, and observation discontinuities in the multimodal standard dataset. To address this, a weighted time interpolation function is constructed to timestamp-align 10Hz mobile terminal data, 100Hz fixed monitoring data, and medium-frequency vehicle-mounted observation data. A sliding window synchronization strategy is employed to uniformly resample these data into a standard reference time axis. Furthermore, a Dynamic Time Warping (DTW) algorithm is introduced to dynamically pair the similarities between different source signals to ensure time synchronization. Subsequently, the synchronized spatiotemporal data stream is projected onto a two-dimensional regular grid based on its geographical location, with a grid granularity of 10 meters by 10 meters, constructing an equivalent spatial data container. Within this grid, the Euclidean distance and spatial relative direction are calculated using the latitude and longitude coordinates of each sampling point and the known spatial reference coordinates of the fixed monitoring station. A spatial topology matrix is ​​generated by constructing a weighting function based on spatial adjacency rules (such as a Gaussian decay kernel or a distance-based inverse proportional function). This matrix records the spatial coupling relationship between each grid cell and its surrounding fixed reference points, providing structural priors for subsequent feature modeling. In the feature extraction stage, a three-level extraction strategy is employed to perform tensor transformation on the GNSS spatial topology matrix. The first level of extraction constructs the original feature vector using local signal statistics (mean, standard deviation, extrema, etc.). The second level introduces frequency domain and waveform variation features (such as spectral centroid, bandwidth expansion, modulation mode features, etc.), and enhances these features through Fast Fourier Transform (FFT) or Empirical Mode Decomposition (EMD). The third level introduces topological correlation and temporal variation trends, extracting higher-order spatial distribution features based on graph convolution or multi-channel convolution kernel mechanisms. Finally, the signal features processed through these three levels are encoded into a structurally unified third-order tensor, whose dimensions typically correspond to the time step, spatial location, and feature dimension, used to comprehensively express the variation patterns and potential interference characteristics of GNSS signals in the spatiotemporal domain.

[0038] Of particular importance, step S22 includes the following steps: Step S221: The adaptive Kriging interpolation method is used to map the GNSS synchronized spatiotemporal data stream to a 10m*10m grid, and the data smoothness is evaluated to generate the GNSS synchronized grid. Step S222: Combine the spatial location of the GNSS synchronization grid with the high-precision data of the fixed monitoring station to perform spatial grid calibration and generate a calibrated GNSS synchronization grid; Step S223: Construct the spatial topology matrix based on the calibrated GNSS synchronization grid to obtain the GNSS spatial topology matrix.

[0039] In this embodiment of the invention, adaptive Kriging interpolation is used to perform gridded mapping processing on the GNSS synchronized spatiotemporal data stream. Kriging interpolation, as a geostatistical interpolation method, uses a spatial variogram model between data points to estimate the values ​​of unknown points. Its "adaptive" characteristic is reflected in its ability to adaptively adjust the semi-variogram model parameters based on local data density, spatial variance changes, and measurement errors, achieving spatial continuity estimation of GNSS observations within a 10m × 10m grid. After interpolation, the smoothness of the interpolation results is evaluated by calculating the residual sum of squares, interpolation error distribution, and spatial autocorrelation index, ensuring that the spatial grid has sufficient stability and analytical capability. Considering the systematic drift or offset errors of GNSS moving observation points, the interpolated GNSS synchronized grid is spatially registered and calibrated with the high-precision spatial location data of fixed monitoring stations. Calibration methods include spatial nearest neighbor mapping, least squares fitting registration, and boundary weighting adjustment to ensure that the interpolated grid is aligned with the actual physical station in the reference coordinate system, generating a spatially consistent calibrated GNSS synchronized grid. Based on the calibrated grid, a spatial topology matrix is ​​constructed according to conditions such as spatial adjacency between grids, similarity thresholds of GNSS measurements, and spatial gradient change patterns. This matrix adopts a weighted adjacency matrix representation, and the matrix elements reflect the correlation between grid nodes, potential interference propagation paths, or spatiotemporal coupling strength, providing clear data structure support for subsequent multidimensional signal feature extraction and network analysis. The entire process achieves an effective conversion from unstructured GNSS spatiotemporal flow to a structured spatial matrix, possessing a rigorous foundation for data parsing and spatial logical organization.

[0040] Preferably, step S23 includes the following steps: Step S231: Calculate the pseudorange change rate within a window of 1s for the GNSS spatial topology matrix, and perform time-domain carrier phase jitter calculation to generate time-domain signal feature data; Step S232: Perform S-transform on the GNSS spatial topology matrix to obtain the S-transform decomposed signal; extract the energy entropy of the 1-20MHz sub-band from the S-transform decomposed signal to obtain frequency domain signal feature data; Step S233: Calculate the signal strength gradient of the GNSS spatial topology matrix to obtain signal strength gradient data; calculate the Pearson spatial correlation coefficient based on the signal strength gradient data to generate spatial-level signal feature data; Step S234: Construct a three-dimensional tensor from the time-domain signal feature data, frequency-domain signal feature data, and spatial-domain signal feature data according to time × space × feature, and generate the GNSS interference detection feature tensor.

[0041] In this embodiment of the invention, for pseudorange data in GNSS signals, a sliding window size of 1 second is selected. The pseudorange change rate at adjacent time points is calculated within each sampling window to capture potential abrupt changes and interference behaviors. Simultaneously, combined with high-frequency sampled carrier phase data, the carrier phase jitter between adjacent carrier cycles is calculated. This parameter reflects signal propagation stability and measurement accuracy, exhibiting high sensitivity in GNSS interference identification. The obtained indicators, after standardization, constitute a time-domain feature dataset. The Stockwell Transform is used to perform time-frequency decomposition on the GNSS pseudorange or signal strength sequence, obtaining the frequency amplitude distribution corresponding to each time point. Based on the decomposition results, sub-band separation is further performed in the 1–20 MHz frequency band, and the energy entropy value is calculated for each sub-band signal. Energy entropy, as a complex measure of frequency domain power distribution, can reveal frequency domain disturbance characteristics such as spectral spread and power dispersion, thus forming frequency-domain signal feature data. Based on the signal strength values ​​of each grid cell in the spatial grid, the system performs gradient calculations to capture the spatial variation trend of the signal field strength and constructs a spatial difference structure through gradient amplitude. Furthermore, by selecting the signal intensity variation values ​​between adjacent grids and calculating their Pearson correlation coefficients, the spatial correlation between each location and its neighborhood is obtained, thus characterizing the signal similarity and interference propagation behavior in the spatial domain. The above three stages output feature vector data in the time domain, frequency domain, and spatial domain, respectively. Based on a unified GNSS spatiotemporal reference, the system aligns and normalizes the data in the above three dimensions according to time (T), spatial location (L), and feature type (F), and constructs a three-dimensional tensor of dimension T×L×F to achieve a highly structured expression of GNSS interference detection features, providing data support for subsequent interference identification, classification, and modeling. Throughout the process, hierarchical deconstruction of data, index reduction, and spatial consistency mapping are emphasized to ensure that the feature tensor possesses multi-scale and multi-dimensional integrity and coherence.

[0042] Preferably, step S3, which involves constructing a baseline network based on the GNSS interference detection feature tensor and optimizing the coordinate DEM output, includes: A baseline network for interference detection features is constructed based on the GNSS interference detection feature tensor to generate an interference detection feature baseline network. The interference detection feature baseline network is used to calculate the ratio of signal arrival time difference to arrival angle to generate TDOA-AOA ratio data. Obtain historical electromagnetic environment maps; perform prior distribution analysis on the TDOA-AOA ratio data and historical electromagnetic environment maps, and update the distribution by resampling to generate interference-resistant high-precision DEM data.

[0043] In this embodiment of the invention, high-precision anti-interference DEM (Digital Elevation Model) data is generated by constructing an interference feature baseline network and fusing spatiotemporal statistical features with historical prior distributions. In the first stage, the system uses the GNSS interference detection feature tensor as input and performs structured indexing based on the time dimension and spatial topological relationship to construct a feature baseline network. The nodes of this network represent GNSS monitoring points, and the node attributes are extracted from the three-dimensional feature tensor, including carrier phase jitter, energy entropy distribution, and spatial correlation coefficient; the edge weights combine the timestamp differences and spatial distances between nodes to reflect the potential coupling characteristics of the interference propagation path, thereby constructing a GNSS interference detection feature baseline network based on topology and spatiotemporal dependence. Next, the system performs signal propagation parameter estimation on the network. First, it extracts the signal propagation delay information carried on each edge and calculates the Time Difference of Arrival (TDOA). Then, it extracts the Angle of Arrival (AOA) using carrier phase directivity data from adjacent nodes or azimuth information measured by directional antenna arrays. The TDOA and AOA are normalized and their ratio is calculated to construct the TDOA-AOA ratio data for each edge. This ratio provides a composite representation of time delay and directionality in the signal propagation path at the data level, helping to describe the signal propagation behavior and spatial distortion degree under interference source emission. Subsequently, in the third stage, the system introduces a historical electromagnetic environment map as prior knowledge, performs rasterized distribution modeling on it, and obtains the probability density function of interference or anomalous propagation at different spatiotemporal locations. A mapping relationship is established between the TDOA-AOA ratio data and the historical electromagnetic map, and the posterior distribution is calculated using a Bayesian prior distribution framework. This process involves particle filtering or Monte Carlo sampling for resampling and distribution updates, allowing current observations and historical distributions to jointly constrain the spatial distribution estimation of interference, thereby correcting interference distortion in GNSS terrain elevation. Finally, the updated distribution data is used to reconstruct the existing GNSS DEM model, outputting high-precision DEM data with interference correction capabilities, providing spatial geographic support for subsequent electromagnetic propagation modeling. The entire process uses feature tensors to drive feature network construction, combining propagation geometry parameters and prior scene distributions to complete spatiotemporal interference identification and terrain inversion processing.

[0044] Preferably, step S3, which involves simulating the path of the three-dimensional electromagnetic propagation substrate in the city and calculating the loss, includes: Obtain the building BIM model; construct a three-dimensional propagation base by combining the anti-interference high-precision DEM data with the building BIM model to generate a three-dimensional electromagnetic propagation base for the city; The path of the three-dimensional electromagnetic propagation substrate in the city is determined and the path loss is simulated to obtain three-dimensional simulated path data; Based on the three-dimensional simulated path data, the interference intensity coverage area is statistically analyzed to generate an interference probability density distribution map. Critical infrastructure is identified by analyzing the interference probability density distribution map, resulting in a priority response list.

[0045] In this embodiment of the invention, the Building Information Modeling (BIM) model of the building is imported in a structured format, and the three-dimensional spatial geometric parameters, material properties, and hierarchical structural information contained therein are extracted and fused with previously generated anti-interference high-precision DEM data. During the data fusion process, a unified spatial reference coordinate system is used as the benchmark, and a three-dimensional registration algorithm is used to spatially align and stitch the DEM data and BIM geometry, mapping information such as building boundaries, height, floor thickness, and electromagnetic properties of materials onto the DEM terrain surface to construct a three-dimensional electromagnetic propagation base for the city. Subsequently, electromagnetic path simulation is performed based on this propagation base, and the propagation path of electromagnetic waves in three-dimensional space is modeled using a ray tracing algorithm or a finite difference time-domain method (FDTD) based on ray approximation. In path simulation, the system uses the GNSS signal transmission point as the source and establishes propagation path data for each propagation direction. Each propagation segment in the path records its incident angle, reflection angle, penetration path, and interaction points with buildings. Combining the reflection, refraction, and absorption coefficients of building materials for different frequency bands of electromagnetic waves from BIM data, the system iteratively calculates path loss to obtain the total power attenuation value for each path, thus generating complete 3D simulated path data. Subsequently, the system performs statistical analysis on all 3D simulated path data, counting the number of times each area appears in the simulated path and the corresponding signal strength attenuation value in spatial grid units, and normalizing this to generate a spatial coverage map of interference intensity. Then, through spatial Gaussian kernel density estimation, the system models the interference probability of each grid unit, forming an interference probability density distribution map. Finally, the system incorporates a database of critical infrastructure locations, such as communication base stations, traffic control nodes, and power dispatch centers, matching their spatial locations in the interference probability density distribution map and sorting them according to their interference impact value, outputting a priority response list to provide a ranking and quantitative basis for subsequent response decisions regarding interference exposure. The entire process revolves around spatial fusion, path modeling, signal attenuation analysis, and density estimation, realizing a complete data chain from the construction of three-dimensional spatial information to the identification of interference-affected areas.

[0046] As an example of the present invention, reference is made to... Figure 2 As shown, step S4 in this example includes: Step S41: Design a multi-level alarm strategy from the priority response list to obtain the multi-level alarm strategy; Step S42: Construct the state space based on the multi-level alarm strategy, and construct a model to optimize the reward function threshold to obtain the GNSS interference detection model; Step S43: Construct a holographic three-dimensional interference situation map based on the GNSS interference detection model, and generate a big data GNSS interference analysis report.

[0047] In this embodiment of the invention, a multi-level alarm strategy is designed based on the spatial distribution and influence weights of each interference source in the priority response list. This strategy quantifies and classifies multi-dimensional features such as the severity, frequency, and spatiotemporal distribution of interference events, dividing alarm levels into multiple tiers to distinguish and prepare responses to interference events of different levels. Specifically, the system maps the data in the priority response list into multi-dimensional feature vectors, divides alarm levels by setting threshold ranges, constructs corresponding decision rule sets, and forms a structured multi-level alarm strategy system. Subsequently, based on this multi-level alarm strategy, a state-space model is constructed, dynamically mapping interference states and their changes to a defined set of state variables. A reinforcement learning framework is used to design and optimize the reward function. The reward function adjusts the weights between each state and action based on indicators such as detection accuracy, false alarm rate, and response timeliness, achieving adaptive optimization of model parameters. Through iterative calculations, algorithms such as Bayesian update or gradient descent are used to optimize the reward function threshold, enabling the model to maintain superior detection performance in complex interference environments. After model construction is completed, the system combines the detection results with Geographic Information System (GIS) data to generate a holographic 3D interference situation map based on 3D coordinates and time dimensions. This situation map, based on a 3D grid and incorporating time-series data, dynamically displays the distribution, intensity changes, and spatiotemporal evolution trends of interference sources. Finally, by summarizing the situation map data and historical detection records, big data analytics are used for statistical analysis, trend analysis, and anomaly identification, generating a systematic GNSS interference analysis report. The report includes information such as spatiotemporal distribution characteristics, event frequency, and interference type classification, supporting subsequent decision-making and response strategy optimization. The entire process relies on multi-level alarm design, state-space modeling, reward function optimization, and 3D spatiotemporal situation visualization to form a complete data processing chain for interference detection and situation analysis.

[0048] Preferably, step S41 includes the following: Based on the judgment of interference intensity and influence radius of the priority response list, a multi-level alarm strategy is generated, which includes a first-level local alarm strategy, a second-level regional alarm strategy, and a third-level wide-area alarm strategy. When the interference intensity is greater than 10dB·Hz and the influence radius is less than 1km, it is judged as a level one local alarm strategy, a level one local command is generated, and it is transmitted to the motion terminal through the spare channel. When the interference intensity is greater than 15dB·Hz and the influence radius is less than 5km, it is judged as a level 2 area alarm strategy, a level 2 area command is generated, and it is transmitted to the motion terminal and vehicle terminal through the normal cloud channel. When the interference intensity is greater than 20 dB·Hz and the influence radius is greater than 5 km, it is judged as a level three wide-area alarm strategy, and a level three wide-area command is generated and transmitted to the motion terminal, vehicle terminal and fixed monitoring station through the fiber optic fast channel for rapid early warning.

[0049] In this embodiment of the invention, the intensity and spatial impact range of each interference event are quantitatively calculated based on the interference source data in the priority response list. Interference intensity is determined by analyzing the signal power spectral density and calibrating it using decibel-hertz (dB·Hz) units, achieving accurate measurement of different interference signal energy levels. The impact radius is calculated by analyzing the spatial propagation characteristics and path loss model of the interference signal, combined with propagation path simulation data in a three-dimensional electromagnetic propagation substrate in the city, to determine the effective spatial coverage of the interference signal. Based on these two key parameters, the system defines a multi-level alarm strategy. The first-level local alarm strategy corresponds to events with interference intensity greater than 10 dB·Hz and an impact radius less than 1 km, primarily focusing on high-intensity interference within a small local area. The second-level regional alarm strategy corresponds to events with interference intensity exceeding 15 dB·Hz and an impact radius less than 5 km, representing medium-to-high intensity interference within a medium range. The third-level wide-area alarm strategy covers wide-area impact events with interference intensity exceeding 20 dB·Hz and an impact radius greater than 5 km. For each record in the priority response list, the system inputs its interference intensity and impact radius data into the alarm determination module, and automatically assigns it to the corresponding alarm level through threshold comparison and conditional logic judgment. This determination process is based on structured data indexing and conditional filtering algorithms to achieve the fusion processing of multi-source heterogeneous data. The alarm strategy level classification results serve as the basis for subsequent alarm management and response priority adjustment, forming a hierarchical expression of the spatiotemporal characteristics and intensity features of interference events, providing data support for triggering subsequent alarm actions. The entire process relies on accurate signal strength measurement, spatial impact range estimation, and hierarchical threshold determination algorithms to ensure the hierarchical identification and classification of interference events, constructing a multi-level alarm strategy framework that conforms to spatiotemporal distribution patterns.

[0050] Of particular importance, step S42 includes the following steps: Step S421: Define the state space for the multi-level alarm strategy and perform discrete feature encoding to generate an interference multi-level state coding space; Step S422: Design a multi-objective function for the interference multi-level state coding space, and train the model and optimize the reward function threshold to obtain the interference model reward function; Step S423: Construct a model based on the interference model reward function to obtain a GNSS interference detection model.

[0051] In this embodiment of the invention, a state space for GNSS interference events is constructed based on a predetermined multi-level alarm strategy. The definition of this state space is based on key dimensions such as interference intensity (e.g., in dB·Hz), influence radius (in km), interference frequency band distribution, interference event duration, and interference source location confidence level. The original interference scenario is discretized in a data structure. Discrete feature encoding of multi-dimensional features in the state space is achieved through segmented mapping and conditional threshold labeling, thereby constructing a multi-level interference state encoding space. This ensures that the states of various interference events have clear and distinguishable coded representations in the numerical space. This encoding space is input into a multi-objective function optimization framework. The designed objective functions include minimizing the false alarm rate, maximizing the high-intensity interference recognition rate, improving the accuracy of response decisions, and controlling the model convergence speed. The objective function is iteratively optimized using policy gradient methods or deep Q-learning algorithms in reinforcement learning, and the reward value distribution range is dynamically adjusted through an adaptive reward mechanism, thereby establishing an interference event response mechanism covering different alarm levels. Through model training driven by a large amount of state-action-result triplet data, the system determines the interference model reward function. The formula for the reward function is: ; Wherein, TPR (True Positive Rate) is the percentage of correctly detected interference events; FPR (False Positive Rate) is the false alarm rate; Resource Cost is the computational resource consumption (CPU utilization × time); and the weighting coefficients (α=0.7, β=0.2, γ=0.1) are dynamically adjusted through Pareto optimization. Based on the optimized reward function, the final GNSS interference detection model is constructed. The model uses state encoding as input, which is mapped to alarm response actions through a policy function, and outputs the optimal interference identification and response path. Throughout the construction process, the structured state representation, the targeted and learnable nature of the reward function, and the optimality of the model's output actions are ensured, providing a highly consistent logical foundation for subsequent 3D situation rendering and alarm policy execution.

[0052] Preferably, step S43 includes the following: Step S431: Perform interference detection based on the GNSS interference detection model and output GNSS interference detection output data; Step S432: Map the GNSS interference detection output data and multi-level alarm strategy to the 3D engine for visualization rendering to obtain a three-dimensional interference detection scene; perform interference intensity gradient coloring on the three-dimensional interference detection scene to obtain a holographic three-dimensional interference situation map; Step S433: Generate a report from the holographic 3D interference situation map to obtain a big data GNSS interference analysis report.

[0053] In this embodiment of the invention, a GNSS interference detection model is used as the core. The model performs joint analysis and processing on the input GNSS signal feature tensor and multi-level alarm strategy data. By identifying signal anomaly patterns and determining their states, the model outputs structured GNSS interference detection results data. This data includes multi-dimensional information such as the location coordinates, intensity level, timestamp, and associated alarm level of the interference event. Subsequently, the system maps and associates this interference detection output data with a preset multi-level alarm strategy, forming a composite dataset containing multi-dimensional attributes such as spatiotemporal, intensity, and alarm level. This composite dataset is input to a visualization module based on a 3D graphics engine. The interference location is displayed in three dimensions in a spatial coordinate system. Combined with alarm level information, dynamic visual encoding of the interference intensity is achieved through color gradient mapping, generating a three-dimensional interference detection scene with spatial awareness and intensity expression. Furthermore, GPU-accelerated shading technology is used to render the interference intensity gradient in real time, forming a holographic three-dimensional interference situation map. This situation map not only displays spatial distribution characteristics but also expresses the intensity gradient changes of the interference signal, supporting interactive perspective changes and multi-scale observation. Ultimately, based on the spatiotemporal data and alarm strategies in the holographic 3D interference situation map, the system automatically invokes a big data analysis framework, integrating historical and real-time monitoring data. Through multidimensional statistical analysis, trend mining, and event correlation techniques, it generates a detailed GNSS interference analysis report. This report covers the spatiotemporal distribution of interference events, intensity evolution, alarm response records, and potential risk assessment. The data structure is rigorous and possesses temporal logic, providing a scientific basis for subsequent decision support and interference mitigation. The entire process relies on spatiotemporal mapping of data, multidimensional attribute fusion, and real-time, efficient graphics rendering technology, achieving a closed-loop data-driven system from model output to visualization and then to the generation of in-depth data analysis reports.

[0054] This specification provides a big data-based GNSS interference analysis system for performing the aforementioned big data-based GNSS interference analysis method. The big data-based GNSS interference analysis system includes: The signal acquisition and standardization module is used to deploy multi-source signal acquisition sensors to acquire GNSS signals, perform standardization processing, and generate a multimodal standard dataset. The spatiotemporal alignment and feature construction module is used to perform non-uniform sampling alignment on the multimodal standard dataset to obtain the GNSS synchronized spatiotemporal data stream; the GNSS synchronized spatiotemporal data stream is mapped to a three-dimensional spatial grid to construct a multidimensional tensor to obtain the GNSS interference detection feature tensor; The high-precision DEM construction and propagation modeling module is used to construct a baseline network based on the GNSS interference detection feature tensor, optimize the coordinate DEM output, and generate anti-interference high-precision DEM data; construct a three-dimensional propagation base from the anti-interference high-precision DEM data to generate a three-dimensional electromagnetic propagation base for the city; perform path simulation on the three-dimensional electromagnetic propagation base for the city, calculate the loss, and obtain a priority response list. The interference detection and situation assessment module is used to design multi-level alarm strategies from the priority response list and optimize the reward function threshold to obtain a GNSS interference detection model; based on the GNSS interference detection model, a holographic three-dimensional interference situation map is constructed, and a big data GNSS interference analysis report is generated.

[0055] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0056] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A GNSS interference analysis method based on big data, characterized in that, Includes the following steps: Step S1: Deploy multi-source signal acquisition sensors to collect GNSS signals, and perform standardization processing to generate a multimodal standard dataset; Step S2: Perform non-uniform sampling alignment on the multimodal standard dataset to obtain the GNSS synchronized spatiotemporal data stream; The GNSS synchronized spatiotemporal data stream is mapped to a three-dimensional spatial grid to construct a multidimensional tensor, thus obtaining the GNSS interference detection feature tensor. Step S2 includes the following steps: Step S21: Perform non-uniform sampling alignment on the multimodal standard dataset to obtain the GNSS synchronized spatiotemporal data stream; Step S22: Map the GNSS synchronized spatiotemporal data stream to a 10m×10m three-dimensional spatial grid, and construct a spatial topology matrix by combining the spatial location of high-precision data from fixed monitoring stations to obtain the GNSS spatial topology matrix; Step S23: Perform three-level extraction processing on the GNSS spatial topology relation matrix to obtain the three-level extracted signal feature tensor; Step S23 includes the following steps: Step S231: Calculate the pseudorange change rate within a window of 1s for the GNSS spatial topology matrix, and perform time-domain carrier phase jitter calculation to generate time-domain signal feature data; Step S232: Perform S-transform on the GNSS spatial topology matrix to obtain the S-transform decomposed signal; extract the energy entropy of the 1-20MHz sub-band from the S-transform decomposed signal to obtain frequency domain signal feature data; Step S233: Calculate the signal strength gradient of the GNSS spatial topology matrix to obtain signal strength gradient data; calculate the Pearson spatial correlation coefficient based on the signal strength gradient data to generate spatial-level signal feature data; Step S234: Construct a three-dimensional tensor from the time-domain signal feature data, frequency-domain signal feature data, and spatial-domain signal feature data according to time × space × feature, and generate the GNSS interference detection feature tensor; Step S3: Construct a baseline network based on the GNSS interference detection feature tensor, and output optimized coordinate DEM to generate anti-interference high-precision DEM data; construct a three-dimensional propagation base from the anti-interference high-precision DEM data to generate a three-dimensional electromagnetic propagation base for the city; perform path simulation on the three-dimensional electromagnetic propagation base for the city, calculate the loss, and obtain a priority response list. Step S4: Design a multi-level alarm strategy for the priority response list and optimize the reward function threshold to obtain the GNSS interference detection model; construct a holographic three-dimensional interference situation map based on the GNSS interference detection model and generate a big data GNSS interference analysis report.

2. The GNSS interference analysis method based on big data according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Deploy lightweight GNSS signal acquisition software on the mobile terminal, configure a sampling rate of 10Hz, capture signal strength, carrier-to-noise ratio and pseudorange residual in real time, and generate raw data from the mobile terminal. Step S12: Deploy an INS / GNSS integrated navigation system with an attitude angle accuracy of 0.05° on the vehicle terminal to collect Doppler frequency shift, vehicle acceleration, and GNSS data, and generate raw vehicle observation data; Step S13: Deploy a multi-frequency multi-mode GNSS receiver at the fixed monitoring station to collect reference data of phase center deviation at a sampling rate of 100Hz, and generate high-precision data for the fixed monitoring station; Step S14: Adaptive wavelet packet transform is used to jointly reduce noise from the original data of the mobile terminal, the original observation data of the vehicle, and the high-precision data of the fixed monitoring station. The wavelet basis function of Daubechies-8 is selected according to the signal frequency bands L1: 1575.42MHz and B1: 1561.098MHz to generate the GNSS noise reduction dataset. Step S15: Unify the geographic location benchmark of the GNSS noise reduction dataset to generate a multimodal standard dataset.

3. The GNSS interference analysis method based on big data according to claim 1, characterized in that, Step S3, which involves constructing a baseline network based on the GNSS interference detection feature tensor and optimizing the coordinate DEM output, includes: A baseline network is constructed based on the GNSS interference detection feature tensor to generate an interference detection feature baseline network. The interference detection feature baseline network is used to calculate the ratio of signal arrival time difference to arrival angle to generate TDOA-AOA ratio data. Obtain historical electromagnetic environment maps; perform prior distribution analysis on the TDOA-AOA ratio data and historical electromagnetic environment maps, and update the distribution by resampling to generate interference-resistant high-precision DEM data.

4. The GNSS interference analysis method based on big data according to claim 1, characterized in that, Step S3, which involves simulating the path of electromagnetic propagation on the urban three-dimensional electromagnetic base and calculating the loss, includes: Obtain the building BIM model; construct a three-dimensional propagation base by combining the anti-interference high-precision DEM data with the building BIM model to generate a three-dimensional electromagnetic propagation base for the city; The path of the three-dimensional electromagnetic propagation substrate in the city is determined and the path loss is simulated to obtain three-dimensional simulated path data; Based on the three-dimensional simulated path data, the interference intensity coverage area is statistically analyzed to generate an interference probability density distribution map. Critical infrastructure is identified by analyzing the interference probability density distribution map, resulting in a priority response list.

5. The GNSS interference analysis method based on big data according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: Design a multi-level alarm strategy from the priority response list to obtain the multi-level alarm strategy; Step S42: Construct the state space based on the multi-level alarm strategy, and construct a model to optimize the reward function threshold to obtain the GNSS interference detection model; Step S43: Construct a holographic three-dimensional interference situation map based on the GNSS interference detection model, and generate a big data GNSS interference analysis report.

6. The GNSS interference analysis method based on big data according to claim 5, characterized in that, Step S41 includes the following: Based on the judgment of interference intensity and influence radius of the priority response list, a multi-level alarm strategy is generated, which includes a first-level local alarm strategy, a second-level regional alarm strategy, and a third-level wide-area alarm strategy. When the interference intensity is greater than 10dB·Hz and the influence radius is less than 1km, it is judged as a level one local alarm strategy, a level one local command is generated, and it is transmitted to the motion terminal through the spare channel. When the interference intensity is greater than 15dB·Hz and the influence radius is less than 5km, it is judged as a level 2 area alarm strategy, a level 2 area command is generated, and it is transmitted to the motion terminal and vehicle terminal through the normal cloud channel. When the interference intensity is greater than 20 dB·Hz and the influence radius is greater than 5 km, it is judged as a level three wide-area alarm strategy, and a level three wide-area command is generated and transmitted to the motion terminal, vehicle terminal and fixed monitoring station through the fiber optic fast channel for rapid early warning.

7. The GNSS interference analysis method based on big data according to claim 5, characterized in that, Step S43 includes the following steps: Step S431: Perform interference detection based on the GNSS interference detection model and output GNSS interference detection output data; Step S432: Map the GNSS interference detection output data and multi-level alarm strategy to the 3D engine for visualization rendering to obtain a three-dimensional interference detection scene; perform interference intensity gradient coloring on the three-dimensional interference detection scene to obtain a holographic three-dimensional interference situation map; Step S433: Generate a report from the holographic 3D interference situation map to obtain a big data GNSS interference analysis report.

8. A GNSS interference analysis system based on big data, characterized in that, For performing the big data-based GNSS interference analysis method as described in claim 1, the big data-based GNSS interference analysis system comprises: The signal acquisition and standardization module is used to deploy multi-source signal acquisition sensors to acquire GNSS signals, perform standardization processing, and generate a multimodal standard dataset. The spatiotemporal alignment and feature construction module is used to perform non-uniform sampling alignment on the multimodal standard dataset to obtain the GNSS synchronized spatiotemporal data stream; the GNSS synchronized spatiotemporal data stream is mapped to a three-dimensional spatial grid to construct a multidimensional tensor to obtain the GNSS interference detection feature tensor; The high-precision DEM construction and propagation modeling module is used to construct a baseline network based on the GNSS interference detection feature tensor, optimize the coordinate DEM output, and generate anti-interference high-precision DEM data; construct a three-dimensional propagation base from the anti-interference high-precision DEM data to generate a three-dimensional electromagnetic propagation base for the city; perform path simulation on the three-dimensional electromagnetic propagation base for the city, calculate the loss, and obtain a priority response list. The interference detection and situation assessment module is used to design multi-level alarm strategies from the priority response list and optimize the reward function threshold to obtain a GNSS interference detection model; based on the GNSS interference detection model, a holographic three-dimensional interference situation map is constructed, and a big data GNSS interference analysis report is generated.

Images