Slope stability prediction method and system based on high-resolution optical satellite image

Abstract

The invention discloses a slope stability prediction method and system based on a high-resolution optical satellite image, and belongs to the technical field of rock soil, and the system comprises a database module, a database logic control module, a data processing module and a prediction module. According to the invention, human activity factors are introduced into slope stability analysis for the first time, so that prediction accuracy and robustness are improved; according to the method, a scientific and accurate prediction result can be provided, more effective land resource planning and management are facilitated, disaster risks are reduced, sustainable utilization of land resources is promoted, and the method has remarkable social and economic values.

Description

Technical Field

[0001] The present invention relates to the field of geotechnical technology, and in particular to a slope stability prediction method and system based on high-resolution optical satellite images. Background Art

[0002] Global climate change has increased the frequency and severity of extreme climate events (such as earthquakes, typhoons, and floods), and has led to a large number of geological disasters around the world, causing a large number of casualties and economic losses every year. Although local governments have tried to prevent slope disasters by establishing geological disaster early warning systems and weather forecast systems, annual economic losses are still increasing. One of the major reasons for this situation is that human engineering activities are constantly changing the geological environment.

[0003] To address this problem, researchers often use land use and land cover (LULC) to describe the impact of human engineering activities, which include natural elements (such as water bodies, forests, cultivated land, and grasslands) and human-modified surfaces (such as construction activity areas). To date, many studies have analyzed the relationship between LULC and slope failure and regional landslide susceptibility from different perspectives, but most studies have limitations in the dynamic simulation of LULC's spatiotemporal changes. In addition, one point that is often overlooked in existing studies is that road distance is usually regarded as a static factor without considering its dynamic characteristics. However, in reality, road construction directly affects the stability of slopes by changing the cut slopes or surface water runoff patterns. Few studies have explicitly explored how human engineering activities specifically affect this relationship.

[0004] Human engineering activities, especially those related to the construction of transportation networks, may significantly change land use patterns, thereby affecting the temporal and spatial distribution of slopes. However, through the above analysis, it can be found that most of the existing slope assessment methods focus on static analysis, and few methods can achieve dynamic analysis of slope stability considering human engineering activities. Especially if future changes are taken into account, how to predict the temporal and spatial impact of human engineering activities on slope stability is still a technical challenge. Summary of the invention

[0005] In order to solve the above problems, the purpose of the present invention is to provide a slope stability prediction technology based on high-resolution optical satellite images, use high-precision Planet optical remote sensing images to obtain a list of landslides within a specific time period, use the patch generation land use simulation (PLUS) model to predict future land use, and finally calculate the spatial areas prone to landslides under future land use conditions based on intelligent algorithms, thereby providing a reference for subsequent government decision-making and land use planning.

[0006] In order to achieve the above technical objectives, the present invention provides a slope stability prediction method based on high-resolution optical satellite images, comprising:

[0007] A database module for storing high-resolution optical satellite images, data sets affecting slope stability generated based on the high-resolution optical satellite images, and data generated by the system;

[0008] Database logic control module, used to control the execution of logic operations between the database module and other functional modules of the system;

[0009] A data processing module for obtaining a data set affecting slope stability based on the selected LULC data according to high-resolution optical satellite images;

[0010] The prediction module is used to predict the slope stability according to the intelligent algorithm model based on the stacking strategy after training with the data set.

[0011] Preferably, the data processing module is also used to call the high-resolution optical satellite image of the database module through the database logic control module, obtain LULC data according to the called high-resolution optical satellite image, and analyze and obtain slope influencing factors, including static factors and dynamic shadows, and generate a data set according to the data corresponding to the static factors and the dynamic factors, and send it to the database module, wherein the static factors include elevation, slope, aspect, plane curvature, profile curvature, lithology and soil type; the dynamic factors include land use and distance to the road.

[0012] Preferably, the data processing module is further used to analyze and obtain dynamic factors and static factors based on the LULC data and driving factor data of the image corresponding to the landslide occurrence list.

[0013] Preferably, the data processing module is also used to generate LULC data based on land use, GDP, population density, distance to primary roads, distance to secondary roads, distance to tertiary roads, distance to governments, distance to water bodies, soil type, average annual rainfall, average annual temperature and elevation.

[0014] Preferably, the data processing module is also used to obtain a predicted land use distribution map based on the PLUS model according to land development probability, land demand, conversion matrix and neighborhood weight, wherein the conversion matrix represents the conversion rules between land use types, and the neighborhood factor parameters represent the expansion intensity of each land type, reflecting the expansion capacity of each land type under the influence of spatial driving factors.

[0015] Preferably, the data processing module is further used to construct a data set based on LULC data of multiple different years, wherein the multiple different years include at least two adjacent years.

[0016] Preferably, the data processing module is also used to obtain a list of landslides within a specific time period based on the called high-resolution optical satellite images and the Planet optical remote sensing images, determine the landslide area within the time period by visual interpretation, and obtain the selected LULC data based on the landslide area.

[0017] Preferably, the data processing module is also used to select driving factors based on the land use distribution map, according to the land use expansion in a specific time period, and according to the land expansion analysis strategy, to obtain the development probability and factor contribution of each land use as driving factor data.

[0018] Preferably, the prediction module is also used to construct an intelligent algorithm model based on one or more of the k-nearest neighbor algorithm, random forest model, support vector machine model, gradient boosting tree model, Gaussian mixture model, and logistic regression algorithm.

[0019] The present invention discloses the following technical effects:

[0020] The present invention uses high-precision Planet optical remote sensing images to obtain a list of landslides within a certain period of time, uses the PLUS model to accurately predict future land use changes, and then accurately calculates the slope prone areas by integrating random forests, extreme gradient boosting and lightweight gradient boosting stacking technology. This method incorporates human activity factors into slope stability analysis for the first time, improving prediction accuracy and robustness. In addition, the present invention can provide scientific and accurate prediction results, which is conducive to more effective land resource planning and management, reducing disaster risks, and promoting the sustainable use of land resources, and has significant social and economic value. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0022] Figure 1 is the test result of the smart model shown in the embodiment of the present invention;

[0023] Figure 2 It is a schematic diagram of the system structure of the present invention. DETAILED DESCRIPTION

[0024] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0025] like Figure 1-2 As shown, the invention provides a slope stability prediction system based on high-resolution optical satellite images, comprising:

[0026] A database module for storing high-resolution optical satellite images, data sets affecting slope stability generated based on the high-resolution optical satellite images, and data generated by the system;

[0027] Database logic control module, used to control the execution of logic operations between the database module and other functional modules of the system;

[0028] A data processing module for obtaining a data set affecting slope stability based on the selected LULC data according to high-resolution optical satellite images;

[0029] The prediction module is used to predict the slope stability according to the intelligent algorithm model based on the stacking strategy after training with the data set.

[0030] Further preferably, the data processing module of the spatiotemporal prediction system mentioned in the present invention is also used to call the high-resolution optical satellite image of the database module through the database logic control module, obtain LULC data according to the called high-resolution optical satellite image, and analyze and obtain slope influencing factors, including static factors and dynamic shadows, and generate a data set according to the data corresponding to the static factors and the dynamic factors, and send it to the database module, wherein the static factors include elevation, slope, aspect, plane curvature, profile curvature, lithology and soil type; the dynamic factors include land use and distance to the road.

[0031] Further preferably, the data processing module of the spatiotemporal prediction system mentioned in the present invention is also used to analyze and obtain dynamic factors and static factors based on the LULC data and driving factor data of the images corresponding to the landslide occurrence list.

[0032] Further preferably, the data processing module of the spatiotemporal prediction system mentioned in the present invention is also used to generate LULC data based on land use, GDP, population density, distance to primary roads, distance to secondary roads, distance to tertiary roads, distance to governments, distance to water bodies, soil type, average annual rainfall, average annual temperature and elevation.

[0033] Further preferably, the data processing module of the spatiotemporal prediction system mentioned in the present invention is also used to obtain a predicted land use distribution map based on the PLUS model, according to land development probability, land demand, conversion matrix and neighborhood weight, wherein the conversion matrix represents the conversion rules between land use types, and the neighborhood factor parameters represent the expansion intensity of each land type, reflecting the expansion capacity of each land type under the influence of spatial driving factors.

[0034] Further preferably, the data processing module of the spatiotemporal prediction system mentioned in the present invention is also used to construct a data set based on LULC data of multiple different years, wherein the multiple different years include at least two adjacent years.

[0035] Further preferably, the data processing module of the spatiotemporal prediction system mentioned in the present invention is also used to obtain a list of landslides within a specific time period based on the called high-resolution optical satellite images and the Planet optical remote sensing images, determine the landslide area within the time period by visual interpretation, and obtain the selected LULC data based on the landslide area.

[0036] Further preferably, the data processing module of the spatiotemporal prediction system mentioned in the present invention is also used to select driving factors based on the land use distribution map, according to the land use expansion in a specific time period, and according to the land expansion analysis strategy, to obtain the development probability and factor contribution of each land use as driving factor data.

[0037] Further preferably, the prediction module of the spatiotemporal prediction system mentioned in the present invention is also used to construct an intelligent algorithm model based on one or more of the k-nearest neighbor algorithm, random forest model, support vector machine model, gradient boosting tree model, Gaussian mixture model, and logistic regression algorithm.

[0038] Example: In coastal mountainous areas, which are affected by rainfall-induced landslides every year, land use changes (e.g., deforestation, agricultural and urban area expansion) and road network expansion have exacerbated the occurrence of slope hazards. In addition, with the rapid development of urbanization and engineering construction, human engineering activities have caused significant changes in land use and cover, further changing the spatial distribution characteristics of landslides.

[0039] In addition, human engineering activities not only change land use and land cover patterns, but also change surface water flow paths and soil moisture conditions, which may be key factors affecting slope stability. Therefore, this invention will also explore how human engineering activities, especially road construction, further affect slope stability by changing surface water flow and soil moisture conditions. This invention aims to fill this research gap and propose a novel method for predicting the spatiotemporal impact of human engineering activities on slope stability, in order to provide more effective technical support for preventing and mitigating slope disasters.

[0040] Based on this, the present invention provides a method for predicting the temporal and spatial impact of human engineering activities on slope stability, comprising the following steps:

[0041] S1: Use high-precision optical remote sensing images to obtain a list of landslides within a certain period of time

[0042] S2: Based on the PLUS model and historical LULC data, the LULC change pattern of the study area was established and verified;

[0043] S3: Based on the LULC change pattern of the study area obtained in S2, the stacking-basedRF-XGB-LGBM model is used to predict the spatial areas prone to landslides under future land use conditions.

[0044] Step S1 uses two high-resolution optical satellite images (3-meter resolution) from Planet to identify landslides in the time period between the two images by visual interpretation. The two images cover the entire study area with low cloud coverage.

[0045] Step S2 includes the following steps:

[0046] ① Changes in land use and land cover in a region are inseparable from the local natural environment, humanities and social factors. Therefore, to predict land use and land cover, it is necessary to first determine its important driving factors. Prepare LULC data and driving factor data for many years based on ArcGIS software, where LULC data must be from at least two different years, including: land use, GDP, population density, distance to primary roads, distance to secondary roads, distance to tertiary roads, distance to government, distance to water bodies, soil type, average annual rainfall, average annual temperature and elevation.

[0047] ② The land use expansion in a certain period of time was extracted according to the land expansion module in the PLUS software. Then, combined with the land expansion analysis strategy module, 11 driving factors were selected to generate the development probability and factor contribution of each land use.

[0048] ③ Based on the land use data of a certain time period mentioned above, the Markov chain in the PLUS software predicts the demand for various types of land in the subsequent time period; the CARS module combines land development probability, land demand, conversion matrix, and neighborhood weight to obtain a predicted land use distribution map. The conversion matrix represents the conversion rules between land use types. When one land use type cannot be replaced by another, the corresponding value of the matrix is ​​0, otherwise it is 1. The conversion matrix takes into account the characteristics of the socio-economic development and natural environment of the study area, and restricts the conversion of water systems to other land use development scenarios. The neighborhood factor parameter represents the expansion intensity of each land type, reflecting the expansion capacity of each land type under the influence of spatial driving factors. The parameter range is 0-1. The closer it is to 1, the stronger its expansion capacity. It is calculated by the expansion area ratio of each land type, and the formula is as follows:

[0049]

[0050] In the formula, X i is the neighborhood weight parameter of type i; ΔTA i is the change in land use type i; ΔTA max and ΔTA max are the maximum and minimum values ​​of land use extension area respectively.

[0051] ④ The present invention selects the initial time of a certain time period as the benchmark starting time, and uses the development probability of each land use during a certain time period to predict the land use at the end of a certain time period. The verification module in the PLUS software is used to input the actual land use data and predicted land use data at the end of a certain time period, and verify the results through the Kappa coefficient to evaluate the simulation results. Generally, a Kappa coefficient > 0.8 is considered to provide almost perfect simulation results. On this basis, the present invention further simulates the land use of another time period following a certain time period.

[0052] Step S3 includes the following steps:

[0053] ① The factors affecting the slope preparation include:

[0054] (1) Static factors: elevation, slope, aspect, plan curvature, profile curvature, lithology, soil type;

[0055] (2) Dynamic factors: land use and distance to roads. The land use data used the land use predicted by the PLUS model for another time period; the road network data used the road network distribution of the study area for a certain time period. The Euclidean distance in the ArcGIS platform was used to calculate the road distance. All spatial variables were resampled to a resolution of 30 m.

[0056] ②The impact factor dataset was randomly divided into 70% training set and 30% test set. Then the stacking strategy was used in Python environment to combine RF, XGBoost and LightGBM models. The LogisticRegression model was used as a secondary learner. The area under the subject curve was used to test the final stacking hybrid model prediction results.

[0057] ③ After determining the optimal model, the present invention uses dynamic factors to analyze the temporal and spatial impact of human engineering activities on slope stability.

[0058] Embodiment: (1) Obtain time-series Planet optical remote sensing images related to the typhoon "Megi" event. Ensure that these images include time points before and after the typhoon event (such as August 2016 and November 2016) and have low cloud coverage. Use the acquired optical remote sensing images to compare the surface changes before and after the typhoon event, identify and calibrate the slope area, and thus generate a slope list. This list includes the location, area, and shape attributes of each slope.

[0059] (2) Patch Generation Land,Use Simulation (PLUS) Model. This model proposes a new land expansion analysis strategy that can avoid the exponential growth of transformation types with categories. Specifically, a stochastic model is used to analyze the driving relationship between land use expansion and conditional factors in two periods, thereby deriving the development probability of each land use type and the weight of each conditional factor. The PLUS model is used to better evaluate the evolutionary process of land use and predict future changes in land use distribution, which includes the following steps:

[0060] ① Prepare driving factor data including: land use, GDP, population density, distance to primary roads, distance to secondary roads, distance to tertiary roads, distance to government, distance to water bodies, soil type, average annual rainfall, average annual temperature and elevation.

[0061] ② The land use expansion from 2010 to 2020 was extracted according to the land expansion module in the PLUS software. Then, combined with the land expansion analysis strategy module, 11 driving factors were selected to generate the development probability and factor contribution of each land use.

[0062] ③ Based on the land use data in 2010 and 2020, the Markov chain in the PLUS software predicts the demand for various types of land in 2030; the CARS module combines land development probability, land demand, conversion matrix, and neighborhood weight to obtain the predicted land use distribution map. The conversion matrix represents the conversion rules between land use types. When one land use type cannot be replaced by another, the corresponding value of the matrix is ​​0, otherwise it is 1. The conversion matrix takes into account the characteristics of the socio-economic development and natural environment of the study area, and restricts the transformation of water systems into other land use development scenarios (Table 1). The neighborhood factor parameter represents the expansion intensity of each land type, reflecting the expansion capacity of each land type under the influence of spatial driving factors. The parameter range is 0-1. The closer it is to 1, the stronger its expansion capacity. It is calculated by the expansion area ratio of each land type, and the formula is as follows:

[0063]

[0064] In the formula, X i is the neighborhood weight parameter of type i; ΔTA i is the change in land use type i; ΔTA max and ΔTA max are the maximum and minimum values ​​of land use extension area respectively.

[0065] Table 1 Land use conversion matrix

[0066] arable land forest grassland Waters Building land arable land 1 1 1 0 1 forest 1 1 1 0 1 grassland 1 1 1 0 1 Waters 0 0 0 1 0 Building land 0 0 0 0 1

[0067] ④ The present invention selects 2010 as the benchmark starting time, and uses the development probability of each land use from 2010 to 2020 to predict the land use in 2020. The verification module in the PLUS software is used to input the actual land use data and predicted land use data in 2020, and verify its results through the Kappa coefficient to evaluate the simulation results. Generally, a Kappa coefficient of >0.8 is considered to provide almost perfect simulation results. On this basis, the present invention further simulates the land use in 2030 and 2060.

[0068] (3) Human engineering activities, especially those related to the construction of transportation networks, may significantly change land use patterns, thereby affecting the spatiotemporal distribution of slopes. In addition, most of the existing slope susceptibility assessment methods focus on static analysis, and few methods can achieve the prediction of the dynamic relationship between human engineering activities and slope susceptibility. With the rapid development of artificial intelligence and computer science, machine learning models have been widely used for slope spatial prediction. However, these models have a significant limitation: the optimal model may be different in different study areas. Due to differences in model operation strategies and sensitivity to data sets, it is difficult to find a model with strong generalization ability that performs well in multiple study areas. The "strong and different" RF, XGBoost, and LightGBM models are very suitable for further model integration to achieve better models, which specifically include the following steps:

[0069] ① The slope influencing factors include: (1) static factors, elevation, slope, aspect, plan curvature, profile curvature, lithology, soil type; (2) dynamic factors, land use and distance to roads. The land use data used the land use predicted by the PLUS model in 2030 and 2060; the road network data used the road network distribution in the study area from 2018 to 2020. The Euclidean distance in the ArcGIS platform was used to calculate the road distance. All spatial variables were resampled to a resolution of 30m.

[0070] ②The impact factor dataset was randomly divided into 70% training set and 30% test set. Then the stacking strategy was used in Python environment to combine RF, XGBoost and LightGBM models. The Logistic Regression model was used as a secondary learner. The area under the subject curve was used to test the final stacking hybrid model prediction results, such as Figure 1 shown.

[0071] ③The stacking-based RF-XGB-LGBM model analyzes the spatiotemporal impact of human engineering activities on slope stability by changing dynamic factors. Specifically, the long-term prediction takes into account the simultaneous changes in roads and land use. The area within 200m of the urbanized area is defined as the direct impact area. The probability of occurrence is divided into very low probability (0≤S<0.05), low probability (0.05≤S<0.35), medium probability (0.35≤S<0.75) and high probability (0.75≤S≤1.00).

[0072] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0073] In the description of the present invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0074] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A slope stability prediction system based on high-resolution optical satellite images, characterized in that: include: A database module, used to store high-resolution optical satellite images, data sets affecting slope stability generated based on the high-resolution optical satellite images, and data generated by the system; A database logic control module, used to control the execution of logic operations between the database module and other functional modules of the system; A data processing module, for obtaining a data set affecting slope stability according to the high-resolution optical satellite image and the selected LULC data; The prediction module is used to predict the slope stability according to the intelligent algorithm model based on the stacking strategy after being trained with the data set.

2. The slope stability prediction system based on high-resolution optical satellite images according to claim 1 is characterized by: The data processing module is also used to call the high-resolution optical satellite image of the database module through the database logic control module, obtain LULC data based on the called high-resolution optical satellite image, and analyze and obtain slope influencing factors, including static factors and dynamic shadows, and generate the data set based on the data corresponding to the static factors and the dynamic factors, and send it to the database module, wherein the static factors include elevation, slope, aspect, plane curvature, profile curvature, lithology and soil type; the dynamic factors include land use and distance to the road.

3. The slope stability prediction system based on high-resolution optical satellite images according to claim 2 is characterized by: The data processing module is also used to analyze and obtain dynamic factors and static factors based on the LULC data and driving factor data of the image corresponding to the landslide occurrence list.

4. The slope stability prediction system based on high-resolution optical satellite images according to claim 3 is characterized by: The data processing module is also used to generate LULC data based on land use, GDP, population density, distance to primary roads, distance to secondary roads, distance to tertiary roads, distance to governments, distance to water bodies, soil type, average annual rainfall, average annual temperature and elevation.

5. The slope stability prediction system based on high-resolution optical satellite images according to claim 4 is characterized by: The data processing module is also used to obtain a predicted land use distribution map based on the PLUS model according to land development probability, land demand, conversion matrix and neighborhood weight, wherein the conversion matrix represents the conversion rules between land use types, and the neighborhood factor parameters represent the expansion intensity of each land type, reflecting the expansion capacity of each land type under the influence of spatial driving factors.

6. The slope stability prediction system based on high-resolution optical satellite images according to claim 5 is characterized by: The data processing module is further used to construct the data set according to LULC data of multiple different years, wherein the multiple different years include at least two adjacent years.

7. The slope stability prediction system based on high-resolution optical satellite images according to claim 6 is characterized by: The data processing module is also used to obtain a list of landslides within a specific time period based on the called high-resolution optical satellite images and the Planet optical remote sensing images, determine the landslide area within the time period by visual interpretation, and obtain the selected LULC data based on the landslide area.

8. The slope stability prediction system based on high-resolution optical satellite images according to claim 7 is characterized by: The data processing module is also used to select driving factors based on the land use distribution map, according to the land use expansion in a specific time period, and according to the land expansion analysis strategy, to obtain the development probability and factor contribution of each land use as the driving factor data.

9. The slope stability prediction system based on high-resolution optical satellite images according to claim 8, characterized in that: The prediction module is also used to construct an intelligent algorithm model based on one or more of a k-nearest neighbor algorithm, a random forest model, a support vector machine model, a gradient boosting tree model, a Gaussian mixture model, and a logistic regression algorithm.

10. The slope stability prediction method of the slope stability prediction system based on high-resolution optical satellite images according to any one of claims 1 to 9, characterized in that: The following steps are involved: Based on high-resolution optical satellite images, a data set that affects slope stability is obtained, and the slope stability is predicted through a stacking strategy.

Images