Artificial intelligence large model driven spatio-temporal information pair identification method and system

Abstract

The invention discloses a spatio-temporal information pair identification method and system driven by an artificial intelligence large model, and relates to the technical field of multi-source information identification, and the method comprises the steps: obtaining mine multi-source spatio-temporal data, carrying out the cleaning, denoising and spatio-temporal alignment processing, and generating a standard spatio-temporal data sequence; inputting the standard spatio-temporal data sequence into a pre-trained spatio-temporal correlation large model, and performing multi-modal spatio-temporal feature extraction to obtain comprehensive spatio-temporal feature representation; based on the comprehensive spatial-temporal feature representation, constructing a dynamic anomaly recognition network, and respectively performing equipment abnormal displacement recognition, personnel trajectory deviation safety area recognition and geographic area and safety production data anomaly association recognition to generate an initial anomaly information pair set; and performing space-time confidence coefficient compensation and risk level fusion on the initial abnormal information pair set, and outputting an identification result containing an abnormal position coordinate, a time node and a risk level. According to the method, the technical problem of insufficient association analysis of the mine multi-source spatio-temporal data in the prior art is solved.

Description

Technical Field

[0001] This invention relates to the field of multi-source information recognition technology, specifically to a method and system for recognizing spatiotemporal information pairs driven by large artificial intelligence models. Background Technology

[0002] Mine safety is a core concern in the mining industry. With the widespread adoption of the Internet of Things (IoT) and sensor technologies, various monitoring devices, such as environmental sensors and personnel positioning systems, are deployed in mine environments, generating massive amounts of multi-source spatiotemporal data. Current mine safety monitoring systems typically rely on independent analysis of this data, employing simple threshold alarms or rule engines for anomaly detection, such as triggering equipment displacement alarms or personnel boundary crossing warnings by setting fixed thresholds. However, these methods have significant drawbacks: firstly, multi-source data struggles to form a comprehensive safety situational awareness; secondly, due to the complexity and dynamic changes of the mine environment, rules based on simple thresholds are ill-suited to adapt to varying conditions, easily leading to false alarms and missed alarms. Summary of the Invention

[0003] This application provides a method and system for identifying spatiotemporal information pairs driven by artificial intelligence large models, which is used to address the technical problem of insufficient correlation analysis of multi-source spatiotemporal data in mines in the existing technology.

[0004] In view of the above problems, this application provides a method and system for identifying spatiotemporal information pairs driven by artificial intelligence large models.

[0005] Firstly, this application provides a method for identifying spatiotemporal information pairs driven by a large artificial intelligence model, the method comprising: Acquire multi-source spatiotemporal data from the mine, and perform cleaning, noise reduction, and spatiotemporal alignment processing on the multi-source spatiotemporal data to generate a standard spatiotemporal data sequence; The standard spatiotemporal data sequence is input into a pre-trained spatiotemporal correlation model to perform multimodal spatiotemporal feature extraction and obtain a comprehensive spatiotemporal feature representation. The comprehensive spatiotemporal feature representation includes a trend feature sequence of equipment location changes over time, a spatial matching feature sequence of personnel trajectory and roadway environment, and a coupling feature sequence of safety production data and geographical area. Based on the comprehensive spatiotemporal feature representation, a dynamic anomaly identification network is constructed to identify abnormal equipment displacement, personnel trajectory deviation from safe areas, and abnormal correlation between geographical areas and safety production data, thereby generating an initial set of anomaly information pairs. The initial abnormal information set is subjected to spatiotemporal confidence compensation and risk level fusion to output the identification result containing the abnormal location coordinates, time nodes and risk levels.

[0006] Secondly, this application provides a spatiotemporal information pair recognition system driven by a large artificial intelligence model, including: The data acquisition module is used to acquire multi-source spatiotemporal data from the mine, clean and denoise the multi-source spatiotemporal data and perform spatiotemporal alignment processing to generate a standard spatiotemporal data sequence. The feature extraction module is used to input the standard spatiotemporal data sequence into a pre-trained spatiotemporal correlation model to perform multimodal spatiotemporal feature extraction and obtain a comprehensive spatiotemporal feature representation, wherein the comprehensive spatiotemporal feature representation includes a trend feature sequence of equipment location changes over time, a spatial matching feature sequence of personnel trajectory and roadway environment, and a coupling feature sequence of safety production data and geographical area. The anomaly identification module is used to construct a dynamic anomaly identification network based on the comprehensive spatiotemporal feature representation, and to identify abnormal equipment displacement, personnel trajectory deviation from the safe area, and abnormal correlation between geographical area and safety production data, thereby generating an initial set of anomaly information pairs. The identification result output module is used to perform spatiotemporal confidence compensation and risk level fusion on the initial abnormal information set, and output the identification result containing the abnormal location coordinates, time nodes and risk levels.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application proposes an AI-driven spatiotemporal information pair identification method and system. It acquires multi-source spatiotemporal data from mines, cleans and denoises it, and aligns it spatiotemporally to generate standard sequences. Then, a pre-trained spatiotemporal correlation model is used to extract comprehensive spatiotemporal feature representations including equipment location trends, personnel trajectory matching, and coupling relationships between safety production data. Based on this, a dynamic anomaly identification network is constructed to perform parallel identification of abnormal equipment displacement, personnel trajectory deviation from safe areas, and abnormal correlations between geographical areas and safety production data, generating an initial set of anomaly information pairs. Finally, spatiotemporal confidence compensation and risk level fusion are performed on this set, significantly improving the accuracy and comprehensiveness of multi-source anomaly situation perception in mine safety production monitoring. Compared to traditional methods, the technical solution provided in this application significantly overcomes the false alarm and missed alarm problems caused by isolated data source analysis and static rules, achieving the technical effect of improving the real-time monitoring accuracy of mine safety monitoring systems for multi-dimensional and cross-scale anomaly events in complex dynamic environments. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1This is a flowchart illustrating the spatiotemporal information pair recognition method driven by a large artificial intelligence model, as provided in an embodiment of this application.

[0010] Figure 2 This is a schematic diagram of the structure of the artificial intelligence large model-driven spatiotemporal information pair recognition system provided in the embodiments of this application.

[0011] The components represented by each number in the attached diagram are explained below: The system includes a data acquisition module 100, a feature extraction module 200, an anomaly detection module 300, and a detection result output module 400. Detailed Implementation

[0012] This application provides a method and system for identifying spatiotemporal information pairs driven by artificial intelligence large models, which is intended to address the technical problem of insufficient correlation analysis of multi-source spatiotemporal data in mines in existing technologies.

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0014] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to these processes, methods, products, or devices.

[0015] Example 1, as Figure 1 As shown, this application provides a spatiotemporal information pair recognition method driven by a large artificial intelligence model, wherein the method includes: S10: Acquire multi-source spatiotemporal data from the mine, perform cleaning, noise reduction, and spatiotemporal alignment processing on the multi-source spatiotemporal data, and generate a standard spatiotemporal data sequence.

[0016] In mine safety monitoring, the effective utilization of multi-source spatiotemporal data is fundamental to accurate anomaly identification. However, in practical applications, raw data directly collected from sensors, positioning equipment, and monitoring systems suffers from significant quality issues, hindering the reliability of subsequent analysis. Multi-source data originates from heterogeneous systems with inconsistent acquisition timestamps and spatial coordinate systems. For example, equipment location data, personnel trajectories, and environmental parameters each use independent spatiotemporal reference standards, making direct correlation and comparison between data difficult. This inconsistency at the data level introduces inherent biases into subsequent feature extraction and anomaly identification model inputs, not only failing to capture true dynamic change patterns but potentially amplifying errors and generating erroneous warning signals.

[0017] Step S10 in the method provided in this application embodiment includes: Acquire multi-source spatiotemporal data of the mine, wherein the multi-source spatiotemporal data includes geographic coordinate sequences, environmental monitoring data, equipment operating status data, and personnel positioning trajectory data; Noise filtering and outlier removal are performed on each type of data, and spatiotemporal interpolation is used to fill in missing data to obtain a cleaned data sequence. Based on a unified spatiotemporal reference system, coordinate alignment and timestamp synchronization are performed on the cleaned data sequence to generate a standard spatiotemporal data sequence.

[0018] In this embodiment of the application, multi-source spatiotemporal data of the mine is acquired, and the multi-source spatiotemporal data is cleaned, denoised, and spatiotemporally aligned to generate a standard spatiotemporal data sequence.

[0019] Specifically, the first step is to collect multi-source spatiotemporal data from the mine. This multi-source spatiotemporal data includes geographic coordinate sequences, environmental monitoring data, equipment operating status data, and personnel location trajectory data. For example, multiple latitude and longitude coordinates of mine roadways and equipment are collected to obtain geographic coordinate sequences; temperature, humidity, and gas concentration in mine roadways are collected using temperature and humidity sensors and gas sensors to obtain environmental monitoring data; the on / off status and operating power of mine equipment such as coal mining machines are recorded to obtain equipment operating status data; and personnel location trajectory data is collected using a GPS positioning system.

[0020] For each data category, noise filtering and outlier removal were performed separately. Spatiotemporal interpolation was used to impute missing data, resulting in a cleaned data sequence. Specifically, for fluctuating data, such as environmental monitoring data sequences, moving average filtering was used for noise filtering. Furthermore, box plots were used to identify and remove outliers that significantly deviated from the normal range; for example, records in equipment operating status data with speeds far exceeding the rated speed were identified as outliers and removed. Additionally, for missing values ​​in the data sequence due to transmission failures, spatiotemporal kriging interpolation was used to impute them. Finally, a data sequence was obtained that was free of noise interference, had no obvious outliers, and was continuous and complete.

[0021] Based on a unified spatiotemporal reference system, the cleaned data sequence is aligned by coordinates and synchronized with timestamps to generate a standard spatiotemporal data sequence. Specifically, spatially, all collected geographic coordinates, equipment location coordinates, and personnel trajectory coordinates are transformed using the GDAL library to unify all geographic coordinates to the WGS84 coordinate system. Temporally, the timestamps of all data records are uniformly converted to UTC time, and time matching is performed on data from different devices that represent the same moment to ensure that all data are synchronized on the timeline, ultimately obtaining a well-organized standard spatiotemporal data sequence that is completely consistent in both spatial and temporal references.

[0022] By performing targeted cleaning, denoising, and outlier removal operations on multi-source spatiotemporal data, noise introduced by environmental interference and equipment failures was effectively filtered out, data gaps were repaired, and the integrity of the original data was significantly improved. Furthermore, by employing spatiotemporal interpolation methods and synchronizing coordinates and timestamps with a unified spatiotemporal reference system, the originally scattered and heterogeneous data streams were integrated into a standard sequence with a consistent spatiotemporal dimension. This enables precise correlation and comparison of data from different modalities, such as equipment coordinates, personnel trajectories, and environmental parameters, within a unified spatiotemporal framework, greatly enhancing the inherent consistency and comparability of the data.

[0023] S20: Input the standard spatiotemporal data sequence into the pre-trained spatiotemporal correlation model to perform multimodal spatiotemporal feature extraction and obtain a comprehensive spatiotemporal feature representation, wherein the comprehensive spatiotemporal feature representation includes the trend feature sequence of equipment location changes over time, the spatial matching feature sequence of personnel trajectory and roadway environment, and the coupling feature sequence of safety production data and geographical area.

[0024] Existing analytical methods are typically limited to independent examination of single data sources or shallow features, such as analyzing only equipment displacement trajectories or monitoring environmental parameters in isolation. They lack mechanisms for the collaborative integration of multimodal spatiotemporal information. This approach fails to capture the interaction between equipment operating trends and roadway geographical constraints, the degree of matching between personnel movement paths and environmental safety status, and the coupling characteristics between safety production data and spatial location in a specific area. It is difficult to identify potential risks caused by the synergistic effects of multiple factors, such as complex faults caused by abnormal equipment displacement accompanied by changes in environmental parameters in a specific area, or abnormal clustering of personnel trajectories in key geographical areas.

[0025] Step S20 in the method provided in this application embodiment includes: The construction steps of a large-scale spatiotemporal correlation model include: Collect multi-source spatiotemporal data samples from historical mines and construct a set of sample spatiotemporal data sequences; Label data on equipment location trends, personnel trajectory matching degree, and correlation between safety production data and geographical areas in the sample are used to obtain a set of sample feature sequences; Based on machine learning, a network architecture for a spatiotemporal correlation model is constructed. The model is trained under supervision using the set of sample spatiotemporal data sequences and the set of sample feature sequences. After training convergence, the spatiotemporal correlation model is obtained.

[0026] In this embodiment of the application, standard spatiotemporal data sequences are input into a pre-trained spatiotemporal correlation model to extract multimodal spatiotemporal features and obtain a comprehensive spatiotemporal feature representation. The comprehensive spatiotemporal feature representation includes a trend feature sequence of equipment location changes over time, a spatial matching feature sequence of personnel trajectory and roadway environment, and a coupling feature sequence of safety production data and geographical area.

[0027] Specifically, firstly, a large-scale spatiotemporal correlation model is constructed.

[0028] Historical mine multi-source spatiotemporal data samples were collected to construct a sample spatiotemporal data sequence set. These historical mine multi-source spatiotemporal data samples were pre-processed samples, for example, based on equipment operation logs and personnel location records, historical geographic coordinate sequences, historical environmental monitoring data, historical equipment operating status data, and historical personnel location trajectory data were obtained and pre-processed to obtain historical mine multi-source spatiotemporal data samples, thus constructing a sample spatiotemporal data sequence set.

[0029] Furthermore, label data is used to annotate the equipment location trends, personnel trajectory matching degrees, and the correlation between safety production data and geographical areas in the samples, thereby obtaining a set of sample feature sequences. Specifically, equipment location trends can be labeled with "stable" or "offset," personnel trajectory matching degrees can be labeled with "match" or "mismatch," and the correlation between safety production data and geographical areas can be labeled with "safe" or "risk exists," thus obtaining a set of sample feature sequences.

[0030] Based on machine learning, a network architecture for a large-scale spatiotemporal correlation model is constructed. For example, a spatiotemporal correlation model with a three-layer network structure is built. The first layer is a temporal feature extraction layer, constructed using a Long Short-Term Memory (LSTM) network with 128 neurons, used to capture temporal dependencies in the data. The second layer is a spatial feature extraction layer, using a two-dimensional convolutional neural network with 3×3 convolutional kernels and 64 filters, used to extract spatial features. The third layer is a multimodal fusion layer, a fully connected layer containing 256 neurons, which fuses the spatiotemporal features. The model's input dimension is set to (sequence length, feature dimension), where the sequence length can be set to 24 based on the actual data, representing 24 time points, and the feature dimension is set according to coordinates, environmental parameters, etc. Furthermore, supervised training is performed using a sample spatiotemporal data sequence set and a sample feature sequence set. The sample spatiotemporal data sequence set is used as input, and the sample feature sequence set is used as the target output. The Adam optimizer is used, with a learning rate of 0.001, a batch size of 32, and a mean squared error loss function. The loss value = average value × (model predicted value - target value). 2 The training process continues until the loss function converges. If the loss function does not decrease significantly after 100 training rounds, the completed spatiotemporal correlation model is obtained.

[0031] Furthermore, the preprocessed standard spatiotemporal data sequence is input into a trained spatiotemporal correlation model. The model first extracts time-series features through an LSTM layer to obtain a trend feature sequence of equipment location changes over time; then, it processes spatial data through a CNN layer to obtain a spatial matching feature sequence of personnel trajectories and the roadway environment; finally, it fuses multimodal information through a fully connected layer to obtain a coupled feature sequence of safety production data and geographical area. These feature sequences together constitute a comprehensive spatiotemporal feature representation, where the trend feature sequence reflects the movement patterns of equipment, the spatial matching feature sequence reveals the degree of adaptation between personnel activities and the environment, and the coupled feature sequence demonstrates the correlation strength between safety parameters and regional features.

[0032] By inputting standard spatiotemporal data sequences into a pre-trained large model, a comprehensive feature representation is extracted from complex data. This representation encompasses the dynamic trends of equipment locations, the matching relationship between personnel trajectories and roadway spatial structures, and the coupling characteristics of safety production indicators with geographical regions. This comprehensive spatiotemporal feature representation overcomes the limitations of traditional single-feature descriptions, integrating previously isolated data dimensions into an information carrier that comprehensively reflects the operational status of the mine system. It not only depicts the dynamic changes of individual elements but, more importantly, reveals the intrinsic connections and interactions between multiple elements such as equipment, personnel, and the environment in the spatiotemporal dimension. This high-level, integrated feature representation provides richer and more discriminative information input for subsequent anomaly detection networks, enabling the detection model to perceive and judge anomalies from a holistic system perspective rather than a conventional local perspective. This lays a crucial feature foundation for significantly improving the detection capability of complex anomaly patterns, especially multi-factor coupled risks.

[0033] S30: Based on the comprehensive spatiotemporal feature representation, a dynamic anomaly identification network is constructed to identify abnormal equipment displacement, personnel trajectory deviation from safe areas, and abnormal correlation between geographical areas and safety production data, thereby generating an initial set of anomaly information pairs.

[0034] Existing technologies often employ serial or isolated anomaly detection modules, such as detecting equipment anomalies first and then analyzing personnel trajectories. This approach struggles to capture potential correlations between different anomaly types and is inefficient. More importantly, conventional methods often rely on fixed thresholds or simple rules to define anomalies, failing to adapt to the complex and nonlinear interactions between equipment displacement, personnel behavior, and regional safety conditions in the dynamic environment of mines. For example, a minor change in equipment position may be considered normal under certain circumstances, but when coupled with specific personnel trajectories or regional environmental data, it may pose a high risk.

[0035] Step S30 in the method provided in this application embodiment includes: Based on the comprehensive spatiotemporal feature representation, a dynamic anomaly identification network is constructed, wherein the dynamic anomaly identification network includes a sub-network for identifying abnormal equipment displacement, a sub-network for identifying personnel trajectory deviation, and a sub-network for identifying anomaly associated regions. The device abnormal displacement identification subnetwork is used to fit the device motion trajectory based on the trend feature sequence, calculate the deviation between the actual position and the fitted device motion trajectory, and identify the device abnormal displacement. The personnel trajectory deviation identification subnetwork is used to extract the real-time trajectory point sequence of personnel based on the spatial matching feature sequence, and to identify trajectory deviations from the safe area based on predefined electronic fence data. The abnormal association region identification sub-network is used to calculate the association strength between geographical regions and safety production data based on the coupled feature sequence, and to identify abnormal association regions. Specifically, based on the coupled feature sequence, the correlation strength between geographical regions and safety production data is calculated, and abnormal correlation regions are identified, including: Extract the real-time correlation strength index between geographical regions and safety production data from the coupled feature sequence, and compare it with the correlation strength benchmark under historical normal working conditions to calculate the correlation deviation. Specifically, the real-time correlation strength index between geographical regions and safety production data in the coupled feature sequence is extracted and compared with the correlation strength benchmark under historical normal working conditions to calculate the correlation deviation, including: Based on the coupling feature sequence, extract the real-time correlation strength index between the target geographical area and the corresponding safety production data in the current time period; Obtain the correlation strength benchmark sequence corresponding to the same geographical area under historical normal operating conditions, wherein the correlation strength benchmark sequence contains correlation strength data for multiple historical normal periods; Calculate the statistical distribution difference between the real-time correlation strength index and multiple correlation strength benchmark sequences; Based on the sliding time window mechanism, the overall deviation of the real-time correlation strength index of the current time period and adjacent time periods from the correlation strength benchmark sequence is calculated. The statistical distribution differences are weighted and fused with the overall deviation to generate a comprehensive correlation deviation. When the correlation deviation continuously exceeds the preset abnormal threshold, it is determined that there is an abnormal correlation in the current geographical area and it is marked as an abnormal correlation area. The identification results of the equipment abnormal displacement identification subnetwork, the personnel trajectory deviation identification subnetwork, and the abnormal associated area identification subnetwork are combined to generate an initial abnormal information pair set.

[0036] In this embodiment of the application, a dynamic anomaly identification network is constructed based on comprehensive spatiotemporal feature representation to identify abnormal equipment displacement, personnel trajectory deviation from safe area, and abnormal correlation between geographical area and safety production data, thereby generating an initial set of anomaly information pairs.

[0037] Specifically, firstly, a dynamic anomaly identification network is constructed based on the comprehensive spatiotemporal feature representation. The dynamic anomaly identification network includes a sub-network for identifying abnormal equipment displacement, a sub-network for identifying personnel trajectory deviation, and a sub-network for identifying anomaly-related regions.

[0038] The equipment abnormal displacement identification subnetwork is used to fit the equipment's motion trajectory based on a trend feature sequence, calculate the deviation between the actual position and the fitted trajectory, and identify abnormal equipment displacement. Specifically, a linear regression method is used to fit the equipment's motion trajectory, obtaining a fitted position sequence, and the deviation between the actual position and the fitted position at each time point is calculated, such as by calculating the absolute distance between the actual and fitted positions as the deviation magnitude. Simultaneously, anomaly thresholds are dynamically adjusted based on historical displacement data; for example, the historical average deviation magnitude multiplied by a safety factor is used as the dynamic threshold. For instance, if the historical average deviation magnitude is 0.2 meters and the safety factor is set to 1.1, then the dynamic threshold = 0.2 × 1.1 = 0.22. When the deviation magnitude exceeds the dynamic anomaly threshold, the equipment position is marked as an abnormal displacement. For example, if the deviation magnitude is 0.4 meters, which is greater than the dynamic threshold, it is determined to be abnormal. The final output is the equipment abnormal displacement identification result, which includes the abnormal location and time point.

[0039] The personnel trajectory deviation recognition subnetwork is used to extract real-time trajectory point sequences of personnel based on spatial matching feature sequences, and to identify trajectory deviations from safe zones based on predefined electronic fence data. Specifically, the predefined electronic fence data is first acquired. Electronic fence data is boundary data pre-set to define safe zones to prevent personnel from entering dangerous areas; it can be defined, for example, as a set of polygonal boundary points. From the spatial matching feature sequence, the real-time trajectory point sequence of personnel is extracted. The trajectory points in the sequence are filtered to obtain those not within the electronic fence, and their shortest Euclidean distance to the electronic fence boundary is calculated. This distance is compared with a preset safe distance threshold. If the Euclidean distance of a trajectory point continuously or repeatedly exceeds the safe distance threshold, it is determined that the trajectory has deviated from the safe zone. For example, if the preset safe distance threshold is 1 meter, and the distance from a trajectory point not within the electronic fence to the boundary is 1.2 meters, and the Euclidean distance from the trajectory point to the boundary at three consecutive time points exceeds the safe distance threshold, then it is determined to be a deviation. Finally, the personnel trajectory deviation recognition result is output, including the deviation location and time sequence. This screening process eliminates occasional deviations and identifies trajectory deviations of personnel that truly pose safety risks.

[0040] An abnormal association region identification subnetwork is used to calculate the association strength between geographical regions and safety production data based on coupled feature sequences, and to identify abnormal association regions. Specifically, it extracts real-time association strength indicators between geographical regions and safety production data from the coupled feature sequences, compares them with the association strength benchmark under historical normal operating conditions, and calculates the association deviation.

[0041] Based on the coupled feature sequence, a real-time correlation strength index between the target geographic area and the corresponding safety production data within the current time period is extracted. For example, a pre-acquired feature decoder can be used for mapping to obtain the real-time correlation strength index between the target geographic area and the corresponding safety production data. The feature decoder can be constructed based on a linear decoder, using a linear regression method for feature decoding. Finally, the coupled feature sequence is input, and the decoder can accurately output the real-time correlation strength index.

[0042] Furthermore, the correlation strength benchmark sequence corresponding to the same geographical area under historical normal operating conditions is obtained, wherein the correlation strength benchmark sequence contains correlation strength data for multiple historical normal periods.

[0043] Calculate the statistical distribution difference between the real-time correlation strength index and multiple correlation strength benchmark sequences. For example, the statistical distribution difference = |(current real-time correlation strength index - mean of the correlation strength benchmark sequence) ÷ standard deviation of the correlation strength benchmark sequence|.

[0044] Based on a sliding time window mechanism, the overall deviation of the real-time correlation strength index relative to the correlation strength benchmark sequence for the current and adjacent time periods is calculated. For example, using a sliding time window of size 5, the absolute value of the difference between the real-time correlation strength index at the current time point and the mean of the correlation strength benchmark sequence is calculated to obtain the absolute deviation value: Absolute deviation value = |current real-time correlation strength index - mean of the correlation strength benchmark sequence|. The absolute deviation values ​​for all time points within the time window are then summed to obtain the total absolute deviation value for the time window. The average absolute deviation is obtained by dividing the total absolute deviation value by the sliding time window size. Based on the average absolute deviation, the overall deviation is obtained: Overall deviation = Average absolute deviation ÷ Standard deviation of the correlation strength benchmark sequence.

[0045] The statistical distribution difference and the overall deviation are weighted and integrated to generate a comprehensive correlation deviation. Correlation deviation = W1 × statistical distribution difference + W2 × overall deviation, where the weights W1 + W2 = 1. W1 and W2 can be obtained based on the importance of the correlation deviation analysis according to the instantaneous deviation and recent trend. For example, if more emphasis is placed on analyzing the instantaneous deviation, W1 can be set to 0.6 and W2 = 1 - 0.6 = 0.4.

[0046] When the correlation deviation continuously exceeds the preset abnormal threshold, such as when the correlation exceeds the preset abnormal threshold for three consecutive time points, it is determined that there is an abnormal correlation in the current geographic area, and the current geographic area is marked as an abnormal correlation area.

[0047] Furthermore, the recognition results of the equipment abnormal displacement recognition subnetwork, the personnel trajectory deviation recognition subnetwork, and the abnormal association area recognition subnetwork are integrated to generate an initial set of abnormal information pairs. Each abnormal information pair includes the equipment abnormal displacement recognition result, the personnel trajectory deviation recognition result set, and the abnormal association area recognition result.

[0048] This application, based on comprehensive spatiotemporal feature representation, uses dedicated sub-networks to simultaneously analyze abnormal equipment displacement, personnel trajectory deviations from safe zones, and abnormal correlations between geographical regions and safety production data. This parallel processing architecture not only improves identification efficiency but also ensures the specialization and depth of anomaly detection for different types of anomalies. It provides structured, multi-source anomaly event inputs for subsequent comprehensive risk assessment, significantly enhancing the ability to capture complex, cross-dimensional risks.

[0049] S40: Perform spatiotemporal confidence compensation and risk level fusion on the initial abnormal information set, and output the identification result containing the abnormal location coordinates, time nodes and risk levels.

[0050] Existing methods typically output initial identification results directly or perform simple weighted averaging, lacking consideration of the spatiotemporal context of abnormal events, such as historical patterns and the state of surrounding areas. This can lead to risk assessment results that are overly sensitive to short-term fluctuations or insufficiently aware of gradual risks. Furthermore, the failure to uniformly quantify and classify the potential risks of different pairs of abnormal information makes it difficult for managers to quickly determine the severity and priority of events, thus affecting emergency response efficiency.

[0051] Step S40 in the method provided in this application embodiment includes: Extract the spatiotemporal context features of each anomaly information pair in the initial anomaly information pair set, calculate the matching degree with historical anomaly patterns, and generate spatiotemporal confidence weights. Specifically, the spatiotemporal context features of each anomaly pair in the initial anomaly information pair set are extracted, and the matching degree with historical anomaly patterns is calculated to generate spatiotemporal confidence weights, including: Obtain the spatiotemporal context features corresponding to the anomaly information, wherein the spatiotemporal context features include time periodic features and spatial distribution features; Calculate the time period matching degree between the time period features and historical anomaly patterns to obtain the time period matching degree; Calculate the spatial matching degree between the spatial distribution features and historical anomaly patterns to obtain the spatial distribution matching degree; The time period matching degree and the spatial distribution matching degree are weighted and fused to generate the spatiotemporal confidence weight; The degree of anomaly of the corresponding abnormal information pairs is compensated and corrected based on the spatiotemporal confidence weight to obtain a weighted anomaly score, wherein the degree of anomaly is the weighted fusion value of the abnormal displacement deviation of the equipment, the distance of the trajectory from the safe area, and the correlation deviation. The weighted anomaly score is matched with a preset risk level threshold range to determine the final risk level; Combining the location coordinates and time nodes of the anomaly information pair, the output includes the anomaly location coordinates, time nodes, and the final risk level.

[0052] In this embodiment of the application, the initial abnormal information set is subjected to spatiotemporal confidence compensation and risk level fusion to output the identification result containing the abnormal location coordinates, time nodes and risk levels.

[0053] Specifically, the spatiotemporal context features of each anomaly pair in the initial anomaly information pair set are extracted, and the matching degree with historical anomaly patterns is calculated to generate spatiotemporal confidence weights.

[0054] First, the spatiotemporal context features corresponding to the anomaly information are obtained. These features include time periodicity features and spatial distribution features. For example, time periodicity features characterize whether the anomaly occurs during peak operating hours or maintenance periods, while spatial distribution features characterize whether the anomaly location is in the main area or peripheral area of ​​the main mine roadway. The time periodicity matching degree is determined by calculating the similarity between the current anomaly occurrence time period and historically high-frequency anomaly time periods. For example, if the current time is 9:00 AM, and historical data shows a high frequency of anomalies during this time period, then the time periodicity matching degree is high: Time periodicity matching degree = 1 - |(current time point - historical high-frequency time point) ÷ 24|. The spatial distribution matching degree is determined by calculating the Euclidean distance between the current anomaly location and historically high-frequency anomaly areas. For example, if the current anomaly location is close to the center point of a historically high-frequency anomaly area, then the spatial distribution matching degree is high: Spatial distribution matching degree = 1 - (distance between the current location and the center of the historically high-frequency area ÷ maximum effective distance). The temporal period matching degree and spatial distribution matching degree are weighted and fused to generate a spatiotemporal confidence weight, which is calculated as: Spatiotemporal Confidence Weight = A1 × Temporal Period Matching Degree + A2 × Spatial Distribution Matching Degree. Here, A1 + A2 = 1. The default value can be set to A1 = A2 = 0.5. A1 and A2 are dynamically adjusted based on the importance of the temporal period matching degree and spatial distribution matching degree in the spatiotemporal confidence weight analysis. For example, if the spatial distribution matching degree is more important to the spatiotemporal confidence weight analysis, then A2 can be set to 0.6, and A1 = 1 - 0.6 = 0.4.

[0055] The anomaly severity of corresponding anomaly information pairs is compensated and corrected based on spatiotemporal confidence weights to obtain a weighted anomaly score. The anomaly severity is a weighted fusion value of the amplitude of abnormal equipment displacement deviation, the distance of trajectory deviation from the safe zone, and the degree of correlation deviation. Specifically, for abnormal equipment displacement, the anomaly severity is the Euclidean distance between the actual and expected positions of the equipment, i.e., the amplitude of the abnormal equipment displacement deviation; for personnel trajectory deviation, the anomaly severity is the shortest Euclidean distance from the personnel position to the boundary of the safe zone, i.e., the distance of trajectory deviation from the safe zone; for regional anomaly correlation, the anomaly severity is the degree of correlation deviation. The anomaly severity is obtained by weighted fusion: Anomaly Severity = B1 × Amplitude of Abnormal Equipment Displacement Deviation + B2 × Distance of Trajectory Deviation from the Safe Zone + B3 × Degree of Correlation Deviation. Where B1 + B2 + B3 = 1. The weights B1, B2, and B3 can be obtained based on the importance of the anomaly. For example, if the danger level of personnel trajectory deviation is high, the anomaly importance is high, so B2 can be set to 0.4, and B1 and B3 can both be set to 0.3 for weighted calculation to obtain the original anomaly severity.

[0056] The original anomaly severity is compensated using spatiotemporal confidence weights to obtain a weighted anomaly score. Weighted anomaly score = Original anomaly severity × Spatiotemporal confidence weight. For example, if the original anomaly severity is 0.8 and its spatiotemporal confidence weight is 0.9, then the weighted anomaly score = 0.8 × 0.9 = 0.72.

[0057] The weighted anomaly score is matched against preset risk level threshold ranges to determine the final risk level. For example, four risk level threshold ranges are set: low risk (0-0.2), medium risk (0.2-0.6), high risk (0.6-0.8), and very high risk (0.8-1). The weighted anomaly score for each anomaly pair is compared with these threshold ranges to determine its corresponding risk level. For example, an anomaly pair with a weighted anomaly score of 0.72 is classified as high risk.

[0058] Combining the location coordinates and time points of the anomaly information pair, the output includes the anomaly location coordinates, time points, and the final risk level. For example, the identification result could be [Anomaly displacement of equipment, (112.5, 35.8), 2024-05-01 14:30:00, Medium risk].

[0059] By extracting the spatiotemporal context features of each anomaly pair and matching them with historical anomaly patterns, a spatiotemporal confidence weight representing the credibility of each anomaly is generated. This compensates for and corrects the initial anomaly level, effectively suppressing false alarms caused by accidental interference. Simultaneously, it enhances the significance of real risks consistent with historical anomaly patterns, improving the accuracy of the results in the spatiotemporal dimension. Furthermore, by mapping the weighted anomaly scores to preset risk level thresholds, unified quantification and grading of anomaly risks from different sources and of different types are achieved, generating a clear final risk level. The final output identification result not only includes the location coordinates and time of the anomaly but also adds risk level information. This transforms the output from a simple list of raw signals into intelligently refined decision support information, greatly improving the early warning accuracy and decision support capability of the mine safety monitoring system.

[0060] Example 2, as Figure 2 As shown, based on the same inventive concept as the AI ​​large model-driven spatiotemporal information pair recognition method provided in Embodiment 1, this embodiment of the invention also provides an AI large model-driven spatiotemporal information pair recognition system, including: The data acquisition module 100 is used to acquire multi-source spatiotemporal data from the mine, clean and denoise the multi-source spatiotemporal data and perform spatiotemporal alignment processing to generate a standard spatiotemporal data sequence. The feature extraction module 200 is used to input the standard spatiotemporal data sequence into a pre-trained spatiotemporal correlation model to perform multimodal spatiotemporal feature extraction and obtain a comprehensive spatiotemporal feature representation, wherein the comprehensive spatiotemporal feature representation includes a trend feature sequence of equipment location changes over time, a spatial matching feature sequence of personnel trajectory and roadway environment, and a coupling feature sequence of safety production data and geographical area. Anomaly identification module 300 is used to construct a dynamic anomaly identification network based on the comprehensive spatiotemporal feature representation, and to identify abnormal equipment displacement, personnel trajectory deviation from safe area, and abnormal correlation between geographical area and safety production data, thereby generating an initial set of anomaly information pairs. The identification result output module 400 is used to perform spatiotemporal confidence compensation and risk level fusion on the initial abnormal information set, and output the identification result containing the abnormal location coordinates, time nodes and risk levels.

[0061] In one embodiment, the data acquisition module 100 is further configured to: Acquire multi-source spatiotemporal data of the mine, wherein the multi-source spatiotemporal data includes geographic coordinate sequences, environmental monitoring data, equipment operating status data, and personnel positioning trajectory data; Noise filtering and outlier removal are performed on each type of data, and spatiotemporal interpolation is used to fill in missing data to obtain a cleaned data sequence. Based on a unified spatiotemporal reference system, coordinate alignment and timestamp synchronization are performed on the cleaned data sequence to generate a standard spatiotemporal data sequence.

[0062] In one embodiment, the feature extraction module 200 is further configured to: The construction steps of a large-scale spatiotemporal correlation model include: Collect multi-source spatiotemporal data samples from historical mines and construct a set of sample spatiotemporal data sequences; Label data on equipment location trends, personnel trajectory matching degree, and correlation between safety production data and geographical areas in the sample are used to obtain a set of sample feature sequences; Based on machine learning, a network architecture for a spatiotemporal correlation model is constructed. The model is trained under supervision using the set of sample spatiotemporal data sequences and the set of sample feature sequences. After training convergence, the spatiotemporal correlation model is obtained.

[0063] In one embodiment, the anomaly detection module 300 is further configured to: Based on the comprehensive spatiotemporal feature representation, a dynamic anomaly identification network is constructed, wherein the dynamic anomaly identification network includes a sub-network for identifying abnormal equipment displacement, a sub-network for identifying personnel trajectory deviation, and a sub-network for identifying anomaly associated regions. The device abnormal displacement identification subnetwork is used to fit the device motion trajectory based on the trend feature sequence, calculate the deviation between the actual position and the fitted device motion trajectory, and identify the device abnormal displacement. The personnel trajectory deviation identification subnetwork is used to extract the real-time trajectory point sequence of personnel based on the spatial matching feature sequence, and to identify trajectory deviations from the safe area based on predefined electronic fence data. The abnormal association region identification sub-network is used to calculate the association strength between geographical regions and safety production data based on the coupled feature sequence, and to identify abnormal association regions. Specifically, based on the coupled feature sequence, the correlation strength between geographical regions and safety production data is calculated, and abnormal correlation regions are identified, including: Extract the real-time correlation strength index between geographical regions and safety production data from the coupled feature sequence, and compare it with the correlation strength benchmark under historical normal working conditions to calculate the correlation deviation. Specifically, the real-time correlation strength index between geographical regions and safety production data in the coupled feature sequence is extracted and compared with the correlation strength benchmark under historical normal working conditions to calculate the correlation deviation, including: Based on the coupling feature sequence, extract the real-time correlation strength index between the target geographical area and the corresponding safety production data in the current time period; Obtain the correlation strength benchmark sequence corresponding to the same geographical area under historical normal operating conditions, wherein the correlation strength benchmark sequence contains correlation strength data for multiple historical normal periods; Calculate the statistical distribution difference between the real-time correlation strength index and multiple correlation strength benchmark sequences; Based on the sliding time window mechanism, the overall deviation of the real-time correlation strength index of the current time period and adjacent time periods from the correlation strength benchmark sequence is calculated. The statistical distribution differences are weighted and fused with the overall deviation to generate a comprehensive correlation deviation. When the correlation deviation continuously exceeds the preset abnormal threshold, it is determined that there is an abnormal correlation in the current geographical area and it is marked as an abnormal correlation area. The identification results of the equipment abnormal displacement identification subnetwork, the personnel trajectory deviation identification subnetwork, and the abnormal associated area identification subnetwork are combined to generate an initial abnormal information pair set.

[0064] In one embodiment, the recognition result output module 400 is further configured to: Extract the spatiotemporal context features of each anomaly information pair in the initial anomaly information pair set, calculate the matching degree with historical anomaly patterns, and generate spatiotemporal confidence weights. Specifically, the spatiotemporal context features of each anomaly pair in the initial anomaly information pair set are extracted, and the matching degree with historical anomaly patterns is calculated to generate spatiotemporal confidence weights, including: Obtain the spatiotemporal context features corresponding to the anomaly information, wherein the spatiotemporal context features include time periodic features and spatial distribution features; Calculate the time period matching degree between the time period features and historical anomaly patterns to obtain the time period matching degree; Calculate the spatial matching degree between the spatial distribution features and historical anomaly patterns to obtain the spatial distribution matching degree; The time period matching degree and the spatial distribution matching degree are weighted and fused to generate the spatiotemporal confidence weight; The degree of anomaly of the corresponding abnormal information pairs is compensated and corrected based on the spatiotemporal confidence weight to obtain a weighted anomaly score, wherein the degree of anomaly is the weighted fusion value of the abnormal displacement deviation amplitude of the equipment, the distance of the trajectory from the safe zone, and the correlation deviation. The weighted anomaly score is matched with a preset risk level threshold range to determine the final risk level; Combining the location coordinates and time nodes of the anomaly information pair, the output includes the anomaly location coordinates, time nodes, and the final risk level.

[0065] In summary, the embodiments of this application have at least the following technical effects: This application proposes a method and system for identifying spatiotemporal information pairs driven by a large artificial intelligence model. It acquires multi-source spatiotemporal data from mines, cleans and denoises it, and aligns it spatiotemporally to generate standard sequences. Then, a pre-trained spatiotemporal correlation model is used to extract comprehensive spatiotemporal feature representations including equipment location trends, personnel trajectory matching, and coupling relationships between safety production data. Based on this, a dynamic anomaly identification network is constructed to perform parallel identification of abnormal equipment displacement, personnel trajectory deviation from safe areas, and abnormal correlations between geographical areas and safety production data, generating an initial set of anomaly information pairs. Finally, spatiotemporal confidence compensation and risk level fusion are performed on this set, significantly improving the accuracy and comprehensiveness of multi-source anomaly situation perception in mine safety production monitoring. Specifically, through deep feature extraction driven by a large model and dynamic network recognition, adaptive learning and accurate characterization of the complex spatiotemporal relationships between equipment, personnel, and the environment are achieved. This enables the capture of potential trend deviations and collaborative risks from massive heterogeneous data. Simultaneously, by introducing a spatiotemporal confidence compensation mechanism, historical anomaly patterns and real-time contextual features are effectively combined, enhancing the ability to distinguish between instantaneous anomalies and progressive risks. This avoids judgment distortion caused by environmental fluctuations or data noise, making risk assessment results more consistent with actual working conditions. Compared to traditional methods, the technical solution provided in this application significantly overcomes the problems of false alarms and missed alarms caused by isolated data source analysis and static rules, achieving the technical effect of improving the real-time monitoring accuracy of mine safety monitoring systems for multi-dimensional and cross-scale anomalies in complex dynamic environments.

[0066] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0067] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0068] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A spatiotemporal information pair recognition method driven by a large artificial intelligence model, characterized in that, The method includes: Acquire multi-source spatiotemporal data from the mine, and perform cleaning, noise reduction, and spatiotemporal alignment processing on the multi-source spatiotemporal data to generate a standard spatiotemporal data sequence; The standard spatiotemporal data sequence is input into a pre-trained spatiotemporal correlation model to perform multimodal spatiotemporal feature extraction and obtain a comprehensive spatiotemporal feature representation. The comprehensive spatiotemporal feature representation includes a trend feature sequence of equipment location changes over time, a spatial matching feature sequence of personnel trajectory and roadway environment, and a coupling feature sequence of safety production data and geographical area. Based on the comprehensive spatiotemporal feature representation, a dynamic anomaly identification network is constructed to identify abnormal equipment displacement, personnel trajectory deviation from safe areas, and abnormal correlation between geographical areas and safety production data, thereby generating an initial set of anomaly information pairs. The initial abnormal information set is subjected to spatiotemporal confidence compensation and risk level fusion to output the identification result containing the abnormal location coordinates, time nodes and risk levels.

2. The method for identifying spatiotemporal information pairs driven by a large artificial intelligence model according to claim 1, characterized in that, Acquire multi-source spatiotemporal data from the mine, perform cleaning, denoising, and spatiotemporal alignment processing on the multi-source spatiotemporal data, and generate a standard spatiotemporal data sequence, including: Acquire multi-source spatiotemporal data of the mine, wherein the multi-source spatiotemporal data includes geographic coordinate sequences, environmental monitoring data, equipment operating status data, and personnel positioning trajectory data; Noise filtering and outlier removal are performed on each type of data, and spatiotemporal interpolation is used to fill in missing data to obtain a cleaned data sequence. Based on a unified spatiotemporal reference system, coordinate alignment and timestamp synchronization are performed on the cleaned data sequence to generate a standard spatiotemporal data sequence.

3. The method for identifying spatiotemporal information pairs driven by a large artificial intelligence model according to claim 1, characterized in that, The construction steps of a large-scale spatiotemporal correlation model include: Collect multi-source spatiotemporal data samples from historical mines and construct a set of sample spatiotemporal data sequences; Label data on equipment location trends, personnel trajectory matching degree, and correlation between safety production data and geographical areas in the sample are used to obtain a set of sample feature sequences; Based on machine learning, a network architecture for a spatiotemporal correlation model is constructed. The model is trained under supervision using the set of sample spatiotemporal data sequences and the set of sample feature sequences. After training convergence, the spatiotemporal correlation model is obtained.

4. The method for identifying spatiotemporal information pairs driven by a large artificial intelligence model according to claim 1, characterized in that, Based on the comprehensive spatiotemporal feature representation, a dynamic anomaly identification network is constructed to identify abnormal equipment displacement, personnel trajectory deviation from safe areas, and anomaly correlation between safety production data and geographical areas, including: Based on the comprehensive spatiotemporal feature representation, a dynamic anomaly identification network is constructed, wherein the dynamic anomaly identification network includes a sub-network for identifying abnormal equipment displacement, a sub-network for identifying personnel trajectory deviation, and a sub-network for identifying anomaly associated regions. The device abnormal displacement identification subnetwork is used to fit the device motion trajectory based on the trend feature sequence, calculate the deviation between the actual position and the fitted device motion trajectory, and identify the device abnormal displacement. The personnel trajectory deviation identification subnetwork is used to extract the real-time trajectory point sequence of personnel based on the spatial matching feature sequence, and to identify trajectory deviations from the safe area based on predefined electronic fence data. The abnormal association region identification sub-network is used to calculate the association strength between geographical regions and safety production data based on the coupled feature sequence, and to identify abnormal association regions. The identification results of the equipment abnormal displacement identification subnetwork, the personnel trajectory deviation identification subnetwork, and the abnormal associated area identification subnetwork are combined to generate an initial abnormal information pair set.

5. The method for identifying spatiotemporal information pairs driven by a large artificial intelligence model according to claim 4, characterized in that, Based on the aforementioned coupling feature sequence, the correlation strength between geographical regions and safety production data is calculated, and abnormal correlation regions are identified, including: Extract the real-time correlation strength index between geographical regions and safety production data from the coupled feature sequence, and compare it with the correlation strength benchmark under historical normal working conditions to calculate the correlation deviation. When the correlation deviation continuously exceeds the preset abnormal threshold, it is determined that there is an abnormal correlation in the current geographical area and it is marked as an abnormal correlation area.

6. The method for identifying spatiotemporal information pairs driven by a large artificial intelligence model according to claim 5, characterized in that, Extract the real-time correlation strength index between geographical regions and safety production data from the coupled feature sequence, and compare it with the correlation strength benchmark under historical normal working conditions to calculate the correlation deviation, including: Based on the coupling feature sequence, extract the real-time correlation strength index between the target geographical area and the corresponding safety production data in the current time period; Obtain the correlation strength benchmark sequence corresponding to the same geographical area under historical normal operating conditions, wherein the correlation strength benchmark sequence contains correlation strength data for multiple historical normal periods; Calculate the statistical distribution difference between the real-time correlation strength index and multiple correlation strength benchmark sequences; Based on the sliding time window mechanism, the overall deviation of the real-time correlation strength index of the current time period and adjacent time periods from the correlation strength benchmark sequence is calculated. The statistical distribution differences are weighted and fused with the overall deviation to generate a comprehensive correlation deviation.

7. The method for identifying spatiotemporal information pairs driven by a large artificial intelligence model according to claim 1, characterized in that, The initial anomaly information set is subjected to spatiotemporal confidence compensation and risk level fusion to output an identification result containing anomaly location coordinates, time nodes, and risk levels, including: Extract the spatiotemporal context features of each anomaly information pair in the initial anomaly information pair set, calculate the matching degree with historical anomaly patterns, and generate spatiotemporal confidence weights. The degree of anomaly of the corresponding abnormal information pairs is compensated and corrected based on the spatiotemporal confidence weight to obtain a weighted anomaly score, wherein the degree of anomaly is the weighted fusion value of the abnormal displacement deviation amplitude of the equipment, the distance of the trajectory from the safe zone, and the correlation deviation. The weighted anomaly score is matched with a preset risk level threshold range to determine the final risk level; Combining the location coordinates and time nodes of the anomaly information pair, the output includes the anomaly location coordinates, time nodes, and the final risk level.

8. The method for identifying spatiotemporal information pairs driven by a large artificial intelligence model according to claim 7, characterized in that, Extract the spatiotemporal context features of each anomaly pair in the initial anomaly information pair set, calculate the matching degree with historical anomaly patterns, and generate spatiotemporal confidence weights, including: Obtain the spatiotemporal context features corresponding to the anomaly information, wherein the spatiotemporal context features include time periodic features and spatial distribution features; Calculate the time period matching degree between the time period features and historical anomaly patterns to obtain the time period matching degree; Calculate the spatial matching degree between the spatial distribution features and historical anomaly patterns to obtain the spatial distribution matching degree; The time period matching degree and the spatial distribution matching degree are weighted and fused to generate the spatiotemporal confidence weight.

9. A spatiotemporal information pair recognition system driven by a large artificial intelligence model, characterized in that, The system is used to implement the artificial intelligence large model-driven spatiotemporal information pair identification method according to any one of claims 1-8, the system comprising: The data acquisition module is used to acquire multi-source spatiotemporal data from the mine, clean and denoise the multi-source spatiotemporal data and perform spatiotemporal alignment processing to generate a standard spatiotemporal data sequence. The feature extraction module is used to input the standard spatiotemporal data sequence into a pre-trained spatiotemporal correlation model to perform multimodal spatiotemporal feature extraction and obtain a comprehensive spatiotemporal feature representation, wherein the comprehensive spatiotemporal feature representation includes a trend feature sequence of equipment location changes over time, a spatial matching feature sequence of personnel trajectory and roadway environment, and a coupling feature sequence of safety production data and geographical area. The anomaly identification module is used to construct a dynamic anomaly identification network based on the comprehensive spatiotemporal feature representation, and to identify abnormal equipment displacement, personnel trajectory deviation from the safe area, and abnormal correlation between geographical area and safety production data, thereby generating an initial set of anomaly information pairs. The identification result output module is used to perform spatiotemporal confidence compensation and risk level fusion on the initial abnormal information set, and output the identification result containing the abnormal location coordinates, time nodes and risk levels.

Images