Artificial intelligence-based positioning interference detection method and system

Abstract

This invention discloses a location interference detection method and system based on artificial intelligence. The method includes multi-source data acquisition, signal data optimization, construction of a two-level collaborative interference detection model, and intelligent detection of location interference. This invention relates to the field of data processing technology, specifically to a location interference detection method and system based on artificial intelligence. This solution innovatively proposes a two-level collaborative detection model structure of the signal layer and the location layer, improving the accuracy of location interference detection and enhancing its environmental adaptability and comprehensiveness. By introducing an improved variational mode decomposition algorithm, fuzzy entropy combined with multi-scale coarse-grained mode classification, improved wavelet threshold function for noise reduction, and a multi-step closed-loop signal data optimization method with mode fusion reconstruction mechanism, the accuracy of signal data processing and its noise resistance stability are improved. An improved Transformer design is used for the signal interference detection layer, improving the accuracy of location signal interference type identification.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a location interference detection method and system based on artificial intelligence. Background Technology

[0002] The AI-based positioning interference detection method and system refers to a technical system that uses artificial intelligence technology to intelligently identify, classify, and detect potential interference sources in the positioning environment through real-time acquisition and analysis of multi-source heterogeneous signal data. This enables automated detection of interference behavior, promoting the transformation of positioning security detection from traditional manual analysis to real-time, intelligent, and high-precision identification, thus ensuring the reliability and stability of positioning services.

[0003] However, traditional positioning interference detection methods suffer from several technical problems. Firstly, they rely solely on signal layer features for judgment, failing to effectively reflect the actual impact of interference on positioning results, leading to low accuracy in positioning interference identification and incomplete detection results. Secondly, traditional positioning interference detection methods suffer from limitations in signal data processing, including the inability to effectively remove interference noise and retain effective signal features. Thirdly, traditional positioning interference detection methods rely on single time-domain or frequency-domain analysis for interference type identification, failing to effectively capture the dynamic characteristics of interference signals changing over time and lacking sensitivity to sudden temporal characteristics of interference, thus resulting in low accuracy in identifying positioning signal interference types. Summary of the Invention

[0004] To address the above issues and overcome the shortcomings of existing technologies, this invention provides an artificial intelligence-based method and system for detecting positioning interference. Addressing the technical problems of traditional positioning interference detection methods, which rely solely on signal layer features for judgment and fail to effectively reflect the actual impact of interference on positioning results, leading to low accuracy and incomplete detection results, this solution innovatively proposes a two-level collaborative detection model structure at the signal and positioning layers. The signal layer is responsible for detecting the existence and type of interference, and for early identification and classification of multiple interference sources. The positioning layer is responsible for evaluating changes in positioning quality and reflecting the true impact of interference on positioning performance. This achieves end-to-end comprehensive detection of positioning interference from the signal feature level to the positioning result level, significantly improving the accuracy, completeness, and depth of positioning interference detection. It can operate stably in complex environments and under multi-source interference conditions, improving the environmental adaptability and detection comprehensiveness of positioning interference detection, and effectively supporting intelligent positioning interference detection and early warning. Furthermore, addressing the technical problems of traditional positioning interference detection methods in signal data processing, such as the inability to effectively remove interference noise and retain effective signal features, this solution innovatively… This paper proposes a multi-step closed-loop signal data optimization method by introducing an improved variational mode decomposition algorithm, fuzzy entropy combined with multi-scale coarse-grained mode classification, improved wavelet threshold function for denoising, and mode fusion reconstruction mechanism. Specifically, by introducing an improved variational mode decomposition algorithm that does not require pre-setting the number of modes and constructs a multi-constraint objective function, it can adaptively decompose complex non-stationary signals, automatically splitting the original signal into different modal components. The fuzzy entropy combined with multi-scale coarse-grained mode classification enhances the quantification and mode selection capabilities of short-time-series signals, accurately distinguishing between effective positioning signal modes and interference noise modes. The improved wavelet threshold function for denoising uses a piecewise exponential threshold function to denoise interference noise modes, preserving interference features while removing high-frequency noise, achieving precise separation of interference noise and effective signals. This effectively identifies and preserves interference features in positioning signals, accurately reflects effective signal features, improves the accuracy and noise resistance stability of signal data processing, ensures the usability and consistency of the optimized signal, and enhances the accuracy, stability, and environmental adaptability of positioning interference detection.To address the technical problem of low accuracy in identifying location interference types in traditional methods, which rely on single time-domain or frequency-domain analysis and fail to effectively capture the dynamic characteristics of interference signals over time, and lack sensitivity to sudden temporal features of interference, this solution innovatively proposes an improved Transformer that incorporates relative position encoding and activation gating mechanisms. Relative position encoding accurately identifies the occurrence time and duration of interference signals, while activation gating effectively suppresses the propagation of irrelevant features, significantly reduces the impact of temporal noise, and significantly enhances the detection model's time response capability to sudden interference changes and non-stationary signal variations. This enables accurate identification of location interference signals in multi-source interference scenarios, significantly improving the accuracy of location interference type identification.

[0005] The technical solution adopted by this invention is as follows: The positioning interference detection method and system based on artificial intelligence provided by this invention includes the following steps:

[0006] Step S1: Acquisition of multi-source data;

[0007] Step S2: Signal data optimization;

[0008] Step S3: Construct a two-level cooperative interference detection model;

[0009] Step S4: Intelligent detection of location interference.

[0010] Further, in step S1, the multi-source data acquisition specifically involves performing a data acquisition operation to obtain the original interference detection data, and performing data optimization processing on the original interference detection data to obtain optimized interference detection data.

[0011] The raw interference detection data includes historical interference detection data and real-time interference detection data;

[0012] The data optimization process is used to optimize the original data, specifically including data cleaning, data standardization, and data encoding to obtain interference detection optimized data.

[0013] Both the historical interference detection data and the real-time interference detection data include signal data, positioning data, and environmental data;

[0014] The historical interference detection data also includes historical signal interference detection results and location anomaly detection results.

[0015] Further, in step S2, the signal data optimization specifically involves decomposing, filtering, denoising, and fusing the signal data to obtain optimized signal data, including the following steps:

[0016] Step S21: Adaptive signal decomposition, specifically, using an improved variational mode decomposition algorithm to adaptively decompose the original signal data to obtain the final set of mode components; including the following steps:

[0017] Step S211: Input the original signal, specifically setting the original signal to be composed of the superposition of three parts: the extracted modal component set, the current modal component to be extracted, and the unprocessed signal;

[0018] Step S212: Construct a multi-constraint objective function, specifically by constructing a multi-constraint objective function that includes a modal bandwidth minimization constraint, a residual signal minimization constraint, and a modal differentiation constraint; the formula used is as follows:

[0019] ;

[0020] In the formula, This represents the modal bandwidth minimization constraint. Indicates the center frequency of the current mode. This represents the constraint term for minimizing the residual signal. Indicates the modal differentiation constraint term. Represents the balance coefficient. Indicates residual signal, This represents the original signal at time t. This represents the Lth modal component to be extracted.

[0021] Step S213: Solving the modal components, specifically, based on the multi-constraint objective function, the constrained optimization problem is transformed into an unconstrained iterable form by using the augmented Lagrangian function, and iteratively solved using the alternating direction multiplier method. When the difference between adjacent iterations of the objective function is less than the preset iteration termination threshold, the iteration is terminated, and the modal components in the Lth decomposition are obtained.

[0022] Step S214: Modal component adaptive decomposition terminates, specifically by obtaining the modal components in the Lth decomposition after each round of decomposition and updating the residual signal. The modal component decomposition is stopped when the residual signal energy is lower than the set ratio threshold of the original signal energy, and the final modal component set is obtained; otherwise, the modal component solution is continued.

[0023] Step S22: Key mode identification and screening, specifically, involves evaluating the complexity of each mode component based on fuzzy entropy and multi-scale coarse-grained processing, and classifying the mode components using the Pearson correlation coefficient to obtain the effective localized mode components and interference noise mode components; including the following steps:

[0024] Step S221: Multi-scale coarsening processing, specifically, using multi-scale coarsening technology, each modal component in the final modal component set is coarsened according to a preset scale factor to generate non-overlapping coarsened sequences.

[0025] Step S222: Calculate the comprehensive fuzzy entropy of the modal components. Specifically, calculate the fuzzy entropy for each scale of coarse-grained sequence points, and take the average of the fuzzy entropies at the same scale to obtain the comprehensive fuzzy entropy value of the modal components; the formula used is as follows:

[0026] ;

[0027] In the formula, express The fuzzy entropy value, Representing scale The kth coarse-grained sequence point below, Indicates the length of the coarse-grained sequence. Indicates the embedding dimension, which controls the length of similar vectors in the sequence. Indicates the tolerance for similarity. Represents the gradient of an exponential function. Indicates in In the middle, from the point in time Starting from the first The sequence values ​​of each sampling point Indicates in In the middle, from the point in time Starting from the first The sequence values ​​of each sampling point Represents the fuzzy membership function. This represents the maximum difference between two embedding vectors at the l-th element;

[0028] Step S223: Modal component classification, specifically, first calculate the j-th modal component. and the original signal Pearson correlation coefficient between Then, based on the modal component comprehensive fuzzy entropy value and Pearson correlation coefficient, the modal components are classified. If the Pearson correlation coefficient of the modal components is... And the comprehensive fuzzy entropy value If the modal component is valid, it is a valid positioning modal component; otherwise, it is an interference noise modal component. Thresholds for Pearson correlation coefficients are assigned to modal components. This represents the threshold of the comprehensive fuzzy entropy for modal component partitioning. This represents the combined fuzzy entropy value of the j-th modal component;

[0029] Step S23: Interference noise mode denoising. Specifically, first, wavelet transform is performed on the interference noise mode components to decompose them into multiple frequency sub-bands. Then, piecewise exponential functions are used to denoise the high-frequency detail coefficients of the interference modes. Finally, inverse wavelet transform is performed to reconstruct the interference noise mode components after removing high-frequency noise into time-domain signals, thus obtaining the denoised interference noise mode components. The formula used is as follows:

[0030] ;

[0031] In the formula, This represents the wavelet coefficients after thresholding. Represents the original wavelet coefficients after wavelet transform. Indicates the threshold parameter. This represents the regularization parameter, used to control the smoothness of the threshold function. and Represents the exponential smoothing parameter. Represents a symbolic function;

[0032] Step S24: Effective signal fusion and reconstruction, specifically, the effective positioning mode component and the noise reduction interference mode component are fused and reconstructed to obtain optimized signal data that fully reflects the interference characteristics while maintaining the effective positioning characteristics.

[0033] Furthermore, in step S3, the construction of the two-level cooperative interference detection model specifically includes the following steps:

[0034] Step S31: Design a signal interference detection layer, specifically by converting the signal interference detection feature set into a feature sequence through a linear projection layer, and inputting it into an improved Transformer model that incorporates relative position encoding and activation gating mechanisms for processing to obtain interference detection features; including the following steps:

[0035] Step S311: Constructing relative position encoding, specifically based on the index difference between position i and position j in the feature sequence. Construct the relative distance function value and calculate the bond-side position factor. Sum-side position factor The formula used is as follows:

[0036] ;

[0037] In the formula, This represents the relative distance function value between the i-th position and the j-th position. This represents the amplitude factor, used to control the overall amplitude of the distance function. Indicates the attenuation coefficient;

[0038] Step S312: Multi-head attention calculation, specifically, constructing query matrix, key matrix and value matrix respectively through linear projection operation on input feature sequence, then introducing activation gating mechanism to perform gating linear transformation operation on the three matrices to obtain query gating matrix, key gating matrix and value gating matrix, and performing single-head attention weight calculation in combination with key side position factor and value side position factor to obtain the output corresponding to each attention head, and finally concatenating the outputs of each attention head and obtaining key interference features through linear mapping;

[0039] Step S313: Obtaining interference detection features, specifically by inputting key interference features into the feedforward network layer, performing nonlinear transformation on the key interference features, and combining residual connections and layer normalization operations to obtain interference detection features;

[0040] Step S32: Design a localization anomaly detection layer, specifically by inputting the localization anomaly feature set into a bidirectional long short-term memory network to obtain localization anomaly detection features;

[0041] Step S33: Output layer of the dual-level detection model, specifically, the interference detection features and the localization anomaly detection features are input into the signal interference classification output layer and the localization anomaly state classification output layer respectively to obtain the signal interference detection results and the localization anomaly detection results;

[0042] Step S34: Detection model training, specifically, establishing an initial dual-level cooperative interference detection model through the designed signal interference detection layer, the designed positioning anomaly detection layer, and the dual-level detection model output layer, inputting historical interference detection training data into the initial dual-level cooperative interference detection model, and training the model to obtain the trained dual-level cooperative interference detection model.

[0043] Furthermore, in step S4, the intelligent detection of positioning interference specifically involves inputting real-time detection data into the trained dual-level collaborative interference detection model to obtain real-time signal interference detection results and real-time positioning anomaly detection results. Based on the dual detection results, positioning interference early warning information is generated to realize intelligent detection and early warning of positioning interference.

[0044] The technical solution adopted by the present invention is as follows: The positioning interference detection system based on artificial intelligence provided by the present invention includes a multi-source data acquisition module, a signal data optimization module, a module for constructing a two-level collaborative interference detection model, and a positioning interference intelligent detection module;

[0045] The multi-source data acquisition module obtains interference detection optimization data through data acquisition operations and data optimization processing, and sends the data to the signal data optimization module, the two-level collaborative interference detection model construction module, and the intelligent interference detection module.

[0046] The signal data optimization module receives data sent by the multi-source data acquisition module. Specifically, it first uses an improved variational mode decomposition algorithm that does not require a preset number of modes and constructs a multi-constraint objective function to adaptively decompose the original signal data. Then, it introduces fuzzy entropy combined with multi-scale coarsening technology to classify the mode components. Subsequently, it uses a piecewise exponential improved wavelet threshold function to denoise the interference noise mode components. Finally, it fuses the effective positioning mode components with the denoised interference noise mode components to obtain optimized signal data, and sends the data to the module for constructing a two-level collaborative interference detection model and the intelligent positioning interference detection module.

[0047] The module for constructing a dual-level collaborative interference detection model receives data from the multi-source data acquisition module and the signal data optimization module. Specifically, it first designs a signal interference detection layer based on an improved Transformer model that incorporates relative position encoding and activation gating mechanism. Secondly, it designs a positioning anomaly detection layer based on a bidirectional long short-term memory network. Subsequently, in the output layer of the dual-level detection model, it classifies and distinguishes the signal interference status and the positioning anomaly status, generating signal interference detection results and positioning anomaly detection results. Finally, it trains the model to obtain the trained dual-level collaborative interference detection model and sends the data to the positioning interference intelligent detection module.

[0048] The intelligent positioning interference detection module receives data from the multi-source data acquisition module, the signal data optimization module, and the module for constructing a two-level collaborative interference detection model. Specifically, it inputs real-time detection data into the trained detection model to obtain real-time signal interference detection results and real-time positioning anomaly detection results.

[0049] The beneficial effects achieved by the present invention using the above solution are as follows:

[0050] (1) In view of the technical problems in traditional positioning interference detection methods, which rely solely on signal layer features for single judgment and cannot effectively reflect the actual impact of interference on positioning results, resulting in low accuracy of positioning interference identification and incomplete detection results, this solution innovatively proposes a dual-level collaborative detection model structure of signal layer and positioning layer. The signal layer is responsible for detecting the existence and type of interference and early identification and classification of multiple interference sources. The positioning layer is responsible for evaluating changes in positioning quality and reflecting the real impact of interference on positioning performance. This achieves end-to-end comprehensive positioning interference detection from signal feature level to positioning result level, significantly improving the accuracy, completeness and depth of positioning interference detection. It can operate stably in complex environments and under multi-source interference backgrounds, improving the environmental adaptability and detection comprehensiveness of positioning interference detection, and effectively supporting intelligent positioning interference detection and early warning.

[0051] (2) To address the technical problem in traditional positioning interference detection methods that fail to effectively remove interference noise and retain effective signal features during signal data processing, this solution innovatively introduces a multi-step closed-loop signal data optimization method that combines an improved variational mode decomposition algorithm, fuzzy entropy with multi-scale coarse-grained mode classification, and an improved wavelet threshold function for noise reduction and mode fusion reconstruction. Specifically, by introducing an improved variational mode decomposition algorithm that does not require a pre-defined number of modes and constructs a multi-constraint objective function, it can adaptively decompose complex non-stationary signals, automatically splitting the original signal into different modal components. This is achieved by combining fuzzy entropy with multi-scale coarse-grained mode classification. This mode classification enhances the quantification and mode selection capabilities of complex short-time-series signals, accurately distinguishing between effective positioning signal modes and interference noise modes. By improving the wavelet threshold function for noise reduction, a piecewise exponential threshold function is used to denoise the interference noise modes, preserving interference features while removing high-frequency noise. This achieves precise separation of interference noise from effective signals, effectively identifying and preserving interference features in positioning signals, accurately reflecting the characteristics of effective signals, improving the accuracy and noise resistance stability of signal data processing, ensuring the usability and consistency of optimized signals, and enhancing the accuracy, stability, and environmental adaptability of positioning interference detection.

[0052] (3) In view of the technical problem that the traditional positioning interference detection method relies on a single time domain or frequency domain analysis for positioning interference type identification, which cannot effectively capture the dynamic characteristics of the interference signal changing over time and is not sensitive enough to the sudden temporal characteristics of interference, resulting in low accuracy of positioning signal interference type identification, this solution innovatively proposes an improved Transformer that introduces relative position coding and activation gating mechanism. Through relative position coding, the occurrence time and duration of interference signal can be accurately identified. Through activation gating, irrelevant feature propagation can be effectively suppressed, the influence of temporal noise can be significantly reduced, and the time response capability of the detection model to interference abrupt changes and non-stationary signal changes can be significantly enhanced. In the multi-source interference scenario, accurate identification of positioning interference signal is achieved, and the accuracy of positioning signal interference type identification is significantly improved. Attached Figure Description

[0053] Figure 1 A flowchart illustrating the AI-based location interference detection method provided by this invention;

[0054] Figure 2 A schematic diagram of the modules of the AI-based positioning interference detection system provided by the present invention;

[0055] Figure 3 This is a flowchart illustrating step S2;

[0056] Figure 4 This is a flowchart illustrating step S21;

[0057] Figure 5 This is a flowchart illustrating step S22;

[0058] Figure 6 This is a flowchart illustrating step S3;

[0059] Figure 7 This is a flowchart illustrating step S31;

[0060] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0061] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0062] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the system or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0063] Example 1, see Figure 1 The technical solution adopted by this invention is as follows: The positioning interference detection method based on artificial intelligence provided by this invention includes the following steps:

[0064] Step S1: Multi-source data acquisition. Through data acquisition operations and data optimization processing, interference detection optimized data is obtained.

[0065] Step S2: Signal data optimization, used to optimize the signal data in the original interference detection data, retaining the effective positioning signal features while suppressing noise interference; specifically, firstly, an improved variational mode decomposition algorithm that does not require a preset number of modes and constructs a multi-constraint objective function is used to adaptively decompose the original signal data; then, fuzzy entropy combined with multi-scale coarsening technology is introduced to classify the mode components; subsequently, the interference noise mode components are denoised using a piecewise exponential improved wavelet threshold function; finally, the effective positioning mode components and the denoised interference noise mode components are fused to obtain the optimized signal data.

[0066] Step S3: Construct a two-level collaborative interference detection model to comprehensively analyze the signal layer and the positioning layer, and realize two-level collaborative detection from interference source identification to positioning impact assessment. Specifically, firstly, a signal interference detection layer is designed based on an improved Transformer model that introduces relative position encoding and activation gating mechanism. Secondly, a positioning anomaly detection layer is designed based on a bidirectional long short-term memory network. Subsequently, in the output layer of the two-level detection model, the signal interference status and the positioning anomaly status are classified and judged respectively to generate signal interference detection results and positioning anomaly detection results. Finally, the model is trained to obtain the trained two-level collaborative interference detection model.

[0067] Step S4: Intelligent detection of localization interference, specifically, inputting real-time detection data into the trained detection model to obtain real-time signal interference detection results and real-time localization anomaly detection results.

[0068] By performing the above operations, this solution addresses the technical problems of traditional positioning interference detection methods, which rely solely on signal layer features for judgment and cannot effectively reflect the actual impact of interference on positioning results, leading to low accuracy in positioning interference identification and incomplete detection results. It innovatively proposes a two-level collaborative detection model structure with signal and positioning layers. The signal layer is responsible for detecting the existence and type of interference, and for the early identification and classification of multiple interference sources. The positioning layer is responsible for evaluating changes in positioning quality and reflecting the true impact of interference on positioning performance. This achieves end-to-end comprehensive positioning interference detection from the signal feature level to the positioning result level, significantly improving the accuracy, completeness, and depth of positioning interference detection. It can operate stably in complex environments and under multi-source interference conditions, improving the environmental adaptability and detection comprehensiveness of positioning interference detection, and effectively supporting intelligent positioning interference detection and early warning.

[0069] Example 2, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In step S1, the multi-source data acquisition specifically involves first performing data acquisition operations through multiple sensors and detection devices to obtain multi-source data related to positioning interference, obtaining raw interference detection data, and then performing data optimization processing on the raw interference detection data to obtain optimized interference detection data.

[0070] The raw interference detection data includes historical interference detection data and real-time interference detection data;

[0071] The data optimization process is used to optimize the original data, specifically including data cleaning, data standardization, and data encoding to obtain interference detection optimized data.

[0072] Both the historical interference detection data and the real-time interference detection data include signal data, positioning data, and environmental data;

[0073] The historical interference detection data also includes historical signal interference detection results and location anomaly detection results;

[0074] The interference detection optimization data includes historical interference detection optimization data and real-time interference detection optimization data;

[0075] The signal interference detection results include no interference, narrowband interference, broadband noise interference, and spoofing signal interference;

[0076] The positioning anomaly detection results include normal, error exceeding the limit, drift, and signal loss;

[0077] The error exceeding the limit specifically means that the positioning accuracy exceeds the predetermined tolerance range;

[0078] The drift specifically refers to the gradual decrease in the accuracy of the positioning system and its gradual deviation from the correct trajectory.

[0079] The signal loss specifically refers to the inability to receive sufficient satellite signals, resulting in the inability to obtain valid location information;

[0080] The historical interference detection data refers to the data previously recorded by the positioning system or signal receiving device, which is used as training data for model training.

[0081] The signal data is specifically collected through radio receiving equipment or base stations;

[0082] The positioning data specifically refers to positioning data collected through GNSS receiver equipment, including location coordinates, number of satellites, HDOP, PDOP, direction of movement of the positioning device, speed, acceleration, and positioning signal strength;

[0083] The positioning device refers to a device that needs to use positioning technology to perform navigation, positioning or other location-related operations.

[0084] The environmental data specifically refers to the environmental data around the positioning device used to analyze the impact of the environment on signal propagation, including building density, height, temperature, humidity, air pressure, and weather conditions.

[0085] Data cleaning is used to ensure the accuracy, completeness, and consistency of the original data. Specifically, it involves cleaning the original data for interference detection to remove invalid, missing, or abnormal data, including handling missing data and removing outliers. The handling of missing data specifically involves filling in missing values ​​in the original data using the mean imputation method. The removal of outliers specifically involves detecting and removing extreme values ​​and logical anomalies in the original data using the Z-Score algorithm.

[0086] Data standardization is used to eliminate the dimensional differences between different data features. Specifically, it maps all continuous variables to the [0,1] interval through the min-max normalization method to unify the scale of the data.

[0087] Data encoding processing is used to successfully convert categorical variables into numerical data. Specifically, it uses one-hot encoding to encode the categorical fields in the original data, converting discrete fields into sparse numerical vectors.

[0088] Example 3, see Figure 1 , Figure 2 , Figure 3 , Figure 4 and Figure 5 This embodiment is based on the above embodiment. In step S2, the signal data optimization is used to optimize the signal data, retaining effective positioning signal features while suppressing noise interference. Specifically, it involves decomposing, filtering, denoising, and fusing the signal data to obtain optimized signal data, including the following steps:

[0089] Step S21: Adaptive signal decomposition, used to separate the effective components and interference components in the original signal at the modal level, solving the strong non-stationarity and aliasing problems of different frequency components in the positioning signal, thereby achieving fine-grained modal-level processing. Specifically, it involves adaptively decomposing the original signal data using an improved variational mode decomposition algorithm to obtain the final set of modal components; including the following steps:

[0090] Step S211: Raw signal input, used to process the signal data. Specifically, the raw signal is set to be composed of the superposition of three parts: the extracted modal component set, the current modal component to be extracted, and the unprocessed signal; the formula used is as follows:

[0091] ;

[0092] In the formula, This represents the original signal at time t, i.e., the signal data in the interference detection optimization data. This represents the set of modal components that have been successfully extracted in the first L−1 decompositions. This represents the i-th modal component, i.e., the modal component extracted in the i-th iteration. This represents the Lth modal component to be extracted. This indicates an unprocessed signal;

[0093] Step S212: Construct a multi-constraint objective function to ensure the stability of mode decomposition, avoid mode aliasing, and extract the true physical components of the signal. Specifically, this involves constructing a multi-constraint objective function that includes a mode bandwidth minimization constraint, a residual signal minimization constraint, and a mode discrimination constraint. The formula used is as follows:

[0094] ;

[0095] ;

[0096] ;

[0097] ;

[0098] ;

[0099] ;

[0100] In the formula, This represents a modal bandwidth minimization constraint, used to limit the frequency bandwidth of the modes. This represents the first-order derivative operator in the time domain, used to quantify the smoothness of mode changes over time. This represents the impulse function, used to construct the analytical form of a signal. Represents the Hilbert transform kernel. This represents the convolution operation. This represents the rotation factor, used to modulate the signal to the vicinity of the corresponding center frequency. Indicates the center frequency of the current mode. This represents a constraint term that minimizes the residual signal. It is used to ensure the integrity of signal decomposition and controls the extracted modes to fit the original signal as closely as possible, thereby minimizing the residual. This represents a mode discrimination constraint, used to avoid frequency aliasing between adjacent modes and ensure that interference noise mode components and valid signal mode components can be correctly distinguished. This represents the center frequency variable, used to describe the filter's response across the entire frequency axis. This represents the center frequency of the i-th extracted modal component. This represents the regularization coefficient, used to adjust the smoothness of the filter. The frequency domain response function of the discriminative filter for the i-th extracted mode component is represented by the function of... The frequency domain representation is obtained through Fourier transform, which describes the response characteristics of the orthogonal filter function in the frequency domain. This represents the time-domain filtering function for the i-th extracted modal component. Represents the square of the L2 norm. This represents the balance coefficient, set to 0.25. This indicates that optimization should be performed under these conditions; it is a constraint symbol. Indicates residual signal;

[0101] Step S213: Modal component solution, specifically, based on a multi-constraint objective function, the constrained optimization problem is transformed into an unconstrained iterable form using an augmented Lagrangian function, and the alternating direction multiplier method is employed. , , and Lagrange multipliers The iteration is performed, and the iteration is terminated when the difference between adjacent iterations of the objective function is less than a preset iteration termination threshold, thus obtaining the modal components in the Lth decomposition; the formula used is as follows:

[0102] The iterative solution specifically involves performing the following steps sequentially in each iteration: renew, Correction, Update and Lagrange multipliers Correction;

[0103] ;

[0104] ;

[0105] In the formula, This represents the augmented Lagrange function. This represents the penalty parameter, used to control the convergence speed and stability of the constraint terms during the optimization process. This represents the iterative difference of the objective function, used to measure the degree of convergence of the iterations. This indicates the preset iteration termination threshold. This represents the value of the augmented Lagrange function at the nth iteration. Indicates the first The augmented Lagrange function value at the next iteration Indicates inner product operation;

[0106] Step S214: Adaptive termination of modal component decomposition. This step is used to determine the energy of the remaining undecomposed signal, thereby achieving adaptive termination of the decomposition process and preventing over-decomposition or under-decomposition, ensuring a reasonable number of modes and complete signal features. Specifically, after each round of decomposition, the optimal modal components in the Lth decomposition are obtained, and the residual signal is updated. It calculates the ratio of residual signal energy to the total energy of the original signal, and determines whether the residual signal energy is lower than a set threshold ratio of the original signal energy. If the condition is met, then the modal component decomposition stops, and the final set of modal components is obtained; otherwise, the modal component solution continues. The formulas used are as follows:

[0107] ;

[0108] ;

[0109] ;

[0110] In the formula, Represents the original signal energy. Represents residual signal energy. Indicates the total duration of the signal;

[0111] Step S22: Key mode identification and screening, used to identify effective localization modes and interference noise modes from the multiple modes obtained by decomposition, avoiding irrelevant modes from affecting the accuracy of interference identification. Specifically, it involves evaluating the complexity of each mode component based on fuzzy entropy and multi-scale coarse-grained processing, and classifying the mode components using Pearson correlation coefficient to obtain effective localization mode components and interference noise mode components; including the following steps:

[0112] Step S221: Multi-scale coarsening processing is used to generate coarse-grained sequences at different scales, reducing the impact of high-frequency noise and enhancing the stability of complexity assessment. Specifically, multi-scale coarsening technology is used to coarse-grain each modal component in the final modal component set according to a preset scale factor, generating non-overlapping coarse-grained sequences. The formula used is as follows:

[0113] ;

[0114] In the formula, Representing scale The k-th coarse-grained sequence point is used to represent the average eigenvalue of the signal at this time scale. The scale factor, representing multi-scale coarsening, determines the size of each window to adapt to the temporal characteristics of the positioning signal. This represents the value of the j-th modal component in the final modal component set at time te. This represents the j-th modal component;

[0115] Step S222: Calculate the modal component comprehensive fuzzy entropy, which is used to quantify the complexity of each coarse-grained sequence. Specifically, this involves calculating the coarse-grained sequence points at each scale. The fuzzy entropy is calculated, and the average value of the fuzzy entropy at the same scale is taken to obtain the comprehensive fuzzy entropy value of the modal components; the formula used is as follows:

[0116] ;

[0117] ;

[0118] In the formula, express The fuzzy entropy value, Indicates the length of the coarse-grained sequence. This represents the embedding dimension, controlling the length of similar vectors in the sequence; it is set to 2. This represents the similarity tolerance, used to define the threshold for sequence similarity. This represents the gradient of the exponential function, used to smooth the membership function, and has a value of 2. Indicates in In the middle, from the point in time Starting from the first The sequence values ​​of each sampling point Indicates in In the middle, from the point in time Starting from the first The sequence values ​​of each sampling point Represents the fuzzy membership function. This represents the maximum difference between two embedding vectors at the l-th element. This represents the combined fuzzy entropy value of the j-th modal component;

[0119] Step S223: Modal component classification, used to filter out effective localization modes and interference noise modes; specifically, first calculate the j-th modal component. and the original signal Pearson correlation coefficient between Then, based on the modal component comprehensive fuzzy entropy value and Pearson correlation coefficient, the modal components are classified. If the Pearson correlation coefficient of the modal components is... And the comprehensive fuzzy entropy value If the modal component is valid, it is a valid positioning modal component; otherwise, it is an interference noise modal component. The Pearson correlation coefficient threshold for the modal components was set to 0.75. The threshold for the comprehensive fuzzy entropy of modal component partitioning is set to 1.5;

[0120] Step S23: Interference noise mode denoising, used to perform targeted denoising on the identified interference noise modes, suppressing broadband noise while retaining key features of narrowband interference and spoofing signals, avoiding the loss of interference information. Specifically, firstly, wavelet transform is performed on the interference noise mode components to decompose them into multiple frequency sub-bands. Then, piecewise exponential functions are used to denoise the high-frequency detail coefficients of the interference modes. Finally, inverse wavelet transform is performed to reconstruct the interference noise mode components after removing high-frequency noise into time-domain signals, obtaining the denoised interference noise mode components. ;

[0121] The wavelet transform specifically involves selecting the db6 wavelet basis and performing a 3-level wavelet decomposition.

[0122] The piecewise exponential function is used to suppress high-frequency noise while retaining the effective components of the signal, avoiding the signal oscillation problem caused by traditional hard thresholding and the fixed deviation problem of soft thresholding;

[0123] The formula used is as follows:

[0124] ;

[0125] In the formula, This represents the wavelet coefficients after thresholding. Represents the original wavelet coefficients after wavelet transform. This represents a threshold parameter used to control which frequency components in a signal are considered noise. This represents the regularization parameter, used to control the smoothness of the threshold function. Setting it to 2 determines the degree of denoising of the wavelet coefficients, balancing the relationship between noise removal and signal feature preservation. and This represents the exponential smoothing parameter, set to 0.8 and 8 respectively. Represents a symbolic function;

[0126] Step S24: Effective signal fusion and reconstruction, used to reconstruct the signal from the effective positioning mode and the noise-reduced interference mode to recover complete signal data. Specifically, the effective positioning mode components and the noise-reduced interference mode components are fused and reconstructed to obtain optimized signal data that fully reflects the interference characteristics while maintaining the effective positioning features. The formula used is as follows:

[0127] ;

[0128] In the formula, This represents the optimized signal data at time t. Indicates the effective positioning mode components, Indicates the number of effective positioning modal components. This indicates the number of modal components of the interference noise after noise reduction. An index representing the number of valid signal modes.

[0129] By performing the above operations, this solution addresses the technical problem in traditional positioning interference detection methods where signal data processing cannot effectively remove interference noise and retain effective signal features. It innovatively introduces a multi-step closed-loop signal data optimization method that incorporates an improved variational mode decomposition algorithm, fuzzy entropy combined with multi-scale coarse-grained mode classification, and an improved wavelet threshold function for noise reduction and mode fusion reconstruction. Specifically, by introducing an improved variational mode decomposition algorithm that does not require a pre-defined number of modes and constructs a multi-constraint objective function, it can adaptively decompose complex non-stationary signals, automatically splitting the original signal into different modal components. This is achieved through fuzzy entropy combined with multi-scale coarse-grained mode classification. Granular mode classification enhances the quantification and mode selection capabilities of complex short-time-series signals, accurately distinguishing between effective positioning signal modes and interference noise modes. By improving the wavelet threshold function for noise reduction, a piecewise exponential threshold function is used to denoise the interference noise modes, preserving interference features while removing high-frequency noise. This achieves precise separation of interference noise from effective signals, effectively identifying and preserving interference features in positioning signals, accurately reflecting the characteristics of effective signals, improving the accuracy and noise resistance stability of signal data processing, ensuring the usability and consistency of optimized signals, and enhancing the accuracy, stability, and environmental adaptability of positioning interference detection.

[0130] Example 4, see Figure 1 , Figure 2 , Figure 6 and Figure 7 This embodiment is based on the above embodiment. In step S3, the construction of the two-level cooperative interference detection model specifically includes the following steps:

[0131] Step S31: Design a signal interference detection layer to detect the type and presence of interference in the signal and provide early warning; specifically, the signal interference detection feature set is converted into a feature sequence through a linear projection layer, and then input into an improved Transformer model that incorporates relative position encoding and activation gating mechanisms for processing to obtain interference detection features; including the following steps:

[0132] The signal interference detection feature set specifically involves performing time-domain analysis, frequency-domain analysis, and time-frequency analysis on optimized signal data to obtain time-domain features, frequency-domain features, and time-frequency features. These features are then concatenated under a discrete-time index to form a structured signal interference detection feature set. The time-domain features include signal strength, envelope mean, variance, kurtosis, and instantaneous energy.

[0133] The frequency domain features include power spectral density, spectral peak position, effective bandwidth, and spectral entropy;

[0134] The time-frequency features include a time-frequency plot and a wavelet scaling plot;

[0135] Step S311: Relative position encoding construction, used to enhance the model's ability to analyze the relative position information of the input feature sequence in the time dimension, specifically by using the index difference between position i and position j in the feature sequence. The relative distance function values ​​are constructed, and the key-side position factor and value-side position factor are calculated; the formulas used are as follows:

[0136] ;

[0137] ;

[0138] In the formula, This represents the relative distance function value between the i-th position and the j-th position. and These represent the bond-side relative position factor and the value-side relative position factor, respectively. This represents the amplitude factor, with a value range of [value range missing]. Used to control the overall magnitude of the distance function. Denotes the attenuation coefficient, and satisfies This is used to achieve the weighting characteristic that decreases exponentially as the position difference increases. and These represent the learnable positional weight parameters on the key side and the value side, respectively. Indicates the base of the natural index;

[0139] Step S312: Multi-head attention calculation, specifically, involves constructing a query matrix, key matrix, and value matrix from the input feature sequence through linear projection operations. Then, activation gating mechanisms are introduced into each of the three matrices to perform gated linear transformation operations, resulting in query-gated matrices, key-gated matrices, and value-gated matrices. Single-head attention weights are then calculated using key-side and value-side position factors to obtain the outputs of each attention head. Finally, the outputs of each attention head are concatenated and linearly mapped to obtain the key interference features. The formula used is as follows:

[0140] ;

[0141] ;

[0142] ;

[0143] ;

[0144] ;

[0145] In the formula, , and Representing the query matrix Key matrix Sum matrix The linear transformation weight matrix, , and Representing the query matrix Key matrix Sum matrix The linear transformation bias term parameters, , and Representing the query matrix Key matrix Sum matrix The gated transform bias term parameters, , and These represent the gating transformation weight matrices for the query matrix, key matrix, and value matrix, respectively. This represents the Sigmoid activation function. This represents element-wise multiplication. Indicates querying the gated matrix. Represents the key-gated matrix. Value-gated matrix, This indicates a gated linear transformation operation. This represents the query vector corresponding to the i-th time position in the query matrix. This represents the key vector corresponding to the j-th time position in the key matrix. This represents the value vector corresponding to the j-th time position in the value matrix. The dimension of the key vector. This represents the attention relevance score between position i and position j. Let represent the attention score between the i-th position and the z-th position. This represents the normalized attention weights. Indicates the length of the input feature sequence. This represents the output vector of the h-th attention head. This represents the output vector of the first attention head. This indicates a splicing operation. Indicates key interference characteristics. Indicates the output mapping matrix;

[0146] Step S313: Obtaining interference detection features, specifically, inputting key interference features into the feedforward network layer, performing nonlinear transformation on the key interference features, and combining residual connections and layer normalization operations to obtain interference detection features for interference type identification.

[0147] Step S32: Design a localization anomaly detection layer to evaluate the quality of localization; specifically, input the localization anomaly feature set into a bidirectional long short-term memory network, model the feature sequence through a forward long short-term memory network and a backward long short-term memory network respectively, obtain the forward hidden state and the backward hidden state at each time step, and concatenate the features of the forward hidden state and the backward hidden state at the same time step to form a localization anomaly detection feature sequence. Finally, perform max pooling on the localization anomaly detection feature sequence to obtain the localization anomaly detection features.

[0148] The location anomaly feature set is specifically formed by aligning and splicing the location data, environmental data and interference detection features in the interference detection optimization data on the same time axis to form a location anomaly feature set organized according to time series.

[0149] Step S33: The dual-level detection model output layer is used to classify and distinguish the signal interference status and the positioning anomaly status respectively, forming a dual-level positioning quality detection result. Specifically, the interference detection features and the positioning anomaly detection features are input into the signal interference classification output layer and the positioning anomaly status classification output layer respectively to obtain the signal interference detection result and the positioning anomaly detection result.

[0150] The signal interference classification output layer takes interference detection features as input, calculates the probability distribution of each type of signal interference through a fully connected layer and a Softmax classifier, and selects the category index corresponding to the type with the highest probability as the signal interference detection result.

[0151] The location anomaly state classification output layer specifically takes the location anomaly detection features as input, calculates the probability distribution of each location anomaly state through a fully connected layer and a Softmax classifier, and selects the category index corresponding to the state with the highest probability as the location anomaly detection result.

[0152] Step S34: Detection model training, used to supervise the learning of the constructed two-level cooperative interference detection model. Specifically, an initial two-level cooperative interference detection model is established through the designed signal interference detection layer, the designed localization anomaly detection layer and the output layer of the two-level detection model. Historical interference detection training data is input into the initial two-level cooperative interference detection model for model training to obtain the trained two-level cooperative interference detection model.

[0153] The historical interference detection training data specifically consists of the location data and environmental data from the historical interference detection optimization data, along with the historical optimized signal data.

[0154] The historical optimized signal data is specifically the signal data obtained by optimizing the signal data in the historical interference detection optimized data.

[0155] The model training process involves first using the cross-entropy loss function to measure the prediction errors of signal interference classification and anomaly localization classification, respectively. The losses of the two tasks are then weighted and summed according to set weights to construct the total loss function. The weight matrix and bias parameters in the model are then iteratively updated using the backpropagation algorithm and gradient descent optimization method. During the training process, the model parameters are continuously optimized through multiple iterations. The iterative training stops when the preset maximum number of training iterations is reached or the total loss function value converges to the set iteration termination threshold.

[0156] By performing the above operations, this solution addresses the technical problem in traditional positioning interference detection methods that rely on single time-domain or frequency-domain analysis for positioning interference type identification, failing to effectively capture the dynamic characteristics of interference signals changing over time and lacking sensitivity to sudden temporal features of interference, thus resulting in low accuracy in identifying positioning signal interference types. This solution innovatively proposes an improved Transformer that introduces relative position coding and activation gating mechanisms. Relative position coding accurately identifies the occurrence time and duration of interference signals, while activation gating effectively suppresses the propagation of irrelevant features, significantly reduces the impact of temporal noise, and significantly enhances the detection model's time response capability to sudden interference changes and non-stationary signal variations. This enables accurate identification of positioning interference signals in multi-source interference scenarios, significantly improving the accuracy of positioning signal interference type identification.

[0157] Example 5, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In step S4, the intelligent detection of positioning interference specifically involves inputting real-time detection data into the trained dual-level collaborative interference detection model to obtain real-time signal interference detection results and real-time positioning anomaly detection results. Based on the dual detection results, positioning interference early warning information is generated to realize intelligent detection and early warning of positioning interference.

[0158] The real-time detection data specifically consists of positioning data, environmental data, and real-time optimized signal data from the real-time interference detection optimization data.

[0159] The real-time optimized signal data is specifically the signal data obtained by optimizing the signal data in the real-time interference detection optimization data.

[0160] The specific rule for generating the positioning interference warning information is as follows: if the real-time signal interference detection result is no interference and the real-time positioning anomaly detection result is normal, then no positioning interference warning information is generated; otherwise, positioning interference warning information is generated.

[0161] The location interference warning information includes the current signal interference type and the current location anomaly status.

[0162] Example 6, see Figure 1 and Figure 2 Based on the above embodiments, the technical solution adopted by the present invention is as follows: The positioning interference detection system based on artificial intelligence provided by the present invention includes a multi-source data acquisition module, a signal data optimization module, a module for constructing a two-level collaborative interference detection model, and a positioning interference intelligent detection module;

[0163] The multi-source data acquisition module obtains interference detection optimization data through data acquisition operations and data optimization processing, and sends the data to the signal data optimization module, the two-level collaborative interference detection model construction module, and the intelligent interference detection module.

[0164] The signal data optimization module receives data sent by the multi-source data acquisition module and optimizes the signal data to retain effective positioning signal features while suppressing noise interference. Specifically, it first uses an improved variational mode decomposition algorithm that does not require a preset number of modes and constructs a multi-constraint objective function to adaptively decompose the original signal data. Then, it introduces fuzzy entropy combined with multi-scale coarsening technology to classify the mode components. Subsequently, it uses a piecewise exponential improved wavelet threshold function to denoise the interference noise mode components. Finally, it fuses the effective positioning mode components with the denoised interference noise mode components to obtain optimized signal data, which is then sent to the dual-level collaborative interference detection model module and the positioning interference intelligent detection module.

[0165] The module for constructing a dual-level collaborative interference detection model receives data from the multi-source data acquisition module and the signal data optimization module. This data is used to comprehensively analyze the signal layer and the positioning layer, achieving dual-level collaborative detection from interference source identification to positioning impact assessment. Specifically, firstly, a signal interference detection layer is designed based on an improved Transformer model that incorporates relative position encoding and activation gating mechanisms. Secondly, a positioning anomaly detection layer is designed based on a bidirectional long short-term memory network. Subsequently, in the output layer of the dual-level detection model, the signal interference status and positioning anomaly status are classified and judged respectively, generating signal interference detection results and positioning anomaly detection results. Finally, the model is trained to obtain the trained dual-level collaborative interference detection model, and the data is sent to the positioning interference intelligent detection module.

[0166] The intelligent positioning interference detection module receives data from the multi-source data acquisition module, the signal data optimization module, and the module for constructing a two-level collaborative interference detection model. Specifically, it inputs real-time detection data into the trained detection model to obtain real-time signal interference detection results and real-time positioning anomaly detection results.

[0167] It should be noted that, in this document, the terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0168] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

[0169] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A location interference detection method based on artificial intelligence, characterized in that: The method includes the following steps: Step S1: Multi-source data acquisition. Through data acquisition operations and data optimization processing, interference detection optimized data is obtained. Step S2: Signal data optimization. Specifically, the signal data is first adaptively decomposed using an improved variational mode decomposition algorithm. Then, fuzzy entropy combined with multi-scale coarsening technology is introduced to classify the mode components. Subsequently, the interference noise mode components are denoised using a piecewise exponential improved wavelet threshold function. Finally, the effectively located mode components are fused with the denoised interference noise mode components to obtain optimized signal data. The improved variational mode decomposition algorithm is specifically improved by eliminating the need to preset the number of modes and constructing a multi-constraint objective function. Step S3: Construct a two-level cooperative interference detection model. Specifically, firstly, design a signal interference detection layer based on an improved Transformer model that incorporates relative position encoding and activation gating mechanism. Secondly, design a localization anomaly detection layer based on a bidirectional long short-term memory network. Subsequently, in the output layer of the two-level detection model, classify and discriminate the signal interference status and the localization anomaly status respectively, and generate signal interference detection results and localization anomaly detection results. Finally, train the model to obtain the trained two-level cooperative interference detection model. Step S4: Intelligent detection of localization interference, specifically, inputting real-time detection data into the trained detection model to obtain real-time signal interference detection results and real-time localization anomaly detection results.

2. The location interference detection method based on artificial intelligence according to claim 1, characterized in that: In step S2, the signal data optimization specifically includes the following steps: Step S21: Signal adaptive decomposition, specifically, the original signal data is adaptively decomposed using an improved variational mode decomposition algorithm to obtain the final set of mode components; Step S22: Key mode identification and screening, specifically, the complexity of each mode component is evaluated based on fuzzy entropy and multi-scale coarse-grained processing, and the mode components are classified in combination with Pearson correlation coefficient to obtain effective localization mode components and interference noise mode components. Step S23: Interference noise mode denoising. Specifically, first, wavelet transform is performed on the interference noise mode components to decompose them into multiple frequency sub-bands. Then, piecewise exponential functions are used to denoise the high-frequency detail coefficients of the interference modes. Finally, inverse wavelet transform is performed to reconstruct the interference noise mode components after removing high-frequency noise into time-domain signals, obtaining the denoised interference noise mode components. The formulas used are as follows: ； In the formula, This represents the wavelet coefficients after thresholding. Represents the original wavelet coefficients after wavelet transform. Indicates the threshold parameter. This represents the regularization parameter, used to control the smoothness of the threshold function. and Represents the exponential smoothing parameter. Represents a symbolic function; Step S24: Effective signal fusion and reconstruction, specifically, the effective positioning mode component and the noise reduction interference mode component are fused and reconstructed to obtain optimized signal data that fully reflects the interference characteristics while maintaining the effective positioning characteristics.

3. The location interference detection method based on artificial intelligence according to claim 2, characterized in that: In step S21, the adaptive decomposition of the signal specifically includes the following steps: Step S211: Input the original signal, specifically setting the original signal to be composed of the superposition of three parts: the extracted modal component set, the current modal component to be extracted, and the unprocessed signal; Step S212: Construct a multi-constraint objective function, specifically by constructing a multi-constraint objective function that includes a modal bandwidth minimization constraint, a residual signal minimization constraint, and a modal differentiation constraint; the formula used is as follows: ； In the formula, This represents the modal bandwidth minimization constraint. Indicates the center frequency of the current mode. This represents the constraint term for minimizing the residual signal. Indicates the modal differentiation constraint term. Represents the balance coefficient. Indicates residual signal, This represents the original signal at time t. This represents the Lth modal component to be extracted. Step S213: Solving the modal components, specifically, based on the multi-constraint objective function, the constrained optimization problem is transformed into an unconstrained iterable form by using the augmented Lagrangian function, and iteratively solved using the alternating direction multiplier method. When the difference between adjacent iterations of the objective function is less than the preset iteration termination threshold, the iteration is terminated, and the modal components in the Lth decomposition are obtained. Step S214: Modal component adaptive decomposition terminates. Specifically, after each round of decomposition, the modal components in the Lth decomposition are obtained, the residual signal is updated, and the ratio of the residual signal energy to the total energy of the original signal is calculated. When the residual signal energy is lower than the set ratio threshold of the original signal energy, the modal component decomposition is stopped, and the final set of modal components is obtained. Otherwise, the modal component solution continues.

4. The location interference detection method based on artificial intelligence according to claim 2, characterized in that: In step S22, the key modality identification and screening specifically includes the following steps: Step S221: Multi-scale coarsening processing, specifically, using multi-scale coarsening technology, each modal component in the final modal component set is coarsened according to a preset scale factor to generate non-overlapping coarsened sequences. Step S222: Calculate the comprehensive fuzzy entropy of the modal components. Specifically, calculate the fuzzy entropy for each scale of coarse-grained sequence points, and take the average of the fuzzy entropies at the same scale to obtain the comprehensive fuzzy entropy value of the modal components; the formula used is as follows: ； In the formula, express The fuzzy entropy value, Representing scale Next, the kth coarse-grained sequence point, Indicates the length of the coarse-grained sequence. Indicates the embedding dimension, which controls the length of similar vectors in the sequence. Indicates the tolerance for similarity. Represents the gradient of an exponential function. Indicates in In the middle, from the point in time Starting from the first The sequence values ​​of each sampling point Indicates in In the middle, from the point in time Starting from the first The sequence values ​​of each sampling point Represents the fuzzy membership function. This represents the maximum difference between two embedding vectors at the l-th element; Step S223: Modal component classification, specifically, first calculate the j-th modal component. and the original signal Pearson correlation coefficient between Then, based on the modal component comprehensive fuzzy entropy value and Pearson correlation coefficient, the modal components are classified. If the Pearson correlation coefficient of the modal components is... And the comprehensive fuzzy entropy value If the modal component is valid, it is a valid positioning modal component; otherwise, it is an interference noise modal component. Thresholds for Pearson correlation coefficients are assigned to modal components. This represents the threshold of the comprehensive fuzzy entropy for modal component partitioning. This represents the comprehensive fuzzy entropy value of the j-th modal component.

5. The location interference detection method based on artificial intelligence according to claim 1, characterized in that: In step S3, the construction of the two-level cooperative interference detection model specifically includes the following steps: Step S31: Design a signal interference detection layer, specifically by converting the signal interference detection feature set into a feature sequence through a linear projection layer, and inputting it into an improved Transformer model that incorporates relative position encoding and activation gating mechanisms for processing to obtain interference detection features; Step S32: Design a localization anomaly detection layer, specifically by inputting the localization anomaly feature set into a bidirectional long short-term memory network to obtain localization anomaly detection features; Step S33: Output layer of the dual-level detection model, specifically, the interference detection features and the localization anomaly detection features are input into the signal interference classification output layer and the localization anomaly state classification output layer respectively to obtain the signal interference detection results and the localization anomaly detection results; Step S34: Detection model training, specifically, establishing an initial dual-level cooperative interference detection model through the designed signal interference detection layer, the designed positioning anomaly detection layer, and the dual-level detection model output layer, inputting historical interference detection training data into the initial dual-level cooperative interference detection model, and training the model to obtain the trained dual-level cooperative interference detection model.

6. The location interference detection method based on artificial intelligence according to claim 5, characterized in that: In step S31, the design of the signal interference detection layer specifically includes the following steps: Step S311: Constructing relative position encoding, specifically based on the index difference between position i and position j in the feature sequence. Construct the relative distance function value and calculate the bond-side position factor. Sum-side position factor The formula used is as follows: ； In the formula, This represents the relative distance function value between the i-th position and the j-th position. This represents the amplitude factor, used to control the overall amplitude of the distance function. Indicates the attenuation coefficient; Step S312: Multi-head attention calculation, specifically, constructing query matrix, key matrix and value matrix respectively through linear projection operation on input feature sequence, then introducing activation gating mechanism to perform gating linear transformation operation on the three matrices to obtain query gating matrix, key gating matrix and value gating matrix, and performing single-head attention weight calculation in combination with key side position factor and value side position factor to obtain the output corresponding to each attention head, and finally concatenating the outputs of each attention head and obtaining key interference features through linear mapping; Step S313: Obtaining interference detection features. Specifically, the key interference features are input into the feedforward network layer, nonlinear transformation is performed on the key interference features, and interference detection features are obtained by combining residual connections and layer normalization operations.

7. The location interference detection method based on artificial intelligence according to claim 1, characterized in that: In step S4, the intelligent detection of positioning interference specifically involves inputting real-time detection data into a trained dual-level collaborative interference detection model to obtain real-time signal interference detection results and real-time positioning anomaly detection results. Based on the dual detection results, positioning interference early warning information is generated to achieve intelligent detection and early warning of positioning interference.

8. The location interference detection method based on artificial intelligence according to claim 1, characterized in that: In step S1, the multi-source data acquisition specifically involves performing a data acquisition operation to obtain raw interference detection data, and performing data optimization processing on the raw interference detection data to obtain optimized interference detection data. The raw interference detection data includes historical interference detection data and real-time interference detection data; The data optimization process specifically includes data cleaning, data standardization, and data encoding to obtain optimized interference detection data. Both the historical interference detection data and the real-time interference detection data include signal data, positioning data, and environmental data. The historical interference detection data also includes historical signal interference detection results and positioning anomaly detection results.

9. An artificial intelligence-based positioning interference detection system, used to implement the artificial intelligence-based positioning interference detection method as described in any one of claims 1-8, characterized in that: It includes a multi-source data acquisition module, a signal data optimization module, a two-level collaborative interference detection model construction module, and a local interference intelligent detection module; The multi-source data acquisition module obtains interference detection optimization data through data acquisition operations and data optimization processing, and sends the data to the signal data optimization module, the two-level collaborative interference detection model construction module, and the intelligent interference detection module. The signal data optimization module receives data sent by the multi-source data acquisition module. Specifically, it first uses an improved variational mode decomposition algorithm that does not require a preset number of modes and constructs a multi-constraint objective function to adaptively decompose the original signal data. Then, it introduces fuzzy entropy combined with multi-scale coarsening technology to classify the mode components. Subsequently, it uses a piecewise exponential improved wavelet threshold function to denoise the interference noise mode components. Finally, it fuses the effective positioning mode components with the denoised interference noise mode components to obtain optimized signal data, and sends the data to the module for constructing a two-level collaborative interference detection model and the intelligent positioning interference detection module. The module for constructing a dual-level collaborative interference detection model receives data from the multi-source data acquisition module and the signal data optimization module. Specifically, it first designs a signal interference detection layer based on an improved Transformer model that incorporates relative position encoding and activation gating mechanism. Secondly, it designs a positioning anomaly detection layer based on a bidirectional long short-term memory network. Subsequently, in the output layer of the dual-level detection model, it classifies and distinguishes the signal interference status and the positioning anomaly status respectively, generating signal interference detection results and positioning anomaly detection results. Finally, it trains the model to obtain the trained dual-level collaborative interference detection model and sends the data to the positioning interference intelligent detection module. The intelligent positioning interference detection module receives data from the multi-source data acquisition module, the signal data optimization module, and the module for constructing a two-level collaborative interference detection model. Specifically, it inputs real-time detection data into the trained detection model to obtain real-time signal interference detection results and real-time positioning anomaly detection results.

Images