A seismic landslide hazard zoning method based on subtractive clustering ANFIS

Abstract

This invention discloses a method for earthquake landslide hazard zoning based on subtractive clustering ANFIS, belonging to the field of earthquake landslide hazard zoning technology. The method includes the following steps: collecting earthquake landslide monitoring data including seismic data, topographic data, meteorological and hydrological data, and image data; extracting features from the image data to obtain feature-extracted data; based on the earthquake landslide monitoring data, using a subtractive clustering algorithm to determine the initial cluster centers of the earthquake landslide area, thereby obtaining initial clustering information for the earthquake landslide; and combining the earthquake landslide hazard levels and their monitoring data within each geographic grid unit to draw an earthquake landslide hazard zoning map. The data acquisition technology, subtractive clustering algorithm application technology, ANFIS model construction technology, and data calibration technology in this method are closely integrated with modern information technology. The hazard calibration coefficient is used to calibrate the hazard index, improving the accuracy of earthquake landslide hazard assessment.

Description

Technical Field

[0001] This invention relates to the field of earthquake landslide hazard zoning technology, specifically to an earthquake landslide hazard zoning method based on subtractive clustering ANFIS. Background Technology

[0002] Earthquake landslides, as a highly destructive geological hazard, seriously threaten human life and property. In areas with frequent plate activity, earthquakes and landslides often occur together, causing huge losses to the local area. Early assessments of earthquake landslide hazard relied heavily on empirical qualitative analysis and simple mathematical models, which were difficult to accurately consider the complex interactions of multiple factors such as earthquakes, topography, geology, and meteorology. The accuracy and reliability of the assessment results were poor. With the rise of GIS and machine learning technologies, technical support has been provided to improve the accuracy of earthquake landslide hazard assessments. By fully exploring the potential information of multi-source earthquake landslide data, accurate identification of earthquake landslide risks can be achieved, providing a more targeted and reliable basis for earthquake landslide prevention and planning, and offering solutions to reduce the losses caused by earthquake landslides.

[0003] Although existing technologies have made significant progress in earthquake landslide hazard zoning, some issues still need optimization. Traditional earthquake landslide hazard zoning methods rely on historical data samples, requiring high completeness and accuracy of the historical data. When faced with complex geological conditions and earthquake interactions, they struggle to accurately describe the nonlinear relationships involved. Furthermore, machine learning methods for earthquake landslide hazard zoning also have limitations. Therefore, this paper proposes an earthquake landslide hazard zoning method based on subtractive clustering ANFIS to address the limitations of traditional earthquake landslide hazard zoning methods and achieve effective hazard zoning of earthquake landslides. Summary of the Invention

[0004] To achieve the above objectives, the present invention provides the following technical solution: a seismic landslide hazard zoning method based on subtractive clustering ANFIS, comprising the following steps:

[0005] Step 1: Collect earthquake and landslide monitoring data, including earthquake data, topographic data, meteorological and hydrological data, and image data, to provide a data foundation for subsequent steps.

[0006] Step 2: Feature extraction is performed on the image data to obtain feature extraction data, which lays the foundation for correcting the earthquake landslide hazard index;

[0007] Step 3: Based on earthquake landslide monitoring data, a subtractive clustering algorithm is used to determine the initial cluster centers of the earthquake landslide area, thereby obtaining the initial clustering information of the earthquake landslide and providing data preparation for building the ANFIS model;

[0008] Step 4: Based on the acquired initial clustering information of earthquake landslides, construct an ANFIS model, train the ANFIS model, and obtain the final sub-clustering ANFIS model.

[0009] Step 5: Divide the earthquake landslide monitoring area into various geographic grids, assign corresponding earthquake landslide monitoring data to each geographic grid unit, and obtain the earthquake landslide hazard index for each geographic grid unit by combining the subtractive clustering ANFIS model.

[0010] Step 6: Combining meteorological and hydrological data, feature extraction data, and neural network algorithms, construct an earthquake landslide calibration model to obtain the hazard calibration coefficient;

[0011] Step 7: Using the hazard calibration coefficient, calibrate the earthquake landslide hazard index of each geographic grid unit and classify the earthquake landslide hazard level within each geographic grid unit.

[0012] Step 8: Combine the earthquake landslide hazard levels and earthquake landslide monitoring data within each geographic grid unit to draw an earthquake landslide hazard zoning map.

[0013] A further improvement to the technical solution of this invention lies in the following: In step one, the process of collecting earthquake landslide monitoring data includes:

[0014] Different types of data acquisition equipment are deployed, combined with geological monitoring stations, data acquisition equipment and regional databases, to collect seismic data, topographic data, meteorological and hydrological data and image data of the earthquake and landslide monitoring area. The data acquisition equipment includes total station, altitude measuring instrument, temperature sensor, ultrasonic anemometer, optical rain gauge, float-type water level gauge, electromagnetic flow meter, sediment concentration meter, electromagnetic current meter and optical camera.

[0015] Specifically, earthquake data is collected through geological monitoring stations; topographic data, meteorological and hydrological data, and image data are collected through acquisition equipment.

[0016] The earthquake data includes earthquake intensity, earthquake frequency, earthquake duration, peak ground acceleration, and epicentral distance; the topographic data includes vegetation cover, lithology, soil composition, slope, altitude, plan curvature, and profile curvature; the meteorological and hydrological data includes air temperature, wind speed, precipitation, and the water level, flow rate, sediment content, and flow velocity of surrounding water bodies; the image data consists of remote sensing images of the earthquake-landslide monitoring area.

[0017] Data cleaning and standardization were performed on the collected earthquake data, topographic data, and meteorological and hydrological data. Image enhancement and denoising were performed on the collected image data. The preprocessed earthquake data, topographic data, and meteorological and hydrological data were integrated to generate an earthquake landslide monitoring dataset. The earthquake landslide monitoring dataset was divided into a training set and a test set with a ratio of 7:3.

[0018] A further improvement to the technical solution of this invention lies in the fact that, in step two, the process of extracting feature data includes:

[0019] Remote sensing images of the earthquake-slip monitoring area were imported into ENVI software. Using RNVI software, gray values ​​of the green band and near-infrared band were extracted from the remote sensing images of the earthquake-slip monitoring area. Then, the gray values ​​of the green band and near-infrared band were converted into reflectance values ​​of the green band and near-infrared band, respectively.

[0020] Based on the reflectance values ​​in the green and near-infrared bands, the normalized water index is calculated using the multispectral index method. The calculation formula is as follows: ;

[0021] Set a normalized water index threshold, and based on the normalized water index threshold, identify water areas and non-water areas within the earthquake landslide monitoring area;

[0022] Using GIS software, the coordinates of each point within the earthquake landslide monitoring area are obtained. Taking the boundary point of the earthquake landslide monitoring area as the starting point and the boundary point of the water body as the ending point, the distance from the starting point to the ending point is calculated using the Euclidean distance formula. This distance is then used as feature extraction data, and the obtained feature extraction data is integrated into the earthquake landslide monitoring dataset. The calculation process is as follows:

[0023]

[0024] in, The distance between the earthquake and landslide monitoring area and the surrounding water area; , ( ) represents the coordinates of the boundary point of the earthquake-slip monitoring area, i.e., the coordinates of the starting point; , ) represents the coordinates of the boundary point of the water body, i.e., the endpoint coordinates.

[0025] A further improvement to the technical solution of this invention lies in the following: In step three, the process of obtaining the initial clustering information of earthquake landslides includes:

[0026] The initial clustering information of earthquake landslides includes earthquake data, topographic data, and meteorological and hydrological data of the initial cluster centers of earthquake landslides;

[0027] Earthquake landslide monitoring data points are set up, and the earthquake landslide monitoring area is divided into several small monitoring areas. Each small monitoring area is represented by an earthquake landslide monitoring data point.

[0028] Based on the seismic data, topographic data, and meteorological and hydrological data of the earthquake landslide monitoring area, obtain the seismic data, topographic data, and meteorological and hydrological data of the corresponding earthquake landslide monitoring data points;

[0029] The seismic data, topographic data, and meteorological and hydrological data of each earthquake and landslide monitoring data point are integrated to generate a feature vector of the earthquake and landslide monitoring data point.

[0030] The density of each earthquake landslide monitoring data point is calculated using a subtractive clustering algorithm. The earthquake landslide monitoring data point with the highest density is taken as the first cluster center of the earthquake landslide, and the density of the earthquake landslide monitoring data points is updated.

[0031] The density of the updated earthquake landslide monitoring data points is compared with the density of the first cluster center point of the earthquake landslide. If the density of the updated earthquake landslide monitoring data points is less than the density of the first cluster center point of the earthquake landslide, then the first cluster center point of the earthquake landslide is used as the initial cluster center point of the earthquake landslide. If the density of the updated earthquake landslide monitoring data points is greater than the density of the first cluster center point of the earthquake landslide, then the updated earthquake landslide monitoring data points are used as the initial cluster center point of the earthquake landslide. The density of the earthquake landslide monitoring data points is updated repeatedly until the earthquake landslide monitoring data point with the highest density is obtained, and this is used as the initial cluster center point of the earthquake landslide.

[0032] The calculation process for the density of earthquake-slip monitoring data points and the density of updated earthquake-slip monitoring data points includes:

[0033]

[0034]

[0035] in, Density of earthquake and landslide monitoring data points; To update the density of earthquake and landslide monitoring data points; and These are the feature vectors of earthquake landslide monitoring data point i and earthquake landslide monitoring data point j, respectively. and For parameters affecting the range of influence; The feature vector of the h-th cluster center point; The density value of the h-th cluster center;

[0036] By using the feature vectors of earthquake landslide monitoring data points, we can obtain the earthquake data, topographic data, and meteorological and hydrological data of the initial cluster center points of earthquake landslides.

[0037] A further improvement to the technical solution of this invention lies in the following: In step four, the construction process of the ANFIS model includes:

[0038] The initial clustering information of earthquake landslides is used as the input variable of the ANFIS model. A fuzzy membership function and a fuzzy set are defined for each input variable. The actual input variable data are substituted into the corresponding fuzzy membership function to transform each input variable into a fuzzy membership value, thereby realizing the fuzzification of the input data.

[0039] Specifically, for earthquake intensity, the corresponding fuzzy set is defined as low intensity, medium intensity, and high intensity; for earthquake frequency, the corresponding fuzzy set is defined as low frequency, medium frequency, and high frequency; for earthquake duration, the corresponding fuzzy set is defined as very short, short, medium, long, and very long; for peak ground acceleration, the corresponding fuzzy set is defined as very small, small, medium, relatively large, and very large; for epicentral distance, the corresponding fuzzy set is defined as very near, near, relatively near, relatively far, and very far; for slope, the corresponding fuzzy set is defined as very gentle, gentle, relatively steep, steep, and very steep; for elevation, the corresponding fuzzy set is defined as very low, low, medium, relatively high, and very high; for plane curvature, the corresponding fuzzy set is defined as very concave, concave, nearly horizontal, and... For convex and very convex; for profile curvature, the corresponding fuzzy set is defined as very concave, concave, nearly horizontal, convex, and very convex; for air temperature, the corresponding fuzzy set is defined as very cold, cold, moderate, hot, and very hot; for wind speed, the corresponding fuzzy set is defined as very small, small, moderate, large, and very large; for rainfall, the corresponding fuzzy set is defined as light rain, moderate rain, heavy rain, and torrential rain; for water level, the corresponding fuzzy set is defined as very low, relatively low, moderate, relatively high, and high; for water flow, the corresponding fuzzy set is defined as very small, small, moderate, large, and very large; for water sediment content, the corresponding fuzzy set is defined as very little, little, moderate, much, and much; for water velocity, the corresponding fuzzy set is defined as very slow, slow, moderate, fast, and very fast.

[0040] Based on the initial cluster centers of earthquake landslides obtained by the subtractive clustering algorithm, the number of fuzzy rules is determined. By combining each input variable with the corresponding fuzzy set logically, and using AND and OR relations, a fuzzy rule base is constructed.

[0041] Based on the fuzzy rule base, inference operations are performed on the fuzzification results of the input variables to obtain the output results of each fuzzy rule. The outputs of all fuzzy rules are then combined to obtain the total fuzzy rule output results.

[0042] The total fuzzy rule output is defuzzified using the maximum membership degree method. The universe of discourse value corresponding to the maximum membership degree is selected as the output value of the ANFIS model, and then the ANFIS model is constructed. The universe of discourse value corresponding to the maximum membership degree is the interval of the fuzzy membership degree function when the membership degree of the input variable in the fuzzy set reaches the maximum value. The output value of the ANFIS model is the earthquake landslide hazard index.

[0043] A further improvement to the technical solution of this invention lies in the following: Step four, the process of training the ANFIS model and obtaining the reduced-clustering ANFIS model, includes:

[0044] Earthquake data, topographic data, and meteorological and hydrological data are extracted from the earthquake and landslide monitoring dataset. Using the training set data and the backpropagation algorithm, the earthquake data, topographic data, and meteorological and hydrological data are used as inputs, and the earthquake landslide hazard index is used as the output. The nonlinear relationship between earthquake data, topographic data, meteorological and hydrological data and the earthquake landslide hazard index is learned, and the ANFIS model is trained.

[0045] The test set data is input into the ANFIS model, and the output results of the ANFIS model are compared with the actual earthquake landslide hazard index. The parameters of the ANFIS model are adjusted, the ANFIS model is optimized, and a reduced clustering ANFIS model is obtained.

[0046] A further improvement to the technical solution of this invention lies in the following: In step five, the process of obtaining the earthquake landslide hazard index for each geographic grid unit includes:

[0047] Based on the small monitoring areas corresponding to the earthquake landslide monitoring data points, the earthquake landslide monitoring area is divided into various geographical grids, and then the earthquake landslide monitoring data of each geographical grid unit is obtained.

[0048] By combining earthquake data, topographic data, meteorological and hydrological data, and the reduced clustering ANFIS model, the earthquake landslide hazard index of each geographic grid unit is obtained.

[0049] A further improvement to the technical solution of this invention lies in the following: In step six, the process of obtaining the hazard calibration coefficient includes:

[0050] Feature extraction data, water level, flow rate, sediment content and flow velocity of water bodies around the earthquake landslide monitoring area are extracted from the earthquake landslide monitoring dataset. The vegetation coverage, lithology and soil composition of the monitoring area are comprehensively analyzed as input parameters. The extracted input parameters are integrated to construct the input dataset.

[0051] By combining the input dataset with the multiple linear regression algorithm, the extracted input parameters are used as input data, and the hazard calibration coefficient is used as output data. The linear relationship between the input data and the output data is learned to train the earthquake landslide calibration model.

[0052] Input the input dataset into the earthquake landslide calibration model, adjust the intercept term and regression coefficient of the earthquake landslide calibration model to optimize the performance of the earthquake landslide calibration model, deploy the optimized earthquake landslide calibration model into the system, and output the corresponding hazard calibration coefficients based on the current input parameters;

[0053] The expression for the earthquake-induced landslide calibration model is as follows:

[0054]

[0055] in, This is the hazard calibration factor; , , , These are the regression coefficients for each input parameter; , , These are the respective input parameters; and These are the intercept term and error term of the earthquake landslide calibration model, respectively.

[0056] A further improvement to the technical solution of this invention lies in the following: Step seven, the process of calibrating the earthquake landslide hazard index of each geographic grid unit and classifying the earthquake landslide hazard level within each geographic grid unit, includes:

[0057] The seismic landslide hazard index of each geographic grid unit is calibrated using the hazard calibration coefficient, and the seismic landslide hazard index of each geographic grid unit after calibration is obtained.

[0058] A first threshold and a second threshold for the earthquake landslide hazard index are set. When the earthquake landslide hazard index of each geographic grid unit is lower than the first threshold, the corresponding geographic grid unit is classified as having a low earthquake landslide hazard level. When the earthquake landslide hazard index of each geographic grid unit is between the first threshold and the second threshold, the corresponding geographic grid unit is classified as having a medium earthquake landslide hazard level. When the earthquake landslide hazard index of each geographic grid unit is higher than the first threshold, the corresponding geographic grid unit is classified as having a high earthquake landslide hazard level.

[0059] A further improvement to the technical solution of this invention lies in the following: In step eight, the process of drawing the earthquake landslide hazard zoning map includes:

[0060] The earthquake landslide hazard levels within each geographic grid unit are imported into the GIS software. Using the GIS software, green, yellow, and red icons are set for low, medium, and high earthquake landslide hazard levels, respectively, thereby obtaining the earthquake landslide hazard level layer.

[0061] The seismic data, topographic data, and meteorological and hydrological data of each geographic grid unit are converted into raster data. The raster data is then visualized, and the seismic data layer, topographic data layer, and meteorological and hydrological data layer are obtained respectively.

[0062] An earthquake landslide hazard zoning map is drawn by overlaying earthquake landslide hazard level layers, earthquake data layers, topographic data layers, and meteorological and hydrological data layers.

[0063] The beneficial effects of this invention are as follows: The earthquake landslide hazard zoning method based on subtractive clustering ANFIS, compared to traditional earthquake landslide hazard assessment methods, integrates data acquisition technology, subtractive clustering algorithm application technology, ANFIS model construction technology, and data calibration technology with modern information technology. This method accurately captures earthquake data, topographic data, meteorological and hydrological data, and image data, thereby obtaining initial clustering information, hazard index, and hazard calibration coefficients for earthquake landslides. Based on the initial clustering information, an ANFIS model is constructed. The model uses a hazard calibration coefficient to calibrate the hazard index, thereby achieving a more accurate assessment of earthquake landslide hazard. This enables effective monitoring during the earthquake landslide hazard assessment process and solves the problem that traditional earthquake landslide hazard zoning methods rely on historical data samples, have high requirements for the completeness and accuracy of historical data, and struggle to address the nonlinear relationship between complex geological conditions and seismic action. This ensures that the method in this invention can refine the dynamic monitoring standards for earthquake landslide hazard within a more precise range, making the monitored data a more accurate indicator under the same conditions. The development and application of this method significantly enhances the intelligence level of the earthquake landslide hazard assessment process. Attached Figure Description

[0064] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0065] Figure 1 This is a flowchart of an earthquake landslide hazard zoning method based on subtractive clustering ANFIS according to the present invention. Detailed Implementation

[0066] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0067] like Figure 1 As shown, this invention provides a seismic landslide hazard zoning method based on subtractive clustering ANFIS, which consists of the following steps:

[0068] Step 1: Collect earthquake and landslide monitoring data, including earthquake data, topographic data, meteorological and hydrological data, and image data, to provide a data foundation for subsequent steps.

[0069] Step 2: Feature extraction is performed on the image data to obtain feature extraction data, which lays the foundation for correcting the earthquake landslide hazard index;

[0070] Step 3: Based on earthquake landslide monitoring data, a subtractive clustering algorithm is used to determine the initial cluster centers of the earthquake landslide area, thereby obtaining the initial clustering information of the earthquake landslide and providing data preparation for building the ANFIS model;

[0071] Step 4: Based on the acquired initial clustering information of earthquake landslides, construct an ANFIS model, train the ANFIS model, and obtain the final sub-clustering ANFIS model.

[0072] Step 5: Divide the earthquake landslide monitoring area into various geographic grids, assign corresponding earthquake landslide monitoring data to each geographic grid unit, and obtain the earthquake landslide hazard index for each geographic grid unit by combining the subtractive clustering ANFIS model.

[0073] Step 6: Combining meteorological and hydrological data, feature extraction data, and neural network algorithms, construct an earthquake landslide calibration model to obtain the hazard calibration coefficient;

[0074] Step 7: Using the hazard calibration coefficient, calibrate the earthquake landslide hazard index of each geographic grid unit and classify the earthquake landslide hazard level within each geographic grid unit.

[0075] Step 8: Combine the earthquake landslide hazard levels and earthquake landslide monitoring data within each geographic grid unit to draw an earthquake landslide hazard zoning map.

[0076] Step one, the process of collecting earthquake landslide monitoring data includes:

[0077] Different types of data acquisition equipment are deployed, combined with geological monitoring stations, data acquisition equipment and regional databases, to collect seismic data, topographic data, meteorological and hydrological data and image data of the earthquake and landslide monitoring area. The data acquisition equipment includes total station, altitude measuring instrument, temperature sensor, ultrasonic anemometer, optical rain gauge, float-type water level gauge, electromagnetic flow meter, sediment concentration meter, electromagnetic current meter and optical camera.

[0078] Specifically, earthquake data is collected through geological monitoring stations; topographic data, meteorological and hydrological data, and image data are collected through acquisition equipment.

[0079] Specifically, the earthquake data includes earthquake intensity, earthquake frequency, earthquake duration, peak ground acceleration, and epicentral distance; the topographic data includes vegetation cover, lithology, soil composition, slope, altitude, plan curvature, and profile curvature; the meteorological and hydrological data includes air temperature, wind speed, precipitation, and the water level, flow rate, sediment content, and flow velocity of surrounding water bodies; and the image data consists of remote sensing images of the earthquake-landslide monitoring area.

[0080] Data cleaning and standardization were performed on the collected earthquake data, topographic data, and meteorological and hydrological data. Image enhancement and denoising were performed on the collected image data. The preprocessed earthquake data, topographic data, and meteorological and hydrological data were integrated to generate an earthquake landslide monitoring dataset. The earthquake landslide monitoring dataset was divided into a training set and a test set with a ratio of 7:3.

[0081] Step two, the process of feature extraction data includes:

[0082] Specifically, remote sensing images of the earthquake landslide monitoring area are imported into ENVI software. Using RNVI software, gray values ​​of the green band and near-infrared band are extracted from the remote sensing images of the earthquake landslide monitoring area, respectively. Then, the gray values ​​of the green band and near-infrared band are converted into reflectance values ​​of the green band and near-infrared band, respectively.

[0083] Based on the reflectance values ​​in the green and near-infrared bands, the normalized water index is calculated using the multispectral index method. The calculation formula is as follows: ;

[0084] Set a normalized water index threshold, and based on the normalized water index threshold, identify water areas and non-water areas within the earthquake landslide monitoring area;

[0085] Using GIS software, the coordinates of each point within the earthquake landslide monitoring area are obtained. Taking the boundary point of the earthquake landslide monitoring area as the starting point and the boundary point of the water body as the ending point, the distance from the starting point to the ending point is calculated using the Euclidean distance formula. This distance is then used as feature extraction data, and the obtained feature extraction data is integrated into the earthquake landslide monitoring dataset. The calculation process is as follows:

[0086]

[0087] in, The distance between the earthquake and landslide monitoring area and the surrounding water area; , ( ) represents the coordinates of the boundary point of the earthquake-slip monitoring area, i.e., the coordinates of the starting point; , ) represents the coordinates of the boundary point of the water body, i.e., the endpoint coordinates.

[0088] Step three involves obtaining the initial clustering information for earthquake-induced landslides, including:

[0089] The initial clustering information for earthquake-induced landslides includes seismic data, topographic data, and meteorological and hydrological data of the initial cluster centers of earthquake-induced landslides.

[0090] Earthquake landslide monitoring data points are set up, and the earthquake landslide monitoring area is divided into several small monitoring areas. Each small monitoring area is represented by an earthquake landslide monitoring data point.

[0091] Based on the seismic data, topographic data, and meteorological and hydrological data of the earthquake landslide monitoring area, obtain the seismic data, topographic data, and meteorological and hydrological data of the corresponding earthquake landslide monitoring data points;

[0092] The seismic data, topographic data, and meteorological and hydrological data of each earthquake and landslide monitoring data point are integrated to generate a feature vector of the earthquake and landslide monitoring data point.

[0093] The density of each earthquake landslide monitoring data point is calculated using a subtractive clustering algorithm. The earthquake landslide monitoring data point with the highest density is taken as the first cluster center of the earthquake landslide, and the density of the earthquake landslide monitoring data points is updated.

[0094] The density of the updated earthquake landslide monitoring data points is compared with the density of the first cluster center point of the earthquake landslide. If the density of the updated earthquake landslide monitoring data points is less than the density of the first cluster center point of the earthquake landslide, then the first cluster center point of the earthquake landslide is used as the initial cluster center point of the earthquake landslide. If the density of the updated earthquake landslide monitoring data points is greater than the density of the first cluster center point of the earthquake landslide, then the updated earthquake landslide monitoring data points are used as the initial cluster center point of the earthquake landslide. The density of the earthquake landslide monitoring data points is updated repeatedly until the earthquake landslide monitoring data point with the highest density is obtained, and this is used as the initial cluster center point of the earthquake landslide.

[0095] The calculation process for the density of earthquake-slip monitoring data points and the density of updated earthquake-slip monitoring data points includes:

[0096]

[0097]

[0098] in, Density of earthquake and landslide monitoring data points; To update the density of earthquake and landslide monitoring data points; and These are the feature vectors of earthquake landslide monitoring data point i and earthquake landslide monitoring data point j, respectively. and For parameters affecting the range of influence; The feature vector of the h-th cluster center point; The density value of the h-th cluster center;

[0099] By using the feature vectors of earthquake landslide monitoring data points, we can obtain the earthquake data, topographic data, and meteorological and hydrological data of the initial cluster center points of earthquake landslides.

[0100] Step four, the construction process of the ANFIS model includes:

[0101] The initial clustering information of earthquake landslides is used as the input variable of the ANFIS model. A fuzzy membership function and a fuzzy set are defined for each input variable. The actual input variable data are substituted into the corresponding fuzzy membership function to transform each input variable into a fuzzy membership value, thereby realizing the fuzzification of the input data.

[0102] Specifically, for earthquake intensity, the corresponding fuzzy set is defined as low intensity, medium intensity, and high intensity; for earthquake frequency, the corresponding fuzzy set is defined as low frequency, medium frequency, and high frequency; for earthquake duration, the corresponding fuzzy set is defined as very short, short, medium, long, and very long; for peak ground acceleration, the corresponding fuzzy set is defined as very small, small, medium, relatively large, and very large; for epicentral distance, the corresponding fuzzy set is defined as very near, near, relatively near, relatively far, and very far; for slope, the corresponding fuzzy set is defined as very gentle, gentle, relatively steep, steep, and very steep; for elevation, the corresponding fuzzy set is defined as very low, low, medium, relatively high, and very high; for plane curvature, the corresponding fuzzy set is defined as very concave, concave, nearly horizontal, and... For convex and very convex; for profile curvature, the corresponding fuzzy set is defined as very concave, concave, nearly horizontal, convex, and very convex; for air temperature, the corresponding fuzzy set is defined as very cold, cold, moderate, hot, and very hot; for wind speed, the corresponding fuzzy set is defined as very small, small, moderate, large, and very large; for rainfall, the corresponding fuzzy set is defined as light rain, moderate rain, heavy rain, and torrential rain; for water level, the corresponding fuzzy set is defined as very low, relatively low, moderate, relatively high, and high; for water flow, the corresponding fuzzy set is defined as very small, small, moderate, large, and very large; for water sediment content, the corresponding fuzzy set is defined as very little, little, moderate, much, and much; for water velocity, the corresponding fuzzy set is defined as very slow, slow, moderate, fast, and very fast.

[0103] Based on the initial cluster centers of earthquake landslides obtained by the subtractive clustering algorithm, the number of fuzzy rules is determined. By combining each input variable with the corresponding fuzzy set logically, and using AND and OR relations, a fuzzy rule base is constructed.

[0104] Based on the fuzzy rule base, inference operations are performed on the fuzzification results of the input variables to obtain the output results of each fuzzy rule. The outputs of all fuzzy rules are then combined to obtain the total fuzzy rule output results.

[0105] The total fuzzy rule output is defuzzified using the maximum membership method. The universe of discourse value corresponding to the maximum membership is selected as the output value of the ANFIS model, and then the ANFIS model is constructed. The universe of discourse value corresponding to the maximum membership is the interval of the fuzzy membership function when the membership of the input variable in the fuzzy set reaches the maximum value. The output value of the ANFIS model is the earthquake landslide hazard index.

[0106] Step four, the process of training the ANFIS model and obtaining the subtractive clustering ANFIS model, includes:

[0107] Earthquake data, topographic data, and meteorological and hydrological data are extracted from the earthquake and landslide monitoring dataset. Using the training set data and the backpropagation algorithm, the earthquake data, topographic data, and meteorological and hydrological data are used as inputs, and the earthquake landslide hazard index is used as the output. The nonlinear relationship between earthquake data, topographic data, meteorological and hydrological data and the earthquake landslide hazard index is learned, and the ANFIS model is trained.

[0108] The test set data is input into the ANFIS model, and the output results of the ANFIS model are compared with the actual earthquake landslide hazard index. The parameters of the ANFIS model are adjusted, the ANFIS model is optimized, and a reduced clustering ANFIS model is obtained.

[0109] Step five involves obtaining the earthquake landslide hazard index for each geographic grid cell, including:

[0110] Based on the small monitoring areas corresponding to the earthquake landslide monitoring data points, the earthquake landslide monitoring area is divided into various geographical grids, and then the earthquake landslide monitoring data of each geographical grid unit is obtained.

[0111] By combining earthquake data, topographic data, meteorological and hydrological data, and the reduced clustering ANFIS model, the earthquake landslide hazard index of each geographic grid unit is obtained.

[0112] Step six, the process of obtaining the hazard calibration coefficient, includes:

[0113] Feature extraction data, water level, flow rate, sediment content and flow velocity of water bodies around the earthquake landslide monitoring area are extracted from the earthquake landslide monitoring dataset. The vegetation coverage, lithology and soil composition of the monitoring area are comprehensively analyzed as input parameters. The extracted input parameters are integrated to construct the input dataset.

[0114] By combining the input dataset with the multiple linear regression algorithm, the extracted input parameters are used as input data, and the hazard calibration coefficient is used as output data. The linear relationship between the input data and the output data is learned to train the earthquake landslide calibration model.

[0115] Input the input dataset into the earthquake landslide calibration model, adjust the intercept term and regression coefficient of the earthquake landslide calibration model to optimize the performance of the earthquake landslide calibration model, deploy the optimized earthquake landslide calibration model into the system, and output the corresponding hazard calibration coefficients based on the current input parameters;

[0116] The expression for the earthquake-induced landslide calibration model is as follows:

[0117]

[0118] in, This is the hazard calibration factor; , , , These are the regression coefficients for each input parameter; , , These are the respective input parameters; and These are the intercept term and error term of the earthquake landslide calibration model, respectively.

[0119] Step seven, which involves calibrating the earthquake landslide hazard index for each geographic grid cell and classifying the earthquake landslide hazard level within each geographic grid cell, includes:

[0120] The seismic landslide hazard index of each geographic grid unit is calibrated using the hazard calibration coefficient, and the seismic landslide hazard index of each geographic grid unit after calibration is obtained.

[0121] A first threshold and a second threshold for the earthquake landslide hazard index are set. When the earthquake landslide hazard index of each geographic grid unit is lower than the first threshold, the corresponding geographic grid unit is classified as having a low earthquake landslide hazard level. When the earthquake landslide hazard index of each geographic grid unit is between the first threshold and the second threshold, the corresponding geographic grid unit is classified as having a medium earthquake landslide hazard level. When the earthquake landslide hazard index of each geographic grid unit is higher than the first threshold, the corresponding geographic grid unit is classified as having a high earthquake landslide hazard level.

[0122] Step eight, the process of drawing the earthquake landslide hazard zoning map, includes:

[0123] The earthquake landslide hazard levels within each geographic grid unit are imported into the GIS software. Using the GIS software, green, yellow, and red icons are set for low, medium, and high earthquake landslide hazard levels, respectively, thereby obtaining the earthquake landslide hazard level layer.

[0124] The seismic data, topographic data, and meteorological and hydrological data of each geographic grid unit are converted into raster data. The raster data is then visualized, and the seismic data layer, topographic data layer, and meteorological and hydrological data layer are obtained respectively.

[0125] An earthquake landslide hazard zoning map is drawn by overlaying earthquake landslide hazard level layers, earthquake data layers, topographic data layers, and meteorological and hydrological data layers.

[0126] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for seismic landslide hazard zonation based on reduced clustering ANFIS, characterized by: Includes the following steps: Step 1: Collect earthquake and landslide monitoring data, including earthquake data, topographic data, meteorological and hydrological data, and image data; Step 2: Extract features from the image data to obtain feature extraction data; Step 3: Based on earthquake landslide monitoring data, a subtractive clustering algorithm is used to determine the initial cluster centers of the earthquake landslide area, thereby obtaining the initial clustering information of the earthquake landslide. Step 4: Based on the acquired initial clustering information of earthquake landslides, construct an ANFIS model, train the ANFIS model, and obtain the final sub-clustering ANFIS model. Step 5: Divide the earthquake landslide monitoring area into various geographic grids, assign corresponding earthquake landslide monitoring data to each geographic grid unit, and obtain the earthquake landslide hazard index for each geographic grid unit by combining the subtractive clustering ANFIS model. Step 6: Combining meteorological and hydrological data, feature extraction data, and neural network algorithms, construct an earthquake landslide calibration model to obtain the hazard calibration coefficient; Step 7: Using the hazard calibration coefficient, calibrate the earthquake landslide hazard index of each geographic grid unit and classify the earthquake landslide hazard level within each geographic grid unit. Step 8: Based on the earthquake landslide hazard levels and earthquake landslide monitoring data within each geographic grid unit, draw an earthquake landslide hazard zoning map; In step one, the data acquisition process for earthquake landslide monitoring includes: deploying different types of acquisition equipment, combining geological monitoring stations, acquisition equipment, and regional databases to collect earthquake data, topographic data, meteorological and hydrological data, and image data of the earthquake landslide monitoring area. The acquisition equipment includes a total station, an altimeter, a temperature sensor, an ultrasonic anemometer, an optical rain gauge, a float-type water level gauge, an electromagnetic flowmeter, a sediment concentration meter, an electromagnetic current meter, and an optical camera. The earthquake data includes earthquake intensity, earthquake frequency, earthquake duration, peak ground acceleration, and epicentral distance. The topographic data includes vegetation data. The data includes: coverage, lithology, soil composition, slope, altitude, planar curvature, and profile curvature; meteorological and hydrological data including air temperature, wind speed, precipitation, and water level, flow rate, sediment content, and flow velocity of surrounding water bodies; and image data consisting of remote sensing images of the earthquake-landslide monitoring area. The collected earthquake data, topographic data, and meteorological and hydrological data are cleaned and standardized. The collected image data undergoes image enhancement and denoising. The preprocessed earthquake data, topographic data, and meteorological and hydrological data are integrated to generate an earthquake-landslide monitoring dataset, which is then divided into a training set and a test set. In step two, the feature extraction process includes: importing remote sensing images of the earthquake landslide monitoring area into ENVI software; using ENVI software to extract grayscale values ​​of the green band and near-infrared band from the remote sensing images of the earthquake landslide monitoring area; then converting the grayscale values ​​of the green band and near-infrared band into reflectance values ​​of the green band and near-infrared band, respectively; based on the reflectance values ​​of the green band and near-infrared band, calculating the normalized water index using the multispectral index method; setting a threshold for the normalized water index, and identifying water and non-water areas within the earthquake landslide monitoring area based on the threshold; obtaining the coordinates of each point within the earthquake landslide monitoring area using GIS software; using the boundary point of the earthquake landslide monitoring area as the starting point and the boundary point of the water body as the ending point, calculating the distance from the starting point to the ending point using the Euclidean distance formula, which is the distance between the earthquake landslide monitoring area and the surrounding water body area, and then using this distance as feature extraction data, and integrating the obtained feature extraction data into the earthquake landslide monitoring dataset. In step three, the process of obtaining the initial clustering information of earthquake landslides includes: the initial clustering information of earthquake landslides includes seismic data, topographic data, and meteorological and hydrological data of the initial cluster center points of earthquake landslides; setting earthquake landslide monitoring data points, dividing the earthquake landslide monitoring area into several small monitoring areas, and representing each small monitoring area by each earthquake landslide monitoring data point; obtaining the seismic data, topographic data, and meteorological and hydrological data of the corresponding earthquake landslide monitoring data points based on the seismic data, topographic data, and meteorological and hydrological data of the earthquake landslide monitoring area; integrating the seismic data, topographic data, and meteorological and hydrological data within each earthquake landslide monitoring data point to generate a feature vector of the earthquake landslide monitoring data point; calculating the density of each earthquake landslide monitoring data point using a subtraction clustering algorithm, and selecting the earthquake landslide monitoring data point with the highest density as the cluster center point. The first cluster center point of the earthquake-slip landslide is determined, and the density of the earthquake-slip landslide monitoring data points is updated. The updated density of the earthquake-slip landslide monitoring data points is compared with the density of the first cluster center point. If the updated density of the earthquake-slip landslide monitoring data points is less than the density of the first cluster center point, then the first cluster center point is used as the initial cluster center point. If the updated density of the earthquake-slip landslide monitoring data points is greater than the density of the first cluster center point, then the updated earthquake-slip landslide monitoring data points are used as the initial cluster center point. The density of the earthquake-slip landslide monitoring data points is updated repeatedly until the earthquake-slip landslide monitoring data point with the highest density is obtained, and this is used as the initial cluster center point. The earthquake data, topographic data, and meteorological and hydrological data of the initial cluster center point are obtained through the feature vector of the earthquake-slip landslide monitoring data points. The calculation process for the density of earthquake-slip monitoring data points and the density of updated earthquake-slip monitoring data points includes: wherein, is a density of the seismic landslide monitoring data points; is a density of the updated seismic landslide monitoring data points; and are feature vectors of the seismic landslide monitoring data point i and the seismic landslide monitoring data point j, respectively; and is an influence range parameter; is a feature vector of the h-th cluster center point; is a density value of the h-th cluster center point; In step six, the process of obtaining the hazard calibration coefficient includes: extracting feature data from the earthquake landslide monitoring dataset, water level, flow rate, sediment content, and flow velocity of water bodies surrounding the earthquake landslide monitoring area, and comprehensively analyzing the vegetation cover, lithology, and soil composition of the monitoring area as input parameters; integrating the extracted input parameters to construct an input dataset; combining the input dataset with a multiple linear regression algorithm, using the extracted input parameters as input data and the hazard calibration coefficient as output data, learning the linear relationship between the input and output data, and training the earthquake landslide calibration model; inputting the input dataset into the earthquake landslide calibration model, adjusting the intercept term and regression coefficients of the earthquake landslide calibration model, optimizing the performance of the earthquake landslide calibration model, deploying the optimized earthquake landslide calibration model into the system, and outputting the corresponding hazard calibration coefficient based on the current input parameters.

2. The earthquake landslide hazard zoning method based on subtractive clustering ANFIS according to claim 1, characterized in that: In step four, the construction process of the ANFIS model includes: using the initial clustering information of earthquake landslides as input variables of the ANFIS model; defining fuzzy membership functions and fuzzy sets for each input variable; substituting the actual input variable data into the corresponding fuzzy membership functions; converting each input variable into a fuzzy membership value to achieve fuzzification of the input data; determining the number of fuzzy rules based on the initial clustering centers of earthquake landslides obtained by the sub-clustering algorithm; constructing a fuzzy rule base by using AND and OR relations through logical combinations between each input variable and its corresponding fuzzy set; and then, according to the fuzzy rules... The system performs inference operations on the fuzzification results of the input variables to obtain the output results of each fuzzy rule. It then synthesizes the outputs of all fuzzy rules to obtain the total fuzzy rule output result. Using the maximum membership degree method, it defuzzifies the total fuzzy rule output result and selects the universe of discourse value corresponding to the maximum membership degree as the output value of the ANFIS model. This allows the construction of the ANFIS model. The universe of discourse value corresponding to the maximum membership degree is the interval of the fuzzy membership function corresponding to the maximum membership degree of the input variables in the fuzzy set. The output value of the ANFIS model is the earthquake landslide hazard index.

3. The earthquake landslide hazard zoning method based on subtractive clustering ANFIS according to claim 2, characterized in that: Step four, the process of training the ANFIS model and obtaining the subtractive clustering ANFIS model, includes: extracting earthquake data, topographic data, and meteorological and hydrological data from the earthquake landslide monitoring dataset; using the training set data and combining the backpropagation algorithm, taking the earthquake data, topographic data, and meteorological and hydrological data as inputs and the earthquake landslide hazard index as output, learning the nonlinear relationship between the earthquake data, topographic data, meteorological and hydrological data, and the earthquake landslide hazard index, and training the ANFIS model; inputting the test set data into the ANFIS model, comparing the output results of the ANFIS model with the actual earthquake landslide hazard index, adjusting the ANFIS model parameters, optimizing the ANFIS model, and obtaining the subtractive clustering ANFIS model.

4. The earthquake landslide hazard zoning method based on subtractive clustering ANFIS according to claim 3, characterized in that: In step five, the process of obtaining the earthquake landslide hazard index of each geographic grid unit includes: dividing the earthquake landslide monitoring area into various geographic grids according to the small monitoring areas corresponding to the earthquake landslide monitoring data points, and then obtaining the earthquake landslide monitoring data of each geographic grid unit; combining earthquake data, topographic data, meteorological and hydrological data and the reduced clustering ANFIS model to obtain the earthquake landslide hazard index of each geographic grid unit.

5. The earthquake landslide hazard zoning method based on subtractive clustering ANFIS according to claim 1, characterized in that: Step seven, the process of calibrating the earthquake landslide hazard index of each geographic grid unit and classifying the earthquake landslide hazard level within each geographic grid unit, includes: calibrating the earthquake landslide hazard index of each geographic grid unit using a hazard calibration coefficient to obtain the calibrated earthquake landslide hazard index of each geographic grid unit; setting a first threshold and a second threshold for the earthquake landslide hazard index; when the earthquake landslide hazard index of each geographic grid unit is lower than the first threshold, the corresponding geographic grid unit is classified as having a low earthquake landslide hazard level; when the earthquake landslide hazard index of each geographic grid unit is between the first threshold and the second threshold, the corresponding geographic grid unit is classified as having a medium earthquake landslide hazard level; and when the earthquake landslide hazard index of each geographic grid unit is higher than the second threshold, the corresponding geographic grid unit is classified as having a high earthquake landslide hazard level.

6. The earthquake landslide hazard zoning method based on subtractive clustering ANFIS according to claim 5, characterized in that: In step eight, the process of drawing the earthquake landslide hazard zoning map includes: importing the earthquake landslide hazard levels within each geographic grid unit into GIS software; using the GIS software, setting green, yellow, and red icons for low, medium, and high earthquake landslide hazard levels respectively, thereby obtaining earthquake landslide hazard level layers; converting the earthquake data, topographic data, and meteorological and hydrological data of each geographic grid unit into raster data; visualizing the raster data; obtaining earthquake data layers, topographic data layers, and meteorological and hydrological data layers respectively; and overlaying the earthquake landslide hazard level layers, earthquake data layers, topographic data layers, and meteorological and hydrological data layers to draw the earthquake landslide hazard zoning map.

Images