Method and system for automatically generating potential landslide slip range based on Monte Carlo

Abstract

The invention discloses a method and a system for automatically generating a potential landslide slip range based on Monte Carlo. The method comprises the following steps: S1, determining a potential initial slip point grid; s2, based on the potential initial slip point grid, constructing a Monte Carlo slope random slip algorithm; s3, carrying out Monte Carlo slope slip simulation; and S4, counting the frequency selected by slip simulation of each grid in the slip path, converting grid frequency statistical data into potential landslide probability values, and classifying the potential landslide probability values to generate a potential landslide range. According to the method for automatically generating the potential landslide slippage range based on the Monte Carlo thought, mountain movement precursor information detected by the InSAR technology is ingeniously utilized, potential landslide hidden dangers are accurately positioned, longitudinal and lateral movement of the potential landslide is simulated by referring to the Monte Carlo random walking thought, and the potential landslide slippage range is automatically generated. And automatic prediction of the large-scale potential landslide slip range is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic prediction of potential landslide sliding ranges, and particularly to a method and system for automatically generating potential landslide sliding ranges based on Monte Carlo. Background Art

[0002] Once a landslide occurs, its influence range is large, and the sliding movements such as collapses and debris flows caused by it will seriously threaten the people, houses, roads and other infrastructure located on the sliding path of the landslide body. Landslide disasters usually occur in the sliding area along the slope rather than the starting area of the slide. Therefore, it is crucial to quickly and automatically predict the potential landslide sliding range in advance for the monitoring and early warning of landslide disasters and the work of disaster prevention and mitigation.

[0003] With the in-depth study of the landslide movement mechanism and the rapid development of computer data processing capabilities, significant progress has been made in the technology of simulating potential landslide sliding ranges. Currently, the methods for simulating landslide sliding ranges are mainly divided into empirical methods and modeling methods. The empirical methods use the farthest sliding distance and accessible angle as the sliding calculation indicators, and predict the sliding range by fitting the empirical equation with the relevant statistical data in the historical landslide data. The modeling methods use geophysical models to simulate the landslide movement process to achieve the prediction of the sliding range of a single landslide. However, there are still many problems to be overcome in the current automatic simulation technology of landslide sliding ranges: First, the prediction accuracy of the empirical method is low, and it depends on the existing statistical data. There is a lack of large-scale historical landslide statistical data, making it difficult to be effectively applied to engineering scenarios. Second, the modeling method needs to consider the local complex geological conditions, and the calculation process depends on complete geotechnical mechanics parameters. However, in actual engineering tasks, reliable relevant parameters are lacking, and it is impossible to quickly generate the landslide sliding range in a large-scale area. Third, the current methods all assume that the landslide has occurred, and the prediction results are more the reconstruction of the sliding process of historical landslide points rather than the simulation of the potential sliding range of the mountain body. Landslides are highly sudden and difficult to prevent. Applying these methods directly to the monitoring and early warning work of potential landslide disasters is difficult to achieve good results. Summary of the Invention

[0004] In view of the deficiencies in the prior art, the present invention provides a method for automatically generating potential landslide sliding ranges based on Monte Carlo. The method includes:[[]]

[0005] S1. Based on multi-temporal radar image data, using radar interferometry and deep learning models to identify landslide hazards, constructing a landslide hazard database, and determining potential initial sliding point grids from the landslide hazards by adopting a differential sampling strategy considering the hazard scale;

[0006] S2. Based on the potential initial slip point grid, construct a Monte Carlo ramp random slip algorithm, including: designing a rambling parameter considering the terrain and the movement direction, simulating the longitudinal and lateral movement processes of the landslide, and constructing a random probability calculation formula for the current slip grid to slip to the target grid;

[0007] S3. Based on the Monte Carlo ramp random slip algorithm, conduct Monte Carlo ramp slip simulation, including: constructing the farthest slip constraint rule considering the slip height difference and area as the slip truncation condition, optimizing the slip truncation condition according to historical landslide data, judging the slip terrain according to the terrain undulation factor, and when the potential slip point enters the bottom of the slope, designing a Monte Carlo flat ground random rambling rule to continue simulating the landslide movement process until the slip truncation condition is reached or there is no target grid;

[0008] S4. Adjust the rambling parameter considering the terrain and the movement direction based on the historical landslide statistical data of the current area, count the frequency of each grid in the slip path being selected in the slip simulation, convert the grid frequency statistical data into potential landslide slip probability values and classify them to generate a potential landslide slip range.

[0009] Preferably, step S1 includes:

[0010] Using radar interferometry technology and a deep learning model to identify landslide hazards in multi-temporal radar image data, and the calculation formula is:

[0011] L = E(F(D1)) + E(F(D2)) (1)

[0012] Where D1 and D2 respectively represent the radar satellite ascending orbit and descending orbit data in the multi-temporal radar image data, F represents calculating the surface deformation rate of the ascending orbit and descending orbit data using radar interferometry technology, E represents automatically identifying landslide hazards for the surface deformation rate using a deep learning model, and L represents the landslide hazard identification result integrating the ascending orbit and descending orbit data;

[0013] Construct a landslide hazard database, and conduct vectorization, data statistics, and geometric screening and normalization processing on the landslide hazard identification results;

[0014] Adopt a differential sampling strategy considering the hazard scale to determine potential initial slip point grids from the landslide hazards, and the calculation formula is:

[0015]

[0016] x i (m) = x i (mpR) (3)

[0017] Where X represents the potential initial slip point grid, x iIt represents the i-th hazard grid, n represents the number of identified hazards, m represents the data sequence of a single hazard grid, R represents the resolution of the DEM grid data, p represents the sampling interval, and i and n are positive integers.

[0018] Preferably, step S2 includes:

[0019] Design the random walk parameter L considering the movement direction in the Monte Carlo random walk process ctrl , this parameter is used to define the backward search distance, and the direction control grid of the current slip grid is located according to the backward search distance. The distance between the next slip grid and the direction control grid must be greater than the random walk parameter L ctrl , through the random walk parameter L ctrl Constrain the random walk direction of potential slip points in the DEM grid;

[0020] Design the random walk parameter R considering the terrain in the Monte Carlo random walk process max , this parameter is used to define the maximum rising elevation in the slip simulation process. The elevation difference between each slip grid and the previous slip grid during the slip simulation process cannot be greater than the random walk parameter R max , the random walk parameter R max is usually 1 / 5 of the resolution of the DEM grid data. Through the random walk parameter R max Constrain the unreasonable upward movement during the slip simulation process.

[0021] Preferably, in step S2, construct the random probability calculation formula for the current slip grid to slip to the target grid. The calculation formula is:

[0022]

[0023] where P px is the probability that the target grid around the current slip grid becomes the next slip grid, n is the number of target grids, β is the slope between the current slip grid and the target grid, f d and f β are the weight factors for adjusting the sliding direction and slope respectively. When the direction of the next slip grid is the same as the previous one, f d is d 2 . When the sliding direction changes by 45°, f d is d. When the sliding direction changes by 90°, f d is 1, and both f β and d are initialized input parameters.

[0024] Preferably, step S3 includes:

[0025] The design undulation factor is used to judge the sliding terrain. If the undulation factor is less than the slope threshold, based on the random probability calculation formula, a Monte Carlo flat ground random walk rule is constructed, and the control of the walk parameters considering the terrain is released, so as to realize the conversion of the Monte Carlo sliding simulation from the slope to the flat ground terrain.

[0026] Preferably, in step S3, a farthest sliding constraint rule considering the sliding height difference and area is constructed as the sliding truncation condition, and the sliding truncation condition is optimized according to historical landslide data, including:

[0027] Construct a sliding distance truncation formula considering the sliding height difference and the hidden danger area. Assume that the farthest sliding distance is a continuous smooth function of the sliding height difference and the hidden danger area, that is, the sliding truncation condition F. The Taylor expansion of this function at zero can obtain the multiple regression formula of the farthest sliding distance as:

[0028]

[0029] where x1 is the sliding height difference, x2 is the hidden danger area, Q n is the influence coefficient of x n and Q m is the cross influence coefficient of the sliding height difference and the hidden danger area on the sliding distance;

[0030] According to historical landslide data, the fitting formula of the sliding truncation condition F is obtained by multiple regression as:

[0031] F(x1,x2) = 3.089x1 0.47 + 4.258x2 0.43 - 173.476 (9)

[0032] Design the distance segmentation parameter L seg Calculate the farthest sliding distance during the sliding process, and divide the sliding path into sub-paths according to the distance segmentation parameter L seg The farthest sliding distance is the sum of the sub-paths, and the calculation formula is:

[0033]

[0034] where L is the current farthest sliding distance and n is the number of sub-paths divided.

[0035] Set the maximum number of random walks n max , and perform n max times of random sliding on the potential initial sliding points at the same time. If the current farthest sliding distance L exceeds the sliding truncation condition F or there is no target grid, end a sliding and store the number of times each grid is randomly selected for sliding.

[0036] Preferably, step S4 includes:

[0037] Using geospatial analysis technology, the frequency of each grid selected in the landslide path is counted. The random walk parameters considering terrain and movement direction are adjusted according to the historical landslide statistical data of the current area. This process is continuously iterated until the optimal effect is achieved. The optimal simulation result is output in GIS visualization. The grid frequency statistical data is converted into potential landslide sliding probability values, and the natural break method is used to classify the potential landslide sliding probability values to generate the potential landslide sliding range.

[0038] Preferably, step S4 includes:

[0039] S41. For the sliding influence grids triggered by the same hidden danger and all hidden danger sliding points, different methods are used to count the selected frequencies, and the calculation formula is:

[0040] S = max(s i ), i ∈ n (11)

[0041]

[0042] Among them, S represents the selected frequency of the same hidden danger, s i represents the selected frequency of the i-th sliding point in this hidden danger, S sum represents the selected frequency of all hidden danger sliding points, m represents the number of hidden dangers affecting the target grid, and i, j, n, m are positive integers;

[0043] S42. Convert the grid frequency statistical data into sliding influence probability values, and the calculation formula is:

[0044]

[0045] Among them, P represents the sliding influence probability value, S sum represents the selected frequency of all hidden danger sliding points, n max represents the maximum number of random walks;

[0046] S43. Use the natural break classification method to divide the sliding influence probability values into three influence range levels of high, medium, and low to generate the potential landslide sliding range.

[0047] Based on the same inventive concept, the present invention also provides an automatic generation system for potential landslide sliding range based on the Monte Carlo idea. The system includes:

[0048] An initial sliding point determination module, which is used to identify landslide hidden dangers based on multi-temporal radar image data, using radar interferometry and deep learning models, construct a landslide hidden danger database, and determine potential initial sliding point grids from the landslide hidden dangers by using a differential sampling strategy considering the scale of hidden dangers;

[0049] A sliding algorithm construction module for constructing a Monte Carlo ramp random sliding algorithm based on the potential initial sliding point grid, including: designing a walking parameter considering terrain and movement direction, simulating the longitudinal and lateral movement processes of landslides, and constructing a random probability calculation formula for the current sliding grid to slide to the target grid;

[0050] A ramp sliding simulation module for performing Monte Carlo ramp sliding simulation based on the Monte Carlo ramp random sliding algorithm, including: constructing a farthest sliding constraint rule considering sliding height difference and area as a sliding truncation condition, optimizing the sliding truncation condition according to historical landslide data, judging the sliding terrain according to the terrain undulation factor, and when the potential sliding point enters the bottom of the slope, designing a Monte Carlo flat random walking rule to continue simulating the landslide movement process until the sliding truncation condition is reached or there is no target grid;

[0051] A sliding range generation module for adjusting the walking parameter considering terrain and movement direction based on the historical landslide statistical data of the current area, counting the frequency of each grid in the sliding path being selected by the sliding simulation, converting the grid frequency statistical data into potential landslide sliding probability values and classifying them to generate a potential landslide sliding range.

[0052] Preferably, the initial sliding point determination module is specifically used for:

[0053] Using radar interferometry and deep learning models to identify landslide hazards in multi-temporal radar image data, and the calculation formula is:

[0054] L = E(F(D1)) + E(F(D2)) (1)

[0055] Where D1 and D2 respectively represent the radar satellite ascending orbit and descending orbit data in the multi-temporal radar image data, F represents calculating the surface deformation rate of the ascending orbit and descending orbit data using radar interferometry, E represents automatically identifying landslide hazards for the surface deformation rate using a deep learning model, and L represents the landslide hazard identification result integrating the ascending orbit and descending orbit data;

[0056] Constructing a landslide hazard database and performing vectorization, data statistics, and geometric screening normalization processing on the landslide hazard identification results;

[0057] Adopting a differential sampling strategy considering hazard scale to determine potential initial sliding point grids from the landslide hazards, and the calculation formula is:

[0058]

[0059] x i (m) = x i (mpR) (3)

[0060] Among them, X represents the grid of potential initial slip points, and x i represents the i-th hidden danger grid, n represents the number of identified hidden dangers, m represents the data sequence of a single hidden danger grid, R represents the resolution of the DEM grid data, p represents the sampling interval, and i and n are positive integers.

[0061] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0062] The present invention first uses InSAR technology to detect the microscopic deformation of the ground surface, identify the deformation aggregation area in the mountainous area, construct a landslide hidden danger data set, determine the potential landslide slip points, and then formulate the rules for the landslide process and the farthest slip distance constraint based on the Monte Carlo idea. Multiple random walk simulations are carried out on the slip points in the GIS environment to generate the prediction results of the potential slip range. The present invention utilizes the precursor information of mountain movement detected by InSAR technology to accurately locate the potential landslide hidden dangers, simulates the longitudinal and lateral movements of the potential landslide by referring to the Monte Carlo random walk idea, realizes the automatic prediction of the large-scale potential landslide slip range, improves the problems of traditional landslide slip prediction methods that require complex model parameters and have poor applicability in large-scale slip prediction tasks, can be quickly and effectively applied to engineering operations such as landslide hidden danger risk assessment, and is of great significance for improving the intelligent level of landslide disaster monitoring and early warning. Description of the Drawings

[0063] Figure 1 It is a schematic flow chart of a method for automatically generating the potential landslide slip range based on Monte Carlo provided by the present invention;

[0064] Figure 2 It is a technical flow chart of a method for automatically generating the potential landslide slip range based on Monte Carlo provided by the present invention;

[0065] Figure 3 It is a schematic diagram of the potential slip simulation process based on Monte Carlo random walk provided by the present invention;

[0066] Figure 4 It is a schematic diagram of the slip direction constraint based on Monte Carlo random walk provided by the present invention;

[0067] Figure 5 It is a schematic diagram of the slip distance calculation method based on Monte Carlo random walk provided by the present invention;

[0068] Figure 6 It is a schematic diagram of the simulation result of the potential landslide slip range in Zhouqu County provided by the present invention. Detailed Embodiments

[0069] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0070] The present invention will be further described in detail below with reference to the accompanying drawings.

[0071] As Figure 1-2 shown, an automatic generation method for the potential landslide sliding range based on Monte Carlo provided by an embodiment of the present invention includes:

[0072] S1. Based on multi-temporal radar image data, using radar interferometry and deep learning models to identify landslide hazards, constructing a landslide hazard database, and determining potential initial slip point grids from the landslide hazards by adopting a differential sampling strategy considering the hazard scale;

[0073] S2. Based on the potential initial slip point grids, constructing a Monte Carlo slope random sliding algorithm, including: designing a walking parameter considering the terrain and the movement direction, simulating the longitudinal and lateral movement processes of the landslide, and constructing a random probability calculation formula for the current slip grid to slide to the target grid;

[0074] S3. Based on the Monte Carlo slope random sliding algorithm, performing Monte Carlo slope sliding simulation, including: constructing a farthest sliding constraint rule considering the sliding height difference and area as the sliding truncation condition, optimizing the sliding truncation condition according to historical landslide data, judging the sliding terrain according to the terrain undulation factor, and when the potential slip point enters the bottom of the slope, designing a Monte Carlo flat random walking rule to continue simulating the landslide movement process until the sliding truncation condition is reached or there is no target grid;

[0075] S4. Adjusting the walking parameter considering the terrain and the movement direction based on the historical landslide statistical data of the current area, counting the frequency of each grid in the sliding path being selected in the sliding simulation, converting the grid frequency statistical data into potential landslide sliding probability values and classifying them to generate a potential landslide sliding range.

[0076] As Figure 3As shown in the figure, for step S1, the surface deformation rate is calculated based on multi-temporal radar images using radar interferometry (InSAR). A deep learning artificial intelligence model is used to identify landslide hazards. The hazard database is subjected to vectorization, data statistics, geometric screening, and other normalization processes to construct a landslide hazard database. A differential sampling strategy considering the hazard scale is adopted to determine potential landslide initial slip points from the landslide hazards, realizing the migration from InSAR detecting surface micro-motion information to potential slip points, and providing high-quality slip starting point data for simulating the slip range of potential landslides.

[0077] Step S1 includes:

[0078] S11. Use radar interferometry and a deep learning model to identify landslide hazards in multi-temporal radar image data. The calculation formula is:

[0079] L = E(F(D1)) + E(F(D2)) (1)

[0080] Where D1 and D2 respectively represent the radar satellite ascending orbit and descending orbit data in the multi-temporal radar image data, F represents calculating the surface deformation rate of the ascending orbit and descending orbit data using radar interferometry, E represents automatically identifying landslide hazards from the surface deformation rate using a deep learning model, and L represents the landslide hazard identification result integrating the ascending orbit and descending orbit data;

[0081] S12. Construct a landslide hazard database and perform vectorization, data statistics, and geometric screening normalization processes on the landslide hazard identification result;

[0082] S13. Adopt a differential sampling strategy considering the hazard scale to determine potential initial slip point grids from the landslide hazards. The calculation formula is:

[0083]

[0084] x i (m) = x i (mpR) (3)

[0085] Where X represents the potential initial slip point grid, x i represents the i-th hazard grid, n represents the number of identified hazards, m represents the data sequence of a single hazard grid, R represents the resolution of the DEM grid data, p represents the sampling interval, and i and n are positive integers.

[0086] For step S2, a ramp Monte Carlo landslide random walk algorithm is constructed. The longitudinal and lateral movement processes of sliding are simulated by designing the walk parameters and rules that take into account the terrain and movement direction. The random sliding probability calculation formula is used to achieve the random walk of the sliding points among the DEM grid data, and the walk parameters are initialized according to the DEM data resolution and the hidden danger scale information.

[0087] Step S2 includes:

[0088] S21. The direction of the sliding movement usually does not change violently. As Figure 4 shown, design the walk parameter L that takes into account the movement direction in the Monte Carlo random walk process ctrl , which is used to define the backward search distance. According to the backward search distance, the direction control grid of the current sliding grid is located. The distance between the next sliding grid and the direction control grid must be greater than the walk parameter L ctrl . The direction of the potential sliding points randomly walking in the DEM grid is constrained by the walk parameter L ctrl ;

[0089] S22. Design the walk parameter R that takes into account the terrain in the Monte Carlo random walk process max , which is used to define the maximum rising elevation in the sliding simulation process. The elevation difference between each sliding grid and the previous sliding grid during the sliding simulation process cannot be greater than the walk parameter R max . The walk parameter R max is usually 1 / 5 of the resolution of the DEM grid data. The unreasonable upward movement during the sliding simulation process is constrained by the walk parameter R max ;

[0090] S23. Construct the random probability calculation formula for the current sliding grid to slide to the target grid. The calculation formula is:

[0091]

[0092] where P px is the probability that the target grid around the current sliding grid becomes the next sliding grid, n is the number of target grids, β is the slope between the current sliding grid and the target grid, f d and f β are the weight factors for adjusting the sliding direction and slope respectively. When the direction of the next sliding grid is the same as the previous one, f d is d 2 . When the sliding direction changes by 45°, f d is d. When the sliding direction changes by 90°, f d is 1. Both f β and d are initialized input parameters.

[0093] In the embodiment of the present invention, for step S3, a farthest sliding constraint rule considering the sliding height difference and area is designed as the sliding truncation condition. The truncation condition can be continuously optimized through historical landslide data. According to the Monte Carlo random walk algorithm constructed in step two, a large number of random walks are simultaneously performed on the sliding starting point, and the sliding position is judged according to the terrain factor. When the sliding point enters the bottom of the slope, a flat ground random walk rule is designed to continue simulating the landslide movement process until the sliding truncation condition is reached or there are no target sliding grids during the random walk.

[0094] Step S3 includes:

[0095] S31. Design a terrain undulation factor to judge the sliding terrain. If the terrain undulation factor is less than the slope threshold, a Monte Carlo flat ground random walk rule is constructed based on the random probability calculation formula, and the control of the random walk parameters considering the terrain is released to realize the conversion of the Monte Carlo sliding simulation from the slope to the flat ground terrain;

[0096] S32. Construct a sliding distance truncation formula considering the sliding height difference and the hidden danger area. Assume that the farthest sliding distance is a continuous smooth function of the sliding height difference and the hidden danger area, that is, the sliding truncation condition F. Taylor expansion of this function at zero can obtain the multiple regression formula for the farthest sliding distance as:

[0097]

[0098] where x1 is the sliding height difference, x2 is the hidden danger area, Q n is the influence coefficient of x n , and Q m is the cross influence coefficient of the sliding height difference and the hidden danger area on the sliding distance;

[0099] The fitting formula of the sliding truncation condition F is obtained by multiple regression according to historical landslide data as:

[0100] F(x1,x2) = 3.089x1 0.47 + 4.258x2 0.43 - 173.476 (9)

[0101] S33. Since the random walk process will generate uneven sliding grids, directly using the sum of the sliding grid distances will cause a large deviation between the calculation result and the actual distance in the main sliding direction. Design a distance segmentation parameter L seg to calculate the farthest sliding distance during the sliding process. Divide the sliding path into sub-paths according to the distance segmentation parameter L seg . The farthest sliding distance is the sum of the sub-paths, as shown in Figure 5 . The calculation formula is:

[0102]

[0103] Wherein, L is the current farthest slip distance, and n is the number of sub-paths divided.

[0104] S34. Set the maximum number of random walks n max , and perform n max times of random slips on potential initial slip points simultaneously. If the current farthest slip distance L exceeds the slip cut-off condition F or there is no target grid, end one slip and store the number of times each grid is selected by random slip.

[0105] In the embodiment of the present invention, step S4 includes:

[0106] Use geospatial analysis technology to count the frequencies of grid slips selected in the slip path, adjust the random walk parameters considering terrain and movement direction according to the historical landslide statistical data of the current area, continuously iterate this process until the optimal effect is achieved, perform GIS visualization output on the optimal simulation result, convert the grid frequency statistical data into potential landslide slip probability values, and use the natural break point method to classify the potential landslide slip probability values to generate potential landslide slip ranges.

[0107] Step S4 includes:

[0108] S41. For the slip-impacted grids triggered by the same hidden danger and all hidden danger slip points, different methods are used to count the selection frequencies, and the calculation formula is:

[0109] S = max(s i ), i ∈ n (11)

[0110]

[0111] Wherein, S represents the selection frequency of the same hidden danger, s i represents the selection frequency of the i-th slip point in this hidden danger, S sum represents the selection frequency of all hidden danger slip points, m represents the number of hidden dangers affecting the target grid, and i, j, n, m are positive integers;

[0112] S42. Convert the grid frequency statistical data into slip impact probability values, and the calculation formula is:

[0113]

[0114] Wherein, P represents the slip impact probability value, S sum represents the selection frequency of all hidden danger slip points, n max represents the maximum number of random walks;

[0115] S43. Use the natural break point classification method to divide the slip impact probability values into three impact range levels: high, medium, and low.

[0116] As Figure 6 shown, the simulation results of the potential landslide sliding range in Zhouqu County. ①②③ indicated by the boxes in the left figure represent three landslide hazards identified by InSAR. The three upper figures respectively correspond to the influence probabilities of the three landslide hazards, and the three lower figures respectively correspond to the influence levels of the three landslide hazards. The potential landslide sliding range can be obtained intuitively.

[0117] The embodiment of the present invention also provides a system for automatically generating the potential landslide sliding range based on the Monte Carlo idea, including:

[0118] An initial sliding point determination module, configured to identify landslide hazards based on multi-temporal radar image data, using radar interferometry and deep learning models, construct a landslide hazard database, and determine potential initial sliding point grids from the landslide hazards by adopting a differential sampling strategy considering the hazard scale;

[0119] A sliding algorithm construction module, configured to construct a Monte Carlo slope random sliding algorithm based on the potential initial sliding point grids, including: designing a random walk parameter considering the terrain and the movement direction, simulating the longitudinal and lateral movement processes of the landslide, and constructing a random probability calculation formula for the current sliding grid to slide to the target grid;

[0120] A slope sliding simulation module, configured to perform Monte Carlo slope sliding simulation based on the Monte Carlo slope random sliding algorithm, including: constructing a farthest sliding constraint rule considering the sliding height difference and area as the sliding truncation condition, optimizing the sliding truncation condition according to historical landslide data, judging the sliding terrain according to the terrain undulation factor, and when the potential sliding point enters the bottom of the slope, designing a Monte Carlo random walk rule on the flat ground to continue simulating the landslide movement process until the sliding truncation condition is reached or there is no target grid;

[0121] A sliding range generation module, configured to adjust the random walk parameter considering the terrain and the movement direction based on the historical landslide statistical data of the current area, count the frequencies of each grid in the sliding path being selected by the sliding simulation, convert the grid frequency statistical data into potential landslide sliding probability values and classify them, and generate the potential landslide sliding range.

[0122] The present invention integrates the InSAR-detected landslide hazards into the potential landslide sliding simulation algorithm by designing a sampling strategy for the sliding starting point that takes into account the hazard scale, realizes the adaptation of the InSAR surface deformation information to the potential landslide driving force, and effectively solves the problem that the traditional InSAR technology can only detect the potential landslide sliding starting area but cannot evaluate the sliding influence range. By constructing a random sliding algorithm based on the Monte Carlo idea, the longitudinal and lateral sliding movement processes of potential landslides are effectively simulated. Combining the sliding constraint conditions that take into account the sliding area and height difference, the problem that it is difficult for traditional landslide influence range simulation methods to be extended to large-scale applications is solved, and the rapid and automatic prediction of the potential landslide sliding influence range is realized.

[0123] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A Monte Carlo-based method for automatically generating the potential landslide sliding range, characterized in that: The method comprises: S1. Based on multi-temporal radar image data, radar interferometry technology and deep learning models are used to identify landslide hazards, build a landslide hazard database, and use a differentiated sampling strategy that takes into account the hazard scale to determine the potential initial slip point grid from the landslide hazards; S2. Based on the potential initial sliding point grid, construct a Monte Carlo slope random sliding algorithm, including: designing walking parameters that take into account the terrain and movement direction, simulating the longitudinal and lateral movement process of the landslide, and constructing a random probability calculation formula for the current sliding grid sliding to the target grid; S3. Based on the Monte Carlo slope random sliding algorithm, a Monte Carlo slope sliding simulation is performed, including: constructing a farthest sliding constraint rule taking into account the sliding height difference and area as a sliding cutoff condition, optimizing the sliding cutoff condition according to historical landslide data, judging the sliding terrain according to the terrain undulation factor, and when the potential sliding point enters the bottom of the slope, designing a Monte Carlo flat random walk rule, and continuing to simulate the landslide movement process until the sliding cutoff condition is reached or the target grid does not exist; S4. Based on the historical landslide statistical data of the current area, the walking parameters taking into account the terrain and the movement direction are adjusted, the frequency of each grid in the sliding path being selected by the sliding simulation is counted, the grid frequency statistical data is converted into a potential landslide sliding probability value and classified, and the potential landslide sliding range is generated.

2. The method according to claim 1, characterized in that Step S1 includes: The radar interferometry technology and deep learning model are used to identify landslide hazards from multi-temporal radar image data. The calculation formula is: L=E(F(D1))+E(F(D2)) (1) Wherein, D1 and D2 represent the radar satellite ascending and descending orbit data in the multi-temporal radar image data, respectively; F represents the calculation of the surface deformation rate of the ascending and descending orbit data using radar interferometry technology; E represents the automatic identification of landslide hazards based on the surface deformation rate using a deep learning model; and L represents the landslide hazard identification result of the fusion of the ascending and descending orbit data; Constructing a landslide hazard database, and performing vectorization, data statistics, and geometric screening and normalization processing on the landslide hazard identification results; A differentiated sampling strategy that takes into account the scale of the hazard is used to determine the potential initial slip point grid from the landslide hazard. The calculation formula is: x i (m)=x i (mpR) (3) Where X represents the potential initial slip point grid, x i represents the i-th hidden danger grid, n represents the number of identified hidden dangers, m represents a single hidden danger grid data sequence, R represents the resolution of the DEM grid data, p represents the sampling interval, and i and n are positive integers.

3. The method according to claim 2, characterized in that Step S2 includes: Design the walking parameter L that takes into account the direction of movement during the Monte Carlo random walk ctrl This parameter is used to define the backward search distance. The direction control grid of the current sliding grid is located according to the backward search distance. The distance between the sliding grid and the direction control grid must be greater than the walk parameter L for the next sliding grid. ctrl , through the walk parameter L ctrl Constrain the direction of random walk of potential sliding points in the DEM grid; Designing the walk parameter R in the Monte Carlo random walk process to take into account the terrain max This parameter is used to define the highest rise elevation during the sliding simulation process. The elevation difference between each sliding grid and the previous sliding grid during the sliding simulation process cannot be greater than the walking parameter R max , walk parameter R max Usually it is 1 / 5 of the resolution of the DEM raster data. max Unreasonable upward motion during constrained sliding simulation.

4. The method according to claim 1, characterized in that: In step S2, a random probability calculation formula for the current sliding grid sliding to the target grid is constructed, and the calculation formula is: Among them, P px is the probability that the target grid around the current sliding grid becomes the next sliding grid, n is the number of target grids, β is the slope between the current sliding grid and the target grid, and f d and f β are weight factors for adjusting the sliding direction and slope, respectively. When the direction of the next sliding grid is the same as the previous one, f d is d 2 , when the sliding direction changes by 45°, f d is d, when the sliding direction changes by 90°, f d is 1, f β and d are initialization input parameters.

5. The method according to claim 1, characterized in that Step S3 includes: A terrain undulation factor is designed to determine the slip terrain. If the terrain undulation factor is less than the slope threshold, a Monte Carlo flat random walk rule is constructed based on the random probability calculation formula to release the control of the walk parameters taking the terrain into consideration, thereby realizing the conversion of the Monte Carlo slip simulation from the slope to the flat terrain.

6. The method according to claim 1, characterized in that In step S3, a farthest sliding constraint rule taking into account the sliding height difference and the area is constructed as a sliding cutoff condition, and the sliding cutoff condition is optimized according to historical landslide data, including: A slip distance cutoff formula that takes into account the slip height difference and the potential hazard area is constructed. Assuming that the farthest slip distance is a continuous smooth function of the slip height difference and the potential hazard area, that is, the slip cutoff condition F, the Taylor expansion of this function at the zero point can obtain the multivariate regression formula of the farthest slip distance: Among them, x1 is the slip height difference, x2 is the hidden danger area, Q n For x n The influence coefficient, Q m is the cross-influence coefficient of sliding height difference and hidden danger area on sliding distance; According to the historical landslide data, the fitting formula of the slip cutoff condition F is obtained through multivariate regression: <h2 style=";text-align:left;direction:ltr">F(x1,x2) = 3.089x1<h2 style=";text-align:left;direction:ltr"> 0.47 <h2 style=";text-align:left;direction:ltr"> +4.258x2<h2 style=";text-align:left;direction:ltr"> 0.43 <h2 style=";text-align:left;direction:ltr"> -173.476 (9) Design distance segmentation parameter L seg Calculate the longest sliding distance during the sliding process and divide the parameter L by distance seg The sliding path is divided into sub-paths, and the farthest sliding distance is the sum of the sub-paths. The calculation formula is: Among them, L is the current farthest sliding distance, and n is the number of divided sub-paths. Set the maximum number of random walks n max , and perform n simultaneous operations on potential initial slip points max If the current farthest sliding distance L exceeds the sliding cutoff condition F or there is no target grid, the sliding is ended and the number of times each grid is selected by random sliding is stored.

7. The method according to claim 1, characterized in that Step S4 includes: The frequency of each grid in the slip path being selected for slip simulation is counted using geospatial analysis technology. The walking parameters that take into account the terrain and movement direction are adjusted according to the historical landslide statistics in the current area. This process is iterated continuously until the optimal effect is achieved. The optimal simulation results are output through GIS visualization, and the grid frequency statistics are converted into potential landslide slip probability values. The natural breakpoint method is used to classify the potential landslide slip probability values ​​to generate the potential landslide slip range.

8. The method according to claim 1, characterized in that Step S4 includes: S41. For the slippage influence grids triggered by the same hidden danger and all hidden danger slippage points, different methods are used to count the selection frequency, and the calculation formula is: S=max(s i ),i∈n (11) Among them, S represents the selection frequency of the same hidden danger, s i represents the selection frequency of the i-th slip point in the hidden danger, S sum represents the selection frequency of all potential danger slip points, m represents the number of potential dangers affecting the target grid, n represents the number of sampled slip points in a single potential danger, i, j, n are positive integers; S42, converting grid frequency statistical data into slip impact probability values, the calculation formula is: Among them, P represents the slip impact probability value, S sum Indicates the selection frequency of all potential slip points, n max represents the maximum number of random walks; S43. The natural breakpoint classification method is used to divide the slip impact probability value into three impact range levels: high, medium and low, to generate the potential landslide slip range.

9. A system for automatically generating the potential landslide sliding range based on Monte Carlo theory, used to implement the method described in any one of claims 1 to 8, characterized in that: The system comprises: An initial slip point determination module is used to identify landslide hazards based on multi-temporal radar image data using radar interferometry technology and deep learning models, build a landslide hazard database, and determine potential initial slip point grids from the landslide hazards using a differentiated sampling strategy that takes into account the hazard scale; A sliding algorithm construction module is used to construct a Monte Carlo slope random sliding algorithm based on the potential initial sliding point grid, including: designing walking parameters that take into account the terrain and movement direction, simulating the longitudinal and lateral movement process of the landslide, and constructing a random probability calculation formula for the current sliding grid sliding to the target grid; A slope sliding simulation module is used to perform Monte Carlo slope sliding simulation based on the Monte Carlo slope random sliding algorithm, including: constructing a farthest sliding constraint rule that takes into account the sliding height difference and area as a sliding cutoff condition, optimizing the sliding cutoff condition according to historical landslide data, judging the sliding terrain according to the terrain undulation factor, and when the potential sliding point enters the bottom of the slope, designing a Monte Carlo flat random walk rule to continue simulating the landslide movement process until the sliding cutoff condition is reached or the target grid does not exist; The sliding range generation module is used to adjust the walking parameters taking into account the terrain and movement direction based on the historical landslide statistics of the current area, count the frequency of each grid in the sliding path being selected by the sliding simulation, convert the grid frequency statistics into potential landslide sliding probability values ​​and classify them, and generate the potential landslide sliding range.

10. The system according to claim 9, characterized in that The initial slip point determination module is specifically used for: The radar interferometry technology and deep learning model are used to identify landslide hazards from multi-temporal radar image data. The calculation formula is: L=E(F(D1))+E(F(D2)) (1) Wherein, D1 and D2 represent the radar satellite ascending and descending orbit data in the multi-temporal radar image data, respectively; F represents the calculation of the surface deformation rate of the ascending and descending orbit data using radar interferometry technology; E represents the automatic identification of landslide hazards based on the surface deformation rate using a deep learning model; and L represents the landslide hazard identification result of the fusion of the ascending and descending orbit data; Constructing a landslide hazard database, and performing vectorization, data statistics, and geometric screening and normalization processing on the landslide hazard identification results; A differentiated sampling strategy that takes into account the scale of the hazard is used to determine the potential initial slip point grid from the landslide hazard. The calculation formula is: x i (m)=x i (mpR) (3) Where X represents the potential initial slip point grid, x i represents the i-th hidden danger grid, n represents the number of identified hidden dangers, m represents a single hidden danger grid data sequence, R represents the resolution of the DEM grid data, p represents the sampling interval, and i and n are positive integers.

Images