Method for quickly investigating and evaluating geological disasters based on multi-source remote sensing data

Abstract

The invention relates to the field of geological disaster monitoring and evaluation, and discloses a geological disaster rapid investigation and evaluation method based on multi-source remote sensing data, which comprises the following steps of: acquiring optical remote sensing data, SAR (Synthetic Aperture Radar) data and ground sensor data, and establishing unified modeling; consistency verification is carried out on the multi-source data, abnormal data are removed, and the weight is dynamically adjusted; data fusion is realized through a distributed Kalman filtering and variational inference technology; key risk indexes such as the earth surface deformation rate and the rainfall threshold value are extracted, a multi-factor risk evaluation model is constructed, risk grades are divided, and a risk distribution diagram and dynamic early warning information are generated. According to the invention, the problems of single data source, insufficient inconsistency processing and lagging risk assessment in the prior art are solved, the precision and efficiency of geological disaster monitoring are significantly improved, and a scientific basis is provided for disaster early warning and emergency response.

Description

Technical Field

[0001] The present invention relates to the technical field of geological disaster monitoring and assessment, and in particular to a method for rapid investigation and assessment of geological disasters based on multi-source remote sensing data. Background Art

[0002] Geological disasters (such as landslides, debris flows and ground subsidence) are major natural disasters that seriously threaten the safety of human life and property. Their causes are complex, dynamic and affected by a variety of geological and environmental factors. Monitoring and assessment technologies for geological disasters are of great significance in disaster reduction, disaster prevention and disaster emergency response. In existing technologies, monitoring and risk assessment of geological disasters mainly rely on remote sensing data and ground monitoring data. However, with the increase in disaster monitoring needs and the complexity of disaster scenarios, existing technologies still have many shortcomings in monitoring accuracy, data processing efficiency and comprehensiveness of risk assessment.

[0003] First, existing technologies usually use a single data source for monitoring, such as optical remote sensing data or SAR (synthetic aperture radar) data. These single data sources have certain limitations in spatial coverage and information dimensions. For example, optical remote sensing data cannot obtain effective information under bad weather conditions (such as cloudy, rainy and snowy), and although SAR data can capture tiny surface deformations, it is affected by terrain undulations and vegetation coverage, and the results may be noisy and biased. At the same time, although the monitoring scheme that relies solely on ground sensors can provide local high-precision dynamic change information, the monitoring range is limited and the cost is high. Therefore, a single data source cannot fully meet the needs of large-scale geological disaster monitoring, and it is difficult to take into account both macro-scale geological disaster monitoring and local refined analysis.

[0004] Secondly, the fusion processing of multi-source remote sensing data and ground sensor data faces technical difficulties. The resolution, sampling frequency and data format of different data sources vary greatly, resulting in inconsistency of data in time and space. At the same time, the noise levels of different sensors vary, and reliability verification and abnormal data removal are required before fusion. Existing technologies mostly use simple alignment or interpolation methods for data preprocessing, lacking an effective verification mechanism for data noise, inconsistency and dynamic changes, resulting in the inability to guarantee the quality and credibility of fused data.

[0005] In addition, existing technologies usually adopt a centralized processing framework in the process of multi-source data fusion, such as the traditional Kalman filtering method. This method has a heavy computational burden when facing multi-source and multi-level data, and is difficult to adapt to disaster scenarios with high real-time requirements. Especially in dynamic disaster monitoring, centralized processing is difficult to quickly respond to changes in sensor networks and new data sources, and lacks dynamic optimization capabilities for data uncertainty, resulting in insufficient real-time and robustness of the fusion results.

[0006] Therefore, the present invention proposes a method for rapid investigation and evaluation of geological disasters based on multi-source remote sensing data to address the deficiencies of the prior art. Summary of the invention

[0007] In view of the problems in the prior art such as limited monitoring range of a single data source, insufficient consistency verification of multi-source data, low data fusion efficiency, and insufficient dynamic and multi-factor support of risk assessment models, the present invention provides a method for rapid investigation and evaluation of geological disasters based on multi-source remote sensing data. By introducing multi-source data consistency verification, multi-factor dynamic data fusion and a multi-parameter risk assessment model, efficient and accurate investigation and evaluation of geological disasters can be achieved, providing a scientific basis for disaster warning and emergency response.

[0008] To achieve the above objectives, the present invention is implemented by the following technical scheme: A method for rapid investigation and evaluation of geological disasters based on multi-source remote sensing data comprises the following steps: Acquire multi-source remote sensing data, verify the consistency of multi-source remote sensing data, remove abnormal data and dynamically adjust the weight of each data source; After fusion of verified data, the dynamic state of geological hazards is estimated through data fusion technology; Conduct geological disaster risk assessment based on fused data, extract key risk indicators and classify risk levels; Output geological disaster risk assessment results and generate risk distribution maps and warnings for the monitored areas.

[0009] Preferably, the multi-source remote sensing data includes optical remote sensing data, synthetic aperture radar data and ground sensor data, and the ground sensor data includes data from a displacement sensor, an inclination sensor and a rainfall sensor.

[0010] Preferably, the consistency verification comprises the following steps: Calculate observation errors between different data sources; Verify data consistency based on the error covariance matrix; Eliminate abnormal data whose error exceeds the confidence interval; The weight of each data source is dynamically adjusted, and the weight is inversely proportional to the noise covariance of the data.

[0011] Preferably, the data fusion technology is implemented by distributed Kalman filtering, including: Predict the status of each data source node and update the error covariance; Use observation data to update node states and pass local estimation results to the global system; The weights are dynamically adjusted according to the error covariance matrix of each node, and global state estimation is achieved through weighted averaging.

[0012] Preferably, the global state estimation in the distributed Kalman filter is implemented by the following method: The state estimation result of each node is weighted according to the trace value of its error covariance; The weight is inversely proportional to the trace value of the error covariance matrix, and the fusion result is the weighted average of all nodes.

[0013] Preferably, the geological disaster risk assessment includes: Extract key risk indicators, including surface deformation rates, rainfall thresholds, and geodynamic parameters; Divide the risk level of the monitoring area based on the risk assessment model; Generate a risk distribution map to show the risk distribution of the monitored area.

[0014] Preferably, the output risk assessment results include a risk level distribution map, dynamic warning information and a detailed disaster assessment report.

[0015] Preferably, the dynamic adjustment of weights in the consistency verification is achieved by real-time analysis and optimal allocation of the data noise covariance matrix, thereby achieving dynamic adjustment of data source weights.

[0016] Preferably, the multi-source remote sensing data fusion technology is further combined with a variational inference method to optimize data uncertainty, minimize data uncertainty and dynamically update data source weights.

[0017] Preferably, the rapid geological disaster investigation and evaluation system based on multi-source remote sensing data is characterized by comprising: Data acquisition module, used to acquire multi-source remote sensing data, including optical remote sensing data, SAR data and ground sensor data; Data consistency verification module, used to verify the consistency of multi-source remote sensing data, eliminate abnormal data and dynamically adjust data weights; Data fusion module, used to dynamically fuse multi-source data through distributed Kalman filtering and variational inference methods; Risk assessment module, used to extract key risk indicators of geological hazards and classify risk levels; The result output module is used to generate risk distribution maps of geological hazards, dynamic warning information and detailed regional assessment reports.

[0018] The present invention provides a method for rapid investigation and evaluation of geological disasters based on multi-source remote sensing data. It has the following beneficial effects: 1. The present invention adopts a unified modeling technology solution for multi-source remote sensing data, and achieves the technical effect of comprehensively expressing the dynamic state of geological disasters through the comprehensive collection and standardized modeling of optical remote sensing data, SAR data and ground sensor data. Compared with the limitations of single data source monitoring solutions in the prior art in terms of spatial coverage and information richness, it solves the problem of high fusion difficulty caused by inconsistent spatiotemporal resolution and complex data formats between different data sources.

[0019] 2. The present invention adopts a consistency verification technical solution based on error propagation theory and dynamic weight adjustment. It calculates errors through the covariance matrix and removes abnormal data. At the same time, it dynamically allocates data weights according to the noise level, achieving the technical effect of significantly improving data reliability and input quality. Compared with the technical solution in the prior art that only relies on single error control or manual screening, it solves the technical problem of unstable data quality and severe noise interference leading to insufficient accuracy of monitoring results.

[0020] 3. The present invention adopts a multi-source data fusion technology solution that combines distributed Kalman filtering with variational inference optimization, and realizes dynamic state estimation through state prediction, local update and global weighting, achieving a high-precision, real-time data fusion technology effect. Compared with the traditional Kalman filtering technology in the prior art, which has a heavy computational burden and insufficient real-time performance under a centralized processing architecture, it solves the problems of low efficiency of multi-source data fusion and poor real-time dynamic estimation capabilities.

[0021] 4. The present invention adopts a risk assessment technical solution based on a multi-factor weighted evaluation model. By extracting key risk indicators such as surface deformation rate, rainfall threshold, and geodynamic parameters and combining the GIS system to generate a risk distribution map, it achieves the technical effect of accurately dividing risk levels and dynamically updating risk results. Compared with the technical solution of risk assessment using a single factor or static model in the prior art, it solves the problem of low risk assessment accuracy and poor timeliness in the context of multi-hazard coupling. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a flow chart of the rapid investigation and evaluation method of geological hazards based on multi-source remote sensing data; Figure 2 This is the architecture diagram of the geological disaster rapid investigation and evaluation system based on multi-source remote sensing data. DETAILED DESCRIPTION

[0023] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] Please refer to the attached Figure 1 The embodiment of the present invention provides a method for rapid investigation and evaluation of geological disasters based on multi-source remote sensing data. The following is a detailed description of each step of the method of the present invention.

[0025] The method for rapid investigation and evaluation of geological hazards based on multi-source remote sensing data is characterized by comprising the following steps: S1. Obtain multi-source remote sensing data, verify the consistency of multi-source remote sensing data, remove abnormal data and dynamically adjust the weight of each data source First, it is necessary to obtain multi-source remote sensing data as the basic input for the entire geological disaster rapid investigation and evaluation method. Specifically, the multi-source remote sensing data obtained include space remote sensing data and ground sensor data. Space remote sensing data includes optical remote sensing data and SAR (synthetic aperture radar) data, which are used to capture large-scale surface deformation and geological features, while ground sensor data includes displacement sensors, inclination sensors, and rainfall sensors, etc., which are used to supplement local high-precision dynamic monitoring information. The acquired data needs to be uniformly modeled to construct system state equations and observation equations to standardize the expression of multi-source remote sensing data and provide structured support for subsequent data processing.

[0026] In this embodiment, the steps of acquiring and modeling multi-source remote sensing data include the following specific contents.

[0027] In general, the acquisition of space remote sensing data relies on commonly used satellite systems. For example, optical remote sensing data can be derived from high-resolution satellites (such as Sentinel, Landsat or Gaofen series), with a resolution of sub-meter level, which can be used to capture the surface morphology and environmental changes in geological disaster areas; SAR data can be obtained through synthetic aperture radar (such as Sentinel-1 or Radarsat), which has the characteristics of all-weather and high penetration, and can be used for surface deformation monitoring and identification of small deformations. As an option, SAR data can also be used in conjunction with optical remote sensing data to achieve multimodal monitoring of geological disasters.

[0028] Specifically, ground sensor data includes displacement sensors, inclination sensors and rainfall sensors, which are installed in high-risk areas of geological disasters to capture local dynamic change information. Displacement sensors can be used to monitor the horizontal and vertical displacement of slopes or surfaces; inclination sensors are used to measure the changes in the slope angle; and rainfall sensors are used to monitor rainfall intensity in real time and capture external environmental factors that trigger geological disasters.

[0029] In a possible implementation, the acquired multi-source data is structured through unified modeling to form a system state equation and an observation equation, which are used to describe the dynamic change law of geological hazards and the relationship between the observed data and the real state. Specifically, the system state equation can be expressed as:

[0030] in, For the moment The geological hazard state variables include but are not limited to surface deformation rate, strain tensor, slope, etc.; is the state transfer matrix, which describes the evolution of geological hazard state over time; is the process noise, which represents the random error of the system state and satisfies the Gaussian distribution. The noise covariance is .

[0031] In terms of observation modeling, the observation equation can be expressed as:

[0032] in, For the moment From Observations obtained from data sources; The observation matrix defines the mapping relationship between the observation data and the geological hazard status; is the observation noise, which satisfies the Gaussian distribution and has a covariance of , used to characterize the The observation error of a data source.

[0033] In some embodiments, for different types of data sources, the observation matrix may be in different forms. For example, for optical remote sensing data, It can be expressed as a linear relationship between pixel values ​​in high-resolution images and geological hazard state variables (such as slope and vegetation coverage); for SAR data, It can be extracted through interferometric radar technology to describe the relationship between surface displacement and geological state variables; for ground sensor data, It may be a correlation matrix between ground monitoring parameters (such as inclination or rainfall) and the state of geological hazards.

[0034] As an alternative, the process noise covariance in the geohazard state Covariance with observation noise Statistical estimation can be performed through historical data. For example, the noise covariance of SAR data can be estimated by the residual sum of squares of multiple observations; the noise covariance of ground sensors can be calibrated through the calibration data of sensor equipment. These noise covariances will serve as important parameters for subsequent consistency verification and data fusion.

[0035] In the specific implementation process, the modeling parameters can also be dynamically adjusted for different types of geological disaster areas. For example, for landslide areas, the state variables It may be necessary to include the slope height change rate and rainfall accumulation value; for earthquake-induced landslide areas, the model may need to add the earthquake peak acceleration parameter. The flexibility of the above modeling methods ensures the applicability of the system in different application scenarios.

[0036] Through the above acquisition and modeling process, multi-source remote sensing data are unified into a structured input form, providing standardized data expression for subsequent steps (such as consistency verification and data fusion). This structured representation can effectively solve the problem of differences in spatiotemporal resolution of different data sources, laying a solid foundation for the rapid investigation and assessment of geological hazards.

[0037] S2. Fusion of verified data to estimate the dynamic state of geological hazards through data fusion technology After completing the acquisition and modeling of multi-source remote sensing data in step S1, the acquired data may have problems such as data inconsistency or noise interference due to differences in sensor type, spatiotemporal resolution, and observation accuracy. Therefore, it is necessary to analyze and process the multi-source remote sensing data through consistency verification, remove abnormal data, and dynamically adjust the weight of each data source to ensure the reliability and accuracy of the input data. This step is crucial before realizing multi-source data fusion, and provides a high-quality input foundation for subsequent data processing.

[0038] In this embodiment, consistency verification is mainly implemented based on technical steps such as error analysis, covariance propagation and dynamic weight adjustment, and the specific contents are as follows.

[0039] Generally speaking, the consistency verification of multi-source data requires the evaluation of the observation errors between different data sources. The observation data of the data source is , from The observation data of the data source is , then the observation error can be defined as:

[0040] in, Represents the observation difference between two data sources and is used to measure the consistency of multi-source data. As an option, error assessment can be combined with error propagation theory to calculate data consistency through the propagation of the covariance matrix. Specifically, the covariance matrix of the observation error can be expressed as:

[0041] in, and Respectively and The observation noise covariance matrix of the data source describes the observation error characteristics of each data source.

[0042] Specifically, in order to determine whether the data meets the consistency criteria, it is necessary to verify the data based on the confidence interval. Set the confidence interval threshold to , if the observation error The following conditions are met:

[0043] If , the data are considered to be consistent; otherwise, the data are considered to be abnormal and should be removed. Here, is the significance level, which is used to control the error tolerance probability.

[0044] In one possible implementation, data that does not meet the consistency criteria can be directly eliminated to avoid negative impact on subsequent data fusion. At the same time, for data that meets the consistency criteria, its weight needs to be dynamically adjusted according to its observation noise. The weight of the data source can be calculated based on the trace value of the observation noise covariance matrix

[0045] in, For the The weight of the data source, Indicates The trace value of the data source observation noise covariance matrix is ​​used to reflect the reliability of the data source. Generally speaking, the data source with smaller observation noise has a higher weight, thus contributing more to the overall estimation in the subsequent fusion process.

[0046] In some embodiments, in order to improve the robustness of data verification, consistency verification can be performed simultaneously between multiple pairs of data sources. For example, for the surface deformation values ​​from optical remote sensing data and the surface displacement values ​​from SAR data, the consistency errors are calculated based on the temporal and spatial overlap areas of the two. For the dynamic change data of ground sensors (such as displacement sensors), they can also be compared with the high-resolution deformation monitoring results of SAR to further improve the accuracy of consistency verification.

[0047] As an implementation method, the sensor observation noise covariance Ri,kR_{i,k}Ri,k can also be dynamically updated. For example, the real-time noise covariance is calculated using the mean and variance of multiple historical observation residuals, and the weight distribution is dynamically corrected. This dynamic adjustment mechanism can improve the flexibility and real-time performance of consistency verification.

[0048] In certain cases, such as high-risk landslide areas with densely distributed sensors, a local weight adjustment strategy can be introduced. Specifically, for multiple ground sensors in a local area, the weight adjustment can take into account the spatial correlation of geographical locations. For example, when the difference in sensor observations in adjacent areas is small, a higher weight can be assigned, thereby making more effective use of local area information.

[0049] Through the above consistency verification steps, abnormal data can be effectively eliminated, and the weight of the data source can be dynamically adjusted according to its reliability, providing high-quality input for subsequent data fusion. This verification mechanism solves the problems caused by observation errors, inconsistencies or noise interference in multi-source data, thus providing a reliable data foundation for the entire geological disaster investigation and evaluation process.

[0050] S3. Conduct geological disaster risk assessment based on fused data, extract key risk indicators and classify risk levels After completing the consistency verification of multi-source remote sensing data in step S2, the obtained verification data has high reliability, but due to the diversity of data sources and the differences in resolution and sampling frequency of each data source, it is still necessary to uniformly process the multi-source data through data fusion technology to achieve accurate estimation of the dynamic state of geological hazards. Specifically, this step uses distributed Kalman filtering technology to perform data fusion, adopts a distributed processing framework combined with a dynamic weighting strategy, and achieves a unified estimation of multi-source data in time and space, providing dynamic state input for subsequent geological hazard risk assessment.

[0051] In this embodiment, the main steps of data fusion include state prediction, local update and global weighted fusion, and the distributed framework is combined to improve the computing efficiency. The specific implementation method will be described in detail below.

[0052] Generally speaking, distributed Kalman filtering is a common method for processing multi-source dynamic data. Its core is to optimize the estimation of dynamic processes through state prediction and updating. In this embodiment, it is assumed that the state variable of geological disasters is , its dynamic changes can be described by the state equation:

[0053] in, is the state transfer matrix, which describes the evolution of geological state over time; is process noise, satisfying Gaussian distribution is the process noise covariance, which describes the uncertainty of the system state.

[0054] As an option, the data fusion process needs to combine multi-source observation data and describe the relationship between the observation values ​​of each data source and the geological state through the observation equation:

[0055] in, Indicates that from Data sources at time Observed value of is the observation matrix, which describes the mapping relationship between the observation data and the true state; is the observation noise, satisfying the Gaussian distribution For the Noise covariance of the data source.

[0056] In this embodiment, data fusion first performs state prediction and local update at each sensor node. Specifically, at time , for nodes To make a status prediction:

[0057]

[0058] in, For Node The predicted status of is the prediction error covariance matrix; For Node The error covariance at the previous time step.

[0059] After the prediction is completed, the state is updated based on the observed data. The update steps are:

[0060]

[0061]

[0062] in, For Node The Kalman gain is used to dynamically adjust the contribution of observation data to state estimation; For Node Updated state estimate; For Node The updated error covariance of .

[0063] In one possible implementation, in order to improve the global consistency and accuracy of data fusion, it is necessary to perform global weighted fusion on the state estimation results of each node. Global state estimation is achieved through the following formula:

[0064] in, For Node The weight of is defined as:

[0065] here, Representation Node Update the trace value of the error covariance matrix, reflecting the node Uncertainty in state estimation. Generally, nodes with smaller error covariance have higher weights and thus contribute more to the global state estimation.

[0066] In some embodiments, in order to further improve the computing efficiency, a distributed computing framework can be used to assign state prediction, update and weighted fusion tasks to each node for parallel execution. For example, ground sensor data can complete state updates at local nodes, while the processing of optical remote sensing data and SAR data can be performed in parallel through the cloud computing platform. Finally, the local node transmits the updated state estimate and error covariance to the global center to complete the final weighted fusion.

[0067] In a specific implementation, the global state estimation can also be optimized in combination with a time weighting strategy. For example, a time weighting factor can be introduced into the current state estimation in combination with the dynamic change trend of historical data to further improve the stability of the estimation. The time weighting factor can be dynamically calculated based on the error change rate at the most recent multiple moments.

[0068] Through the above steps, data fusion realizes the unified processing of multi-source data in the time and space dimensions, can accurately estimate the dynamic state of geological hazards, and provide accurate input for subsequent risk assessment. This method not only improves the accuracy of data fusion, but also has good real-time performance and computational efficiency, and is suitable for a variety of geological disaster monitoring scenarios.

[0069] S4. Output geological disaster risk assessment results and generate risk distribution maps and early warnings for the monitored areas After completing the fusion of multi-source remote sensing data through distributed Kalman filtering in step S3, a unified and dynamic geological disaster status estimation result is obtained. Based on these fusion results, it is necessary to further quantitatively evaluate the risk level of the geological disaster area. Specifically, this step extracts key risk indicators, constructs a risk assessment model, and divides the monitoring area into risk levels to form a comprehensive geological disaster risk assessment result, providing an important basis for early warning and emergency response.

[0070] In this embodiment, the main contents of geological hazard risk assessment include the extraction of risk indicators, the classification of risk levels and the implementation of risk assessment models, which will be described in detail below in combination with formulas and specific implementation methods.

[0071] In general, geological disaster risk assessment needs to be based on the geological disaster state variables in the fused data to extract key risk indicators that can characterize the potential threat level of the disaster. The selection of key risk indicators is usually determined based on the type of disaster and the geological environment characteristics of the monitoring area. Specifically, for common geological disaster types such as landslides and mudslides, commonly used risk indicators include but are not limited to: surface deformation rate, rainfall threshold and geodynamic parameters.

[0072] In this embodiment, the surface deformation rate can be calculated by using the differential interferometry (DInSAR) technology of SAR data. Specifically, the surface displacement after fusion is assumed to be , then the surface deformation rate It can be expressed as:

[0073] in, is the time interval between two observation times.

[0074] As an option, the rainfall threshold can be estimated by combining the data from the rainfall sensor and the historical rainfall records of the area. For example, suppose the cumulative rainfall in the monitoring area is , and combined with the regional empirical rainfall threshold ,when When the rainfall conditions in the area reach the threshold level for triggering geological disasters, it can be judged that the rainfall conditions in the area have reached the threshold level for triggering geological disasters.

[0075] Specifically, the calculation of geodynamic parameters can be based on the stress-strain state in the fused data. Assuming the shear stress of the geological body is The shear strength is , then the safety factor It can be expressed as:

[0076] Among them, when When , it means that there is a risk of landslide instability in the area.

[0077] In a possible implementation, a geological disaster risk assessment model can be constructed based on the above key risk indicators. The assessment model usually uses a multi-factor superposition method to comprehensively map multiple risk indicators into the risk level of the region. The basic form of the risk assessment model is:

[0078] in, Indicates the risk level; is the standardized value of each risk indicator; is the weight of each indicator, reflecting its contribution to the overall risk.

[0079] In some embodiments, the weights may be determined using the analytic hierarchy process (AHP) or an expert weighting method. For example, in landslide monitoring, the surface deformation rate may have a higher weight, while the rainfall threshold may have a higher weight in debris flow monitoring.

[0080] As an option, the risk level division can be based on the output of the risk assessment model, and the monitoring area can be divided into low-risk, medium-risk and high-risk areas by setting risk level thresholds. satisfy , it is judged as a low-risk area; if , it is determined to be a medium-risk area; if , it is judged as a high-risk area.

[0081] Specifically, the generated risk level results can be spatially mapped through the GIS system to form a risk distribution map of the monitored area. Each regional unit of the risk distribution map will be marked with the corresponding risk level, and different risk levels will be represented by color gradients. For example, low-risk areas can be represented by green, medium-risk areas by yellow, and high-risk areas by red. This visualization result helps to intuitively display the regional risk distribution and facilitate decision makers to quickly understand the spatial pattern of disaster threats.

[0082] In certain cases, a dynamic risk assessment mechanism can also be introduced. Specifically, for areas where monitoring data are frequently updated (such as high-risk areas for landslides during the rainy season), the risk index value can be dynamically updated in combination with time series analysis methods. For example, when continuous monitoring shows that the rate of surface deformation has significantly accelerated, the risk level of the area can be dynamically increased.

[0083] Through the above steps, this embodiment realizes the geological disaster risk assessment based on fused data, forming a complete process from risk indicator extraction to risk level classification and result visualization. All risk assessment processes are based on the dynamic estimation results of the aforementioned multi-source data fusion, ensuring the scientificity and accuracy of risk assessment. The risk assessment results provide important decision-making support for geological disaster warning and emergency response.

[0084] Please refer to the attached Figure 2 The embodiment of the present invention also provides a rapid geological disaster investigation and evaluation system based on multi-source remote sensing data, which is characterized by comprising: Data acquisition module, used to acquire multi-source remote sensing data, including optical remote sensing data, SAR data and ground sensor data; Data consistency verification module, used to verify the consistency of multi-source remote sensing data, eliminate abnormal data and dynamically adjust data weights; Data fusion module, used to dynamically fuse multi-source data through distributed Kalman filtering and variational inference methods; Risk assessment module, used to extract key risk indicators of geological hazards and classify risk levels; The result output module is used to generate risk distribution maps of geological hazards, dynamic warning information and detailed regional assessment reports.

[0085] The embodiment of the present invention provides a rapid investigation and assessment system for geological disasters, which can realize the complete process from the collection and verification of multi-source remote sensing data to risk assessment and result output. This system is implemented based on the aforementioned rapid investigation and assessment method for geological disasters, and constructs a functional modular system architecture, aiming to improve the efficiency and accuracy of geological disaster monitoring, especially in multi-source data fusion and dynamic risk assessment. Specifically, this system includes a data acquisition module, a data consistency verification module, a data fusion module, a risk assessment module and a result output module. The logic between each module is clear and has a clear connection relationship.

[0086] In this embodiment, the functions and specific implementations of each module of the system are as follows: Generally speaking, the data acquisition module is the front-end part of the entire system, which is used to acquire and pre-process multi-source remote sensing data. Optical remote sensing data comes from high-resolution satellites (such as Sentinel and Landsat), which are mainly used to extract surface morphology and geological characteristics; SAR data is obtained through synthetic aperture radar (such as Sentinel-1 or Radarsat), which is particularly suitable for surface deformation monitoring; ground sensor data comes from displacement sensors, tilt sensors and rainfall monitors deployed on site, which are used to capture dynamic geological changes. In one possible implementation, the data acquisition module can also combine external data sources (such as meteorological data) to assist in the analysis of disaster triggering factors.

[0087] Specifically, the data acquisition module also needs to implement data standardization operations. For example, for remote sensing image data with different resolutions, the spatial resolution can be unified through resampling technology; for sensor data with different sampling frequencies, the temporal resolution can be unified by interpolation methods to ensure the temporal and spatial consistency of subsequent analysis.

[0088] The function of the data consistency verification module is to remove outliers from multi-source data and dynamically adjust data weights to ensure the reliability of data input. Specifically, based on the error propagation theory, this module calculates the observation errors of each data source through the covariance matrix and verifies the consistency of the data according to the confidence interval. For example, for the difference in deformation values ​​between optical remote sensing data and SAR data, the module calculates the consistency error based on its noise covariance matrix and removes data that exceeds the threshold.

[0089] As an option, the data consistency verification module can also dynamically adjust the weight of the data source according to the sensor noise level. Calculated by the following formula:

[0090] in, It is the trace value of the observed noise covariance matrix, reflecting the uncertainty of the data source. Generally speaking, data sources with less noise will receive higher weights, thus contributing more in subsequent data fusion.

[0091] The data fusion module uses distributed Valman filtering combined with variational inference methods to achieve dynamic fusion of multi-source data. This module first completes state prediction and update at each sensor node, and then completes global fusion through weighted average method. The local state estimation and error covariance calculation formula of the node are as follows:

[0092]

[0093]

[0094] The global state estimation is based on the weighted average of the states of each node:

[0095] This module achieves parallel processing through distributed computing, which greatly improves the fusion efficiency. As a possible implementation method, the data fusion module also introduces the variational inference method to optimize the fusion results to minimize the uncertainty of state estimation. Specifically, by optimizing the variational free energy , variational distribution is dynamically adjusted to the true posterior distribution .

[0096] The risk assessment module is used to perform quantitative risk assessment on the fusion results. Specifically, this module calculates the risk level of the monitoring area by extracting key risk indicators from the fusion data, such as surface deformation rate (cross-rainfall threshold) and geodynamic parameters. The risk assessment model uses the weighted superposition method to comprehensively calculate the standardized values ​​of different indicators according to the weights. The expression of the model is as follows:

[0097] in, is the risk level; is the standardized value of each risk indicator; is the weight, which reflects the contribution of the indicator to the overall risk.

[0098] Specifically, the risk assessment module maps the risk level results to the GIS platform to generate a risk distribution map. The risk level of each regional unit is visualized through a color gradient, thereby intuitively displaying the disaster risk level of the monitored area.

[0099] The result output module is the terminal part of the system, which is used to generate and output risk assessment reports and warning information. Specifically, the output content of this module includes: geological disaster risk distribution map, dynamic warning information of the monitoring area, and detailed assessment reports for key areas. As an option, the result output module also supports data interaction with the external emergency command platform, and pushes the warning results in real time through the API interface.

[0100] This system has formed a complete set of rapid investigation and evaluation processes for geological disasters through modular design. The logical connection between each module is clear, which can not only realize the automated processing from data collection to result output, but also improve the overall efficiency through the distributed computing framework. This system is particularly suitable for dynamic monitoring and risk assessment scenarios of complex geological disasters such as landslides and mudslides, providing scientific support for emergency decision-making and disaster management.

[0101] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A rapid investigation and assessment method for geological hazards based on multi-source remote sensing data, characterized in that: The following steps are involved: Acquire multi-source remote sensing data, verify the consistency of multi-source remote sensing data, remove abnormal data and dynamically adjust the weight of each data source; After fusion of verified data, the dynamic state of geological hazards is estimated through data fusion technology; Conduct geological disaster risk assessment based on fused data, extract key risk indicators and classify risk levels; Output geological disaster risk assessment results and generate risk distribution maps and warnings for the monitored areas.

2. The method for rapid investigation and assessment of geological hazards based on multi-source remote sensing data according to claim 1 is characterized in that: The multi-source remote sensing data includes optical remote sensing data, synthetic aperture radar data and ground sensor data, and the ground sensor data includes data from a displacement sensor, an inclination sensor and a rainfall sensor.

3. The method for rapid investigation and assessment of geological hazards based on multi-source remote sensing data according to claim 1 is characterized in that: The consistency verification comprises the following steps: Calculate observation errors between different data sources; Verify data consistency based on the error covariance matrix; Eliminate abnormal data whose error exceeds the confidence interval; The weight of each data source is dynamically adjusted, and the weight is inversely proportional to the noise covariance of the data.

4. The method for rapid investigation and assessment of geological hazards based on multi-source remote sensing data according to claim 1 is characterized in that: The data fusion technology is implemented by distributed Kalman filtering, including: Predict the status of each data source node and update the error covariance; Use observation data to update node states and pass local estimation results to the global system; The weights are dynamically adjusted according to the error covariance matrix of each node, and global state estimation is achieved through weighted averaging.

5. The method for rapid investigation and assessment of geological hazards based on multi-source remote sensing data according to claim 4 is characterized in that: The global state estimation in the distributed Kalman filter is achieved by the following method: The state estimation result of each node is weighted according to the trace value of its error covariance; The weight is inversely proportional to the trace value of the error covariance matrix, and the fusion result is the weighted average of all nodes.

6. The method for rapid investigation and assessment of geological hazards based on multi-source remote sensing data according to claim 1 is characterized in that: The geological disaster risk assessment includes: Extract key risk indicators, including surface deformation rates, rainfall thresholds, and geodynamic parameters; Divide the risk level of the monitoring area based on the risk assessment model; Generate a risk distribution map to show the risk distribution of the monitored area.

7. The method for rapid investigation and assessment of geological hazards based on multi-source remote sensing data according to claim 1 is characterized in that: The output risk assessment results include risk level distribution maps, dynamic warning information and detailed disaster assessment reports.

8. The method for rapid investigation and assessment of geological hazards based on multi-source remote sensing data according to claim 1 is characterized in that: The dynamic adjustment of weights in the consistency verification is achieved by real-time analysis and optimal allocation of the data noise covariance matrix, thereby achieving dynamic adjustment of data source weights.

9. The method for rapid investigation and assessment of geological hazards based on multi-source remote sensing data according to claim 1, characterized in that: The multi-source remote sensing data fusion technology is further combined with a variational inference method to optimize data uncertainty, minimize data uncertainty and dynamically update data source weights.

10. A rapid geological disaster investigation and evaluation system based on multi-source remote sensing data for implementing the method described in any one of claims 1 to 9, characterized in that: include: Data acquisition module, used to acquire multi-source remote sensing data, including optical remote sensing data, SAR data and ground sensor data; Data consistency verification module, used to verify the consistency of multi-source remote sensing data, eliminate abnormal data and dynamically adjust data weights; Data fusion module, used to dynamically fuse multi-source data through distributed Kalman filtering and variational inference methods; Risk assessment module, used to extract key risk indicators of geological hazards and classify risk levels; The result output module is used to generate risk distribution maps of geological hazards, dynamic warning information and detailed regional assessment reports.

Images