Landslide early warning method and system

Abstract

The invention discloses a landslide early warning method and system, and the method comprises the steps: carrying out the field testing of a sampling point through a multi-parameter soil rapid detection device, obtaining the index data of the water content, density, particle size distribution and the like of the soil, uploading the data to a cloud server in real time through a wireless transmission module, and achieving the comprehensive evaluation of the soil indexes; according to the soil evolution prediction result, combined with a preset landslide risk threshold value, the landslide risk grade is dynamically evaluated, if the predicted landslide possibility exceeds the threshold value, an early warning signal is generated, and the early warning signal is pushed to related departments and personnel in real time through a wireless transmission module; according to different geological conditions and landslide types, a plurality of soil property-environment factor coupling analysis sub-models are pre-established, and according to the geological features of the current sampling point, the optimal sub-model is dynamically selected to perform landslide risk pre-judgment, so that the applicability and accuracy of early warning are improved.

Description

Technical Field

[0001] The invention belongs to the field of information technology, and in particular relates to a landslide early warning method and system. Background Art

[0002] Soil quality assessment is a key link in landslide early warning. The commonly used soil sampling and analysis methods have some limitations: first, the selection of sampling points often relies on experience and lacks scientific guidance, resulting in unrepresentative sampling; second, the test items of soil quality indicators are single and it is difficult to fully reflect the soil quality conditions; third, the correlation analysis between soil quality data and environmental factors is not in-depth enough to reveal the law of soil quality changes. These problems restrict the accuracy and effectiveness of soil quality assessment.

[0003] In view of the above technical bottlenecks, it is urgent to develop an intelligent soil quality assessment method and system for landslide early warning. Summary of the invention

[0004] In view of the shortcomings of the existing technology, the present invention proposes a landslide early warning method and system. First, based on the geological model and historical disaster data, the machine learning algorithm is used to optimize the sampling point layout plan to improve the pertinence of sampling; second, a multi-parameter soil rapid detection device is integrated to achieve a comprehensive assessment of soil indicators; third, a soil-environmental factor coupling analysis model is constructed to dynamically simulate the soil evolution process and predict the landslide risk level. The system should also have the capabilities of automatic data collection, wireless transmission, and real-time processing to provide efficient technical support for landslide early warning.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A landslide early warning method comprises the following steps:

[0007] According to the existing geological model and historical disaster data, the support vector machine algorithm is used to optimize the layout of sampling points. By calculating the spatial similarity between each candidate sampling point and the known landslide point, the position with the highest similarity is selected as the new sampling point.

[0008] Conduct on-site testing of sampling points to obtain soil quality index data, and upload the soil quality index data to the cloud server in real time for comprehensive evaluation;

[0009] According to the comprehensive assessment results, the random forest algorithm is used to select features and sort the importance of soil index data, screen out the key indicators with the most significant impact on landslide risk, and integrate them with environmental factor data to build a soil-environmental factor coupling analysis model;

[0010] In the soil quality-environmental factor coupling analysis model, the time series analysis method is used to model the dynamic change process of soil quality indicators and environmental factors, and the recursive neural network algorithm is used to predict the soil quality evolution trend;

[0011] According to the soil evolution trend prediction results and the preset landslide risk threshold, the landslide risk level is dynamically evaluated. If the predicted landslide possibility exceeds the threshold, an early warning signal is generated.

[0012] Preferably, based on the existing geological model and historical disaster data, the support vector machine algorithm is used to optimize the layout of the sampling points, and the method of selecting the position with the highest similarity as the new sampling point by calculating the spatial similarity between each candidate sampling point and the known landslide point includes:

[0013] Based on the existing geological model and historical disaster data, a landslide-prone area identification model is constructed;

[0014] Based on the landslide-prone area identification model, determine whether the area where the candidate sampling point is located belongs to the landslide-prone area;

[0015] If the candidate sampling point is located in a landslide-prone area, the geological characteristics of the point are obtained and used as input features of the support vector machine model;

[0016] Select some points from known landslide points as training samples, extract their geological features, normalize all features, and build a support vector machine classification model;

[0017] The trained support vector machine classification model is used to calculate the spatial similarity between the candidate sampling points and the known landslide points, and the similarity score of each candidate point is obtained;

[0018] Sort the candidate sampling points according to their similarity scores, and select the top N points with the highest scores as new sampling points;

[0019] The newly selected sampling points are merged with the original sampling points to obtain an optimized sampling point layout plan.

[0020] Preferably, the method of conducting on-site testing on the sampling points, obtaining soil index data, and uploading the index data to the cloud server in real time for comprehensive evaluation includes:

[0021] The soil moisture content, density and particle size composition parameter data collected by the portable soil testing equipment are uploaded to the cloud server in real time through the wireless transmission module;

[0022] In the cloud server, the uploaded soil parameter data is comprehensively analyzed according to the preset soil assessment model to obtain a comprehensive score of soil quality;

[0023] Based on the comprehensive soil score of each sampling point, the soil quality distribution map of the entire area is generated using an interpolation algorithm;

[0024] For areas where the soil quality score is below the preset threshold, a density clustering algorithm is used to delineate soil degradation clusters;

[0025] For the concentrated areas of soil degradation, the decision tree algorithm is used to recommend corresponding soil improvement measures based on the distribution characteristics of soil parameters in the area;

[0026] The soil quality distribution map, soil degradation cluster division results, and soil improvement measures are displayed to users through a visual interface to assist agricultural production management decisions.

[0027] Preferably, based on the comprehensive evaluation results, a random forest algorithm is used to perform feature selection and importance ranking on soil quality index data, and the key indicators with the most significant impact on landslide risk are screened out, and the indicators are integrated with environmental factor data to construct a soil quality-environmental factor coupling analysis model, including:

[0028] Acquire soil quality index data and environmental factor data, and preprocess the soil quality index data and the environmental factor data to obtain preprocessed soil quality index data and preprocessed environmental factor data;

[0029] According to the comprehensive evaluation results, the random forest algorithm is used to perform feature selection on the pre-processed soil index data to obtain the selected key soil index data, wherein the key soil index data is the soil index data with the greatest impact on the landslide risk selected according to a preset importance threshold by calculating the importance score of each feature;

[0030] The screened key soil quality index data are integrated with the pre-processed environmental factor data to construct a soil quality-environmental factor coupling analysis model.

[0031] Preferably, in the soil quality-environmental factor coupling analysis model, a time series analysis method is used to model the dynamic change process of soil quality indicators and environmental factors, and a method for predicting soil quality evolution trends through a recursive neural network algorithm includes:

[0032] Based on the historical monitoring data of soil quality indicators and environmental factors, a time series data set is constructed, and each time point includes multi-dimensional quantitative values ​​of soil quality indicators and environmental factors;

[0033] Use data preprocessing techniques to clean, align and normalize time series data to eliminate the impact of outliers and missing values;

[0034] A recursive neural network model was constructed based on the long short-term memory network (LSTM). The preprocessed time series dataset was used as the model input, and the soil evolution law was learned through model training.

[0035] The trained recursive neural network model is used to predict soil indicators and environmental factors in the future and obtain the trend curve of soil evolution.

[0036] Preferably, according to the soil evolution trend prediction result, combined with the preset landslide risk threshold, the landslide risk level is dynamically evaluated, and if the predicted landslide possibility exceeds the threshold, the method of generating an early warning signal includes:

[0037] Based on the soil-environmental factor coupling data set, a landslide risk analysis model was constructed using the support vector machine algorithm;

[0038] Obtain soil evolution prediction result data and pass it to the landslide risk assessment model;

[0039] According to the preset landslide risk threshold, the landslide risk assessment model is used to dynamically evaluate the soil evolution prediction results to obtain the current landslide risk level;

[0040] Determine whether the current landslide risk level exceeds the preset landslide risk threshold. If it exceeds the threshold, the generation of an early warning signal is triggered.

[0041] The present invention also provides a landslide early warning system, including: a sampling point optimization layout module, a soil rapid detection module, a data analysis modeling module, a landslide risk prediction module and an early warning signal generation module;

[0042] The sampling point optimization layout module is used to optimize the layout of sampling points based on the existing geological model and historical disaster data using a support vector machine algorithm, and to select the position with the highest similarity as a new sampling point by calculating the spatial similarity between each candidate sampling point and the known landslide point;

[0043] The soil quality rapid detection module is used to conduct on-site testing of sampling points, obtain soil quality index data, and upload the soil quality index data to the cloud server in real time for comprehensive evaluation;

[0044] The data analysis and modeling module is used to select features and sort the importance of soil index data using a random forest algorithm based on the comprehensive assessment results, screen out the key indicators that have the most significant impact on landslide risk, and integrate them with environmental factor data to build a soil-environmental factor coupling analysis model;

[0045] The landslide risk prediction module is used to model the dynamic change process of soil indicators and environmental factors in the soil quality-environmental factor coupling analysis model by using a time series analysis method, and to predict the soil quality evolution trend by using a recursive neural network algorithm;

[0046] The warning signal generation module is used to dynamically evaluate the landslide risk level according to the soil evolution trend prediction results and the preset landslide risk threshold, and generate a warning signal if the predicted landslide possibility exceeds the threshold.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] The present invention discloses an intelligent early warning method and system for landslide risk based on soil-environmental factor coupling analysis. The sampling point layout is optimized by a support vector machine algorithm, and a multi-parameter soil rapid detection device is used to obtain soil index data and upload it in real time. In the cloud server, a random forest algorithm is used to screen key indicators and construct a soil-environmental factor coupling analysis model. The model combines time series analysis and recursive neural network algorithms to achieve soil evolution trend prediction and dynamic assessment of landslide risk. The present invention also deploys a wireless sensor network to achieve automated monitoring, and dynamically selects the optimal sub-model according to geological conditions. The present invention improves the pertinence, dynamism and accuracy of landslide risk warning, and provides an intelligent solution for landslide disaster prevention and control. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0050] Figure 1 A schematic diagram of a landslide early warning method according to an embodiment of the present invention;

[0051] Figure 2 The figure is a schematic diagram of a landslide early warning system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments 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.

[0053] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0054] Embodiment 1

[0055] like Figure 1 As shown, the present invention provides a landslide early warning method, comprising the following steps:

[0056] S101: Based on the existing geological model and historical disaster data, the support vector machine algorithm is used to optimize the layout of sampling points. By calculating the spatial similarity between each candidate sampling point and the known landslide point, the position with the highest similarity is selected as the new sampling point to improve the pertinence and representativeness of the sampling.

[0057] S102: Use a multi-parameter soil rapid detection device to conduct on-site testing of the sampling points to obtain soil moisture content, density, particle size distribution and other indicator data, and upload the data to the cloud server in real time through a wireless transmission module to achieve a comprehensive assessment of soil indicators.

[0058] S103: Based on the comprehensive assessment results, the random forest algorithm is used in the cloud server to perform feature selection and importance ranking of soil index data, screen out the key indicators that have the most significant impact on landslide risk, and integrate them with the environmental factor data to construct a soil-environmental factor coupling analysis model.

[0059] S104: In the soil-environmental factor coupling analysis model, the time series analysis method is used to model the dynamic change process of soil indicators and environmental factors, and the recursive neural network algorithm is used to predict the soil evolution trend and determine the possibility of landslides in the future.

[0060] S105: Based on the soil evolution prediction results and the preset landslide risk threshold, the landslide risk level is dynamically evaluated. If the predicted landslide possibility exceeds the threshold, an early warning signal is generated and pushed to relevant departments and personnel in real time through the wireless transmission module.

[0061] In this embodiment, based on the existing geological model and historical disaster data, the support vector machine algorithm is used to optimize the layout of the sampling points. By calculating the spatial similarity between each candidate sampling point and the known landslide point, the method of selecting the position with the highest similarity as the new sampling point includes:

[0062] Based on the existing geological model and historical disaster data, a landslide-prone area identification model is constructed to determine whether the area where the candidate sampling point is located belongs to a landslide-prone area; if the candidate sampling point is located in a landslide-prone area, the geological characteristics of the point, including terrain slope, lithology, soil type, etc., are obtained as input features of the support vector machine model; some points are selected from the known landslide points as training samples, their geological characteristics are extracted, and the characteristics are normalized to construct a support vector machine classification model; using the trained support vector machine model, the spatial similarity between the candidate sampling points and the known landslide points is calculated to obtain the similarity score of each candidate point; the candidate sampling points are sorted according to the similarity score, and the top N points with the highest score are selected as new sampling points, and the value of N is determined according to actual needs; the newly selected sampling points are merged with the original sampling points to obtain an optimized sampling point layout plan; according to the optimized sampling point layout plan, field sampling work is carried out to obtain the geological data of each sampling point for subsequent landslide risk assessment and early warning.

[0063] Specifically, the existing geological model and historical disaster data are used to construct a landslide-prone area identification model. The logistic regression algorithm can be used to take 10 geological features such as terrain slope, lithology, and soil type as input variables, and whether landslides have occurred in the area in history as a binary output variable to perform parameter estimation on the training set. When the model accuracy reaches more than 85%, it can be used to determine whether the area where the candidate sampling point is located is prone to landslides. For the candidate points located in the landslide-prone area, eight main geological features such as terrain slope, lithology, and soil type are extracted. 70 of the 100 known landslide points are randomly selected as training samples, and the 8-dimensional features of these 70 points are normalized to the maximum and minimum values. Then, the support vector machine with Gaussian kernel function is used to train the binary classification model, and the model parameters are optimized using grid search. The average accuracy of 5-fold cross validation reaches 92%. Then, the trained support vector machine model is used to calculate the kernel function similarity between each candidate point and the known landslide point, and the average value is taken as the similarity score of the candidate point. All candidate points are sorted from high to low according to the similarity score, and the top 20% of the points are taken as new sampling points, which are merged with the original sampling points to obtain the optimized sampling point layout plan. Finally, field sampling work is carried out according to the optimization plan, and the longitude and latitude coordinates of each sampling point are accurately recorded with a GPS locator, the slope is measured with a laser rangefinder, and rock and soil samples are collected and brought back to the laboratory for analysis. The data obtained can be used for subsequent landslide risk assessment and early warning system development.

[0064] In this embodiment, the method of conducting on-site testing on sampling points, obtaining soil index data, and uploading the index data to a cloud server in real time for comprehensive evaluation includes:

[0065] The soil moisture content, density, particle size composition and other parameter data collected by the portable soil testing equipment are uploaded to the cloud server in real time through the wireless transmission module. In the cloud server, the uploaded soil parameter data are comprehensively analyzed according to the preset soil evaluation model to obtain a comprehensive score of soil quality. If the comprehensive soil score is lower than the preset threshold, the soil quality of the sampling point is judged to be unqualified, and the location information of the sampling point is marked as abnormal. According to the comprehensive soil score of each sampling point, the interpolation algorithm is used to generate a soil quality distribution map of the entire area. For areas with low soil quality scores, the density clustering algorithm is used to divide the clustered areas of soil degradation. For the soil degradation clustered areas, combined with the soil parameter distribution characteristics in the area, the decision tree algorithm is used to recommend corresponding soil improvement measures. The soil quality distribution map, the results of the degraded clustering area division, and the improvement measures suggestions are displayed to the user through a visual interface to assist agricultural production management decisions.

[0066] Specifically, the portable soil testing equipment uploads the collected soil moisture content, density, particle size composition and other parameter data to the cloud server in real time through the wireless transmission module. For example, the equipment collects soil moisture content of 18% and density of 35g / cm at a certain sampling point. 3 , the silt content in the particle size composition is 30%, and the clay content is 25%. These data are transmitted to the cloud server in real time through the 4G network. After receiving the data, the cloud server conducts a comprehensive analysis of the uploaded soil parameter data according to the preset soil evaluation model, and uses the weighted average method to calculate the comprehensive score of soil quality, in which the water content weight is 3, the density weight is 2, the silt content weight is 2, and the clay content weight is 3. The soil comprehensive score of the sampling point is 78 points. If the soil score threshold is set to 60 points, the soil quality of the sampling point is qualified. If the soil comprehensive score of a sampling point is 55 points, which is lower than the threshold, the soil quality of the point is judged to be unqualified, and its location information is marked as an abnormal point. The Kriging interpolation algorithm is used to spatially interpolate the soil comprehensive score of each sampling point to generate a soil quality distribution map covering the entire area. The spatial variation characteristics of soil quality can be intuitively seen from the figure. For areas with soil scores below 60 points, the DBSCAN density clustering algorithm is used to divide the clustering area. The clustering radius is set to 500 meters and the minimum number of points is 5, which can identify the clustering area of ​​soil degradation. According to the soil characteristics in the degradation cluster area, the decision tree algorithm is used to recommend corresponding soil improvement measures. For example, for soil density greater than 4g / cm 3 In degraded areas with silt content below 20%, deep loosening and application of organic fertilizer are recommended to improve soil aggregate structure. Finally, the soil quality distribution map, the results of the degraded cluster division, and the improvement measures are visualized through the Web GIS platform. Agricultural managers can intuitively view the regional soil quality status and provide data support for the formulation of precise soil management measures.

[0067] In this embodiment, based on the comprehensive evaluation results, the random forest algorithm is used to select features and sort the importance of soil index data, screen out the key indicators with the most significant impact on landslide risk, and integrate them with environmental factor data. The method for constructing a soil-environmental factor coupling analysis model includes:

[0068] Acquire soil quality index data and environmental factor data, and preprocess the soil quality index data and the environmental factor data to obtain preprocessed soil quality index data and preprocessed environmental factor data;

[0069] According to the comprehensive evaluation results, the random forest algorithm is used to perform feature selection on the pre-processed soil index data to obtain the selected key soil index data, wherein the key soil index data is the soil index data with the greatest impact on the landslide risk selected according to a preset importance threshold by calculating the importance score of each feature;

[0070] The screened key soil quality index data are integrated with the pre-processed environmental factor data to construct a soil quality-environmental factor coupling analysis model.

[0071] Specifically, the acquired soil quality index data and environmental factor data are first preprocessed, including data cleaning, normalization and other operations. Among them, soil quality index data include soil moisture content, density, particle size distribution, etc., and environmental factor data include rainfall, slope, vegetation coverage, etc. After preprocessing, the random forest algorithm is used to select features of soil quality index data. By calculating the Gini importance score of each feature, soil quality indicators with scores greater than 1 are selected as key indicators, such as soil moisture content, clay content, etc. The selected key soil quality index data are integrated with the preprocessed environmental factor data to construct a soil quality-environmental factor coupling analysis model.

[0072] In this embodiment, in the soil quality-environmental factor coupling analysis model, a time series analysis method is used to model the dynamic change process of soil quality indicators and environmental factors, and a method for predicting soil quality evolution trends through a recursive neural network algorithm includes:

[0073] Based on the historical monitoring data of soil indicators and environmental factors, a time series data set is constructed, and each time point includes multi-dimensional quantitative values ​​of soil indicators and environmental factors. Data preprocessing technology is used to clean, align, and normalize the time series data to eliminate the influence of outliers and missing values ​​and improve data quality. A recursive neural network model is constructed based on the long short-term memory network (LSTM), and soil indicators and environmental factors are used as model inputs to learn the laws of soil evolution through model training. During the model training process, the cross-validation method is used to optimize the model hyperparameters to improve the generalization ability and prediction accuracy of the model. The trained recursive neural network model is used to predict soil indicators and environmental factors in the future to obtain the trend curve of soil evolution.

[0074] Specifically, based on the historical monitoring data of soil indicators and environmental factors, a time series data set is constructed. Each time point includes 10 indicators such as soil moisture content, soil density, soil cohesion, and 5 environmental factors such as rainfall, temperature, and wind speed, forming a 15-dimensional quantized value vector. Data preprocessing technology is used to clean the time series data, remove outliers beyond the physical threshold range, interpolate and fill missing values, and then perform maximum and minimum normalization operations to map the data to the [0,1] interval to eliminate the influence of different indicator dimensions and improve data quality. A recursive neural network model with 2 hidden layers is constructed based on the long short-term memory network (LSTM), each hidden layer contains 128 neurons, and the normalized 15-dimensional soil indicators and environmental factors are used as model inputs. The model is trained by setting the time step to 30 days and the sliding window size to 90 days to allow the model to learn the laws of soil evolution. During the model training process, the 5-fold cross validation method was used, and the root mean square error (RMSE) was used as the evaluation index to optimize the model hyperparameters. The optimal learning rate was 01, the regularization coefficient was 001, and the number of training iterations was 500 times through grid search to improve the generalization ability and prediction accuracy of the model. Using the trained recursive neural network model, the monitoring data of the last 90 days was used as input to predict the trend of soil indicators and environmental factors in the next 30 days, and the evolution curves of indicators such as soil moisture content and soil density were obtained.

[0075] In this embodiment, according to the soil evolution trend prediction result, combined with the preset landslide risk threshold, the landslide risk level is dynamically evaluated. If the predicted landslide possibility exceeds the threshold, the method of generating an early warning signal includes:

[0076] For the soil-environmental factor coupling data set, the support vector machine algorithm is used to construct a landslide risk analysis model. The radial basis kernel function is used to optimize the model parameters through grid search, such as the penalty coefficient C is 10 and the kernel function parameter γ is 1. The 5-fold cross-validation method is used to evaluate the model performance. If the average accuracy of the model is higher than 85%, it is deployed to the cloud server. When new data is input, it is input into the deployed model after preprocessing to obtain the landslide risk probability of the area. Combined with the preset risk level threshold (low risk: <3, medium risk: 3~7, high risk: >7), the landslide risk level of the area is judged, a risk analysis report is generated, and the analysis results are returned through the cloud interface to provide a decision-making basis for disaster prevention and mitigation.

[0077] The soil evolution prediction result data is obtained and passed as an input parameter to the landslide risk assessment model. According to the preset landslide risk threshold, the landslide risk assessment model is used to dynamically evaluate the soil evolution prediction results to obtain the current landslide risk level. It is determined whether the current landslide risk level exceeds the preset landslide risk threshold. If it exceeds the threshold, the generation of the early warning signal is triggered. After determining that the early warning signal needs to be issued, the early warning signal is converted into a data format suitable for wireless transmission through data encapsulation and encoding. A wireless transmission module is used to establish a communication link with the preset receiving end, and the encapsulated early warning signal data is sent to the receiving end. After the receiving end receives the early warning signal data, the original early warning signal content is restored through data parsing and decoding. According to the pre-set early warning signal push rules, the restored early warning signal content is pushed to relevant departments and personnel so that timely response measures can be taken.

[0078] Specifically, the evolution prediction data of soil moisture content, soil density and other factors in the next 7 days are obtained from the recursive neural network model and passed as input parameters to the landslide risk assessment model based on the support vector machine. According to the preset landslide risk threshold of 6, the model is used to dynamically evaluate the soil evolution prediction results. By calculating the similarity between the soil parameters and the threshold, the current landslide risk level is 75. It is judged that the current landslide risk level 75 exceeds the preset landslide risk threshold of 6, triggering the generation of the early warning signal. The early warning signal content "Landslide risk level 75, exceeds the early warning threshold of 6, please take measures in time" is Base64 encoded and converted into a data format suitable for wireless transmission. A 433MHz wireless transmission module is used to establish a communication link with the preset receiving end, and the encapsulated early warning signal data is sent to the receiving end at a baud rate of 9600bps through the UART serial communication protocol. After the receiving end receives the Base64-encoded early warning signal data through the serial port, it performs Base64 decoding to restore the original early warning signal content. According to the pre-set warning signal push rules, the SMS gateway service is used to push the restored warning signal content to the geological disaster monitoring department and local government leaders in the form of SMS, so that timely response measures such as evacuating residents and reinforcing the mountain can be taken to minimize the losses caused by landslide disasters.

[0079] Embodiment 2

[0080] like Figure 2 As shown, the present invention also provides a landslide early warning system, including: a sampling point optimization layout module, a soil rapid detection module, a data analysis modeling module, a landslide risk prediction module and an early warning signal generation module;

[0081] The sampling point optimization layout module is used to optimize the layout of sampling points based on the existing geological model and historical disaster data using the support vector machine algorithm. By calculating the spatial similarity between each candidate sampling point and the known landslide point, the position with the highest similarity is selected as the new sampling point.

[0082] The soil rapid detection module is used to conduct on-site testing of sampling points, obtain soil index data, and upload the soil index data to the cloud server in real time for comprehensive evaluation;

[0083] The data analysis and modeling module is used to select features and sort the importance of soil index data based on the comprehensive assessment results using the random forest algorithm, screen out the key indicators that have the most significant impact on landslide risk, and integrate them with environmental factor data to build a soil-environmental factor coupling analysis model;

[0084] The landslide risk prediction module is used to model the dynamic change process of soil indicators and environmental factors in the soil-environmental factor coupling analysis model using the time series analysis method, and to predict the soil evolution trend through the recursive neural network algorithm;

[0085] The warning signal generation module is used to dynamically evaluate the landslide risk level based on the soil evolution trend prediction results and the preset landslide risk threshold. If the predicted landslide possibility exceeds the threshold, a warning signal is generated.

[0086] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A landslide early warning method, characterized in that: The following steps are involved: According to the existing geological model and historical disaster data, the support vector machine algorithm is used to optimize the layout of sampling points. By calculating the spatial similarity between each candidate sampling point and the known landslide point, the position with the highest similarity is selected as the new sampling point. Conduct on-site testing of sampling points to obtain soil quality index data, and upload the soil quality index data to the cloud server in real time for comprehensive evaluation; According to the comprehensive assessment results, the random forest algorithm is used to select features and sort the importance of soil index data, screen out the key indicators with the most significant impact on landslide risk, and integrate them with environmental factor data to build a soil-environmental factor coupling analysis model; In the soil quality-environmental factor coupling analysis model, the time series analysis method is used to model the dynamic change process of soil quality indicators and environmental factors, and the recursive neural network algorithm is used to predict the soil quality evolution trend; According to the soil evolution trend prediction results and the preset landslide risk threshold, the landslide risk level is dynamically evaluated. If the predicted landslide possibility exceeds the threshold, an early warning signal is generated.

2. The landslide early warning method according to claim 1, characterized in that: According to the existing geological model and historical disaster data, the support vector machine algorithm is used to optimize the layout of sampling points. By calculating the spatial similarity between each candidate sampling point and the known landslide point, the method of selecting the position with the highest similarity as the new sampling point includes: Based on the existing geological model and historical disaster data, a landslide-prone area identification model is constructed; Based on the landslide-prone area identification model, determine whether the area where the candidate sampling point is located belongs to the landslide-prone area; If the candidate sampling point is located in a landslide-prone area, the geological characteristics of the point are obtained and used as input features of the support vector machine model; Select some points from known landslide points as training samples, extract their geological features, normalize all features, and build a support vector machine classification model; The trained support vector machine classification model is used to calculate the spatial similarity between the candidate sampling points and the known landslide points, and the similarity score of each candidate point is obtained; Sort the candidate sampling points according to their similarity scores, and select the top N points with the highest scores as new sampling points; The newly selected sampling points are merged with the original sampling points to obtain an optimized sampling point layout plan.

3. The landslide early warning method according to claim 1, characterized in that: The method of conducting on-site testing on sampling points, obtaining soil index data, and uploading the index data to the cloud server in real time for comprehensive evaluation includes: The soil moisture content, density and particle size composition parameter data collected by the portable soil testing equipment are uploaded to the cloud server in real time through the wireless transmission module; In the cloud server, the uploaded soil parameter data is comprehensively analyzed according to the preset soil assessment model to obtain a comprehensive score of soil quality; Based on the comprehensive soil score of each sampling point, the soil quality distribution map of the entire area is generated using an interpolation algorithm; For areas where the soil quality score is below the preset threshold, a density clustering algorithm is used to delineate soil degradation clusters; For the concentrated areas of soil degradation, the decision tree algorithm is used to recommend corresponding soil improvement measures based on the distribution characteristics of soil parameters in the area; The soil quality distribution map, soil degradation cluster division results, and soil improvement measures are displayed to users through a visual interface to assist agricultural production management decisions.

4. The landslide early warning method according to claim 1, characterized in that: According to the comprehensive assessment results, the random forest algorithm is used to select features and sort the importance of soil index data, screen out the key indicators with the most significant impact on landslide risk, and integrate them with environmental factor data. The method of constructing a soil-environmental factor coupling analysis model includes: Acquire soil quality index data and environmental factor data, and preprocess the soil quality index data and the environmental factor data to obtain preprocessed soil quality index data and preprocessed environmental factor data; According to the comprehensive evaluation results, the random forest algorithm is used to perform feature selection on the pre-processed soil index data to obtain the selected key soil index data, wherein the key soil index data is the soil index data with the greatest impact on the landslide risk selected according to a preset importance threshold by calculating the importance score of each feature; The screened key soil quality index data are integrated with the pre-processed environmental factor data to construct a soil quality-environmental factor coupling analysis model.

5. The landslide early warning method according to claim 4, characterized in that: In the soil quality-environmental factor coupling analysis model, the time series analysis method is used to model the dynamic change process of soil quality indicators and environmental factors. The methods for predicting soil quality evolution trends through recursive neural network algorithms include: Based on the historical monitoring data of soil quality indicators and environmental factors, a time series data set is constructed, and each time point includes multi-dimensional quantitative values ​​of soil quality indicators and environmental factors; Use data preprocessing techniques to clean, align and normalize time series data to eliminate the impact of outliers and missing values; A recursive neural network model was constructed based on the long short-term memory network (LSTM). The preprocessed time series dataset was used as the model input, and the soil evolution law was learned through model training. The trained recursive neural network model is used to predict soil indicators and environmental factors in the future and obtain the trend curve of soil evolution.

6. The landslide early warning method according to claim 5, characterized in that: According to the soil evolution trend prediction results, combined with the preset landslide risk threshold, the landslide risk level is dynamically evaluated. If the predicted landslide possibility exceeds the threshold, the method of generating an early warning signal includes: Based on the soil-environmental factor coupling data set, a landslide risk analysis model was constructed using the support vector machine algorithm; Obtain soil evolution prediction result data and pass it to the landslide risk assessment model; According to the preset landslide risk threshold, the landslide risk assessment model is used to dynamically evaluate the soil evolution prediction results to obtain the current landslide risk level; Determine whether the current landslide risk level exceeds the preset landslide risk threshold. If it exceeds the threshold, the generation of an early warning signal is triggered.

7. A landslide early warning system, characterized in that: include: Sampling point optimization layout module, soil rapid detection module, data analysis and modeling module, landslide risk prediction module and early warning signal generation module; The sampling point optimization layout module is used to optimize the layout of sampling points based on the existing geological model and historical disaster data using a support vector machine algorithm, and to select the position with the highest similarity as a new sampling point by calculating the spatial similarity between each candidate sampling point and the known landslide point; The soil quality rapid detection module is used to conduct on-site testing of sampling points, obtain soil quality index data, and upload the soil quality index data to the cloud server in real time for comprehensive evaluation; The data analysis and modeling module is used to select features and sort the importance of soil index data using a random forest algorithm based on the comprehensive assessment results, screen out the key indicators that have the most significant impact on landslide risk, and integrate them with environmental factor data to build a soil-environmental factor coupling analysis model; The landslide risk prediction module is used to model the dynamic change process of soil indicators and environmental factors in the soil quality-environmental factor coupling analysis model by using a time series analysis method, and to predict the soil quality evolution trend by using a recursive neural network algorithm; The warning signal generation module is used to dynamically evaluate the landslide risk level according to the soil evolution trend prediction results and the preset landslide risk threshold, and generate a warning signal if the predicted landslide possibility exceeds the threshold.

Images