Landslide disaster monitoring and early warning method and system based on big data

Abstract

The invention discloses a landslide disaster monitoring and early warning method and system based on big data, and the method comprises the steps: obtaining meteorological element data and slope feature data, unifying the data to the same time and space scale through a space-time alignment algorithm, extracting features related to landslide early warning, carrying out the importance evaluation of the features through a random forest algorithm, and carrying out the early warning of the landslide disaster. Determining the influence weights of the meteorological element features and the slope features on the occurrence of the landslide, and obtaining a feature weight matrix; according to the feature weight matrix, a support vector machine algorithm is adopted to establish an interaction relation model of the meteorological element features and the slope features, and an interaction relation matrix is obtained; and inputting the feature weight matrix and the interaction relation matrix into a landslide early warning model, calculating a landslide early warning value and a risk value at the current moment, and obtaining a landslide early warning result. According to the invention, accurate evaluation and early warning of the landslide risk are realized, the accuracy and timeliness of early warning are improved, and effective technical support is provided for landslide disaster prevention and control.

Description

Technical Field

[0001] The present invention belongs to the field of disaster early warning, and in particular relates to a landslide disaster monitoring and early warning method and system based on big data. Background Art

[0002] In the landslide disaster monitoring and early warning system based on big data, how to effectively integrate meteorological big data and slope characteristic data is a key technical issue. Meteorological big data contains a large number of meteorological elements, such as rainfall, temperature, humidity, etc., while slope characteristic data includes information such as the geological structure, soil characteristics, and vegetation coverage of the slope. There are differences in the temporal and spatial scales and data formats of these two types of data. How to organically combine them and extract meaningful features for landslide early warning is a difficult point. In addition, when calculating the landslide warning value and risk value, it is necessary to consider the influence weights of different meteorological elements and slope characteristics on the occurrence of landslides, as well as the interaction between them. How to adaptively determine these weights and relationships through big data algorithms and realize the dynamic update of landslide warning values ​​and risk values ​​is another technical problem that needs to be solved urgently. At the same time, in the actual landslide monitoring and early warning along the road, the real-time, stability and security of data transmission and processing also need to be considered to ensure that the warning information can be delivered to relevant departments and personnel in a timely and accurate manner, thereby minimizing the losses caused by landslide disasters. Summary of the invention

[0003] In order to solve the above technical problems, the present invention provides a landslide disaster monitoring and early warning method and system based on big data. Among them, a landslide disaster monitoring and early warning method based on big data includes:

[0004] Acquire meteorological element data and slope characteristic data, and unify the meteorological element data and the slope characteristic data to the same time and space scales through a spatiotemporal alignment algorithm to obtain an aligned data set;

[0005] Extracting features related to landslide warning from the aligned data set, using a random forest algorithm to evaluate the importance of the features, determining the weights of the impact of meteorological element features and slope features on landslide occurrence, and obtaining a feature weight matrix;

[0006] According to the feature weight matrix, a support vector machine algorithm is used to establish an interaction relationship model between the meteorological element features and the slope features to obtain an interaction relationship matrix;

[0007] The characteristic weight matrix and the interaction relationship matrix are input into a pre-established landslide early warning model, and the landslide early warning value and risk value at the current moment are calculated to obtain a landslide early warning result.

[0008] Preferably, the process of obtaining meteorological element data and slope characteristic data includes:

[0009] Real-time collection of meteorological element data including rainfall, temperature and humidity through meteorological monitoring stations or meteorological sensors;

[0010] Slope characteristic data including geological structure, soil properties and vegetation cover are obtained through field investigation, geological drilling and geotechnical testing.

[0011] Preferably, the process of unifying the meteorological element data and the slope characteristic data to the same time and space scale by using a time-space alignment algorithm comprises:

[0012] According to the spatiotemporal attribute information of the meteorological element data and the slope characteristic data, it is determined whether there are differences between the two types of data in spatiotemporal scale and data format;

[0013] If there are differences, the time alignment algorithm is used for the time attributes of the meteorological element data and the slope characteristic data to align the two types of data in the time dimension and unify them to the same time scale;

[0014] According to the spatial attributes of meteorological element data and slope characteristic data, a spatial alignment algorithm is used to align the two types of data in spatial dimensions and unify them to the same spatial scale.

[0015] At the same time, in the process of spatiotemporal alignment, interpolation algorithms or resampling algorithms are selected according to the attribute characteristics of the data to estimate or convert missing or inconsistent data; through spatiotemporal alignment processing, meteorological element data and slope characteristic data with unified spatiotemporal scale and data format are obtained, and the aligned meteorological element data and slope characteristic data are fused using a data fusion algorithm to obtain the aligned data set.

[0016] Preferably, the process of extracting features related to landslide early warning from the aligned data set comprises:

[0017] Based on the aligned datasets, the meteorological element characteristics and slope characteristics related to landslide warning are determined;

[0018] According to the characteristics of meteorological elements, the time series data of rainfall, temperature and humidity are extracted respectively, and the changing trend of each element is determined by the time series analysis algorithm;

[0019] According to the slope characteristics, the spatial distribution data of geological structure, soil properties and vegetation cover are obtained, and the stability and change of each feature are determined by using spatial analysis algorithms;

[0020] According to the changing trend of meteorological element characteristics and the stability of slope characteristics, a landslide early warning index system is constructed, and the index weights are determined by comprehensively considering the impact of various characteristics on landslides.

[0021] Preferably, the importance of the features is evaluated by using a random forest algorithm, and the weights of the influence of meteorological element features and slope features on the occurrence of landslides are determined. The process of obtaining a feature weight matrix includes:

[0022] For the acquired meteorological element characteristic data and slope characteristic data, the principal component analysis method is used to reduce the dimension of the characteristics, and the meteorological element characteristics and slope characteristics after dimension reduction are obtained;

[0023] The random forest algorithm is used to evaluate the importance of the meteorological elements and slope characteristics after dimensionality reduction, and the weight of the influence of the characteristics on the occurrence of landslides is obtained by calculating the Gini index of the characteristics.

[0024] A feature weight matrix is ​​constructed according to the influence weights, and features in the feature weight matrix having weights greater than a preset threshold are determined as key features.

[0025] Preferably, according to the feature weight matrix, a support vector machine algorithm is used to establish an interaction relationship model between the meteorological element features and the slope features, and the process of obtaining the interaction relationship matrix includes:

[0026] Inputting the feature weight matrix into a support vector machine algorithm, setting a kernel function and a penalty factor, and then training the interaction relationship model to obtain model parameters;

[0027] According to the trained interaction relationship model, the meteorological element characteristics and slope characteristic data are input to predict the interaction relationship and obtain the interaction relationship matrix;

[0028] The interaction relationship matrix is ​​analyzed to extract key interaction feature pairs, and the mutual influence degree between meteorological elements and slope characteristics is obtained according to the weight values ​​of the feature pairs.

[0029] Preferably, according to the trained interaction relationship model, the process of inputting the meteorological element characteristics and the slope characteristic data, predicting the interaction relationship, and obtaining the interaction relationship matrix includes:

[0030] Perform data cleaning, missing value processing and data normalization preprocessing operations on the meteorological element characteristic data and slope characteristic data to obtain a standardized characteristic data set;

[0031] According to the standardized feature data set, a random forest algorithm is used to train the interaction relationship prediction model, and the model hyperparameters are optimized through grid search and cross-validation to obtain the model with the best performance;

[0032] Apply the best performance model obtained through training to the new meteorological element characteristic data and slope characteristic data, predict the interaction relationship strength between the two, and obtain the interaction relationship matrix;

[0033] Visualize the interaction relationship matrix, judge the interaction strength between different meteorological elements and slope characteristics according to the value of the elements in the matrix, and determine the key interaction relationship;

[0034] According to the key interaction relationships, the association rule mining algorithm is used to analyze the association patterns between meteorological elements and slope characteristics, and the association rules of the built-in connection between meteorological elements and slope characteristics are obtained.

[0035] Preferably, the feature weight matrix and the interaction relationship matrix are input into a pre-established landslide early warning model, and the landslide early warning value and risk value at the current moment are calculated. The process of obtaining the landslide early warning result includes:

[0036] Acquire monitoring data at the current moment, and calculate a feature weight matrix and an interaction relationship matrix according to the monitoring data;

[0037] Input the characteristic weight matrix and interaction relationship matrix of the current monitoring data into the landslide early warning model;

[0038] In the landslide early warning model, a support vector machine algorithm is used to train and predict the input matrix data to obtain the landslide early warning value at the current moment;

[0039] In the landslide early warning model, a decision tree algorithm is used to classify and predict the input matrix data to obtain the landslide risk value at the current moment;

[0040] The calculated landslide warning value is compared with the preset warning threshold. If the warning value is greater than or equal to the warning threshold, it is determined that there is a landslide risk and a warning signal is output;

[0041] The calculated landslide risk value is compared with the preset risk level threshold to determine the current landslide risk level, output the risk level information, and generate a complete landslide warning result.

[0042] Preferably, in the landslide early warning model, the process of using a decision tree algorithm to classify and predict the input matrix data to obtain the landslide risk value at the current moment includes:

[0043] The decision tree algorithm is used to classify and predict the input matrix data to obtain preliminary landslide risk assessment results;

[0044] Compare the preliminary assessment result with the preset risk threshold to determine whether there is a landslide risk at the current moment; if the risk value at the current moment exceeds the preset threshold, trigger a landslide warning signal and send the warning information to relevant departments;

[0045] After the early warning signal is triggered, the changes in the matrix data are continuously monitored and the landslide risk assessment results are dynamically updated. According to the changing trend of the risk assessment results, the time series analysis method is used to predict the landslide risk in the future. The risk assessment results and prediction information are visualized to generate intuitive landslide risk distribution maps and trend maps.

[0046] The present invention also provides a landslide disaster monitoring and early warning system based on big data, comprising:

[0047] A data acquisition module, used to acquire meteorological element data and slope characteristic data;

[0048] A spatiotemporal alignment module, used to unify the meteorological element data and the slope characteristic data to the same time and space scales through a spatiotemporal alignment algorithm to obtain an aligned data set;

[0049] A feature extraction module, used to extract features related to landslide warning from the aligned data set;

[0050] A feature evaluation module is used to evaluate the importance of the features using a random forest algorithm, determine the weights of the impact of meteorological element features and slope features on the occurrence of landslides, and obtain a feature weight matrix;

[0051] A relationship modeling module is used to establish an interaction relationship model between the meteorological element characteristics and the slope characteristics according to the feature weight matrix using a support vector machine algorithm to obtain an interaction relationship matrix;

[0052] The early warning calculation module is used to input the characteristic weight matrix and the interaction relationship matrix into a pre-established landslide early warning model, calculate the landslide early warning value and risk value at the current moment, and obtain the landslide early warning result.

[0053] Compared with the prior art, the present invention has the following advantages and technical effects:

[0054] The present invention obtains meteorological element data and slope feature data, uses a spatiotemporal alignment algorithm to unify the data scale, and extracts landslide warning-related features. Subsequently, the random forest algorithm is used to evaluate the importance of the features, a feature weight matrix is ​​obtained, and a support vector machine algorithm is used to establish a model of the interaction relationship between the features. Finally, the feature weight matrix and the interaction relationship matrix are input into the warning model to calculate the landslide warning value and risk value. By integrating meteorological and slope feature data, accurate assessment and warning of landslide risks are achieved, the accuracy and timeliness of warnings are improved, and effective technical support is provided for landslide disaster prevention and control. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0056] Figure 1 A schematic diagram of a method flow of an embodiment of the present invention;

[0057] Figure 2 Schematic diagram of the system structure of an embodiment of the present invention. DETAILED DESCRIPTION

[0058] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0059] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0060] Embodiment 1

[0061] like Figure 1 As shown, this embodiment provides a landslide disaster monitoring and early warning method based on big data, including:

[0062] Acquire meteorological element data and slope characteristic data, unify the meteorological element data and slope characteristic data to the same time and space scale through the spatiotemporal alignment algorithm, and obtain the aligned data set;

[0063] The features related to landslide warning are extracted from the aligned data set, and the importance of the features is evaluated using the random forest algorithm to determine the weights of the impact of meteorological element features and slope features on landslide occurrence, and the feature weight matrix is ​​obtained.

[0064] According to the characteristic weight matrix, the interaction relationship model between meteorological element characteristics and slope characteristics is established by using the support vector machine algorithm to obtain the interaction relationship matrix.

[0065] The characteristic weight matrix and the interaction relationship matrix are input into the pre-established landslide early warning model, the landslide warning value and risk value at the current moment are calculated, and the landslide early warning result is obtained.

[0066] Furthermore, the process of obtaining meteorological element data and slope characteristic data includes:

[0067] Real-time collection of meteorological element data including rainfall, temperature and humidity through meteorological monitoring stations or meteorological sensors;

[0068] Slope characteristic data including geological structure, soil properties and vegetation cover are obtained through field investigation, geological drilling and geotechnical testing.

[0069] Specifically, the acquisition of meteorological element data is the basis for slope stability assessment. Taking rainfall as an example, this embodiment collects data in real time through an automatic rain gauge station, which is composed of a tipping bucket rain gauge, a data collector and a communication module, and automatically records and transmits rainfall data every hour.

[0070] The acquisition of slope characteristic data in this embodiment involves multiple methods. Geological drilling can reveal the internal structure of the slope. For example, drilling of a slope revealed that the surface layer was 5 meters thick of silty clay, and the underlying layer was 10 meters thick of strongly weathered rock. Geotechnical tests can measure soil strength parameters, such as internal friction angle, cohesion, etc. After these data are sorted and analyzed, a slope characteristic database is constructed to provide a basis for subsequent modeling.

[0071] Furthermore, the process of unifying meteorological element data and slope characteristic data to the same temporal and spatial scales through the spatiotemporal alignment algorithm includes:

[0072] According to the spatiotemporal attribute information of meteorological element data and slope characteristic data, it is determined whether there are differences between the two types of data in spatiotemporal scale and data format;

[0073] If there are differences, the time alignment algorithm is used for the time attributes of the meteorological element data and the slope characteristic data to align the two types of data in the time dimension and unify them to the same time scale;

[0074] According to the spatial attributes of meteorological element data and slope characteristic data, a spatial alignment algorithm is used to align the two types of data in spatial dimensions and unify them to the same spatial scale.

[0075] At the same time, in the process of spatiotemporal alignment, interpolation algorithms or resampling algorithms are selected according to the attribute characteristics of the data to estimate or convert missing or inconsistent data; through spatiotemporal alignment processing, meteorological element data and slope characteristic data with unified spatiotemporal scale and data format are obtained, and the aligned meteorological element data and slope characteristic data are fused using a data fusion algorithm to obtain the aligned data set.

[0076] Specifically, spatiotemporal alignment is a key step in the data processing of this embodiment, and is particularly important for meteorological element data and slope characteristic data. Meteorological data are usually in hours or days, while slope characteristic data may be in months or years. For example, rainfall data may be recorded once an hour, while soil moisture may be measured once a week. The time alignment algorithm can unify these data to the same time scale, such as accumulating hourly rainfall into daily rainfall, or interpolating weekly soil moisture into daily data.

[0077] Spatial alignment is also important because the weather station and the slope monitoring point may not be in the same location. For example, the weather station may cover a large area, while the slope monitoring point is more local. The spatial alignment algorithm of this embodiment uses Kriging interpolation to interpolate the weather data to the location of the slope monitoring point, or uses the inverse distance weighted method to expand the slope feature data to a larger area.

[0078] In the alignment process, it is crucial to choose a suitable interpolation or resampling algorithm. This embodiment uses nearest neighbor interpolation for discrete data such as rainfall, and linear interpolation or spline interpolation for continuous data such as temperature. For example, if temperature data is missing on a certain day, the temperature of the two days before and after is used for linear interpolation estimation. Data fusion is the process of merging aligned data sets. This embodiment uses principal component analysis to reduce the data dimension and extract the main features. For example, it may be found that rainfall and soil moisture are highly correlated, and they can be merged into a new "humidity index" feature. In addition, a Kalman filter can be used to fuse data from different sources, such as fusing weather forecast data and actual observation data.

[0079] Furthermore, the process of extracting features related to landslide warning from the aligned dataset includes:

[0080] Based on the aligned datasets, the meteorological element characteristics and slope characteristics related to landslide warning are determined;

[0081] According to the characteristics of meteorological elements, the time series data of rainfall, temperature and humidity are extracted respectively, and the changing trend of each element is determined by the time series analysis algorithm;

[0082] According to the slope characteristics, the spatial distribution data of geological structure, soil properties and vegetation cover are obtained, and the stability and change of each feature are determined by using spatial analysis algorithms;

[0083] According to the changing trend of meteorological element characteristics and the stability of slope characteristics, a landslide early warning index system is constructed, and the index weights are determined by comprehensively considering the impact of various characteristics on landslides.

[0084] Specifically, obtaining the aligned dataset is the basis for landslide early warning analysis. These data include meteorological element characteristics and slope characteristics. Meteorological element characteristics mainly involve time series data of rainfall, temperature and humidity. For example, the daily rainfall in a certain area in a week is 10mm, 15mm, 20mm, 25mm, 30mm, 35mm, and 40mm, showing a trend of increasing day by day. Through time series analysis algorithms, such as moving average method or exponential smoothing method, it can be judged that the rainfall is on an upward trend, which may increase the risk of landslides.

[0085] Slope characteristics include spatial distribution data of geological structure, soil properties and vegetation cover. For example, in a mountainous area, its geological structure may include different rock layers such as sandstone, shale and granite. Through spatial analysis algorithms, such as Kriging interpolation, the distribution range and stability of each rock layer can be determined. For example, sandstone areas may be more prone to landslides, while granite areas are relatively stable. When constructing a landslide warning indicator system, the impact of each feature on the landslide is comprehensively considered. For example, the weight of rainfall can be set to 0.4, the weight of temperature to 0.2, the weight of humidity to 0.2, the weight of geological structure to 0.1, the weight of soil properties to 0.05, and the weight of vegetation cover to 0.05. This weight distribution reflects the important impact of rainfall on landslide risk, while also taking into account the role of other factors.

[0086] Furthermore, the random forest algorithm is used to evaluate the importance of the features and determine the weights of the meteorological elements and slope characteristics on the occurrence of landslides. The process of obtaining the feature weight matrix includes:

[0087] For the acquired meteorological element characteristic data and slope characteristic data, the principal component analysis method is used to reduce the dimension of the characteristics, and the meteorological element characteristics and slope characteristics after dimension reduction are obtained;

[0088] The random forest algorithm is used to evaluate the importance of the meteorological elements and slope characteristics after dimensionality reduction, and the weight of the influence of the characteristics on the occurrence of landslides is obtained by calculating the Gini index of the characteristics.

[0089] According to the influence weights, a feature weight matrix is ​​constructed, and the features in the feature weight matrix whose weights are greater than a preset threshold are determined as key features.

[0090] Specifically, the core of the landslide early warning system lies in feature extraction and risk prediction. First, meteorological elements and slope feature data are obtained, including rainfall, temperature, humidity, geological structure, etc. In order to reduce the data dimension and improve the computational efficiency, the principal component analysis method is used for feature dimension reduction. For example, the original data contains 20 features, which may be reduced to 5 principal components through principal component analysis, which can explain more than 80% of the variance of the original data.

[0091] The random forest algorithm constructs multiple decision trees and calculates the Gini index of each feature to evaluate its importance. Assume that in a certain evaluation, the Gini index of rainfall is 0.3 and the geological structure stability is 0.25, which indicates that rainfall has a slightly greater impact on landslide occurrence than geological structure stability. According to the importance of the features, a feature weight matrix is ​​constructed and key features are determined. If the preset threshold is 0.2, both of the above features will be considered key features. This method can screen out the features with the most predictive value and improve the accuracy and efficiency of the model.

[0092] Furthermore, based on the feature weight matrix, the support vector machine algorithm is used to establish the interaction relationship model between the meteorological element characteristics and the slope characteristics. The process of obtaining the interaction relationship matrix includes:

[0093] The feature weight matrix is ​​input into the support vector machine algorithm, and the interaction relationship model is trained after the kernel function and penalty factor are set to obtain the model parameters;

[0094] According to the trained interaction relationship model, the meteorological element characteristics and slope characteristic data are input to predict the interaction relationship and obtain the interaction relationship matrix;

[0095] The interaction relationship matrix is ​​analyzed to extract key interaction feature pairs, and the degree of mutual influence between meteorological elements and slope characteristics is obtained based on the weight values ​​of the feature pairs.

[0096] Specifically, this embodiment takes into account that the support vector machine algorithm performs well in dealing with nonlinear relationships and is suitable for establishing an interaction model between meteorological elements and slope characteristics. By selecting appropriate kernel functions (such as radial basis functions) and penalty factors, a balance can be achieved between model complexity and generalization ability. The trained model can capture the complex interaction relationship between features. Using the trained model to predict the interaction relationship, a matrix can be obtained to show the influence intensity between each feature pair. For example, the matrix may show that the weight of rainfall on slope is 0.8, while the weight of temperature on lithology is only 0.2. This quantitative result helps to identify key interaction feature pairs. By analyzing the interaction relationship matrix, some important geological and hydrological processes can be discovered. For example, it may be found that the interaction between rainfall and slope is particularly significant, which suggests that heavy rainfall is more likely to cause landslides on steep terrain. For example, the interaction between temperature change and lithology is weak, indicating that the effect of temperature on slopes of different lithologies is relatively uniform.

[0097] Furthermore, according to the trained interaction relationship model, the meteorological element characteristics and slope characteristic data are input to predict the interaction relationship, and the process of obtaining the interaction relationship matrix includes:

[0098] Perform data cleaning, missing value processing and data normalization preprocessing operations on the meteorological element characteristic data and slope characteristic data to obtain a standardized characteristic data set;

[0099] Based on the standardized feature data set, the random forest algorithm was used to train the interaction relationship prediction model, and the model hyperparameters were optimized through grid search and cross-validation to obtain the model with the best performance;

[0100] Apply the best performance model obtained through training to the new meteorological element characteristic data and slope characteristic data, predict the interaction relationship strength between the two, and obtain the interaction relationship matrix;

[0101] Visualize the interaction relationship matrix, judge the interaction strength between different meteorological elements and slope characteristics according to the value of the elements in the matrix, and determine the key interaction relationship;

[0102] According to the key interaction relationships, the association rule mining algorithm is used to analyze the association patterns between meteorological elements and slope characteristics, and the association rules of the built-in connection between meteorological elements and slope characteristics are obtained.

[0103] Specifically, the data preprocessing in this embodiment includes: For example, for rainfall data, there may be outliers, such as the rainfall suddenly reaches 1000 mm on a certain day, which is obviously unreasonable and needs to be eliminated through data cleaning. In terms of missing value processing, the time series interpolation method is used. For missing temperature data, interpolation estimation can be performed based on the temperature values ​​at the previous and next time points. Data normalization uses the minimum-maximum scaling method to unify features of different dimensions into the [0,1] interval to facilitate subsequent modeling.

[0104] The random forest algorithm is an integrated learning method that predicts the results by constructing multiple decision trees and taking majority votes. When training the interaction relationship prediction model, meteorological elements and slope characteristics can be used as input variables, and the interaction intensity can be used as the target variable. Model hyperparameter optimization is the key to improving model performance. For example, this embodiment uses a grid search method to try different numbers of trees (such as 100, 500, 1000) and the maximum depth of each tree (such as 5, 10, 15), combined with K-fold cross validation (such as K=5) to select the optimal parameter combination. The interaction relationship matrix can be visualized through a heat map. Assume that the interaction intensity value between rainfall and slope in the matrix is ​​0.8, while the intensity value between temperature and slope is 0.3, which indicates that the impact of rainfall on slope stability may be more significant than temperature. Association rule mining can use the Apriori algorithm, set the minimum support and confidence thresholds (such as 0.1 and 0.7), and mine rules such as "If the 24-hour rainfall exceeds 100 mm and the slope is greater than 30 degrees, the probability of slope instability increases by 60%." The slope disaster warning decision support system can be built based on the above models and rules. The system can receive data from meteorological stations and slope monitoring equipment in real time. For example, when it detects that the expected rainfall in the next 24 hours will reach 120 mm and the current slope gradient is 35 degrees, the system will predict a high risk level based on the pre-trained model and mined rules, and automatically send warning information to relevant departments. This data-driven warning system can improve the accuracy and timeliness of warnings, provide a scientific basis for timely evacuation, reinforcement and other prevention and control measures, and effectively reduce casualties and property losses caused by slope disasters.

[0105] Furthermore, the feature weight matrix and the interaction relationship matrix are input into the pre-established landslide early warning model to calculate the landslide early warning value and risk value at the current moment. The process of obtaining the landslide early warning result includes:

[0106] Obtain the monitoring data at the current moment, and calculate the feature weight matrix and the interaction relationship matrix based on the monitoring data;

[0107] Input the characteristic weight matrix and interaction relationship matrix of the current monitoring data into the landslide early warning model;

[0108] In the landslide early warning model, the support vector machine algorithm is used to train and predict the input matrix data to obtain the landslide early warning value at the current moment;

[0109] In the landslide early warning model, the decision tree algorithm is used to classify and predict the input matrix data to obtain the landslide risk value at the current moment;

[0110] The calculated landslide warning value is compared with the preset warning threshold. If the warning value is greater than or equal to the warning threshold, it is determined that there is a landslide risk and a warning signal is output;

[0111] The calculated landslide risk value is compared with the preset risk level threshold to determine the current landslide risk level, output the risk level information, and generate a complete landslide warning result.

[0112] Specifically, the landslide early warning model is a mathematical model for evaluating and predicting landslide risks. The model requires input of a feature weight matrix and an interaction relationship matrix. The feature weight matrix reflects the degree of influence of various factors on landslides, such as rainfall, slope, geological conditions, etc. The interaction relationship matrix describes the mutual influence between these factors. Taking a mountainous area as an example, the monitoring data includes rainfall, soil moisture, slope, etc. By analyzing historical data, it can be concluded that the weight of rainfall is 0.4, the soil moisture is 0.3, and the slope is 0.3. These values ​​constitute the feature weight matrix. At the same time, it is found that there is a strong positive correlation between rainfall and soil moisture, and the correlation coefficient is 0.8, which is an element of the interaction relationship matrix. The application of the support vector machine algorithm in the landslide early warning of this embodiment is mainly to establish a hyperplane that can distinguish between a safe state and a dangerous state by learning historical data. For example, the monitoring data of the past 10 years can be used as a training set, including the situation where a landslide occurred and the situation where a landslide did not occur. The algorithm will find an optimal classification boundary to predict whether a new data point is in a landslide risk state. The decision tree algorithm classifies and predicts landslide risks by building a tree-like decision structure. For example, the first node may be to determine whether the rainfall exceeds 1000 mm / day. If so, it will enter the next node to determine whether the soil moisture content exceeds 80%. This progressive judgment will eventually result in a risk level. The setting of the warning threshold needs to take into account many factors. Assuming that the warning value is set at 0.7 (the full score is 1), this means that when the warning value output by the model reaches or exceeds 0.7, the system will issue a warning signal. The selection of this threshold requires weighing the risks of missed reports and false alarms. Too high may lead to missed reports, and too low may lead to frequent false alarms. The risk level is usually divided into multiple levels. For example, a risk value of 0-0.3 can be defined as low risk, 0.3-0.6 as medium risk, 0.6-0.8 as high risk, and above 0.8 as extremely high risk. Each level corresponds to different response measures, such as strengthening monitoring and evacuating personnel. The advantage of this landslide warning system is that it can process a large amount of complex data in real time and make risk assessments quickly. It comprehensively considers multiple factors and their interactions to improve the accuracy of early warning. However, the effectiveness of the system depends largely on the quality and integrity of historical data. In addition, due to the complexity and unpredictability of the natural environment, the model prediction results still need further judgment and verification by professionals. Through this method, landslide risks can be better understood and predicted, providing a scientific basis for disaster prevention and mitigation. With the accumulation of data and optimization of algorithms, the accuracy of the early warning system will continue to improve, providing more reliable protection for the safety of mountain residents.

[0113] Furthermore, in the landslide early warning model, the decision tree algorithm is used to classify and predict the input matrix data, and the process of obtaining the landslide risk value at the current moment includes:

[0114] The decision tree algorithm is used to classify and predict the input matrix data to obtain preliminary landslide risk assessment results;

[0115] Compare the preliminary assessment results with the preset risk threshold to determine whether there is a landslide risk at the current moment; if the risk value at the current moment exceeds the preset threshold, a landslide warning signal is triggered and the warning information is sent to relevant departments;

[0116] After the early warning signal is triggered, the changes in the matrix data are continuously monitored and the landslide risk assessment results are dynamically updated. According to the changing trend of the risk assessment results, the time series analysis method is used to predict the landslide risk in the future. The risk assessment results and prediction information are visualized to generate intuitive landslide risk distribution maps and trend maps.

[0117] Specifically, the landslide early warning model determines the possibility of landslide by analyzing multiple factors, such as geological conditions, rainfall, surface deformation, etc. In practical applications, the model needs to obtain various monitoring data in real time, such as the measurement results of inclinometers, displacement sensors, rain gauges and other equipment. After processing, these data form a feature matrix as the input of the model. Decision tree algorithm In landslide risk assessment, the decision tree can classify the current state into different risk levels according to the input feature matrix. For example, the risk level can be divided into three categories: low, medium and high. The advantage of the decision tree is that its results are easy to interpret and can clearly show the degree of influence of each factor on risk assessment. The setting of risk thresholds is a key link in the landslide early warning system. Usually, the determination of thresholds requires a combination of historical data and expert experience. For example, a region may use cumulative rainfall exceeding 200 mm and surface displacement rate exceeding 5 mm / day as thresholds for triggering early warnings. When the assessment result exceeds the threshold, the system will automatically send early warning information to the mobile devices of relevant departments or the emergency command center. Dynamic monitoring and updating of landslide risks are important means to ensure the effectiveness of the early warning system. The system needs to continuously receive and process new monitoring data and adjust the risk assessment results in real time. For example, during continuous rainfall, the system may update the risk assessment every hour to capture the risk change trend in time. Time series analysis methods can help predict the change of landslide risk in the future. By analyzing the change pattern of historical data, such as the ARIMA model, the risk trend in the next few days or weeks can be predicted. This prediction is of great significance for the formulation of medium- and long-term disaster prevention and mitigation strategies. The visualization of risk assessment results is crucial for decision makers to understand and respond to landslide risks. Common visualization methods include risk heat maps and trend curves. Risk heat maps can intuitively display the risk levels of different areas, using red, yellow, green and other colors to represent high, medium and low risks. Trend curves can show the changes in risk values ​​over time, helping decision makers to predict the development trend of risks. Through this comprehensive landslide early warning system, the prevention and response capabilities of landslide disasters can be greatly improved. The system can not only detect potential risks in a timely manner, but also provide decision makers with scientific decision-making basis, thereby effectively reducing casualties and property losses caused by landslide disasters. At the same time, the continuous operation of the system and data accumulation also provide valuable data support for further optimizing the early warning model and improving prediction accuracy.

[0118] Embodiment 2

[0119] like Figure 2 As shown, based on the same inventive concept, this embodiment also provides a landslide disaster monitoring and early warning system based on big data, including:

[0120] A data acquisition module, used to acquire meteorological element data and slope characteristic data;

[0121] The spatiotemporal alignment module is used to unify the meteorological element data and slope characteristic data to the same time and space scale through the spatiotemporal alignment algorithm to obtain the aligned data set;

[0122] Feature extraction module, used to extract features related to landslide warning from the aligned dataset;

[0123] The feature evaluation module is used to evaluate the importance of features using the random forest algorithm, determine the weights of the impact of meteorological element features and slope features on landslide occurrence, and obtain a feature weight matrix;

[0124] A relationship modeling module is used to establish an interaction relationship model between meteorological element characteristics and slope characteristics based on a feature weight matrix using a support vector machine algorithm to obtain an interaction relationship matrix;

[0125] The early warning calculation module is used to input the characteristic weight matrix and the interaction relationship matrix into the pre-established landslide early warning model, calculate the landslide early warning value and risk value at the current moment, and obtain the landslide early warning result.

[0126] The landslide disaster monitoring and early warning system based on big data provided in this embodiment has all the advantages of the landslide disaster monitoring and early warning method based on big data provided in the first embodiment.

[0127] Embodiment 3

[0128] This embodiment further discloses a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method described in the first embodiment.

[0129] Embodiment 4

[0130] This embodiment further discloses a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the method described in the first embodiment are implemented.

[0131] Embodiment 5

[0132] This embodiment also discloses a computer program product, including a computer program, which implements the steps of the method described in the first embodiment when executed by a processor.

[0133] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A landslide disaster monitoring and early warning method based on big data, characterized in that: include: Acquire meteorological element data and slope characteristic data, and unify the meteorological element data and the slope characteristic data to the same time and space scales through a spatiotemporal alignment algorithm to obtain an aligned data set; Extracting features related to landslide warning from the aligned data set, using a random forest algorithm to evaluate the importance of the features, determining the weights of the impact of meteorological element features and slope features on landslide occurrence, and obtaining a feature weight matrix; According to the feature weight matrix, a support vector machine algorithm is used to establish an interaction relationship model between the meteorological element features and the slope features to obtain an interaction relationship matrix; The characteristic weight matrix and the interaction relationship matrix are input into a pre-established landslide early warning model, and the landslide early warning value and risk value at the current moment are calculated to obtain a landslide early warning result.

2. The method according to claim 1, characterized in that: The process of obtaining meteorological element data and slope characteristic data includes: Real-time collection of meteorological element data including rainfall, temperature and humidity through meteorological monitoring stations or meteorological sensors; Slope characteristic data including geological structure, soil properties and vegetation cover are obtained through field investigation, geological drilling and geotechnical testing.

3. The method according to claim 1, characterized in that The process of unifying the meteorological element data and the slope characteristic data to the same time and space scale by using a time and space alignment algorithm includes: According to the spatiotemporal attribute information of the meteorological element data and the slope characteristic data, it is determined whether there are differences between the two types of data in spatiotemporal scale and data format; If there are differences, the time alignment algorithm is used for the time attributes of the meteorological element data and the slope characteristic data to align the two types of data in the time dimension and unify them to the same time scale; According to the spatial attributes of meteorological element data and slope characteristic data, a spatial alignment algorithm is used to align the two types of data in spatial dimensions and unify them to the same spatial scale. At the same time, in the process of spatiotemporal alignment, interpolation algorithms or resampling algorithms are selected according to the attribute characteristics of the data to estimate or convert missing or inconsistent data; through spatiotemporal alignment processing, meteorological element data and slope characteristic data with unified spatiotemporal scale and data format are obtained, and the aligned meteorological element data and slope characteristic data are fused using a data fusion algorithm to obtain the aligned data set.

4. The method according to claim 1, characterized in that: The process of extracting features related to landslide early warning from the aligned dataset includes: Based on the aligned datasets, the meteorological element characteristics and slope characteristics related to landslide warning are determined; According to the characteristics of meteorological elements, the time series data of rainfall, temperature and humidity are extracted respectively, and the changing trend of each element is determined by the time series analysis algorithm; According to the slope characteristics, the spatial distribution data of geological structure, soil properties and vegetation cover are obtained, and the stability and change of each feature are determined by using spatial analysis algorithms; According to the changing trend of meteorological element characteristics and the stability of slope characteristics, a landslide early warning index system is constructed, and the index weights are determined by comprehensively considering the impact of various characteristics on landslides.

5. The method according to claim 1, characterized in that The importance of the features is evaluated by using the random forest algorithm to determine the weights of the meteorological elements and slope characteristics on the occurrence of landslides. The process of obtaining the feature weight matrix includes: For the acquired meteorological element characteristic data and slope characteristic data, the principal component analysis method is used to reduce the dimension of the characteristics, and the meteorological element characteristics and slope characteristics after dimension reduction are obtained; The random forest algorithm is used to evaluate the importance of the meteorological elements and slope characteristics after dimensionality reduction, and the weight of the influence of the characteristics on the occurrence of landslides is obtained by calculating the Gini index of the characteristics. A feature weight matrix is ​​constructed according to the influence weights, and features in the feature weight matrix having weights greater than a preset threshold are determined as key features.

6. The method according to claim 1, characterized in that According to the feature weight matrix, a support vector machine algorithm is used to establish an interaction relationship model between the meteorological element features and the slope features. The process of obtaining the interaction relationship matrix includes: Inputting the feature weight matrix into a support vector machine algorithm, setting a kernel function and a penalty factor, and then training the interaction relationship model to obtain model parameters; According to the trained interaction relationship model, the meteorological element characteristics and slope characteristic data are input to predict the interaction relationship and obtain the interaction relationship matrix; The interaction relationship matrix is ​​analyzed to extract key interaction feature pairs, and the mutual influence degree between meteorological elements and slope characteristics is obtained according to the weight values ​​of the feature pairs.

7. The method according to claim 6, characterized in that According to the trained interaction relationship model, the process of inputting the meteorological element characteristics and slope characteristic data, predicting the interaction relationship and obtaining the interaction relationship matrix includes: Perform data cleaning, missing value processing and data normalization preprocessing operations on the meteorological element characteristic data and slope characteristic data to obtain a standardized characteristic data set; According to the standardized feature data set, a random forest algorithm is used to train the interaction relationship prediction model, and the model hyperparameters are optimized through grid search and cross-validation to obtain the model with the best performance; Apply the best performance model obtained through training to the new meteorological element characteristic data and slope characteristic data, predict the interaction relationship strength between the two, and obtain the interaction relationship matrix; Visualize the interaction relationship matrix, judge the interaction strength between different meteorological elements and slope characteristics according to the value of the elements in the matrix, and determine the key interaction relationship; According to the key interaction relationships, the association rule mining algorithm is used to analyze the association patterns between meteorological elements and slope characteristics, and the association rules of the built-in connection between meteorological elements and slope characteristics are obtained.

8. The method according to claim 1, characterized in that: The process of inputting the feature weight matrix and the interaction relationship matrix into a pre-established landslide early warning model, calculating the landslide early warning value and risk value at the current moment, and obtaining the landslide early warning result includes: Acquire monitoring data at the current moment, and calculate a feature weight matrix and an interaction relationship matrix according to the monitoring data; Input the characteristic weight matrix and interaction relationship matrix of the current monitoring data into the landslide early warning model; In the landslide early warning model, a support vector machine algorithm is used to train and predict the input matrix data to obtain the landslide early warning value at the current moment; In the landslide early warning model, a decision tree algorithm is used to classify and predict the input matrix data to obtain the landslide risk value at the current moment; The calculated landslide warning value is compared with the preset warning threshold. If the warning value is greater than or equal to the warning threshold, it is determined that there is a landslide risk and a warning signal is output; The calculated landslide risk value is compared with the preset risk level threshold to determine the current landslide risk level, output the risk level information, and generate a complete landslide warning result.

9. The method according to claim 8, characterized in that In the landslide early warning model, the decision tree algorithm is used to classify and predict the input matrix data, and the process of obtaining the landslide risk value at the current moment includes: The decision tree algorithm is used to classify and predict the input matrix data to obtain preliminary landslide risk assessment results; Compare the preliminary assessment result with the preset risk threshold to determine whether there is a landslide risk at the current moment; if the risk value at the current moment exceeds the preset threshold, trigger a landslide warning signal and send the warning information to relevant departments; After the early warning signal is triggered, the changes in the matrix data are continuously monitored and the landslide risk assessment results are dynamically updated. According to the changing trend of the risk assessment results, the time series analysis method is used to predict the landslide risk in the future. The risk assessment results and prediction information are visualized to generate intuitive landslide risk distribution maps and trend maps.

10. A landslide disaster monitoring and early warning system based on big data, characterized in that: include: A data acquisition module, used to acquire meteorological element data and slope characteristic data; A spatiotemporal alignment module, used to unify the meteorological element data and the slope characteristic data to the same time and space scales through a spatiotemporal alignment algorithm to obtain an aligned data set; A feature extraction module, used to extract features related to landslide warning from the aligned data set; A feature evaluation module is used to evaluate the importance of the features using a random forest algorithm, determine the weights of the impact of meteorological element features and slope features on the occurrence of landslides, and obtain a feature weight matrix; A relationship modeling module is used to establish an interaction relationship model between the meteorological element characteristics and the slope characteristics according to the feature weight matrix using a support vector machine algorithm to obtain an interaction relationship matrix; The early warning calculation module is used to input the characteristic weight matrix and the interaction relationship matrix into a pre-established landslide early warning model, calculate the landslide early warning value and risk value at the current moment, and obtain the landslide early warning result.

Images