Internet of Things-InSAR cooperative disaster chain monitoring and early warning method and system

Abstract

The invention provides an Internet of Things-InSAR cooperative disaster chain monitoring and early warning method and system, and the method comprises the steps: carrying out the quality evaluation of obtained Internet of Things data, and obtaining a situation evaluation matrix, the Internet of Things data comprising a voltage value, communication signal intensity, and a data collection timestamp; identifying and extracting abnormal nodes according to the situation assessment matrix to obtain normal node data; calculating a partition data integrity rate based on the normal node data, and determining an area with the data integrity rate lower than a preset threshold value as a monitoring blind area; obtaining InSAR data of the monitoring blind area, extracting earth surface deformation data, and fusing the earth surface deformation data with normal node data to generate a global continuous deformation data set; and performing spatial-temporal feature analysis on the global continuous deformation data set to obtain a fusion feature vector, and inputting the fusion feature vector into a pre-trained disaster chain prediction model to generate a disaster risk thermodynamic diagram. By adopting the method, the prediction precision of disaster chain evolution can be improved.

Description

Technical Field

[0001] The present invention belongs to the field of disaster early warning, and in particular relates to an Internet of Things-InSAR collaborative disaster chain monitoring and early warning method and system. Background Art

[0002] With the development of natural disaster prediction technology and the Internet of Things (IoT), Synthetic Aperture Radar Interferometry (InSAR) has become a key tool for disaster monitoring due to its wide-area, millimeter-level deformation monitoring capabilities. Furthermore, IoT technology, through the deployment of distributed sensor networks, enables high-frequency data collection in localized areas. Currently, disaster chain monitoring primarily utilizes a combination of InSAR wide-area screening and IoT fixed-point monitoring. Traditionally, InSAR is responsible for large-scale deformation identification, while IoT nodes provide real-time data supplementation for key areas. These two technologies operate independently, with the data subsequently fused and processed.

[0003] However, the current approach has obvious shortcomings: on the one hand, the operating status of the IoT network (such as equipment failure and communication interruption) lacks a real-time perception mechanism, making it difficult to detect monitoring blind spots in a timely manner; on the other hand, InSAR data and IoT situation information fail to dynamically coordinate and cannot effectively support the real-time prediction of the evolution path of the disaster chain. Especially in complex terrain conditions, the problems of monitoring blind spots and prediction deviations are prominent. Summary of the Invention

[0004] Based on this, it is necessary to provide an IoT-InSAR collaborative disaster chain monitoring and early warning method and system to address the above technical problems, which can integrate network situational awareness and surface deformation data in real time, dynamically optimize the monitoring network coverage, and improve the prediction accuracy of disaster chain evolution.

[0005] In a first aspect, the present application provides an IoT-InSAR collaborative disaster chain monitoring and early warning method, comprising:

[0006] Perform a quality assessment on the acquired IoT data to obtain a situation assessment matrix. The IoT data includes voltage value, communication signal strength, and data collection timestamp.

[0007] Identify and extract abnormal nodes based on the situation assessment matrix to obtain normal node data;

[0008] Calculate the partition data integrity rate based on normal node data, and identify the area where the data integrity rate is lower than the preset threshold as a monitoring blind spot;

[0009] Acquire InSAR data from monitoring blind areas and extract surface deformation data, then fuse the surface deformation data with normal node data to generate a global continuous deformation dataset.

[0010] The spatiotemporal feature analysis of the global continuous deformation dataset is performed to obtain a fused feature vector, which is then input into a pre-trained disaster chain prediction model to generate a disaster risk heat map.

[0011] In one embodiment, the quality of the acquired IoT data is assessed to obtain a situation assessment matrix, including:

[0012] Clean and standardize IoT data to obtain standardized data sets;

[0013] Calculate the multi-dimensional situation assessment value of each device node in the standardized data set based on the sliding time window;

[0014] The geographical coordinates of each device node are associated with the multi-dimensional situation assessment value to generate a situation assessment matrix stored in row format.

[0015] In one embodiment, identifying and extracting abnormal nodes based on the situation assessment matrix to obtain normal node data includes:

[0016] Obtain terrain elevation data, where the terrain elevation data includes slope value and altitude;

[0017] Based on the equipment health score in the situation assessment matrix, a dynamic health threshold is generated through a slope compensation algorithm, where the slope compensation algorithm adjusts the judgment threshold according to the slope value;

[0018] Use dynamic health thresholds to filter abnormal device nodes and obtain a primary abnormal node set;

[0019] Based on the communication quality score in the situation assessment matrix, a corrected communication quality index is generated through a multipath attenuation correction model;

[0020] The communication abnormal nodes are screened using the corrected communication quality index to obtain a set of secondary abnormal nodes;

[0021] The primary abnormal node set and the secondary abnormal node set are merged as abnormal nodes, and the normal node data is output after filtering the abnormal nodes.

[0022] In one embodiment, calculating the partition data integrity rate based on normal node data and determining the area where the data integrity rate is lower than a preset threshold as a monitoring blind area includes:

[0023] Obtaining a geological structure zoning map, wherein the geological structure zoning map includes lithology distribution characteristics and fault trend information;

[0024] Based on normal node data, the monitoring sub-areas are divided within the same geological unit according to the geological structure zoning map;

[0025] Calculate the data integrity rate variation index of each monitoring sub-area, where the variation index is used to characterize the stability of data collection;

[0026] An adaptive boundary curve is generated according to the variation index, and the area enclosed by the adaptive boundary curve is marked as a monitoring blind area.

[0027] In one embodiment, the training method of the disaster chain prediction model includes:

[0028] Obtain historical disaster event datasets, which include surface deformation time series data, environmental parameter data, and disaster chain records;

[0029] Perform spatiotemporal data enhancement processing on the historical disaster event dataset to generate an enhanced training sample set, where the spatiotemporal data enhancement processing includes random time slicing and spatial affine transformation;

[0030] Build a neural network infrastructure and embed the rock mass mechanics equilibrium equations as physical constraints in the hidden layer;

[0031] Perform multi-region adaptation training on the enhanced training sample set through a meta-learning framework to generate multi-region adaptation model parameters;

[0032] Based on the multi-region adaptation model parameters, the neural network infrastructure is fine-tuned through transfer learning to generate a disaster chain prediction model.

[0033] Secondly, this application also provides an IoT-InSAR collaborative disaster chain monitoring and early warning system, including:

[0034] The quality assessment module is used to perform quality assessment on the acquired IoT data and obtain a situation assessment matrix. The IoT data includes voltage value, communication signal strength, and data collection timestamp;

[0035] The anomaly identification module is used to identify and extract abnormal nodes based on the situation assessment matrix to obtain normal node data;

[0036] The blind zone calibration module is used to calculate the partition data integrity rate based on normal node data and determine the area where the data integrity rate is lower than the preset threshold as a monitoring blind zone;

[0037] The data fusion module is used to obtain InSAR data from monitoring blind areas and extract surface deformation data, and fuse the surface deformation data with normal node data to generate a global continuous deformation dataset;

[0038] The disaster prediction module is used to perform spatiotemporal feature analysis on the global continuous deformation dataset to obtain a fused feature vector, and then input the fused feature vector into the pre-trained disaster chain prediction model to generate a disaster risk heat map.

[0039] In a third aspect, the present application also provides a computer device including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the above-mentioned IoT-InSAR collaborative disaster chain monitoring and early warning method.

[0040] In a fourth aspect, the present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned IoT-InSAR collaborative disaster chain monitoring and early warning method.

[0041] The aforementioned IoT-InSAR collaborative disaster chain monitoring and early warning method and system uses a situation assessment matrix generated by quality assessment of IoT data, including voltage values, communication signal strength, and data acquisition timestamps. This matrix then identifies and extracts abnormal nodes, obtains normal node data, and calculates the partitioned data integrity rate to identify monitoring blind spots with low data integrity. InSAR data from these blind spots is then used to extract surface deformation data, which is then fused with normal node data to form a global continuous deformation dataset. This fused feature vector is then analyzed through spatiotemporal characteristics and input into a pre-trained disaster chain prediction model to generate a disaster risk heat map. This technical approach effectively integrates IoT network situational awareness data with surface deformation data, enabling the monitoring network to dynamically identify coverage blind spots and promptly supplement key data, improving the accuracy of disaster chain predictions. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0043] Figure 1 A schematic diagram of the process of the IoT-InSAR collaborative disaster chain monitoring and early warning method provided by an embodiment of the present invention;

[0044] Figure 2 A schematic diagram of the structure of the IoT-InSAR collaborative disaster chain monitoring and early warning system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0046] First, a brief introduction is given to the terms involved in the embodiments of this application.

[0047] The Internet of Things (IoT) is an intelligent infrastructure system centered around information sensing devices that interconnect everything through communication networks. By deploying terminal devices (such as sensors and controllers) with sensing, computing, and communication capabilities, it collects and transmits real-time status information about physical entities. This information is then integrated into cloud or edge computing platforms for data fusion analysis and intelligent decision-making, ultimately building a collaborative network that deeply interacts between the physical world and the digital space. In the field of disaster monitoring, the IoT serves as the sensory nerve endings that perceive dynamic changes in the physical environment, providing real-time, multi-dimensional data support for risk assessment.

[0048] Synthetic Aperture Radar (InSAR), an emerging active microwave remote sensing technology, leverages its ability to penetrate the atmosphere and acquire deformation information of monitored targets around the clock and in all weather conditions. It has been successfully applied in a variety of fields, including surface deformation monitoring, DEM generation, landslides, volcanic activity, glacial movement, and deformation information extraction of man-made structures. As an emerging spatial geodetic technique, InSAR overcomes the shortcomings of traditional geodetic techniques, such as the need for manual field deployment, low spatial resolution, and high cost. It has become a key technical tool in scientific research and engineering practice in fields such as surveying and mapping, remote sensing, geophysics, geological engineering, environmental engineering, civil engineering, disaster monitoring and assessment, resource exploration, and geographic information engineering.

[0049] Based on the above explanations, the IoT-InSAR collaborative disaster chain monitoring and early warning method provided in the embodiments of this application is described. Schematically, the implementation environment includes: a sensor group, a terminal, and a processor. The terminal is connected to the sensor group, intelligent adjustment device, and processor via a network. The sensor group includes, but is not limited to, voltage monitoring sensors, signal strength monitoring sensors, timestamp recording sensors, and surface deformation monitoring sensors. The processor can be a central processing unit, a neural network processor, or a multi-core processor, without limitation here.

[0050] In combination with the above explanations of terms and implementation environments, the application scenarios of the embodiments of this application are explained. The disaster chain monitoring and early warning method of the Internet of Things-InSAR collaboration provided in the embodiments of this application can be applied to, but not limited to, the following scenarios:

[0051] In areas where earthquakes are frequent or where there are risks of landslides and mudslides, IoT sensors can collect key data such as surface tilt angle and ground stress in real time. At the same time, InSAR can accurately capture tiny surface displacements. The two work together to quickly identify abnormal stratum changes, discover precursors of geological disasters in advance, buy time for evacuation and resettlement, and significantly improve the accuracy of geological disaster monitoring and the timeliness of early warning.

[0052] For large infrastructure such as bridges and dams, IoT sensors are used to monitor structural stress, vibration frequency and other data in real time, while InSAR regularly scans the overall deformation of the structure to promptly detect minor damage caused by natural disasters, ensure the stability and safety of infrastructure under disaster conditions, reduce operation and maintenance costs, and extend the service life cycle.

[0053] After an earthquake, IoT sensors are used to quickly obtain real-time data such as damage to buildings and signs of human activity in the affected area. InSAR can monitor macroscopic surface deformations such as surface collapse and landslides on a large scale. The fusion of the two generates a detailed disaster risk heat map, providing strong support for the scientific deployment of rescue forces and precise assistance to the affected people, greatly improving the efficiency of earthquake emergency response and rescue effectiveness.

[0054] Illustratively, the IoT-InSAR collaborative disaster chain monitoring and early warning method provided in the embodiments of the present application can also be applied to other application scenarios. This is only used as an example and is not limited to specific application scenarios.

[0055] In an exemplary embodiment, Figure 1 As shown, a disaster chain monitoring and early warning method for IoT-InSAR collaboration is provided. This embodiment uses the method as an example of a terminal in the aforementioned implementation environment. It is understandable that the method can also be applied to a server, or to a system including a terminal and a server, and implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps 101 to 105:

[0056] Step 101: perform quality assessment on the acquired IoT data to obtain a situation assessment matrix. The IoT data includes voltage value, communication signal strength, and data collection timestamp.

[0057] Specifically, the acquired IoT data undergoes a quality assessment to generate a situation assessment matrix. IoT data includes voltage values, communication signal strength, and data collection timestamps. Specifically, this IoT data is collected in real time by various sensors deployed in the monitoring area and transmitted to the terminal. The terminal runs a pre-set quality assessment algorithm to quantitatively assess data quality based on multiple metrics, such as voltage stability, communication signal strength, and data collection timestamp continuity. Ultimately, a situation assessment matrix is ​​generated. This matrix visually represents the quality of each node's data, providing a basis for subsequent screening of reliable data. For example, quality assessment can utilize a multi-dimensional indicator fusion approach: for example, by weighting parameters such as voltage stability, signal strength fluctuation, and timestamp continuity, a matrix representing node health is generated. Furthermore, the situation assessment matrix can be combined with entropy weighting or principal component analysis (PCA) to quantify data reliability, providing standardized input for subsequent steps. This step effectively filters out low-quality data and prevents anomalies from interfering with subsequent analysis.

[0058] Step 102: Identify and extract abnormal nodes based on the situation assessment matrix to obtain normal node data.

[0059] Specifically, abnormal node identification uses clustering algorithms (such as DBSCAN or Isolation Forest) to detect nodes that deviate from group characteristics, such as sensors experiencing sudden voltage changes or persistent signal interruptions. Furthermore, dynamic thresholds can be set to segment abnormal data, retaining node data that meets consistency requirements. This step significantly improves data credibility. For example, in slope monitoring, distorted data caused by equipment failure can be eliminated, ensuring that subsequent analysis is based on valid input.

[0060] Step 103 : Calculate the partition data integrity rate based on the normal node data, and determine the area where the data integrity rate is lower than a preset threshold as a monitoring blind area.

[0061] Specifically, the monitoring area is divided into multiple sub-areas based on geographic spatial distribution or network topology. For each sub-area, the completeness of the normal node data is calculated, using a formula that can be calculated as the ratio of the number of normal nodes in that sub-area to the preset number of standard nodes. For example, in a landslide monitoring area, the slope surface, slope foot, and other parts are divided into zones. If the number of standard nodes in a slope zone is 100, but there are only 30 normal nodes, and the preset threshold is 70%, then the slope zone is determined to be a monitoring blind spot.

[0062] Step 104 : Acquire the InSAR data of the monitoring blind area and extract the surface deformation data, and fuse the surface deformation data with the normal node data to generate a global continuous deformation data set.

[0063] Specifically, InSAR data (collected via satellite remote sensing) is acquired from monitoring blind spots. Signal processing algorithms are then used to accurately extract surface deformation characteristic parameters, such as deformation amplitude and deformation rate, from the InSAR data. Data fusion algorithms are then used to fuse this surface deformation data with normal node data. For example, a Kalman filter fusion algorithm is employed, combining the real-time nature of IoT data with the spatial coverage advantages of InSAR data to generate a continuous, seamless deformation dataset across the entire monitoring area. This compensates for the lack of data in monitoring blind spots and enables comprehensive control of surface deformation conditions across the entire monitoring area.

[0064] Step 105 , performing spatiotemporal feature analysis on the global continuous deformation dataset to obtain a fused feature vector, and inputting the fused feature vector into a pre-trained disaster chain prediction model to generate a disaster risk heat map.

[0065] Specifically, the spatiotemporal data analysis algorithm is used to explore the evolution laws of surface deformation in time series and the correlation patterns in spatial distribution from the global continuous deformation data set. For example, the acceleration or deceleration trend of the deformation rate over time, as well as the synergistic or differential characteristics of deformation in adjacent areas are analyzed. Based on this, a fusion feature vector is constructed. The fusion feature vector comprehensively characterizes the potential risk factors of the disaster chain in the monitored area and is input into the disaster chain prediction model trained and optimized with a large amount of historical disaster data. This model is based on machine learning algorithms, such as deep neural networks, which can intelligently identify the mapping relationship between feature vectors and disaster risks, output intuitive disaster risk heat maps, clearly display the disaster risk levels in different regions, and provide a strong basis for disaster prevention and emergency decision-making.

[0066] This IoT-InSAR collaborative disaster chain monitoring and early warning method uses a situation assessment matrix generated by quality-assessing IoT data, including voltage values, communication signal strength, and data acquisition timestamps. This matrix then identifies and extracts abnormal nodes, obtains normal node data, and calculates the data integrity rate for each partition, identifying monitoring blind spots with low data integrity. InSAR data from these blind spots is then used to extract surface deformation data, which is then fused with the normal node data to form a global continuous deformation dataset. This fused feature vector is then analyzed through spatiotemporal characteristics and input into a pretrained disaster chain prediction model to generate a disaster risk heat map. This technical approach effectively integrates IoT network situational awareness data with surface deformation data, enabling the monitoring network to dynamically identify coverage blind spots and promptly supplement key data, improving the accuracy of disaster chain predictions.

[0067] In one embodiment, the quality of the acquired IoT data is assessed to obtain a situation assessment matrix, including:

[0068] Clean and standardize IoT data to obtain standardized data sets;

[0069] Calculate the multi-dimensional situation assessment value of each device node in the standardized data set based on the sliding time window;

[0070] The geographical coordinates of each device node are associated with the multi-dimensional situation assessment value to generate a situation assessment matrix stored in row format.

[0071] Specifically, the raw data is cleaned and standardized: for example, the abnormal fluctuation data is eliminated by the outlier detection algorithm, and the dimensional difference is eliminated by the Z-score standardization method to obtain a standardized data set aligned in time and space. Furthermore, the multi-dimensional situation assessment value of each device node can be calculated based on the sliding time window, for example, the weighted fusion algorithm is used to comprehensively calculate the device health (integrating voltage stability and signal attenuation rate) and data integrity rate (based on timestamp continuity analysis), where the window size is dynamically adjusted according to the device type. The geographic coordinates of the device node are associated with the multi-dimensional situation assessment value to generate a row-stored situation assessment matrix, for example, the GeoHash encoding of the spatial database is used to achieve a fast mapping of coordinates and assessment values. This embodiment systematically processes the IoT data, first cleaning and standardizing it to ensure data quality and comparability, then using the sliding time window to dynamically track the node situation, and combining the geographic coordinates to generate an intuitive situation assessment matrix, providing a high-quality, spatiotemporally correlated data foundation for subsequent precise monitoring and early warning, effectively improving the spatiotemporal accuracy and reliability of disaster monitoring, especially in complex disaster monitoring scenarios with multi-source data fusion.

[0072] In one embodiment, identifying and extracting abnormal nodes based on the situation assessment matrix to obtain normal node data includes:

[0073] Obtain terrain elevation data, where the terrain elevation data includes slope value and altitude;

[0074] Based on the equipment health score in the situation assessment matrix, a dynamic health threshold is generated through a slope compensation algorithm, where the slope compensation algorithm adjusts the judgment threshold according to the slope value;

[0075] Use dynamic health thresholds to filter abnormal device nodes and obtain a primary abnormal node set;

[0076] Based on the communication quality score in the situation assessment matrix, a corrected communication quality index is generated through a multipath attenuation correction model;

[0077] The communication abnormal nodes are screened using the corrected communication quality index to obtain a set of secondary abnormal nodes;

[0078] The primary abnormal node set and the secondary abnormal node set are merged as abnormal nodes, and the normal node data is output after filtering the abnormal nodes.

[0079] Specifically, terrain elevation data, including slope values ​​and altitude, is obtained. This data is derived from a pre-stored geographic information database or real-time remote sensing measurements. Exemplarily, based on the device health score in the situation assessment matrix, the method generates a dynamic health threshold using a slope compensation algorithm. The slope compensation algorithm adaptively adjusts the determination threshold based on the slope value, for example, using a looser threshold in steep areas to compensate for terrain interference. Furthermore, the dynamic health threshold is used to filter abnormal device nodes to obtain a primary abnormal node set. For example, hardware fault nodes are identified by comparing the device health score with the dynamic threshold. Based on the communication quality score in the situation assessment matrix, a corrected communication quality index is generated using a multipath attenuation correction model. Specifically, this model simulates signal propagation path loss using terrain elevation data, for example, using a ray tracing algorithm to eliminate signal distortion caused by terrain obstruction. Furthermore, the corrected communication quality index is used to filter abnormal communication nodes to obtain a secondary abnormal node set. For example, nodes whose corrected index falls below a baseline value are marked as communication failure nodes. The primary and secondary abnormal node sets are combined to form a complete abnormal node set, which is then filtered and output as high-quality normal node data. This implementation uses a slope compensation algorithm and a multipath attenuation correction model, combined with terrain elevation data and a situation assessment matrix, to dynamically adjust health thresholds and calibrate communication quality indicators, thereby more accurately identifying abnormal nodes. This multi-dimensional, dynamic screening method effectively improves the reliability of normal node data and enhances monitoring accuracy in complex terrain and changing environments, providing a solid data foundation for subsequent disaster chain monitoring and early warning.

[0080] In one embodiment, calculating the partition data integrity rate based on normal node data and determining the area where the data integrity rate is lower than a preset threshold as a monitoring blind area includes:

[0081] Obtaining a geological structure zoning map, wherein the geological structure zoning map includes lithology distribution characteristics and fault trend information;

[0082] Based on normal node data, the monitoring sub-areas are divided within the same geological unit according to the geological structure zoning map;

[0083] Calculate the data integrity rate variation index of each monitoring sub-area, where the variation index is used to characterize the stability of data collection;

[0084] An adaptive boundary curve is generated according to the variation index, and the area enclosed by the adaptive boundary curve is marked as a monitoring blind area.

[0085] Specifically, it is crucial to obtain a geological structure zoning map, which is provided by the geological exploration department or generated through professional remote sensing interpretation and contains detailed lithologic distribution characteristics and fault strike information to understand regional geological stability. For example, in earthquake-prone areas, fault strike information can help predict the propagation path and impact range of seismic waves. Based on the collected normal node data, monitoring sub-areas are divided according to the same geological units in the geological structure zoning map. For example, an area with the same lithology and in the same fault block is divided into a monitoring sub-area to ensure that the geological conditions in each sub-area are consistent. The data integrity rate variation index of each monitoring sub-area is calculated. The variation index is obtained by statistically analyzing the continuity and stability of data acquisition in each sub-area. For example, the ratio of the standard deviation to the mean (coefficient of variation) is used to quantify the volatility of data acquisition. The larger the coefficient of variation, the worse the stability of data acquisition. Furthermore, an adaptive boundary curve is generated based on the variation index, and the area enclosed by the adaptive boundary curve is marked as a monitoring blind area. For example, a kernel density estimation method, using the variation index as a weight, calculates the density distribution of each sub-area, generating a smooth, adaptive boundary curve that precisely encloses monitoring blind spots with low data integrity and unstable data collection. This method effectively identifies weak links in the monitoring network, significantly improving its coverage integrity and data reliability. It is particularly suitable for disaster monitoring under complex geological conditions, providing more robust data support for disaster chain warnings.

[0086] In one embodiment, the training method of the disaster chain prediction model includes:

[0087] Obtain historical disaster event datasets, which include surface deformation time series data, environmental parameter data, and disaster chain records;

[0088] Perform spatiotemporal data enhancement processing on the historical disaster event dataset to generate an enhanced training sample set, where the spatiotemporal data enhancement processing includes random time slicing and spatial affine transformation;

[0089] Build a neural network infrastructure and embed the rock mass mechanics equilibrium equations as physical constraints in the hidden layer;

[0090] Perform multi-region adaptation training on the enhanced training sample set through a meta-learning framework to generate multi-region adaptation model parameters;

[0091] Based on the multi-region adaptation model parameters, the neural network infrastructure is fine-tuned through transfer learning to generate a disaster chain prediction model.

[0092] In this embodiment, the method for training the disaster chain prediction model of the scheme includes: obtaining a historical disaster event data set containing surface deformation time series data, environmental parameter data and disaster chain records, such as a ten-year disaster case library collected by geological monitoring stations and satellite remote sensing platforms. Specifically, the data set is subjected to spatiotemporal data enhancement processing: the random time slicing technology is exemplarily used to intercept deformation sequence fragments of different lengths, and spatial affine transformations (such as rotation and scaling) are applied to generate equivalent variants of geological structures, thereby constructing an enhanced training sample set to solve the problem of sample scarcity. Furthermore, when constructing the neural network infrastructure, the rock mechanics equilibrium equation is embedded in the hidden layer as a physical constraint condition, for example, the Moore-Coulomb criterion is converted into a differentiable loss function to force the network output to conform to the rock stress-strain law. Multi-region adaptation training is performed through a meta-learning framework: for example, the model-agnostic meta-learning (MAML) algorithm is used to quickly optimize the initial parameters on sub-sample sets of different geological regions such as volcanic belts and fault zones, so that the model can obtain cross-regional generalization capabilities. Based on the multi-region adaptation model parameters, the network architecture is fine-tuned through transfer learning: for example, the base layer parameters are frozen, and only the fully connected layer is fine-tuned to adapt to the characteristics of the target area, generating a highly robust disaster chain prediction model.

[0093] In summary, the IoT-InSAR collaborative disaster chain monitoring and early warning method provided in this application performs quality assessment on the original data such as voltage value and communication signal strength collected by IoT nodes to generate an assessment matrix reflecting the network operation status; based on the matrix and combined with the terrain compensation mechanism, abnormal nodes are dynamically identified to retain high-quality normal node data; then, the data integrity rate variation index is calculated through the geological structure-guided partitioning algorithm to generate an adaptive boundary curve to accurately calibrate the monitoring blind area; InSAR technology is dispatched to obtain surface deformation data within the blind area range, and a spatiotemporal scale conversion model is used to align and fuse minute-level IoT data with daily InSAR data to form a global continuous deformation data set; by extracting spatiotemporal characteristics such as deformation acceleration and hydrological activity, the physically constrained disaster chain prediction model is input to generate a risk heat map, in which the model achieves cross-regional high-precision prediction through the embedding of rock mechanics equations and the meta-learning framework. This solution integrates IoT situational awareness with InSAR remote sensing monitoring capabilities. On the one hand, through the dynamic elimination of abnormal nodes and the calibration of blind spots guided by geological structures, it achieves real-time optimization of monitoring network coverage, addressing the data loss caused by terrain obstruction or equipment failure in traditional methods. On the other hand, by integrating multi-source spatiotemporal data with a physically constrained prediction model, it overcomes the accuracy limitations of a single data source and improves the reliability of predictions of the evolutionary paths of disaster chains such as landslides and debris flows. Ultimately, it forms a closed-loop architecture of "micro-real-time perception - macro-cyclic scanning - intelligent risk decision-making," providing full-chain technical support for geological disaster prevention and control, from data collection to early warning response.

[0094] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0095] Based on the same inventive concept, embodiments of the present application also provide an IoT-InSAR coordinated disaster chain monitoring and early warning system for implementing the aforementioned IoT-InSAR coordinated disaster chain monitoring and early warning method. The solution provided by this system is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more IoT-InSAR coordinated disaster chain monitoring and early warning system embodiments provided below can be found in the limitations of the IoT-InSAR coordinated disaster chain monitoring and early warning method described above and will not be repeated here.

[0096] In an exemplary embodiment, Figure 2 As shown, a disaster chain monitoring and early warning system 10 based on the collaboration of the Internet of Things and InSAR is provided, comprising:

[0097] The quality assessment module 11 is used to perform quality assessment on the acquired IoT data to obtain a situation assessment matrix. The IoT data includes voltage value, communication signal strength, and data collection timestamp;

[0098] Anomaly identification module 12, used to identify and extract abnormal nodes according to the situation assessment matrix to obtain normal node data;

[0099] The blind spot calibration module 13 is used to calculate the partition data integrity rate based on the normal node data, and determine the area where the data integrity rate is lower than a preset threshold as a monitoring blind spot;

[0100] The data fusion module 14 is used to obtain the InSAR data of the monitoring blind area and extract the surface deformation data, and fuse the surface deformation data with the normal node data to generate a global continuous deformation data set;

[0101] The disaster prediction module 15 is used to perform spatiotemporal feature analysis on the global continuous deformation data set to obtain a fused feature vector, and input the fused feature vector into a pre-trained disaster chain prediction model to generate a disaster risk heat map.

[0102] In one embodiment, the quality assessment module 11 includes:

[0103] A data cleaning unit is used to clean and standardize IoT data to obtain a standardized data set;

[0104] A situation assessment unit, configured to calculate a multi-dimensional situation assessment value for each device node in a standardized data set based on a sliding time window;

[0105] The matrix generation unit is used to associate the geographical coordinates of each device node with the multi-dimensional situation assessment value to generate a situation assessment matrix stored in row format.

[0106] In one embodiment, the anomaly identification module 12 includes:

[0107] A terrain data acquisition unit, used to acquire terrain elevation data, wherein the terrain elevation data includes a slope value and an altitude;

[0108] A health threshold generation unit is configured to generate a dynamic health threshold based on the device health score in the situation assessment matrix by using a slope compensation algorithm, wherein the slope compensation algorithm adjusts the determination threshold according to the slope value;

[0109] A device screening unit is used to screen abnormal device nodes using a dynamic health threshold to obtain a primary abnormal node set;

[0110] a communication correction unit, configured to generate a corrected communication quality indicator through a multipath attenuation correction model based on the communication quality score in the situation assessment matrix;

[0111] a communication screening unit, configured to screen communication abnormal nodes using the corrected communication quality index to obtain a secondary abnormal node set;

[0112] The node filtering unit is used to merge the primary abnormal node set and the secondary abnormal node set as abnormal nodes, filter the abnormal nodes and output normal node data.

[0113] In one embodiment, the blind spot calibration module 13 includes:

[0114] A geological data acquisition unit is used to obtain a geological structure zoning map, wherein the geological structure zoning map includes lithology distribution characteristics and fault trend information;

[0115] Sub-area division unit, used to divide monitoring sub-areas within the same geological unit based on normal node data and geological structure zoning map;

[0116] A variation calculation unit is used to calculate the data integrity rate variation index of each monitoring sub-area, where the variation index is used to characterize the stability of data collection;

[0117] The blind spot calibration unit is used to generate an adaptive boundary curve according to the variation index and mark the area enclosed by the adaptive boundary curve as a monitoring blind spot.

[0118] In one embodiment, the training method of the disaster chain prediction model in the IoT-InSAR coordinated disaster chain monitoring and early warning system 10 includes the following steps:

[0119] Obtain historical disaster event datasets, which include surface deformation time series data, environmental parameter data, and disaster chain records;

[0120] Perform spatiotemporal data enhancement processing on the historical disaster event dataset to generate an enhanced training sample set, where the spatiotemporal data enhancement processing includes random time slicing and spatial affine transformation;

[0121] Build a neural network infrastructure and embed the rock mass mechanics equilibrium equations as physical constraints in the hidden layer;

[0122] Perform multi-region adaptation training on the enhanced training sample set through a meta-learning framework to generate multi-region adaptation model parameters;

[0123] Based on the multi-region adaptation model parameters, the neural network infrastructure is fine-tuned through transfer learning to generate a disaster chain prediction model.

[0124] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the aforementioned IoT-InSAR collaborative disaster chain monitoring and early warning method are implemented.

[0125] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0126] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments. The device embodiments described above are merely illustrative, wherein the components described as separate parts may or may not be physically separated, and the parts displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the disclosed solution. A person of ordinary skill in the art can understand and implement it without expending creative work.

[0127] The above-described embodiments merely represent several implementation methods of the embodiments of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that a person skilled in the art may make various modifications and improvements without departing from the concept of the embodiments of the present application, and these modifications and improvements fall within the scope of protection of the embodiments of the present application.

Claims

1. A disaster chain monitoring and early warning method based on the Internet of Things and InSAR collaboration, characterized in that: The method comprises: Performing a quality assessment on the acquired IoT data to obtain a situation assessment matrix, wherein the IoT data includes voltage value, communication signal strength, and data collection timestamp; Identify and extract abnormal nodes according to the situation assessment matrix to obtain normal node data; Calculate the partition data integrity rate based on the normal node data, and determine the area where the data integrity rate is lower than a preset threshold as a monitoring blind area; Acquire the InSAR data of the monitoring blind area and extract surface deformation data, and fuse the surface deformation data with the normal node data to generate a global continuous deformation data set; A spatiotemporal feature analysis is performed on the global continuous deformation dataset to obtain a fused feature vector, and the fused feature vector is input into a pre-trained disaster chain prediction model to generate a disaster risk heat map.

2. The method according to claim 1, characterized in that The quality assessment of the acquired IoT data is performed to obtain a situation assessment matrix, including: Performing data cleaning and standardization on the IoT data to obtain a standardized data set; Calculating a multidimensional situation assessment value for each device node in the standardized data set based on a sliding time window; The geographical coordinates of each of the device nodes are associated with the multi-dimensional situation assessment value to generate the situation assessment matrix stored in row format.

3. The method according to claim 2, characterized in that The step of identifying and extracting abnormal nodes according to the situation assessment matrix to obtain normal node data includes: Acquiring terrain elevation data, wherein the terrain elevation data includes a slope value and an altitude; Based on the equipment health score in the situation assessment matrix, generating a dynamic health threshold through a slope compensation algorithm, wherein the slope compensation algorithm adjusts the determination threshold according to the slope value; Filter abnormal device nodes using the dynamic health threshold to obtain a primary abnormal node set; Based on the communication quality scores in the situation assessment matrix, generating a corrected communication quality indicator through a multipath attenuation correction model; Using the corrected communication quality indicator to screen communication abnormal nodes to obtain a secondary abnormal node set; The primary abnormal node set and the secondary abnormal node set are merged as the abnormal node, and the normal node data is output after filtering the abnormal nodes.

4. The method according to claim 1, wherein The calculating the partition data integrity rate based on the normal node data and determining the area where the data integrity rate is lower than a preset threshold as a monitoring blind area includes: Obtaining a geological structure zoning map, wherein the geological structure zoning map includes lithology distribution characteristics and fault trend information; Based on the normal node data, the monitoring sub-areas are divided within the same geological unit according to the geological structure zoning map; Calculating a data integrity rate variation index for each monitoring sub-area, wherein the variation index is used to characterize data acquisition stability; An adaptive boundary curve is generated according to the variation index, and an area enclosed by the adaptive boundary curve is marked as the monitoring blind area.

5. The method according to any one of claims 1 to 4, characterized in that The training method of the disaster chain prediction model includes: Acquire a historical disaster event dataset, wherein the historical disaster event dataset includes surface deformation time series data, environmental parameter data, and disaster chain records; Performing spatiotemporal data enhancement processing on the historical disaster event dataset to generate an enhanced training sample set, wherein the spatiotemporal data enhancement processing includes random time slicing and spatial affine transformation; Build a neural network infrastructure and embed the rock mass mechanics equilibrium equations as physical constraints in the hidden layer; Performing multi-region adaptation training on the enhanced training sample set through a meta-learning framework to generate multi-region adaptation model parameters; Based on the multi-region adaptation model parameters, the neural network infrastructure is fine-tuned through transfer learning to generate the disaster chain prediction model.

6. An IoT-InSAR collaborative disaster chain monitoring and early warning system, characterized by: The system comprises: A quality assessment module is used to perform quality assessment on the acquired IoT data to obtain a situation assessment matrix. The IoT data includes voltage value, communication signal strength, and data collection timestamp; An anomaly identification module is used to identify and extract abnormal nodes according to the situation assessment matrix to obtain normal node data; A blind spot calibration module is used to calculate the partition data integrity rate based on the normal node data, and determine the area where the data integrity rate is lower than a preset threshold as a monitoring blind spot; A data fusion module is used to obtain the InSAR data of the monitoring blind area and extract surface deformation data, and fuse the surface deformation data with the normal node data to generate a global continuous deformation data set; The disaster prediction module is used to perform spatiotemporal feature analysis on the global continuous deformation dataset and input a pre-trained disaster chain prediction model to generate a disaster risk heat map.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 5 is implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

Images