CN112200362B - Landslide prediction method, landslide prediction device, landslide prediction equipment and storage medium - Google Patents

Landslide prediction method, landslide prediction device, landslide prediction equipment and storage medium Download PDF

Info

Publication number
CN112200362B
CN112200362B CN202011066224.4A CN202011066224A CN112200362B CN 112200362 B CN112200362 B CN 112200362B CN 202011066224 A CN202011066224 A CN 202011066224A CN 112200362 B CN112200362 B CN 112200362B
Authority
CN
China
Prior art keywords
landslide
prediction
functional area
grid
model
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202011066224.4A
Other languages
Chinese (zh)
Other versions
CN112200362A (en
Inventor
商琪
郑增荣
吴展开
程京凯
江子君
宋杰
胡辉
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hangzhou Ruhr Technology Co Ltd
Original Assignee
Hangzhou Ruhr Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hangzhou Ruhr Technology Co Ltd filed Critical Hangzhou Ruhr Technology Co Ltd
Priority to CN202011066224.4A priority Critical patent/CN112200362B/en
Publication of CN112200362A publication Critical patent/CN112200362A/en
Application granted granted Critical
Publication of CN112200362B publication Critical patent/CN112200362B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biophysics (AREA)
  • Strategic Management (AREA)
  • Economics (AREA)
  • Human Resources & Organizations (AREA)
  • Game Theory and Decision Science (AREA)
  • Development Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Marketing (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Testing Or Calibration Of Command Recording Devices (AREA)

Abstract

The embodiment of the invention discloses a landslide prediction method, a device, equipment and a storage medium, which are used for acquiring remote sensing image information and topographic information of a target area, determining each functional area of the target area according to the remote sensing image information and the topographic information based on a functional area classification model, and realizing accurate partition of the target area; and screening the target functional area according to the labels corresponding to the functional areas, inputting the landslide influence factors of each grid of the target functional area, which are acquired in advance, into a landslide prediction model every day in a set time period, determining the landslide occurrence probability of each grid, carrying out targeted landslide prediction on the target functional area, reducing the calculated amount and improving the landslide prediction efficiency of the target area.

Description

Landslide prediction method, landslide prediction device, landslide prediction equipment and storage medium
Technical Field
The embodiment of the invention relates to the technical field of landslide monitoring, in particular to a landslide prediction method, a landslide prediction device, landslide prediction equipment and a storage medium.
Background
Landslide is one of the most common disastrous natural disasters, has the characteristics of wide distribution range, high occurrence frequency, multiple occurrence, regional property, severity and the like, and can cause a large amount of casualties and serious environmental and infrastructure losses every year. The method has important significance in evaluating the liability of landslide.
The existing landslide susceptibility prediction can be divided into a deterministic method and a non-deterministic method according to the difference of theoretical basis on which the landslide susceptibility prediction is based. The deterministic method is mainly a directional analysis based on expert experience and knowledge and a landslide process or physical model analysis method, and the prediction accuracy is poor. With the rapid development of computer technology and 3S technology in recent years, non-deterministic methods have been widely used, mainly including fuzzy logic methods, analytic hierarchy processes, decision trees, and the like. The method generally analyzes the landslide influence factors of a certain area directly, but the area range related to the area for landslide prediction is often larger, and the method comprises the area where landslide frequently occurs and the area where landslide is difficult to occur, and the landslide influence factors are directly utilized to predict the area to influence the landslide prediction efficiency.
Disclosure of Invention
The invention provides a landslide prediction method, a landslide prediction device, landslide prediction equipment and a storage medium, which realize the effect of improving the landslide prediction efficiency of a target area.
In a first aspect, an embodiment of the present invention provides a method for predicting a landslide, including:
acquiring remote sensing image information and topographic information of a target area;
Determining each functional area of the target area according to the remote sensing image information and the topographic information based on a functional area classification model;
and screening the target functional areas according to the labels corresponding to the functional areas, inputting the landslide influence factors of each grid of the target functional areas, which are acquired in advance, into a landslide prediction model every day, and determining the occurrence probability of landslide of each grid.
In a second aspect, an embodiment of the present invention further provides a landslide prediction apparatus, including:
the information acquisition module is used for acquiring remote sensing image information and topographic information of the target area;
the functional area determining module is used for determining each functional area of the target area according to the remote sensing image information and the topographic information based on a functional area classification model;
the landslide prediction module is used for screening the target functional areas according to the labels corresponding to the functional areas, inputting the landslide influence factors of each grid of the target functional areas, which are acquired in advance, into the landslide prediction model every day, and determining the occurrence probability of landslide of each grid.
In a third aspect, an embodiment of the present invention further provides a landslide prediction apparatus, including:
one or more processors;
A memory for storing one or more programs;
when the one or more programs are executed by the one or more processors, the one or more processors implement the landslide prediction method provided by any embodiment of the present invention.
In a fourth aspect, embodiments of the present invention also provide a storage medium containing computer-executable instructions, which when executed by a computer processor, are used to perform the method of predicting landslide provided by any of the embodiments of the present invention.
According to the technical scheme, remote sensing image information and terrain information of a target area are obtained, each functional area of the target area is determined according to the remote sensing image information and the terrain information based on a functional area classification model, and accurate partition of the target area is achieved; and screening the target functional area according to the labels corresponding to the functional areas, inputting the landslide influence factors of each grid of the target functional area, which are acquired in advance, into a landslide prediction model every day, determining the occurrence probability of landslide of each grid, carrying out targeted landslide prediction on the target functional area, reducing the calculated amount and improving the landslide prediction efficiency of the target area.
Drawings
FIG. 1 is a flow chart of a method for predicting landslide in accordance with a first embodiment of the invention;
FIG. 2 is a flow chart of a method for predicting landslide in accordance with a second embodiment of the present invention;
FIG. 3 is a flowchart of a landslide prediction method in accordance with a third embodiment of the present invention;
fig. 4 is a schematic structural diagram of a landslide prediction device according to a fourth embodiment of the present invention;
fig. 5 is a schematic structural diagram of a landslide prediction apparatus in a fifth embodiment of the invention.
Detailed Description
The invention is described in further detail below with reference to the drawings and examples. It is to be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention. It should be further noted that, for convenience of description, only some, but not all of the structures related to the present invention are shown in the drawings.
Example 1
Fig. 1 is a flowchart of a landslide prediction method according to an embodiment of the present invention, where the method may be performed by a landslide prediction device, and the method includes the following steps:
s110, acquiring remote sensing image information and topographic information of the target area.
The target area is usually an area where landslide occurs, and may be any designated area. The remote sensing image information refers to satellite images, can be obtained through land reflectivity products of Landsat (terrestrial satellite) series, and can reflect land feature types of grids of a target area, such as grassland types, woodland types, rice types and building types; the terrain information may be obtained from a digital elevation model (Digital Elevation Model, DEM) of the target area, reflecting the maximum degree of variation of the grid information of a certain grid and surrounding grids of the target area. The terrain information may include elevation information, gradient information, curvature information, etc., which may reflect a terrain feature Gao Chengdian of a certain grid of the target area, elevation points near an important geographic target, or key elevation points on a distribution range, the gradient information may reflect a degree of steepness of a certain grid of the target area, and the curvature information may reflect a concave-convex condition of a certain grid of the target area.
And S120, determining each functional area of the target area according to the remote sensing image information and the topographic information based on the functional area classification model.
Optionally, the method for determining each functional area of the target area includes: preprocessing the remote sensing image information, wherein the preprocessing comprises at least one of atmospheric correction, radiation correction and edge detection processing; and respectively extracting features of the preprocessed remote sensing image information and the preprocessed topographic information, inputting the features of the remote sensing image and the features of the topographic information into the functional area classification model, and determining each functional area of the target area, wherein the features of the remote sensing image comprise the mean value, the maximum value and the minimum value of the remote sensing image information, the topographic information comprises the elevation information, the gradient information and the curvature information, and the features of the topographic information comprise the mean value, the maximum value and the minimum value of the elevation information, the gradient information and the curvature information.
When a sensor of land surface reflectivity products of Landsat (terrestrial satellite) series images, due to absorption and scattering of atmospheric vapor and aerosol particles, radiation signals received by the sensor cannot normally truly reflect reflection and emission spectral characteristics of ground features, and meanwhile, spectral characteristics of the same ground feature type show larger differences under different gradients for areas with more obvious topographic relief. Therefore, in order to eliminate the interference of the atmosphere and the topographic relief on the remote sensing image, the remote sensing image needs to be subjected to the atmospheric correction and the topographic radiation correction. Optionally, any one mode of a radiation transmission model method, a dark target subtraction method, a constant target method and a histogram matching method can be adopted to carry out atmospheric correction on remote sensing image information, and on the basis, a terrain radiation C correction algorithm is utilized to carry out high-efficiency automatic radiation correction so as to eliminate ground object radiation differences caused by terrain fluctuation.
The edge detection process is to detect pixel points with obvious brightness change in the remote sensing image information, take the points with obvious brightness change as boundary pixel points of the remote sensing image, and determine the spatial distribution of different ground object boundaries of the remote sensing image based on the boundary pixel points, for example, obtain boundaries of ground object types such as water body, road and the like.
Optionally, after the remote sensing image information is processed, a false color synthesis method or google map is used for further preprocessing the remote sensing image information, so that visual interpretation of each functional area is facilitated, and labels corresponding to each functional area are constructed.
In order to reduce the calculation amount of the functional area classification model, the embodiment performs feature extraction on the preprocessed remote sensing image information and the preprocessed topographic information of each wave band, the obtained remote sensing image features comprise the mean value, the maximum value and the minimum value of the remote sensing image information, the obtained topographic features comprise the mean value, the maximum value and the minimum value of the topographic information, the remote sensing image features and the topographic features are respectively input into the functional area classification model, and the labels corresponding to the functional areas are output. The accurate partitioning of the target area is realized according to the remote sensing image information and the topographic information of the target area based on the functional partitioning model.
Optionally, the functional area classification model is a random forest model, and the training method of the functional area classification model includes: acquiring an initial forest model; sample remote sensing image information and sample terrain information of each functional area in a historical time period are input into the initial forest model, and a prediction label and the probability of the prediction label of each functional area are determined; and adjusting parameters of the initial forest model based on the prediction label, the probability of the prediction label and the sample label of each functional area until the prediction label is consistent with the sample label and the probability of the prediction label reaches a set threshold value, so as to obtain the functional area classification model. The probability refers to the probability of a predicted label of the output of the initial forest model, parameters of the initial forest model are adjusted through iteration, if the predicted label is consistent with the sample label and the predicted label probability reaches a set threshold, the probability that the predicted label is the sample label is high, and the initial forest model under the iteration number is used as a functional class classification model. The set threshold may be a larger value, for example, the set threshold takes a value of 0.9. Optionally, before the sample remote sensing image information is input into the initial forest model, the sample remote sensing image information may be preprocessed, for example, atmospheric correction, radiation correction, edge detection, false color synthesis processing, etc. are performed on the sample remote sensing image information, so as to improve the training accuracy of the functional area classification model of the sample remote sensing information.
S130, screening target functional areas according to labels corresponding to the functional areas, inputting landslide influence factors of each grid of the target functional areas, which are acquired in advance, into a landslide prediction model every day, and determining occurrence probability of landslide of each grid.
Optionally, the labels corresponding to the functional areas correspond to historical prediction probabilities of landslide. The method for determining the target function area comprises the following steps: and deleting the functional area corresponding to the label with the history prediction probability smaller than or equal to the probability threshold value, and taking the functional area corresponding to the label with the history prediction probability larger than the probability threshold value as the target functional area. The historical prediction probability is obtained by analyzing landslide detection results of all functional areas of all monitoring areas in advance, and the probability threshold is obtained by analyzing landslide detection results of all functional areas of all monitoring areas in advance, and can be a minimum value, for example, the probability threshold value is 0.2.
It can be understood that each functional area output by the functional area classification model comprises all functional areas of a target area, including functional areas of town buildings, urban green lands, water bodies, farmlands, bare soil, mountain forests and the like, and landslide occurrence probabilities of the functional areas are different. For example, urban buildings and urban green lands have extremely low probability of landslide, and water bodies and mountain forests have relatively high probability of landslide. By comparing the historical prediction probability corresponding to each functional area with the historical probability threshold, the target functional area is screened out, and the target functional area is subjected to targeted landslide prediction, so that the calculated amount can be reduced, and the landslide prediction efficiency of the target area is improved.
The grid data is a data form of dividing a space into regular grids, each grid is formed into a grid or a unit, and corresponding attribute values are given to each unit to represent an entity. The landslide impact factor is raster data. The landslide impact factors mainly include various landslide impact factors of landslide, such as elevation factors and environmental factors. The set period of time may be one day, three days, one week, one month, or other period of time. Optionally, the landslide impact factor includes a dynamic factor and a static factor, wherein the dynamic factor includes at least one of rainfall and soil humidity, and the static factor includes at least one of elevation, slope direction, plane curvature, profile curvature, terrain humidity index, water current intensity index, sediment transport index, terrain roughness index, distance to fault, distance to river, distance to road, lithology, land utilization, vegetation coverage.
The landslide prediction model can be a single neural network model, or can be a complex neural network model formed by sequentially connecting an input end and an output end. Specifically, the training process of the landslide prediction model is as follows: basic information of landslide of a target area or all areas is extracted from files such as landslide field investigation reports, typical landslide monitoring reports and the like, the basic information comprises landslide information such as landslide occurrence time, longitude and latitude, disaster scale and the like and actual measurement rainfall of the landslide areas, landslide influence factors of historical time periods of various landslide occurrence lands are determined according to the landslide information, and the landslide influence factors of the historical time periods form a training set and a verification set according to a set proportion, such as 8:2 or 7:3; and carrying out data correction and registration on each landslide influence factor, carrying out coordinate unified processing and grid unified processing, thus obtaining landslide influence factors with consistent grid sizes, and training a landslide prediction model by using the landslide influence factors processed by the steps. The landslide influence factors are extracted through characteristic engineering to form an input characteristic matrix of a landslide prediction model; initializing parameters of a landslide prediction model, inputting the input feature matrix into the landslide prediction model, performing model training to obtain historical landslide information in a historical time period, adjusting the parameters of the landslide prediction model according to an evaluation result based on an F1-value (F1-Score) and an ROC (Receiver Operating Characteristic) as evaluation indexes, and performing model verification through a verification set when the parameters are satisfied, wherein the trained landslide prediction model is obtained after verification. Wherein the historical landslide information includes a probability of each grid landslide occurrence.
Optionally, before inputting the landslide impact factor of each day in the preset time period of each grid of the target functional area into the landslide prediction model, the method further comprises: and preprocessing the landslide impact factors, wherein the preprocessing comprises at least one of coordinate unification processing, correction processing, data discretization processing and grid unification processing.
Because the landslide impact factors are different in source, the coordinates or the grid sizes of the landslide impact factors are not uniform, and therefore coordinate uniform processing and grid uniform processing are required to be carried out on the landslide impact factors after the landslide impact factors are acquired. Of course, other data preprocessing may be performed according to the specific situation of the landslide impact factor, such as correction processing, outlier removal, non-raster data rasterization processing, and the like. Specifically, the coordinate unification process is mainly used for unifying the coordinate system of the landslide influence factor, for example, the western-style 80 coordinate system is used as the coordinate system of the landslide influence factor, and other coordinate systems can be used as the coordinate system of the landslide influence factor.
According to the technical scheme, remote sensing image information and terrain information of a target area are obtained, each functional area of the target area is determined according to the remote sensing image information and the terrain information based on a functional area classification model, and accurate partition of the target area is achieved; and screening the target functional areas according to the labels corresponding to the functional areas, inputting the landslide influence factors of each grid of the target functional areas, which are acquired in advance, into a landslide prediction model every day in a set time period, determining the occurrence probability of landslide of each grid, carrying out targeted landslide prediction on the target areas, reducing the calculated amount and improving the landslide prediction efficiency of the target areas.
Example two
Fig. 2 is a flowchart of a landslide prediction method according to a second embodiment of the present invention, where the embodiment is a further refinement of the previous embodiment, and the step of inputting a landslide impact factor obtained in advance and within a set period of time of the target functional area into a landslide prediction model to determine occurrence probability of landslide of each grid includes: determining the occurrence probability of landslide every day according to the landslide influence factor based on a first preset model; and if the occurrence probability of the landslide on the current day is larger than a preset probability threshold, determining the occurrence probability of each grid landslide of the target functional area according to the landslide influence factor on the current day based on a second preset model. For parts which are not described in detail in this method embodiment, reference is made to the above-described embodiments. Referring specifically to fig. 2, the method may include the steps of:
s210, acquiring remote sensing image information and topographic information of the target area.
S220, determining each functional area of the target area according to the remote sensing image information and the topographic information based on the functional area classification model.
S230, screening target functional areas according to labels corresponding to the functional areas, and determining the occurrence probability of landslide every day according to landslide influence factors of the target functional areas based on a first preset model.
As described in the foregoing embodiment, the landslide impact factors include data corresponding to each landslide factor, wherein the landslide factors include dynamic factors including rainfall (precipitation) and soil humidity, and static factors including elevation, slope direction, plane curvature, section curvature, terrain humidity index, water flow intensity index, sediment transport index, terrain roughness index, fault distance, river distance, road distance, lithology, land utilization, vegetation coverage, and the like.
The first preset model may be a neural network model, or other learning algorithm. By way of example, the first preset model may be a support vector machine algorithm (Support Vector Machine, SVM), a Long Short Term Memory Network (LSTM), a logistic regression model (Logistics Regression, LR), an XGBoost (Extreme Gradient Boosting, extreme gradient boost decision tree) algorithm, a GBDT (Gradient Boosting Decision Tree, gradient boost decision tree) algorithm, a full convolution Network (Fully Convolutional Networks, FCN), a cyclic convolution Network (Recurrent Neural Network, RNN), a Residual Network (ResNet), a gated loop unit (Gate Recurrent Unit, GRU), and the like. The training method of the first preset model is the same as that of the landslide prediction model in the foregoing embodiment, and will not be described in detail herein.
Optionally, the method for determining the occurrence probability of the daily landslide is as follows: calculating construction characteristics of landslide influence factors of each grid every day, wherein the construction characteristics comprise the sum, average value, maximum value, minimum value, range, quartile and rainfall time of each landslide influence factor of each grid, and obtaining a first characteristic matrix of the first preset model according to the construction characteristics every day; and inputting the first feature matrix into the first preset model to determine the occurrence probability of landslide every day according to the first preset model.
In the foregoing embodiment, the landslide impact factor includes a dynamic factor and a static factor, and features are constructed on the dynamic factor and a first feature matrix is determined in combination with the static factor, and the occurrence probability of the landslide every day is determined based on the first feature matrix and a first preset model.
Illustratively, the landslide impact factor X of each grid acquired daily is: x= { X (1) ,X (2) ,…,X (d) And, wherein the superscript indicates the number of days, X (c) C=1, 2, …, d, representing the landslide impact factor of each grid in the target area on day c, where matrix X (c) The rows of (1) represent a grid, the columns represent landslide factors, and the matrix X (c) Is m x n, i.e. comprises m grids, n landslide factors, where i=1, 2, … n 1 Represents a dynamic factor (rainfall, soil humidity, etc.), i=n 1 +1,n 1 +2, … n, represent the statics factor. Constructing feature set X in days 1 (first feature matrix) constructed as follows: for each X (c) Calculating n by grid 1 The sum, average, maximum, minimum, range, upper quartile, lower quartile and other structural features of each dynamic factor are used to obtain feature set X 1 . For training data, a tag vector Y may also be constructed 1 To indicate whether or not a landslide has occurred every day, wherein 1 indicates that a landslide has occurred and 0 indicates that no landslide has occurred.
S240, if the occurrence probability of the landslide on the current day is larger than a preset probability threshold value, determining the occurrence probability of each grid landslide of the target functional area according to the landslide influence factor of the target functional area on the current day based on a second preset model.
Optionally, the determining, based on the second preset model, the landslide occurrence probability of each grid of the target functional area according to the landslide influence factor of the current day of the target functional area includes: feature extraction is carried out on landslide influence factors of grids corresponding to the current day so as to generate a second feature matrix of the second preset model; normalizing the second feature matrix; and inputting the normalized second feature matrix into the second preset model to determine landslide occurrence probability of each grid of the target functional area on the current day according to the second preset model.
Specifically, assuming that the landslide occurrence probability on the c-th day is greater than the prediction probability threshold, the landslide influence factor X on the c-th day (c) The feature set is sent to a second preset model, and the feature set of the second preset model is constructed in the specific construction mode that: calculating the structural characteristics of the dynamic factors of each grid on the day, such as the total sum, the mean value, the variance, the median, the differential mean value, the variance, the skewness, the kurtosis and the like, corresponding to the days 3, 7, 15 and 30, and considering the characteristics of whether the current grid is the maximum value or the minimum value in the set range, whether the current grid exceeds the mean value corresponding to the set range, whether the grid is landslide, the number of landslide occurrence of the current grid and the like in the set range which is centered on the current grid, such as 3*3 area, and finally obtaining a second characteristic matrix X of a second preset model 2 fea 。
Specifically, the second feature matrix X can be normalized based on max-min 2 fea Is normalized. Of course, other normalization algorithms can be selected for normalization processing. The embodiment of the invention does not limit the normalization algorithm of the first feature matrix and the second feature matrix.
Further, after the normalization process, the method further includes: based on multiple collinearityThe method performs feature selection on the second feature matrix to perform feature screening according to the collinearity degree to obtain a screened second feature matrix
Further, if the feature selection is performed, the above steps are replaced by: the second feature matrix after screeningAnd inputting the second preset model to determine the occurrence probability of each grid landslide on the current day according to the second preset model.
Specifically, the second preset model can be a one-dimensional convolutional neural network model, and comprises a convolutional layer, a batch standardization layer, an activation function and an optimization layer, wherein the value range of the number of convolutional kernels of the convolutional layer is 32-512, and the step length is 16; the activation function may include any one of a ReLU function (Rectified Linear Unit, linear rectification function), a Linear function (Linear function), a Sigmoid function, and a Tanh function (hyperbolic function); the optimization method comprises any one of SGD (Stochastic Gradient Descent, gradient update rule), adam (Adaptive Moment Estimation ), nadam (Nesterov Adaptive Moment Estimation, nesterov acceleration adaptive moment estimation), adagard (Adaptive Gradient Algorithm ), RMSprop (Root Mean Square Prop, forward root mean square gradient descent algorithm) and other optimization algorithms; the initial learning rate may take a value of 0.0001, 0.001, 0.01, or 0.1; the value range of the number of the hidden layer neurons is 4-256, the step length is 4, the value range of the number of the hidden layer neurons is 3-8, and the step length is 1; the random discarding rate is 0-0.8, and the step length is 0.05. The training method of the first preset model is the same as that of the landslide prediction model in the foregoing embodiment, and will not be described in detail herein.
According to the technical scheme provided by the embodiment, the accurate partition of the target area is realized according to the remote sensing image information and the topographic information based on the functional area classification model, the target functional areas are further screened according to the labels corresponding to the functional areas, the prediction is performed by taking the day as a unit, the time sequence information of the dynamic factors is fully considered, and the prediction precision is improved; by setting two preset models, carrying out targeted landslide prediction on the target functional area in two stages, the calculated amount is reduced, if the probability of landslide occurrence in the current day is larger than a set value in the first stage, carrying out second stage landslide prediction, and carrying out landslide occurrence probability of a specific grid by taking the data in the current day as a reference, thereby greatly reducing the waste of calculation resources and improving the landslide prediction efficiency; by constructing the dynamic factor characteristics, the function of the dynamic factor in model prediction is increased, and the accuracy of model prediction is improved; and through feature screening and normalization processing, the efficiency of model prediction is improved. Simultaneously, the two models are respectively subjected to parameter optimization, so that the quality of the models is improved, and the accuracy of prediction is ensured
Example III
Fig. 3 is a flowchart of a landslide prediction method according to a third embodiment of the present invention. The embodiment is used for analyzing the whole landslide prediction flow. For parts which are not described in detail in this method embodiment, reference is made to the above-described embodiments. Referring specifically to fig. 3, the method may include the steps of:
s310, remote sensing image information and topographic information of the target area are obtained.
S320, determining each functional area of the target area according to the remote sensing image information and the topographic information based on the functional area classification model.
S330, screening target functional areas according to labels corresponding to the functional areas, and determining the occurrence probability of landslide every day according to landslide influence factors of the target functional areas based on a first preset model.
In order to improve the characterization effect of the landslide impact factors in the target functional areas, the target size of the grids corresponding to the target functional areas can be determined before the occurrence probability of the landslide every day is determined according to the landslide impact factors based on the first preset model. Alternatively, the target size may be determined by two methods. Optionally, a first method for determining the grid of the target size is to uniformly divide the grid corresponding to the target functional area. The uniform dividing method of the target size comprises the following steps: acquiring landslide information of a target functional area; determining a target size of each grid of the target area based on the landslide information, wherein the landslide information comprises a landslide position and a landslide area; wherein the determining the target size of each grid of the target area based on the landslide information includes: determining the initial size and the initial grid number of each grid of the target area, and determining the landslide grid number and the non-landslide grid number according to the landslide information, wherein the landslide grid number is the grid number of landslide occurrence and the non-landslide grid number is the grid number of non-landslide occurrence; if the ratio of the landslide grid number to the non-landslide grid number is in the set balance interval, taking the initial size as the target size; if the ratio of the landslide grid number to the non-landslide grid number is not in the set balance interval, adjusting the size of each grid of the target area, and re-determining the landslide grid number, the non-landslide grid number and the current grid number; and if the ratio of the redetermined landslide grid number to the non-landslide grid number is in the set balance interval and the ratio of the current grid number to the initial grid number is smaller than a set threshold value, taking the size of each grid after adjustment as the target size.
Wherein determining the target size of each grid of the target area based on the landslide information comprises: acquiring rainfall data of each rainfall station of the target area, the initial size and the initial grid number of each grid of the target area, and determining the landslide grid number and the non-landslide grid number according to the landslide information, wherein the landslide grid number is the number of grids with landslide, and the non-landslide grid number is the number of grids without landslide; calculating rainfall predicted values of the rainfall stations corresponding to any current grid according to rainfall data of the rainfall stations with a specific number, and calculating rainfall errors based on the rainfall predicted values and actually measured rainfall values of the rainfall stations corresponding to the current grid; if the rainfall error is smaller than a set error threshold, and the ratio of the landslide grid number to the non-landslide grid number is in a set balance interval, taking the initial size as the target size; if the rainfall error is larger than a set error threshold, adjusting the size of each grid of the target area, and recalculating the rainfall error according to the rainfall predicted value and the actually measured rainfall value of the rainfall station corresponding to any grid after adjustment; and if the recalculated rainfall error is smaller than the set error threshold, the ratio of the redetermined landslide grid number to the redetermined non-landslide grid number is in a set balance interval, the ratio of the current grid number to the initial grid number is smaller than the set threshold, and the size of the grid after adjustment is taken as the target size.
According to the embodiment, the landslide grid number and the non-landslide grid number are determined in the mode, so that the target size of the grids is determined according to actual landslide information and rainfall, and further the landslide influence factor has the best characterization effect under the target size, and the prediction accuracy of the landslide occurrence probability is improved.
Optionally, the second method for determining the target size is to unevenly divide the grid corresponding to the target functional area. The non-uniform dividing method of the target-sized grid comprises the following steps: acquiring a first grid size and a second grid size of a target area, and acquiring landslide impact factors of the target area under the first grid size and the second grid size respectively, wherein the second grid size is larger than the first grid size; determining whether a second grid size is a desired grid size based on distribution characteristics of a landslide impact factor at the first grid size and a landslide impact factor at a second grid size, respectively, wherein the desired grid size is smaller than the second grid size and greater than or equal to the first grid size; if not, iteratively adjusting the second grid size based on a particular scale factor; and if the current grid size corresponding to the current iteration number is the expected grid size, inputting a landslide influence factor of each day of each grid corresponding to the expected grid size into a landslide prediction model, and determining the occurrence probability of landslide of each grid based on the landslide prediction model.
Wherein the determining whether the second grid size is a desired grid size based on the distribution characteristics of the landslide impact factor at the first grid size and the landslide impact factor at the second grid size, respectively, comprises: performing significance testing on the distribution characteristics of the landslide impact factors under the second grid size based on the distribution characteristics of the landslide impact factors under the first grid size; determining the second grid size as the desired grid size if the landslide imaging factor at the second grid size passes a saliency check, otherwise, the second grid size is not the desired grid size. The determining whether the second grid size is a desired grid size based on the distribution characteristics of the landslide impact factor at the first grid size and the landslide impact factor at the second grid size, respectively, includes: calculating first characteristic data based on the distribution characteristics of the landslide impact factors under the first grid size, and calculating second characteristic data based on the distribution characteristics of the landslide impact factors under the second grid size; wherein the first feature data and the second feature data include at least one of variance and mean; comparing a difference between the first characteristic data and the second characteristic data with a set threshold; and if the difference is smaller than the set threshold, determining the second grid size as the expected grid size, otherwise, determining that the second grid size is not the expected grid size.
The step of iteratively adjusting the second grid size based on the specific scale factor to obtain a current grid size under the current iteration times comprises the following steps: and iteratively reducing the second grid size according to the specific scale factor to obtain the current grid size under the current iteration times.
Through carrying out uneven division on grids with target sizes, landslide influence factors are enabled to simultaneously have the advantages of retaining rich characteristic information and simplifying the redundancy of data under the divided grid sizes; when the landslide probability is predicted subsequently, according to landslide influence factors in grids corresponding to different grid sizes, the landslide influence factors are input into a landslide prediction model in a unit of days, the landslide occurrence probability is predicted, and the landslide prediction accuracy is improved.
In this embodiment, the target size may be obtained by adjusting the initial size multiple times, and the target size may be used as the first grid size. In order to improve accuracy of landslide prediction, resampling can be performed on the first grid size, and landslide prediction is performed by combining characteristic information of the resampled grid size. The specific method comprises the following steps: acquiring a first grid size of a target area, determining a second grid size after resampling the first grid size, dividing an initial grid corresponding to the second grid size into a plurality of grids according to the first grid size, extracting characteristic information of landslide influence factors under the grids of the second grid size based on the first grid size, and inputting the characteristic information and the landslide influence factors into a first preset model and a second preset model to conduct landslide prediction. The feature information extraction method comprises the following steps: acquiring characteristic values of landslide impact factors in eight neighbor grids of the current grid; and determining the characteristic information of the current grid according to the characteristic values in the eight-neighborhood grid.
According to the embodiment, the occurrence probability of each grid landslide is determined by combining the characteristic information and the landslide influence factors, the data size of the landslide influence factors can be increased, the occurrence probability of each grid landslide can be determined more accurately according to the characteristic information, and the accuracy of landslide prediction is improved.
As in the previous embodiment, the landslide impact factor includes a dynamic factor and a static factor. Wherein the dynamic factor includes at least one of rainfall and soil humidity. The rainfall can be determined by inputting the obtained geographical environment data of the target area into a rainfall interpolation model. The geographic environment data includes geographic location data, atmospheric data, terrain data, and underlying data. The rainfall interpolation model can be a back propagation model (BP, back propagation neural network), and is a multi-layer feedforward network consisting of nonlinear transformation units based on an error back propagation algorithm, wherein the BP generally consists of an input layer, an implicit layer and an output layer, each layer further comprises N neurons, the neurons at the same layer are mutually independent, and the output of the neurons between each layer only affects the input of the neurons at the lower layer after a specific excitation function.
Specifically, the training method of the rainfall interpolation model comprises the following steps: acquiring an initial model, determining an initial weight matrix and an initial threshold value of the initial model, and calculating the fitness of the initial model; sample geographic environment information in a historical time period is input into the initial model, predicted rainfall is determined, and according to the predicted rainfall and the actually measured rainfall in the historical time period, an initial weight matrix and an initial threshold value of the initial model are iteratively adjusted based on a genetic algorithm; and adjusting the initial model and calculating the fitness of the adjusted model based on the iteratively adjusted weight matrix threshold value until the recalculated fitness reaches an expected value, and taking the model corresponding to the fitness reaching the expected value as the rainfall interpolation model.
Specifically, the initial weight matrix may include a connection weight between an input layer and an implicit layer and a connection weight between the implicit layer and an output layer; the initial threshold may include a threshold of an hidden layer and a threshold of an output layer. The fitness of the initial model is calculated by the following formula:where n is the number of samples, y (i) is the actual output of the ith sample of the initial model, and t (i) is the expected output of the ith sample of the initial model. Specifically, sample geographic environment information in a historical time period is input into an initial model to obtain a predicted rainfall, a root mean square error between the predicted rainfall and an actually measured rainfall is calculated, if the root mean square error is larger than a preset error threshold, weight coding and threshold coding of the initial model are iteratively adjusted, an adjusted weight matrix and a threshold are determined based on the adjusted weight coding and threshold coding until the root mean square error is not smaller than the preset error threshold, the initial model is adjusted to be in a stable state, the fitness of the adjusted model is calculated by adopting the fitness calculation formula, and if the recalculated fitness reaches a desired value, a model corresponding to the fitness reaching the desired value is used as the rainfall interpolation model.
Optionally, the weight and the threshold are coded by adopting a binary coding, real number coding or Gray code coding mode, and the adjusted weight matrix and the adjusted threshold are determined according to the weight coding and the threshold coding. Optionally, the initial model may be adjusted to a stable state by calculating an average error, an average absolute error or a linear correlation coefficient between the predicted rainfall and the actually measured rainfall, and iteratively adjusting the weight code and the threshold code of the initial model according to the average error, the average absolute error or the linear correlation coefficient until the average error, the average absolute error or the linear correlation coefficient is not less than a preset error threshold, and further calculating the fitness of the adjusted model and determining the rainfall interpolation model.
It should be noted that, the sample geographic environment data in the historical time period is multidimensional, and when the rainfall interpolation model is trained, the precision and the robustness of the rainfall interpolation model can be improved by the multidimensional sample geographic environment data, so that after the daily geographic environment data is obtained, the prediction precision of the rainfall probability can be improved based on the rainfall interpolation model prediction, and the prediction precision of landslide prediction can be improved.
In this embodiment, the target size may be obtained by adjusting the initial size multiple times, and the target size may be used as the first grid size. In order to improve accuracy of landslide prediction, resampling can be performed on the first grid size, and landslide prediction is performed by combining characteristic information of the resampled grid size. The specific method comprises the following steps: acquiring a first grid size of a target area, determining a second grid size after resampling the first grid size, dividing an initial grid corresponding to the second grid size into a plurality of grids according to the first grid size, extracting characteristic information of landslide influence factors under the grids of the second grid size based on the first grid size, and inputting the characteristic information and the landslide influence factors into a first preset model and a second preset model to conduct landslide prediction. The feature information extraction method comprises the following steps: acquiring characteristic values of landslide impact factors in eight neighbor grids of the current grid; and determining the characteristic information of the current grid according to the characteristic values in the eight-neighborhood grid.
According to the embodiment, the occurrence probability of each grid landslide is determined by combining the characteristic information and the landslide influence factors, the data size of the landslide influence factors can be increased, the occurrence probability of each grid landslide can be determined more accurately according to the characteristic information, and the accuracy of landslide prediction is improved.
Unlike the previous embodiments, the first preset model may be a self-classifying learning model that iteratively adjusts a self-step factor and a downsampling ratio determination based on historical landslide data, historical non-landslide data. The self-step factors are determined according to the number of sub-boxes and the iteration times of the historical non-landslide data, and the downsampling proportion is determined according to the self-step factors of each sub-box and the self-step factors of all sub-boxes. Optionally, the method for determining the historical landslide data and the historical non-landslide data includes: determining labels corresponding to grids of a history area, and determining historical rainfall and specific correlation factors corresponding to landslide grids and non-landslide grids at each history time point respectively; and generating a landslide data set comprising the historical rainfall and the specific correlation factor respectively corresponding to the landslide grids and the non-landslide grids based on the labels, and determining the historical landslide data and the historical non-landslide data according to the labels corresponding to the grids in the landslide data set.
Optionally, the method for determining the historical landslide data and the historical non-landslide data includes: determining labels corresponding to grids of a history area, and determining historical rainfall and specific correlation factors corresponding to landslide grids and non-landslide grids at each history time point respectively; and generating a landslide data set comprising the historical rainfall and the specific correlation factor respectively corresponding to the landslide grids and the non-landslide grids based on the labels, and determining the historical landslide data and the historical non-landslide data according to the labels corresponding to the grids in the landslide data set.
Optionally, the training method of the first preset model includes: acquiring an initial prediction model, and acquiring the rainfall of a landslide grid and the rainfall of a non-landslide grid at each historical landslide time point, and specific correlation factors under each label; and inputting the rainfall of the landslide grids at the historical landslide time points, the rainfall of the non-landslide grids and the specific correlation factors under the labels into the initial prediction model according to the days, and adjusting a loss function of the initial prediction model based on the landslide probability output by the initial prediction model and the landslide probability corresponding to the historical landslide time points until the loss function reaches a set threshold value to obtain the first preset model.
The step of adjusting the loss function of the initial prediction model based on the landslide probability output by the initial prediction model and the landslide probability corresponding to the historical landslide time point until the loss function reaches a set threshold value, and obtaining the first preset model comprises the following steps: determining an initial classification hardness of the initial prediction model based on a rainfall of the landslide grid at the historical landslide time point, a rainfall of the non-landslide grid, each of the labels and a specific correlation factor under each of the labels; determining the rainfall capacity of the non-landslide grids and the number of the sub-boxes of the specific correlation factors of the non-landslide grids according to the initial classification hardness, and determining the self-step factors of the sub-boxes of the initial prediction model based on the number of the sub-boxes; determining a downsampling proportion of each sub-box based on the self-step factors, determining the rainfall capacity of downsampled non-landslide grids in each sub-box based on the downsampling proportion, and determining specific correlation factors of the downsampled non-landslide grids; inputting the rainfall of the downsampled non-landslide grids, the specific correlation factors of the downsampled non-landslide grids and the rainfall of the landslide grids into the initial prediction model, determining the loss function based on the landslide probability output by the initial prediction model and the landslide probability corresponding to the historical landslide time point, and iteratively adjusting the self-walking factors and the downsampling proportion of each sub-bin based on the training classification hardness of the loss function on a single sample; and adjusting the initial prediction model according to the self-step factors adjusted in an iteration mode and the downsampling proportion of each sub-bin until the loss function reaches a set threshold value, and obtaining the first preset model.
The downsampling proportion of each sub-box is the ratio of the self-step factor of each sub-box to the self-step factors of all sub-boxes, and the self-step factors are obtained by adding 1 to the sum of the number of sub-boxes, the opposite number of sub-box labels and the iteration number.
The landslide prediction method based on the self-step classification learning model has the advantages that: the self-step factors of the self-step classification learning model are determined according to the number of sub-boxes of the historical non-landslide data and the iteration times, and the down-sampling proportion is determined according to the self-step factors of each sub-box and the self-step factors of all sub-boxes, so that when the self-step factors and the down-sampling proportion are adjusted based on the historical landslide data and the historical non-landslide data in an iteration way, the sampling proportion of each sub-box gradually changes uniformly from the lower down-sampling number of the low hardness to the last lower down-sampling number of each sub-box along with the increase of the iteration times, and the lower down-sampling proportion of each box is uniformly changed, and meanwhile, the lower down-sampling number of the sub-boxes with small hardness can be always higher than the lower down-sampling number of the sub-boxes with large hardness, so that the first preset model with high diversity, robustness and strong inclusion is obtained.
S340, if the occurrence probability of the landslide on the current day is larger than a preset probability threshold value, determining the occurrence probability of each grid landslide of the target functional area according to the landslide influence factor of the target functional area on the current day based on a second preset model.
Alternatively, the second preset model may be a self-organizing learning model. The determination manner of the historical landslide data for training the second preset model and the training method of the second preset model can refer to S330, and this step will not be described in detail.
Optionally, after determining the occurrence probability of each grid landslide in the current day, the occurrence probability of each grid in the current day may be compared with an early warning threshold of the target area, and the occurrence level of the landslide in the target area is determined based on the obtained early warning level corresponding to the occurrence probability of each grid, where the early warning threshold is determined according to the first coefficient and the second coefficient of the grid landslide in each area.
The method for determining the first coefficient and the second coefficient comprises the following steps: acquiring historical probability of landslide occurrence in a set time period of grid landslide of each area; determining a first determined probability interval and a second determined probability interval based on the distribution characteristics of the historical probabilities; respectively calculating the sum of products of probability threshold values and corresponding coefficients of the same risk level in the first determined probability interval and the second determined probability interval to obtain intermediate determined probabilities; and if the intermediate determined probability does not reach the set evaluation index, iteratively adjusting the corresponding coefficient of the probability threshold value of each risk level until the intermediate determined probability reaches the set evaluation index, and determining a first coefficient and a second coefficient according to the corresponding coefficient reaching the set evaluation index.
The method for determining the early warning threshold value comprises the following steps: determining a first prediction probability interval and a second prediction probability interval based on the distribution characteristics of occurrence probability of each grid landslide; calculating a first product of each risk level in the first prediction probability interval and the first coefficient, calculating a second product of each risk level in the second prediction probability interval and the second coefficient, and taking the sum of the first product and the second product as an early warning threshold value of each risk level.
Wherein the determining the first determined probability interval and the second determined probability interval based on the distribution characteristics of the historical probabilities comprises: determining the historical probability and the times corresponding to the historical probability; determining the density distribution characteristics and breakpoint distribution characteristics of the historical probabilities according to the historical probabilities and the landslide occurrence times corresponding to the historical probabilities; the first determined probability interval is determined based on the intensity distribution characteristics and the second determined probability interval is determined based on the breakpoint distribution characteristics.
It can be understood that the above manner is based on the historical probability of the landslide occurring in the set time period of the grid landslide of each region, the first determined probability interval and the second determined probability interval are determined according to the distribution characteristics of the historical probability, the first coefficient and the second coefficient of each region are determined according to the probability threshold value of each risk level in the first determined probability interval and the second determined probability interval and the set evaluation index, and after the occurrence probability of different regions is obtained, the early warning threshold value of the different regions can be flexibly determined according to the first coefficient, the second coefficient and the occurrence probability corresponding to the different regions, so that the risk level of the region can be accurately determined according to the early warning threshold value corresponding to each region.
The determining the occurrence level of the landslide of the target area based on the obtained early warning level corresponding to the occurrence probability of each grid comprises the following steps: screening target grades larger than a first grade in the early warning threshold; calculating the average grade of the grids corresponding to the target grade; and determining the occurrence level of the landslide of the target area based on the average level, the set coefficient and the ratio of the grid number corresponding to the target level to all the grid numbers in the target area.
The calculation formula of the occurrence level of the landslide of the target area is as follows:
wherein alpha is a set coefficient, and is obtained by performing Bayesian calculation on the historical occurrence level of the landslide of the target area,and p is the ratio of the number of grids corresponding to the target grade to the number of all grids in the target zone.
Example IV
Fig. 4 is a schematic diagram of the result of a landslide prediction device according to a fourth embodiment of the present invention, where, as shown in fig. 4, the landslide prediction device includes: an information acquisition module 410, a functional area determination module 420, and a landslide prediction module 430.
The information acquisition module 410 is configured to acquire remote sensing image information and topographic information of a target area;
The functional area determining module 420 is configured to determine each functional area of the target area according to the remote sensing image information and the topographic information based on a functional area classification model;
the landslide prediction module 430 is configured to screen the target functional areas according to the labels corresponding to the functional areas, input the landslide impact factors acquired in advance within the set time period of each grid of the target functional areas to the landslide prediction model, and determine the occurrence probability of landslide of each grid.
According to the technical scheme, remote sensing image information and terrain information of a target area are obtained, each functional area of the target area is determined according to the remote sensing image information and the terrain information based on a functional area classification model, and accurate partition of the target area is achieved; and screening the target functional area according to the labels corresponding to the functional areas, inputting the landslide influence factors of each grid of the target functional area, which are acquired in advance, into a landslide prediction model every day in a set time period, determining the landslide occurrence probability of each grid, carrying out targeted landslide prediction on the target functional area, reducing the calculated amount and improving the landslide prediction efficiency of the target area.
Optionally, the functional area determining module 420 is further configured to determine, based on the functional area classification model, each functional area of the target area according to the remote sensing image information and the topographic information, including:
Preprocessing the remote sensing image information, wherein the preprocessing comprises atmosphere correction, radiation correction and edge detection processing;
and respectively extracting features of the preprocessed remote sensing image information and the preprocessed topographic information, inputting the features of the remote sensing image and the features of the topographic information into the functional area classification model, and determining each functional area of the target area, wherein the features of the remote sensing image comprise the mean value, the maximum value and the minimum value of the remote sensing image information, the topographic information comprises the elevation information, the gradient information and the curvature information, and the features of the topographic information comprise the mean value, the maximum value and the minimum value of the elevation information, the gradient information and the curvature information.
Optionally, the apparatus further comprises: a functional area classification model training module; wherein the functional area classification model is a random forest model;
the functional area classification model training module is specifically used for acquiring an initial forest model;
sample remote sensing image information and sample terrain information of each functional area in a historical time period are input into the initial forest model, and a prediction label and the probability of the prediction label of each functional area are determined;
and adjusting parameters of the initial forest model based on the prediction label, the probability of the prediction label and the sample label of each functional area until the prediction label is consistent with the sample label and the probability of the prediction label reaches a set threshold value, so as to obtain the functional area classification model.
Optionally, the landslide prediction module 430 is further configured to delete a functional area corresponding to a label with a history prediction probability smaller than or equal to a set threshold, and use a functional area corresponding to a label with a history prediction probability greater than the set threshold as the target functional area.
Optionally, the landslide prediction module 430 is further configured to determine, based on a first preset model, a probability of occurrence of a daily landslide according to the landslide impact factor;
and if the occurrence probability of the landslide on the current day is larger than a preset probability threshold, determining the occurrence probability of each grid landslide of the target functional area according to the landslide influence factor on the current day based on a second preset model.
Optionally, the landslide prediction module 430 is further configured to calculate a construction feature of a dynamic factor in the landslide impact factors of each grid every day, where the construction feature includes a sum, an average value, a maximum value, a minimum value, a polar difference, a quartile, and a rainfall time of each dynamic factor of each grid, and obtain a first feature matrix of the first preset model according to the construction feature and the static factor in the landslide impact factors every day;
And inputting the first feature matrix into the first preset model to determine the occurrence probability of landslide every day according to the first preset model.
Optionally, the landslide prediction module 430 is further configured to perform feature extraction on landslide impact factors of each grid corresponding to the current day to generate a second feature matrix of the second preset model;
normalizing the second feature matrix;
and inputting the normalized second feature matrix into the second preset model to determine landslide occurrence probability of each grid of the target functional area on the current day according to the second preset model.
The landslide prediction device provided by the embodiment of the invention can execute the landslide prediction method provided by any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method.
Example five
Fig. 5 is a schematic structural diagram of a landslide prediction apparatus according to a fifth embodiment of the present invention, where, as shown in fig. 5, the apparatus includes a processor 510, a memory 520, an input device 530, and an output device 540; the number of device processors 510 may be one or more, one processor 510 being illustrated in fig. 5; the processor 510, memory 520, input means 530 and output means 540 in the device may be connected by a bus or other means, for example in fig. 5.
The memory 520 is a computer readable storage medium, and may be used to store a software program, a computer executable program, and a module, such as program instructions/modules corresponding to a landslide prediction method in an embodiment of the present invention (for example, the information acquisition module 410, the functional area determination module 420, and the landslide prediction module 430 in a landslide prediction device). The processor 510 executes various functional applications of the apparatus and data processing, i.e., implements the landslide prediction method described above, by running software programs, instructions, and modules stored in the memory 520.
Memory 520 may include primarily a program storage area and a data storage area, wherein the program storage area may store an operating system, at least one application program required for functionality; the storage data area may store data created according to the use of the terminal, etc. In addition, memory 520 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other non-volatile solid-state storage device. In some examples, memory 520 may further include memory located remotely from processor 510, which may be connected to the device/terminal/server via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The input means 530 may be used to receive input numeric or character information and to generate key signal inputs related to user settings and function control of the device. The output 540 may include a display device such as a display screen.
Example six
A sixth embodiment of the invention also provides a storage medium containing computer-executable instructions, which when executed by a computer processor, are for performing a method of predicting a landslide, the method comprising:
acquiring remote sensing image information and topographic information of a target area;
determining each functional area of the target area according to the remote sensing image information and the topographic information based on a functional area classification model;
and screening the target functional areas according to the labels corresponding to the functional areas, inputting the landslide influence factors of each grid of the target functional areas, which are acquired in advance, into a landslide prediction model every day in a set time period, and determining the landslide occurrence probability of each grid.
Of course, the storage medium containing the computer executable instructions provided in the embodiments of the present invention is not limited to the method operations described above, and may also perform the related operations in the landslide prediction method provided in any embodiment of the present invention.
From the above description of embodiments, it will be clear to a person skilled in the art that the present invention may be implemented by means of software and necessary general purpose hardware, but of course also by means of hardware, although in many cases the former is a preferred embodiment. Based on such understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as a floppy disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a FLASH Memory (FLASH), a hard disk or an optical disk of a computer, etc., and include several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method according to the embodiments of the present invention.
It should be noted that, in the embodiment of the landslide prediction apparatus described above, each unit and module included are only divided according to the functional logic, but not limited to the above-described division, so long as the corresponding functions can be implemented; in addition, the specific names of the functional units are also only for distinguishing from each other, and are not used to limit the protection scope of the present invention.
Note that the above is only a preferred embodiment of the present invention and the technical principle applied. It will be understood by those skilled in the art that the present invention is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the invention. Therefore, while the invention has been described in connection with the above embodiments, the invention is not limited to the embodiments, but may be embodied in many other equivalent forms without departing from the spirit or scope of the invention, which is set forth in the following claims.

Claims (7)

1. A method of predicting landslide, comprising:
acquiring remote sensing image information and topographic information of a target area;
determining each functional area of the target area according to the remote sensing image information and the topographic information based on a functional area classification model; the functional areas comprise town buildings, urban green lands, water bodies, farmlands, bare soil and mountain forests;
screening target functional areas according to labels corresponding to the functional areas, inputting landslide influence factors of each grid of the target functional areas, which are acquired in advance, into a landslide prediction model every day in a set time period, and determining occurrence probability of landslide of each grid;
The functional area classification model is a random forest model, and the training method of the functional area classification model comprises the following steps:
acquiring an initial forest model;
sample remote sensing image information and sample terrain information of each functional area in a historical time period are input into the initial forest model, and a prediction label and the probability of the prediction label of each functional area are determined;
based on the prediction label, the probability of the prediction label and the sample label of each functional area, adjusting the parameters of the initial forest model until the prediction label is consistent with the sample label and the probability of the prediction label reaches a set threshold value, so as to obtain the functional area classification model;
the labels corresponding to the functional areas correspond to the historical prediction probability of landslide;
correspondingly, the screening the target functional areas according to the labels corresponding to the functional areas comprises the following steps:
deleting the functional area corresponding to the label with the history prediction probability smaller than or equal to the probability threshold value, and taking the functional area corresponding to the label with the history prediction probability larger than the probability threshold value as the target functional area;
inputting the landslide influence factors of each grid of the target functional area acquired in advance in a set time period to a landslide prediction model, and determining the occurrence probability of landslide of each grid comprises the following steps:
Determining the occurrence probability of landslide every day according to a landslide influence factor of a target functional area based on a first preset model;
and if the occurrence probability of the landslide on the current day is larger than a preset probability threshold, determining the occurrence probability of each grid landslide of the target functional area according to the landslide influence factor on the current day of the target functional area based on a second preset model.
2. The prediction method according to claim 1, wherein the determining each functional area of the target area based on the functional area classification model according to the remote sensing image information and the topographic information comprises:
preprocessing the remote sensing image information, wherein the preprocessing comprises at least one of atmospheric correction, radiation correction and edge detection processing;
and respectively extracting features of the preprocessed remote sensing image information and the preprocessed topographic information, inputting the features of the remote sensing image and the features of the topographic information into the functional area classification model, and determining each functional area of the target area, wherein the features of the remote sensing image comprise the mean value, the maximum value and the minimum value of the remote sensing image information, the topographic information comprises the elevation information, the gradient information and the curvature information, and the features of the topographic information comprise the mean value, the maximum value and the minimum value of the elevation information, the gradient information and the curvature information.
3. The prediction method according to claim 1, wherein the determining the occurrence probability of each landslide according to the landslide influence factor based on the first preset model includes:
calculating construction characteristics of dynamic factors in landslide influence factors of each grid every day, wherein the construction characteristics comprise the sum, average value, maximum value, minimum value, range, quartile and rainfall time of each dynamic factor of each grid, and obtaining a first characteristic matrix of the first preset model according to the construction characteristics every day and the static factors in the landslide influence factors;
and inputting the first feature matrix into the first preset model to determine the occurrence probability of landslide every day according to the first preset model.
4. The prediction method according to claim 1, wherein the determining, based on the second preset model, the occurrence probability of each grid landslide of the target functional area according to the landslide influence factor of the current day includes:
feature extraction is carried out on landslide influence factors of grids corresponding to the current day so as to generate a second feature matrix of the second preset model;
normalizing the second feature matrix;
And inputting the normalized second feature matrix into the second preset model to determine the occurrence probability of each grid landslide of the target functional area on the current day according to the second preset model.
5. A landslide prediction apparatus comprising:
the information acquisition module is used for acquiring remote sensing image information and topographic information of the target area;
the functional area determining module is used for determining each functional area of the target area according to the remote sensing image information and the topographic information based on a functional area classification model; the functional areas comprise town buildings, urban green lands, water bodies, farmlands, bare soil and mountain forests;
wherein the functional area classification model is a random forest model;
the functional area classification model training module is specifically used for:
acquiring an initial forest model;
sample remote sensing image information and sample terrain information of each functional area in a historical time period are input into the initial forest model, and a prediction label and the probability of the prediction label of each functional area are determined;
based on the prediction label, the probability of the prediction label and the sample label of each functional area, adjusting the parameters of the initial forest model until the prediction label is consistent with the sample label and the probability of the prediction label reaches a set threshold value, so as to obtain the functional area classification model;
The landslide prediction module is used for screening the target functional areas according to the labels corresponding to the functional areas, inputting the landslide influence factors of each grid of the target functional areas, which are acquired in advance, into the landslide prediction model every day in a set time period, and determining the occurrence probability of landslide of each grid;
the labels corresponding to the functional areas correspond to the historical prediction probability of landslide;
correspondingly, the landslide prediction module is further configured to delete a functional area corresponding to a label with the history prediction probability smaller than or equal to a set threshold value, and take a functional area corresponding to a label with the history prediction probability larger than the set threshold value as the target functional area;
the landslide prediction module is further used for determining the occurrence probability of landslide every day according to the landslide influence factor based on a first preset model;
and if the occurrence probability of the landslide on the current day is larger than a preset probability threshold, determining the occurrence probability of each grid landslide of the target functional area according to the landslide influence factor on the current day based on a second preset model.
6. A prediction apparatus for landslide, characterized in that the apparatus comprises:
one or more processors;
a memory for storing one or more programs;
When executed by the one or more processors, causes the one or more processors to implement the landslide prediction method of any one of claims 1-4.
7. A storage medium containing computer executable instructions which, when executed by a computer processor, are for performing a method of landslide prediction as claimed in any one of claims 1 to 4.
CN202011066224.4A 2020-09-30 2020-09-30 Landslide prediction method, landslide prediction device, landslide prediction equipment and storage medium Active CN112200362B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011066224.4A CN112200362B (en) 2020-09-30 2020-09-30 Landslide prediction method, landslide prediction device, landslide prediction equipment and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011066224.4A CN112200362B (en) 2020-09-30 2020-09-30 Landslide prediction method, landslide prediction device, landslide prediction equipment and storage medium

Publications (2)

Publication Number Publication Date
CN112200362A CN112200362A (en) 2021-01-08
CN112200362B true CN112200362B (en) 2024-03-29

Family

ID=74012940

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011066224.4A Active CN112200362B (en) 2020-09-30 2020-09-30 Landslide prediction method, landslide prediction device, landslide prediction equipment and storage medium

Country Status (1)

Country Link
CN (1) CN112200362B (en)

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113326724B (en) * 2021-02-07 2024-02-02 海南长光卫星信息技术有限公司 Remote sensing image change detection method, device, equipment and readable storage medium
CN113742438B (en) * 2021-08-19 2022-03-18 中国科学院地理科学与资源研究所 Method and device for determining landslide susceptibility distribution map and storage medium
CN113887515A (en) * 2021-10-28 2022-01-04 中国自然资源航空物探遥感中心 Remote sensing landslide identification method and system based on convolutional neural network
CN114781731B (en) * 2022-04-26 2023-04-18 成都理工大学 Landslide motion distance transcendental probability prediction method and system based on Bayesian theory
CN114863275B (en) * 2022-04-27 2022-12-09 北京良安科技有限公司 Three-dimensional mapping method, system, equipment and storage medium for granary
CN115575584B (en) * 2022-12-06 2023-04-11 江苏旭龙环境科技有限公司 Atmospheric environment cadmium content monitoring and early warning method and system

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106776842A (en) * 2016-11-28 2017-05-31 腾讯科技（上海）有限公司 Multi-medium data detection method and device
CN109613513A (en) * 2018-12-20 2019-04-12 长安大学 A kind of potential landslide automatic identifying method of optical remote sensing for taking InSAR deformation into account
CN111144651A (en) * 2019-12-26 2020-05-12 杭州鲁尔物联科技有限公司 Geological disaster prediction method, device and equipment

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106776842A (en) * 2016-11-28 2017-05-31 腾讯科技（上海）有限公司 Multi-medium data detection method and device
CN109613513A (en) * 2018-12-20 2019-04-12 长安大学 A kind of potential landslide automatic identifying method of optical remote sensing for taking InSAR deformation into account
CN111144651A (en) * 2019-12-26 2020-05-12 杭州鲁尔物联科技有限公司 Geological disaster prediction method, device and equipment

Also Published As

Publication number Publication date
CN112200362A (en) 2021-01-08

Similar Documents

Publication Publication Date Title
CN112200362B (en) Landslide prediction method, landslide prediction device, landslide prediction equipment and storage medium
CN112200358B (en) Landslide level prediction method, landslide level prediction device, landslide level prediction equipment and storage medium
Berberoğlu et al. Cellular automata modeling approaches to forecast urban growth for adana, Turkey: A comparative approach
JP6635038B2 (en) Simulation apparatus, simulation method, and storage medium
Soma et al. Optimization of causative factors using logistic regression and artificial neural network models for landslide susceptibility assessment in Ujung Loe Watershed, South Sulawesi Indonesia
Shahabi et al. Remote sensing and GIS-based landslide susceptibility mapping using frequency ratio, logistic regression, and fuzzy logic methods at the central Zab basin, Iran
CN112200354A (en) Landslide prediction method, device, equipment and storage medium
CN112200356B (en) Landslide prediction method, landslide prediction device, landslide prediction equipment and storage medium
CN111784044A (en) Landslide prediction method, device, equipment and storage medium
Chu et al. Comparison of landslide susceptibility maps using random forest and multivariate adaptive regression spline models in combination with catchment map units
Collini et al. Predicting and understanding landslide events with explainable AI
Riazi et al. Enhancing flood susceptibility modeling using multi-temporal SAR images, CHIRPS data, and hybrid machine learning algorithms
CN115688404B (en) Rainfall landslide early warning method based on SVM-RF model
Biard et al. Automated detection of weather fronts using a deep learning neural network
CN111310898A (en) Landslide hazard susceptibility prediction method based on RNN
CN116151483B (en) Regional rainfall landslide probabilistic prediction method and prediction terminal
Meena et al. Assessing the importance of conditioning factor selection in landslide susceptibility for the province of Belluno (region of Veneto, northeastern Italy)
CN112200355B (en) Landslide prediction method, landslide prediction device, landslide prediction equipment and storage medium
CN112200363B (en) Landslide prediction method, landslide prediction device, landslide prediction equipment and storage medium
CN117708551B (en) Flood disaster influence assessment method and system based on double-precision GDP data distribution
Tang et al. The use of serially complete station data to improve the temporal continuity of gridded precipitation and temperature estimates
Cánibe et al. Assessing the uncertainty arising from standard land‐cover mapping procedures when modelling species distributions
US20200019856A1 (en) System and method for a global digital elevation model
Sardar et al. Ensemble learning with CNN and BMO for drought prediction
CN112200357B (en) Landslide prediction method, landslide prediction device, landslide prediction equipment and storage medium

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant