Landslide monitoring and early warning method and system based on integration of air, space and earthquake

Abstract

The invention provides a landslide monitoring and early warning method based on air-sky-earthquake integration, and belongs to the technical field of landslide disaster monitoring. Air-sky-ground-earthquake monitoring data are integrated, multi-dimensional parameters such as surface deformation, topographic change, internal strain, crack propagation, rainfall intensity and micro-earthquake activity are covered, the limitation of a single monitoring means is solved, and a space-time complementary three-dimensional monitoring system is formed. According to a data analysis processing result, an early warning response rule is set, an early warning model is constructed through a multi-source data fusion technology in combination with an optimal weighted random forest method and a comprehensive scoring model, real-time evaluation and early warning grade judgment of landslide risks are realized, and visual display of monitoring data is realized. The model is iteratively trained by introducing reward and punishment factors, and weight distribution of different data sources is dynamically adjusted, so that the early warning model has adaptive ability, early warning requirements of long-term slow deformation and short-term accelerated deformation are considered, and the early warning accuracy of the model is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of landslide disaster monitoring, and in particular relates to a landslide monitoring and early warning method and system based on air-space earthquake integration. Background Art

[0002] Landslides, a sudden and destructive geological disaster, pose a serious threat to the safety of life and property, as well as infrastructure development. Geological and hydrological conditions, seismic activity, and human activities all contribute to the stability of slopes. Therefore, comprehensive consideration of multiple factors is crucial to ensure the accuracy and effectiveness of early warning systems.

[0003] Traditional landslide monitoring methods often rely on a single monitoring method, making it difficult to fully and accurately reflect the deformation characteristics and evolution patterns of landslide bodies. The accuracy and timeliness of early warnings need to be improved. The satellite navigation system used in space-based monitoring is applicable to various geographical locations, but the signal can be affected by the external environment. The remote sensing technology used in airborne monitoring is highly accurate, but large-scale, long-term monitoring results in high costs and technical complexity. The sensors used in ground-based monitoring can achieve continuous monitoring over long distances of several to tens of kilometers, but their measurement accuracy may be affected by environmental factors. The crack meters used are highly accurate and easy to install and maintain, but they can only monitor changes in crack width. Rain gauges are highly accurate and can monitor in real time, but their accuracy is affected in extreme weather conditions and requires regular maintenance. Seismometers used in microseismic monitoring can provide comprehensive microseismic information, but the data volume is very large and susceptible to interference from environmental noise.

[0004] Therefore, landslide monitoring and early warning still need to be improved. The present invention aims to provide a landslide monitoring and early warning method and system based on the integration of air-space seismic. Summary of the Invention

[0005] The purpose of the present invention is to overcome the above-mentioned shortcomings and provide a landslide monitoring and early warning method and system based on air-space seismic integration.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a landslide monitoring and early warning method based on air-space seismic integration, comprising the following steps: Collect space-based, air-based, ground-based and microseismic landslide monitoring data; Analyze and process the collected landslide monitoring data from space, air, ground and microseismic sources to determine landslide risk areas; Integrating the landslide monitoring data, performing trend analysis and pattern recognition on the integrated multi-source monitoring data, and obtaining a dynamic change law of the landslide body; Based on the dynamic change law of the landslide body, the optimal weighted random forest method is used to select features of multi-source monitoring data. A monitoring and early warning model is constructed based on the selected features. The reward and punishment factors are set to iteratively train the monitoring and early warning model to obtain the trained monitoring and early warning model and the optimal weight. The landslide monitoring data is input into the trained monitoring and early warning model to obtain the predicted monitoring data. A comprehensive scoring model is constructed based on the predicted monitoring data and the optimal weight to obtain the comprehensive scoring result of the landslide risk.

[0007] In the step of collecting space-based, air-based, ground-based and microseismic landslide monitoring data, the space-based data is the terrain data collected by the GNSS monitoring station, the air-based data is the three-dimensional point cloud data of the landslide area, the ground-based data is the strain distribution data of the landslide area, and the microseismic data is the earthquake data of the landslide area.

[0008] In the step of integrating the landslide monitoring data, performing trend analysis and pattern recognition on the integrated multi-source monitoring data, and obtaining the dynamic change law of the landslide body, the specific method is as follows: Integrating the aerospace-earthquake monitoring data into a unified database, and preprocessing the integrated multi-source monitoring data; Based on the pre-processed multi-source monitoring data, the spatial distribution is displayed by overlaying it on the map through GIS visualization tools; Using Matplotlib tools, we conduct trend analysis and pattern recognition on the preprocessed multi-source monitoring data through time series analysis to obtain the dynamic change law of the landslide body.

[0009] In the steps of selecting features of multi-source monitoring data based on the dynamic change law of the landslide body by the optimal weighted random forest method, constructing a monitoring and early warning model based on the selected features, setting reward and punishment factors to iteratively train the monitoring and early warning model, and obtaining a trained monitoring and early warning model and optimal weights, the method for constructing the monitoring and early warning model is as follows: Use air-space-earth-seismic monitoring data as independent basis learners to construct prediction vectors; The optimal weighted random forest method is used to dynamically allocate the weights of each base learner to obtain the feature weights of each learning base; The feature weights of each learning basis are sorted, and the top 80% of the weights are selected to construct a monitoring and early warning model. Real-time landslide displacement prediction and early warning are performed based on the constructed monitoring and early warning model.

[0010] The optimal weighted random forest method is used to dynamically allocate the weights of each base learner to obtain the feature weights of each learning base. The specific formula is as follows:

[0011] in, represents the weight distribution of each base learner, represents the true landslide displacement, represents the weight of the m-th base learner, represents the predicted value of the mth base learner, Represents the square of the average prediction error, which is used to penalize the complexity of the model. is the diagonal element of the mth base learner, reflecting the importance of the feature.

[0012] The monitoring and early warning model constructed is as follows:

[0013] in, Indicates the features selected for the constructed monitoring and early warning model.

[0014] The method of setting the reward and punishment factors to iteratively train the monitoring and early warning model to obtain the trained monitoring and early warning model and the optimal weight is as follows: After each round of warning, the weights are updated using new data:

[0015] in is the weight vector of the tth round, α is the forgetting factor, is the optimal weight calculated based on the new data; Increase the weight of the monitoring data that has performed accurately in the recent period to obtain the optimal weight of the model: Among them, β is the learning rate, is the historical accuracy of the i-th base learner.

[0016] In the step of inputting the landslide monitoring data into the trained monitoring and early warning model to obtain predicted monitoring data, constructing a comprehensive scoring model based on the predicted monitoring data and the optimal weights, and obtaining a comprehensive scoring result of the landslide risk, the specific method is as follows: Perform minimum-maximum normalization on the monitoring data after each prediction:

[0017] in, is the original predicted monitoring data, is the standardized monitoring prediction data, ranging from [0, 1]. For the The maximum value of the monitoring prediction data, For the The minimum value of the monitoring and prediction data; Based on standardized monitoring prediction data and optimal weights , build a comprehensive scoring model.

[0018] The constructed comprehensive scoring model is as follows:

[0019] in, It represents the comprehensive score, which reflects the comprehensive index of landslide risk; Indicates the The optimal weight of each monitoring data.

[0020] In a second aspect, the present invention provides a landslide monitoring and early warning system based on air-space seismic integration, comprising: The acquisition module is used to collect landslide monitoring data from space, air, ground and microseismic sources; The data processing module is used to analyze and process the collected landslide monitoring data from space, air, ground and microseismic sources to determine the landslide risk areas; An analysis and identification module is used to integrate the landslide monitoring data, perform trend analysis and pattern recognition on the integrated multi-source monitoring data, and obtain the dynamic change law of the landslide body; The model building module is used to select features from multi-source monitoring data using the optimal weighted random forest method based on the dynamic change law of the landslide body, build a monitoring and early warning model based on the selected features, set reward and punishment factors to iteratively train the monitoring and early warning model, and obtain a trained monitoring and early warning model and optimal weights; The comprehensive scoring module is used to input the landslide monitoring data into the trained monitoring and early warning model to obtain the predicted monitoring data, and to construct a comprehensive scoring model based on the predicted monitoring data and the optimal weight to obtain the comprehensive scoring result of the landslide risk.

[0021] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides a landslide monitoring and early warning method based on the integration of space, air, ground, and microseismic systems, comprising the following steps: collecting space-based, airborne, ground-based, and microseismic landslide monitoring data; analyzing and processing the collected space-based, airborne, ground-based, and microseismic landslide monitoring data to determine landslide risk areas; integrating the landslide monitoring data, performing trend analysis and pattern recognition on the integrated multi-source monitoring data, and obtaining dynamic change patterns of the landslide body; based on the dynamic change patterns of the landslide body, performing feature selection on the multi-source monitoring data using the optimal weighted random forest method, constructing a monitoring and early warning model based on the selected features, and iteratively training the monitoring and early warning model with reward and penalty factors to obtain a trained monitoring and early warning model and optimal weights; inputting the landslide monitoring data into the trained monitoring and early warning model to obtain predicted monitoring data, and constructing a comprehensive scoring model based on the predicted monitoring data and the optimal weights to obtain a comprehensive landslide risk scoring result. The integration of space, air, ground, and seismic monitoring data covers multi-dimensional parameters such as surface deformation, topographic change, internal strain, crack expansion, rainfall intensity, and microseismic activity, overcoming the limitations of a single monitoring method and forming a three-dimensional monitoring system that complements both time and space. Early warning response rules are set based on the results of data analysis and processing. A warning model is constructed using multi-source data fusion technology, combined with the optimal weighted random forest method and a comprehensive scoring model. This allows for real-time assessment of landslide risk and determination of warning levels, as well as visualization of monitoring data. By iteratively training the model through the introduction of reward and penalty factors and dynamically adjusting the weights assigned to different data sources, the early warning model becomes adaptive, addressing both long-term slow deformation and short-term accelerated deformation warning requirements, thereby improving the accuracy of model warnings.

[0022] Furthermore, a distributed fiber optic sensor network, crack meters, and rain gauges are deployed in the monitoring area, and optical fibers are used as sensors to monitor the strain distribution of the landslide body, and to monitor in real time the changes in parameters such as cracks, rock micro-fractures, and rainfall in the landslide area. By combining fiber optic sensors, crack and rainfall data, the potential risks and development trends of landslides can be further identified.

[0023] Furthermore, trend analysis and pattern recognition techniques are used to extract the dynamic evolution laws of key indicators such as landslide displacement, deformation, and microseismic activity from massive heterogeneous data. Combined with the optimal weighted random forest method, high-contribution features (such as deformation rate and crack expansion trend) are screened to achieve quantitative classification and positioning of risk areas, providing a scientific basis for disaster prevention decisions.

[0024] Furthermore, a comprehensive scoring model constructed based on predicted data and optimal weights converts landslide risks into intuitive numerical indicators, supports multi-level warning threshold settings, is compatible with regional geological condition differences, and facilitates management departments to formulate differentiated emergency plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1Flow chart of the method of the present invention; Figure 2 This is a schematic diagram of the installation of the data collector in the present invention; Figure 3 It is the overall flow chart of the early warning method of the present invention. DETAILED DESCRIPTION

[0026] In order to further understand the content of the present invention, the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the embodiments are only for explaining the present invention and are not intended to limit it.

[0027] like Figure 1 As shown, a landslide monitoring and early warning method based on air-space seismic integration includes the following steps: S1: Collect landslide monitoring data from space, air, ground and microseismic sources; S2: Analyze and process the collected landslide monitoring data from space, air, ground and microseismic sources to determine landslide risk areas; S3: integrating the landslide monitoring data, performing trend analysis and pattern recognition on the integrated multi-source monitoring data, and obtaining the dynamic change law of the landslide body; S4: Based on the dynamic change law of the landslide body, the optimal weighted random forest method is used to select features of multi-source monitoring data. A monitoring and early warning model is constructed based on the selected features. The reward and punishment factors are set to iteratively train the monitoring and early warning model to obtain the trained monitoring and early warning model and the optimal weight. S5: Input the landslide monitoring data into the trained monitoring and early warning model to obtain the predicted monitoring data. Based on the predicted monitoring data and the optimal weight, a comprehensive scoring model is constructed to obtain the comprehensive scoring result of the landslide risk.

[0028] Specifically, in S1, the space-based monitoring data is GNSS (Global Navigation Satellite System) monitoring data; the air-based monitoring data is LiDAR (Light Detection and Ranging) monitoring data; the ground-based monitoring data is distributed optical fiber, crack meter and rain gauge monitoring data; the microseismic monitoring data is seismograph monitoring data; Specifically, in S2, the landslide monitoring data from space, air, ground and microseismic sources are analyzed and processed. The specific methods are as follows: GNSS monitoring stations are deployed in and around the landslide body to obtain real-time three-dimensional surface deformation information and locate large-scale landslide risk areas; LiDAR scanning is carried out regularly to obtain high-precision terrain data and locate medium-scale landslide risk areas; a distributed fiber optic sensing network is deployed inside the landslide body to monitor internal parameters such as strain and temperature in real time, crack meters are placed at potential cracks in the landslide to monitor changes in cracks, and the hydrological conditions in the landslide area are assessed by collecting and recording rainfall; seismometers are deployed in the landslide area to monitor and record small earthquake activities.

[0029] Furthermore, the method for analyzing and processing space-based data is as follows: Monitoring points and benchmark points are set up in the monitoring area, and observations are carried out at different time points. At each observation time point, the geodetic height difference between each monitoring point and the benchmark point is recorded. The coordinate change calculation formula of the monitoring point is:

[0030] Where, are the initial coordinates, is the coordinate after the change, is the coordinate change; The slope displacement is calculated by the coordinate change, and the calculation formula is:

[0031] Where, is the slope displacement, that is, the ground height difference.

[0032] like Figure 2 As shown, GNSS is used to monitor surface displacement and deformation. First, a GNSS base station is set up and a GNSS receiver is installed at the selected monitoring point to receive satellite signals. Solar panels are installed to power the equipment, and the piping system is connected to the GNSS receiver's bracket to ensure that all equipment is functioning properly. A SIM card is inserted into the device to enable data transmission over the network. The GPRS antenna, GNSS antenna, and input voltage cables are properly connected. Professionals then set monitoring parameters to achieve real-time monitoring of surface displacement and deformation. Finally, early warning information is disseminated to the relevant area through various media, such as television and the internet.

[0033] Furthermore, the method for analyzing and processing the air-based data is as follows: LiDAR equipment is deployed in the monitoring area, and several monitoring points are selected for data collection. The data is preprocessed, including filtering, denoising, and data calibration. By obtaining high-precision three-dimensional point cloud data of the landslide area, a digital elevation model (DEM) is generated to achieve digital simulation of the ground terrain. Deformation monitoring and surface feature extraction and calculation are carried out based on the digital elevation model.

[0034] The terrain change is calculated using multi-period DEM data, and the calculation formula is:

[0035] Where, is the first phase DEM elevation value, is the second phase DEM elevation value, is the amount of terrain change; A drone-mounted LIDAR system is used to identify areas at risk of large-scale landslides, set flight paths, and monitor durations. The LIDAR device is operated along the predetermined path, simultaneously emitting laser pulses and receiving reflected signals to calculate distances. Lasers can be fired from different angles to obtain three-dimensional surface information. Processing software is then used to generate point cloud data, extract key features such as cracks and displacements, analyze deformation trends, and visualize the results. Finally, areas where spatial deformation exceeds a preset threshold are selected for further monitoring.

[0036] Furthermore, the method for analyzing and processing the foundation data is as follows: Distributed fiber optic sensors, crack meters and rain gauges are deployed in the monitoring area. Fiber optics are used as sensors to monitor the strain distribution of the landslide body, and changes in parameters such as cracks, rock micro-fractures and rainfall in the landslide area are monitored in real time. By combining fiber optic sensors, crack and rainfall data, potential risks and development trends of landslides can be further identified.

[0037] Monitor the strain distribution of the landslide body, and the calculation formula is:

[0038] Where, For location The fiber length changes at is the initial length of the optical fiber, For location strain at Monitor the changes in the crack expansion rate of the landslide body. The calculation formula is:

[0039] Where, is the difference between the current crack and the initial crack, is the time interval, is the crack expansion rate.

[0040] Monitor the rainfall intensity of the landslide body, and the calculation formula is:

[0041] Where, is the rainfall, is the time interval, The rainfall intensity.

[0042] Distributed fiber optic sensors are used to determine foundation strain, temperature changes, and other factors. First, plan the fiber optic routing to ensure coverage of areas at risk of medium-scale landslides. Install the fiber optic sensors, ensuring a good connection between the sensors and the fiber optic cable. Then, connect the fiber optic cable to a fiber optic interrogator to read the data.

[0043] A crack meter is used to monitor the width, length, and depth of foundation cracks. First, locate the crack and install the crack meter, ensuring it is perpendicular or parallel to the crack. Then, connect the data logger to record the data.

[0044] A rain gauge is used to measure rainfall. Install the rain gauge in an open area, away from obstructions, at a height of 1.5 to 2 meters, with the water collection port facing upward. Connect the rain gauge to a data collector, set monitoring parameters, and transmit the data via a cable to a monitoring center, where the data is analyzed and rainfall conditions are assessed.

[0045] Furthermore, the method for analyzing and processing microseismic data is as follows: A seismograph is installed in the monitoring area to monitor parameters such as frequency, amplitude and energy of microseismic signals, and the deformation trend of the landslide body is determined by the microseismic signals.

[0046] The energy of the microseismic signal of the monitored landslide body is calculated as follows:

[0047] Where, , is the time interval, is the microseismic signal, is the energy of the microseismic signal.

[0048] Seismographs are used to monitor seismic activity in the ground. Installation requires a seismic accelerometer, a data acquisition system, and a power supply. They must be positioned appropriately to capture seismic waves and analyze and process data such as their amplitude, frequency, and duration.

[0049] Specifically, in S3, the landslide monitoring data is integrated, and trend analysis and pattern recognition are performed on the integrated multi-source monitoring data to obtain the dynamic change law of the landslide body. The specific method is as follows: The aerospace-earthquake monitoring data are integrated into a unified database, and the integrated multi-source monitoring data are preprocessed, including data calibration, filtering and denoising operations.

[0050] GIS visualization tools were used to overlay the preprocessed monitoring data, such as the location of monitoring points, crack development, displacement, and other data, on the map to display the spatial distribution and obtain visualization results. Matplotlib tools were used to perform trend analysis and pattern recognition on the preprocessed multi-source monitoring data through time series analysis to further reveal the dynamic change patterns of the landslide body.

[0051] Specifically, in S4, based on the dynamic change law of the landslide body, the optimal weighted random forest method is used to select features of multi-source monitoring data. A monitoring and early warning model is constructed based on the selected features. Reward and punishment factors are set to train the constructed monitoring and early warning model. Rewards are given to samples that are correctly classified or regressed, and penalties are imposed on samples that are misclassified or have large errors, thereby dynamically adjusting the training weight of the model. The specific method is as follows: 1) Using air-space-earth-seismic monitoring data (GNSS displacement, LiDAR terrain change, fiber strain, etc.) as independent basis learners to construct prediction vectors: GNSS (GNSS displacement), LiDAR (GNSS displacement), Fiber (GNSS displacement), Crack (GNSS displacement), Rain (GNSS displacement), Micro (GNSS displacement).

[0052] Each base learner generates a prediction value through an independent data source and model to form a multi-source prediction vector.

[0053] 2) Use the optimal weighted random forest method to dynamically assign weights to each base learner, obtain the feature weights of each learning base, and minimize the prediction error and complexity of the monitoring and early warning model:

[0054] in, represents the weight distribution of each base learner, represents the true landslide displacement, represents the weight of the m-th base learner, represents the predicted value of the mth base learner, Represents the square of the average prediction error, which is used to penalize the complexity of the model. is the diagonal element of the mth base learner, reflecting the importance of the feature.

[0055] 3) Sort the feature weights of each learning basis obtained, select the top 80% of the weights to build a monitoring and early warning model, and eliminate low-contribution basis learners (for example, the weight of microseismic data in non-earthquake periods approaches 0) through weight sparsity: Real-time landslide displacement prediction and early warning are carried out based on the constructed monitoring and early warning model. The constructed monitoring and early warning model is as follows:

[0056] in, Indicates the features selected for the constructed monitoring and early warning model.

[0057] 4) After each round of warning, the weights are updated using new data to achieve dynamic adaptability of the model:

[0058] in is the weight vector of the tth round, α is the forgetting factor (such as 0.95), is the optimal weight calculated based on the new data.

[0059] 5) Set reward and punishment factors to dynamically adjust the training weights of the monitoring and early warning model as follows: Increase the weight of the monitoring data that has performed accurately in the recent period to obtain the optimal weight of the model:

[0060] Among them, β is the learning rate, that is, the reward and punishment factor (such as 0.1), is the historical accuracy of the i-th base learner.

[0061] 6) Calculate the inaccuracy of the comprehensive prediction of the monitoring and early warning model and evaluate the reliability of the model. The specific formula is as follows:

[0062] Among them, M is the number of base learners, is the square of the prediction error, reflecting the uncertainty of the model.

[0063] Furthermore, under certain assumptions (such as a sufficiently high signal-to-noise ratio), the prediction risk of the optimal weighted random forest method satisfies the following convergence properties:

[0064] in, Indicates that the early warning model uses the optimal weight The predicted risk when Represents all possible weights The minimum predicted risk in , that is, the risk of the theoretically optimal model.

[0065] Specifically, in S5, the landslide monitoring data is input into the trained monitoring and early warning model to obtain the predicted monitoring data. The comprehensive scoring result of the landslide is obtained based on the constructed comprehensive scoring model. According to the comprehensive scoring result, the landslide risk is divided into low risk, medium risk and high risk levels, providing a scientific basis and decision-making support for landslide prevention and control.

[0066] The specific method is as follows: In order to eliminate the influence of different dimensions and ranges of the predicted monitoring data, the minimum-maximum standardization is performed on the monitoring data after each prediction:

[0067] in, is the original predicted monitoring data, is the standardized monitoring prediction data, ranging from [0, 1]. For the The maximum value of the monitoring prediction data, For the The minimum value of the monitoring prediction data.

[0068] Based on standardized monitoring prediction data and optimal weights , build a comprehensive scoring model:

[0069] in, It represents the comprehensive score, which reflects the comprehensive index of landslide risk; Indicates the The optimal weight of each monitoring data.

[0070] As a preferred embodiment of this invention, the risk level determination table 1 is as follows: Table 1 Risk level determination table

[0071] Trend analysis and pattern recognition of the integrated monitoring data can effectively enhance the model's sensitivity to key features, reduce misjudgments caused by noise or abnormal data, and further improve the model's ability to identify landslide risks. Then, by constructing multiple decision trees and combining their prediction results, classification or regression analysis of landslide risk can be achieved. In conjunction with a comprehensive scoring model, different monitoring data are subjected to minimum-maximum normalization and assigned weights. A weighted summation method is used to derive a comprehensive score. In this process, a reward and penalty factor mechanism further ensures the rationality of the weight distribution, giving high-risk features a more prominent position in the comprehensive score, thereby achieving quantitative assessment of landslide risk and accurate early warning.

[0072] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.

Claims

1. A landslide monitoring and early warning method based on air-space earthquake integration, characterized in that: The steps include: Collect space-based, air-based, ground-based and microseismic landslide monitoring data; Analyze and process the collected landslide monitoring data from space, air, ground and microseismic sources to determine landslide risk areas; Integrating the landslide monitoring data, performing trend analysis and pattern recognition on the integrated multi-source monitoring data, and obtaining a dynamic change law of the landslide body; Based on the dynamic change law of the landslide body, the optimal weighted random forest method is used to select features of multi-source monitoring data. A monitoring and early warning model is constructed based on the selected features. The reward and punishment factors are set to iteratively train the monitoring and early warning model to obtain the trained monitoring and early warning model and the optimal weight. The landslide monitoring data is input into the trained monitoring and early warning model to obtain the predicted monitoring data. A comprehensive scoring model is constructed based on the predicted monitoring data and the optimal weight to obtain the comprehensive scoring result of the landslide risk.

2. The landslide monitoring and early warning method based on air-space earthquake integration according to claim 1 is characterized in that: In the step of collecting space-based, air-based, ground-based and microseismic landslide monitoring data, the space-based data is the terrain data collected by the GNSS monitoring station, the air-based data is the three-dimensional point cloud data of the landslide area, the ground-based data is the strain distribution data of the landslide area, and the microseismic data is the earthquake data of the landslide area.

3. The landslide monitoring and early warning method based on air-space earthquake integration according to claim 1 is characterized in that: In the step of integrating the landslide monitoring data, performing trend analysis and pattern recognition on the integrated multi-source monitoring data, and obtaining the dynamic change law of the landslide body, the specific method is as follows: Integrating the aerospace-earthquake monitoring data into a unified database, and preprocessing the integrated multi-source monitoring data; Based on the pre-processed multi-source monitoring data, the spatial distribution is displayed by overlaying it on the map through GIS visualization tools; Using Matplotlib tools, we conduct trend analysis and pattern recognition on the preprocessed multi-source monitoring data through time series analysis to obtain the dynamic change law of the landslide body.

4. The landslide monitoring and early warning method based on air-space earthquake integration according to claim 1 is characterized in that: In the steps of selecting features of multi-source monitoring data based on the dynamic change law of the landslide body by the optimal weighted random forest method, constructing a monitoring and early warning model based on the selected features, setting reward and punishment factors to iteratively train the monitoring and early warning model, and obtaining a trained monitoring and early warning model and optimal weights, the method for constructing the monitoring and early warning model is as follows: Use air-space-earth-seismic monitoring data as independent basis learners to construct prediction vectors; The optimal weighted random forest method is used to dynamically allocate the weights of each base learner to obtain the feature weights of each learning base; The feature weights of each learning basis are sorted, and the top 80% of the weights are selected to construct a monitoring and early warning model. Real-time landslide displacement prediction and early warning are performed based on the constructed monitoring and early warning model.

5. The landslide monitoring and early warning method based on air-space earthquake integration according to claim 4 is characterized in that: The optimal weighted random forest method is used to dynamically allocate the weights of each base learner to obtain the feature weights of each learning base. The specific formula is as follows: in, represents the weight distribution of each base learner, represents the true landslide displacement, represents the weight of the m-th base learner, represents the predicted value of the mth base learner, Represents the square of the average prediction error, which is used to penalize the complexity of the model. is the diagonal element of the mth base learner, reflecting the importance of the feature.

6. The landslide monitoring and early warning method based on air-space earthquake integration according to claim 5 is characterized in that: The monitoring and early warning model constructed is as follows: in, Indicates the features selected for the constructed monitoring and early warning model.

7. The landslide monitoring and early warning method based on air-space earthquake integration according to claim 6 is characterized in that: The method of setting the reward and punishment factors to iteratively train the monitoring and early warning model to obtain the trained monitoring and early warning model and the optimal weight is as follows: After each round of warning, the weights are updated using new data: in is the weight vector of the tth round, α is the forgetting factor, is the optimal weight calculated based on the new data; Increase the weight of the monitoring data that has performed accurately in the recent period to obtain the optimal weight of the model: Among them, β is the learning rate, is the historical accuracy of the i-th base learner.

8. The landslide monitoring and early warning method based on air-space earthquake integration according to claim 7 is characterized in that: In the step of inputting the landslide monitoring data into the trained monitoring and early warning model to obtain predicted monitoring data, constructing a comprehensive scoring model based on the predicted monitoring data and the optimal weights, and obtaining a comprehensive scoring result of the landslide risk, the specific method is as follows: Perform minimum-maximum normalization on the monitoring data after each prediction: in, is the original predicted monitoring data, is the standardized monitoring prediction data, ranging from [0, 1]. For the The maximum value of the monitoring prediction data, For the The minimum value of the monitoring and prediction data; Based on standardized monitoring prediction data and optimal weights , build a comprehensive scoring model.

9. The landslide monitoring and early warning method based on air-space earthquake integration according to claim 8 is characterized in that: The constructed comprehensive scoring model is as follows: in, It represents the comprehensive score, which reflects the comprehensive index of landslide risk; Indicates the The optimal weight of each monitoring data.

10. A landslide monitoring and early warning system based on air-space earthquake integration according to any one of claims 1 to 9, characterized in that: include: The acquisition module is used to collect landslide monitoring data from space, air, ground and microseismic sources; The data processing module is used to analyze and process the collected landslide monitoring data from space, air, ground and microseismic sources to determine the landslide risk areas; An analysis and identification module is used to integrate the landslide monitoring data, perform trend analysis and pattern recognition on the integrated multi-source monitoring data, and obtain the dynamic change law of the landslide body; The model building module is used to select features from multi-source monitoring data using the optimal weighted random forest method based on the dynamic change law of the landslide body, build a monitoring and early warning model based on the selected features, set reward and punishment factors to iteratively train the monitoring and early warning model, and obtain a trained monitoring and early warning model and optimal weights; The comprehensive scoring module is used to input the landslide monitoring data into the trained monitoring and early warning model to obtain the predicted monitoring data, and to construct a comprehensive scoring model based on the predicted monitoring data and the optimal weight to obtain the comprehensive scoring result of the landslide risk.

Images