Landslide susceptibility assessment method and system based on second-order environmental similarity constraint

Abstract

The invention relates to a landslide susceptibility evaluation method and system based on second-order environmental similarity constraint. The method comprises the following steps: determining a landslide occurrence area and collecting landslide disaster-causing characteristic factors; determining a slope object of the region through DEM elevation, and extracting slope features; dividing the slope body features into macroscopic environment features and local environment features, calculating a first-order environment similarity between the landslide masses through the macroscopic environment features, determining a landslide mass set with similar macroscopic environments through the first-order environment similarity, and calculating a second-order environment similarity in the landslide mass set through the local environment features; classifying the landslide mass according to the second-order environment similarity, and constructing an intra-class susceptibility database of a corresponding class according to the landslide mass data of each class; and correspondingly constructing a landslide susceptibility evaluation model based on a deep learning model according to the in-class susceptibility database of each category, and performing regional landslide susceptibility evaluation through the landslide susceptibility evaluation model of each category.

Description

TECHNICAL FIELD

[0001] The application relates to a landslide susceptibility assessment method and system based on a second-order environment similarity constraint, and belongs to the technical field of landslide disaster prevention. BACKGROUND

[0002] Landslides are a common type of geological disaster worldwide, and their occurrence not only destroys topography and ecosystems, but also can cause serious casualties and economic losses. Therefore, how to achieve high-precision landslide susceptibility assessment at the regional scale has always been a core problem in the prevention and control of geological disasters. Existing data-driven landslide susceptibility assessment methods usually regard all slope bodies in the study area as homogeneous individuals, and unified modeling and prediction are performed through a single model. However, in the actual natural environment, different slope bodies are often subject to significant differences in topography, hydrology, strata and human activity conditions, and their disaster-pregnant mechanisms differ significantly. If this "environmental heterogeneity" is ignored, unified modeling may not only mask the effects of key trigger factors, but also lead to a significant decrease in prediction accuracy in some areas, making it difficult to meet the needs of refined disaster prevention and mitigation. SUMMARY

[0003] In order to solve the problems existing in the prior art, the application provides a landslide susceptibility assessment method and system based on a second-order environment similarity constraint, which considers both the local and overall environmental characteristics of slope bodies, constructs a second-order environment similarity metric, and classifies landslide bodies that have occurred into different categories based on the similarity. Then, independent landslide susceptibility assessment models are constructed for each category, and a multi-model is used to assess the landslide susceptibility of the region. This method not only enables personalized risk assessment of different environmental categories of slope bodies, but also improves the accuracy and spatial applicability of landslide susceptibility prediction as a whole, providing more scientific technical support for landslide disaster prevention.

[0004] The technical solutions of the application are as follows:

[0005] In one aspect, the application provides a landslide susceptibility assessment method based on a second-order environment similarity constraint, which includes the following steps:

[0006] Determine the landslide occurrence area and collect landslide disaster-causing characteristic factors;

[0007] Determine the slope body objects in the region through DEM elevation, extract the slope body characteristics through the types of landslide disaster-causing characteristic factors, and define the slope body objects intersecting with the landslide region as landslide bodies, and vice versa as non-landslide bodies;

[0008] The slope body features are divided into macro-environmental features and local-environmental features, the first-order environmental similarity between the slope bodies is calculated through the macro-environmental features, the slope body set with the macro-environmental similarity is determined through the first-order environmental similarity, and the second-order environmental similarity is calculated through the local-environmental features in the slope body set;

[0009] The slope bodies are classified according to the second-order environmental similarity, an equal amount of non-slope bodies is randomly extracted according to the number of each type of slope body after division, and the in-class susceptibility database of the corresponding type is constructed according to the slope body data and the extracted non-slope body data of each type;

[0010] The landslide susceptibility evaluation model based on the deep learning model is constructed according to the in-class susceptibility database of each type respectively, and the landslide susceptibility of the region is evaluated through the landslide susceptibility evaluation model of each type.

[0011] As a preferred embodiment, the step of determining the slope body object through the DEM elevation and extracting the hierarchical features of the slope body according to the types of the landslide disaster-causing feature factors comprises:

[0012] The DEM elevation data is preprocessed, the slope and the aspect of each grid cell are calculated after the preprocessing, and the maximum grid cell connected block with the slope and the aspect within a preset tolerance range is taken as a slope body;

[0013] The grid values of the landslide disaster-causing feature factors in the slope body are extracted according to whether the landslide disaster-causing feature factors are discrete or continuous, the maximum frequency in the slope body region is taken for the discrete type, and the mean value in the slope body region is taken for the continuous type.

[0014] As a preferred embodiment, when the slope body object is determined, the step further comprises:

[0015] The flow direction of each grid cell is calculated by introducing a hydrological constraint algorithm;

[0016] The flow direction network is constructed according to the flow direction of each grid cell, and the drainage divide boundary is extracted based on the flow direction network;

[0017] The hydrological subarea is determined according to the drainage divide boundary, and the slope body object is limited to the maximum grid cell connected block within the same hydrological subarea.

[0018] As a preferred embodiment, the slope body features include the slope, the aspect, the terrain humidity index, the rainfall, the terrain relief degree, the elevation, the stratum lithology, the distance from the stratum fracture, the stratum fracture density, the distance from the river, the river density, the vegetation normalized index, the land type, the distance from the road, and the road density;

[0019] Among them, the rainfall, the terrain relief degree, the elevation, the stratum lithology, the distance from the river, the river density, the land type, the distance from the road, and the road density are divided into macro-environmental features.

[0020] The slope, the slope direction, the terrain moisture index, the distance from the stratum fissure, the stratum fissure density and the vegetation normalized index are classified into the local environment characteristics.

[0021] As a preferred embodiment, the step of calculating the first-order environment similarity between the landslide bodies through the macro environment characteristics, determining the landslide body set with the macro environment similarity through the first-order environment similarity and calculating the second-order environment similarity through the local environment characteristics in the landslide body set comprises:

[0022] For each pair of landslide bodies, the first-order environment similarity is calculated according to the macro environment characteristic vectors of the two, and the landslide body pairs with the first-order environment similarity values in the top N% are reserved as the landslide body set with the macro environment similarity;

[0023] The second-order environment similarity is calculated through the local environment characteristics of each pair of landslide bodies in the landslide body set with the macro environment similarity.

[0024] As a preferred embodiment, the step of classifying the landslide bodies according to the second-order environment similarity comprises:

[0025] The landslide body pairs with the second-order environment similarity values in the top M% are reserved for classification;

[0026] For the reserved landslide body pairs, the hierarchical clustering method is used for classification, and the number of clusters is determined through the Gap statistic.

[0027] As a preferred embodiment, the step of evaluating the landslide susceptibility of the region through the landslide susceptibility evaluation models of each category comprises:

[0028] The characteristic vector of the to-be-predicted slope body is obtained, the similarity between the to-be-predicted slope body and each category of landslide body is calculated through the characteristic vector of the to-be-predicted slope body and the characteristic vector of the cluster center point of each category of landslide body;

[0029] The weighting coefficients of the landslide susceptibility evaluation models of each category are respectively calculated through the similarity between the to-be-predicted slope body and each category of landslide body;

[0030] The landslide susceptibility evaluation results of the to-be-predicted slope body are respectively output through the landslide susceptibility evaluation models of each category, and are weighted and fused through the corresponding weighting coefficients to obtain the comprehensive landslide susceptibility evaluation results.

[0031] In another aspect, the present application also provides a landslide susceptibility evaluation system based on the second-order environment similarity constraint, comprising:

[0032] A feature extraction module is configured to determine a landslide occurrence region and collect landslide disaster-causing characteristic factors;

[0033] The slope body determination module is used for determining the slope body object of the region through DEM elevation, extracting the slope body features through the types of landslide disaster-causing characteristic factors, and defining the slope body object intersecting with the landslide region as a landslide body and vice versa as a non-landslide body.

[0034] The second-order environment similarity constraint module is used for dividing the slope body features into macro-environment features and local environment features, calculating the first-order environment similarity between landslide bodies through the macro-environment features, determining the landslide body set with similar macro-environment through the first-order environment similarity, and calculating the second-order environment similarity through the local environment features in the landslide body set.

[0035] The landslide body classification database construction module is used for classifying the landslide bodies according to the second-order environment similarity, randomly extracting an equal amount of non-landslide bodies according to the number of each type of landslide body, and constructing the in-class susceptibility database of the corresponding type according to the landslide body data and the extracted non-landslide body data of each type.

[0036] The landslide susceptibility evaluation module is used for respectively constructing the landslide susceptibility evaluation model based on the deep learning model according to the in-class susceptibility database of each type, and performing the landslide susceptibility evaluation of the region through the landslide susceptibility evaluation model of each type.

[0037] In another aspect, the present application further provides an electronic device, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the landslide susceptibility evaluation method based on the second-order environment similarity constraint as described in any one of the embodiments of the present application.

[0038] In another aspect, the present application further provides a computer readable storage medium, which stores a computer program, and the program is executable on a processor to implement the landslide susceptibility evaluation method based on the second-order environment similarity constraint as described in any one of the embodiments of the present application.

[0039] The landslide susceptibility evaluation method based on the second-order environment similarity constraint has the advantages that it can effectively solve the problem of ignoring the environment difference in traditional unified modeling. The second-order environment similarity between slope bodies is measured under the dual constraints of local environment features and overall environment features, and the landslide body classification is realized by combining the clustering method, so that each type of landslide body can be matched to a dedicated deep learning susceptibility model. On this basis, a multi-model weighted fusion strategy is further adopted to generate a comprehensive susceptibility evaluation result, so as to meet the prediction needs of different landslide body types. This method not only improves the fine degree and spatial applicability of landslide susceptibility evaluation, but also provides more targeted technical support for regional landslide disaster prevention and risk management.

[0040] Additional aspects and advantages of the present application will be set forth in part in the description that follows, and in part will be apparent from the description, or can be learned by practice of the present application. Furthermore, various aspects and advantages of the present application can be realized and attained by means of the instrumentalities and combinations particularly pointed out in the appended claims. BRIEF DESCRIPTION OF DRAWINGS

[0041] Figure 1 A flow chart of a method according to an embodiment of the present application;

[0042] Figure 2 A schematic diagram of clustering by analytic hierarchy process through a second-order environment similarity matrix according to an embodiment of the present application;

[0043] Figure 3 An example of a regional landslide susceptibility map drawn according to a landslide susceptibility evaluation result of a slope body according to an embodiment of the present application. DETAILED DESCRIPTION

[0044] The technical solutions in the embodiments of the present application will be apparently and completely described below with reference to 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 the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without any creative effort fall within the protection scope of the present application.

[0045] It should be understood that the step numbers used herein are only for the convenience of description, and are not limited to the execution sequence of the steps.

[0046] It should be understood that the terms used in the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in the present application specification and the appended claims, the singular forms "a", "an" and "the" are intended to include the plural forms, unless the context clearly indicates otherwise.

[0047] The terms "comprise" and "include" indicate the presence of the described features, integers, steps, operations, elements, and / or components, but do not exclude one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0048] The term "and / or" means any combination of one or more of the associated listed items and all possible combinations thereof, and includes these combinations.

[0049] Reference Figure 1 The present application proposes a landslide susceptibility evaluation method based on second-order environment similarity constraints, comprising the following steps:

[0050] S100, determining the landslide occurrence area by visual interpretation based on regional remote sensing images, and collecting corresponding landslide disaster-causing characteristic factors.

[0051] In an embodiment, the landslide disaster-causing characteristic factors include, but are not limited to, the following 15 factors: slope, slope direction, terrain humidity index, rainfall, terrain relief, elevation, stratum lithology, distance from stratum fissure, stratum fissure density, distance from river, river density, vegetation normalized index, land type, distance from road, and road density.

[0052] S200, determining the slope body object of the region through DEM elevation, extracting the slope body characteristics through the types of landslide disaster-causing characteristic factors, and defining the slope body object intersecting with the landslide region as a landslide body, and vice versa as a non-landslide body.

[0053] In an embodiment, step S200 includes:

[0054] S201, pre-processing the DEM elevation data, specifically using geographic information technology processing software (GrassGis in the present application) for pre-processing (filling and smoothing);

[0055] After pre-processing, the slope and slope direction of each grid cell are calculated, and the specific steps include:

[0056] For each grid cell, the domain elevation values (z1, z2,..., z9) within a 3*3 range (row priority, with the center value being z5) are taken, and the partial derivatives in the x and y directions (i.e., the slope in the east-west and north-south directions) are calculated:

[0057]

[0058] where Δ is the side length of the DEM grid cell. And the slope S(x, y) and the slope direction A(x, y) can be further calculated:

[0059]

[0060] To ensure the geometric consistency of the pixels inside the slope body, the slope body is regarded as a maximum connected sub-block U in the region, and the following formula needs to be satisfied:

[0061]

[0062] where, and respectively represent the average slope and slope direction of the region; δ S and δ A are the slope and slope direction tolerances.

[0063] In one embodiment, to prevent the occurrence of cross-basin slope, a hydrological constraint (D8 algorithm) is introduced to calculate the water flow direction of the grid cell:

[0064]

[0065] where d 5k represents the planar distance between the grid cell z5 and the neighborhood cell z k If it is a diagonal neighbor, take d 5k = Δ, otherwise take d .

[0066] Based on the water flow direction of each grid cell, a flow direction network is constructed, and based on the flow direction network, the watershed boundary is extracted;

[0067] According to the watershed boundary, the hydrological partition is determined, and the slope object is limited to the maximum grid cell connected block in the same hydrological partition B m :

[0068] U m = U∩B m .

[0069] After the above partition, the raster result is vectorized to obtain a set of closed polygons {P i}, which covers the study area, and each P i represents a slope unit.

[0070] S202, according to the type of landslide disaster-causing characteristic factor (discrete type and continuous type), the raster value of the landslide disaster-causing characteristic factor in the slope is extracted based on the generated slope vector shape. For discrete type, take the maximum frequency in the slope area (i.e. the highest frequency value in the complete slope area), and for continuous type, take the mean value in the slope area. For each of the landslide disaster-causing characteristic factors introduced in the above embodiment, the stratum lithology and the land type are discrete type factors, and the others are continuous type factors.

[0071] S300, the slope characteristics are divided into macro-environmental characteristics and local environmental characteristics, and the macro-environmental characteristics and the local environmental characteristics are respectively used to quantify the overall environmental instability condition and the similarity degree of the landslide formation condition. The first-order environmental similarity between landslide bodies is calculated through the macro-environmental characteristics, the landslide body set with similar macro-environment is determined through the first-order environmental similarity, and the second-order environmental similarity is calculated in the landslide body set through the local environmental characteristics.

[0072] In one embodiment, rainfall, terrain relief, elevation, stratum lithology, distance from river, river density, land type, distance from road, and road density are divided into macro-environmental characteristics, denoted as X (env) ;

[0073] Slope, aspect, topographic humidity index, distance from ground fissures, ground fissure density, and vegetation normalization index are classified as local environmental characteristics, denoted as X. (loc) .

[0074] In one embodiment, the steps of calculating the first-order environmental similarity between landslide bodies through macroscopic environmental features, determining a set of landslide bodies with similar macroscopic environments through the first-order environmental similarity, and calculating the second-order environmental similarity within this set of landslide bodies through local environmental features include:

[0075] Let the set of slopes in the region be P = {P1, P2, ..., P}. n For each pair of slopes, a first-order environmental similarity is calculated to quantify the similarity of the macroscopic environment in which the slopes are located.

[0076]

[0077] Among them, Sim (1) (P i ,P j () represents the first-order environmental similarity between the i-th slope and the j-th slope. This represents the macroscopic environmental characteristics of the i-th slope. This represents the macroscopic environmental characteristics of the j-th slope.

[0078] To exclude slope pairs with excessively different environments, this embodiment determines a threshold γ based on first-order environmental similarity results. Specifically, it obtains the first-order similarity set S of all slope pairs. (1) ={Sim (1) (P i ,P j After setting |1≤i≤j≤n}, select the 70th quantile (different quantile thresholds can be selected according to needs) as the threshold γ1, and retain the top 30% of slope pairs with the highest similarity to enter the second-order similarity calculation:

[0079] γ1 = Quantile 0.7 (S (1) ),S (1) ={Sim (1) (P i ,P j )};

[0080]

[0081] Where max represents the maximum value operation, and k and l are the indices of the two least similar slope pairs in the first-order similarity set. This represents the local environmental characteristics of the k-th slope. This represents the local environmental characteristics of the l-th slope.

[0082] S400. Classify landslides based on second-order environmental similarity. Randomly select an equal number of non-landslides based on the number of landslides in each class. Construct an intra-class susceptibility database for each class based on the landslide data and the selected non-landslide data.

[0083] In one embodiment, step S400 includes:

[0084] Landslide pairs with second-order environmental similarity scores in the top M percent are retained for classification. To avoid the impact of landslides with excessively large local feature differences on subsequent clustering results, a relatively relaxed threshold γ2 (selected as the 50th quantile) is introduced in the second-order similarity calculation:

[0085] γ2=Quantile 0.5 (S (2) ),S (2) ={Sim (2) (P i ,P j )};

[0086] After removing 50% of the slope pairs based on this threshold γ2, the final second-order similarity matrix can be obtained:

[0087]

[0088] like Figure 2 As shown, based on the second-order environmental similarity matrix, hierarchical clustering is used to classify landslide bodies, and the optimal number of clusters is determined using the Gap statistic. The clustering criterion of hierarchical clustering is the "shortest distance method," specifically in the form of:

[0089]

[0090] Among them, C p and C q These represent two slope classes to be merged.

[0091] The number of clusters, K, is chosen when the Gap statistic reaches its maximum value. Its mathematical expression is as follows:

[0092]

[0093] Among them, W k This represents the sum of within-class squared errors when the data is divided into k classes. This represents the intra-class error obtained by repeating random sampling B times under a reference uniform distribution.

[0094] Based on the number of landslide bodies in each category, an equal number of non-landslide bodies are randomly selected. For example, if there are 60 landslide bodies in category A after the classification is completed, then 60 non-landslide bodies are randomly selected and combined with the landslide bodies in category A to construct an intra-category susceptibility database for category A.

[0095] S500. Construct landslide susceptibility assessment models based on deep learning models according to the intra-class susceptibility databases for each category. If there are five categories of landslides, construct five corresponding landslide susceptibility assessment models. Divide the slope data in the intra-class susceptibility database for each type of landslide into training and testing sets in a 7:3 ratio. Each category's landslide susceptibility assessment model is used to predict the landslide susceptibility probability of the input slope based on its feature vector.

[0096] In one embodiment, the deep learning model employs a long short-term memory neural network, which includes two long short-term memory neural network feature extraction layers, two dropout layers, and a fully connected layer.

[0097] Finally, the landslide susceptibility of the region was assessed using various landslide susceptibility assessment models.

[0098] In one embodiment, landslide susceptibility assessment of a region is performed using various landslide susceptibility assessment models and a weighted fusion strategy. Specific steps include:

[0099] Obtain the feature vector X of the slope to be predicted i The similarity between the slope to be predicted and various types of landslides is calculated using the feature vectors of the slope to be predicted and the feature vectors of the cluster centers of various landslides.

[0100]

[0101] in, This represents the first-order similarity between the slope i to be predicted and the cluster center of the landslide in the k-th category. This represents the second-order similarity between the slope i to be predicted and the cluster center of the landslide in the k-th category. This represents the macroscopic environmental characteristics of the landslide cluster center point of the k-th category. This represents the local environmental characteristics of the cluster center of the landslide body in the k-th category.

[0102] The results are combined into a comprehensive similarity score, and then normalized proportionally to map it into a weighted coefficient w. i,k :

[0103]

[0104] Among them, S i,k This represents the comprehensive similarity between the slope i to be predicted and the cluster center of the landslides of the kth category, where α is a preset weight. To calculate the overall similarity standard value, w i,kdenoted by , represents the weighting coefficient of the landslide susceptibility assessment model for the k-th category for the slope i to be predicted, where K is the total number of categories.

[0105] The landslide susceptibility assessment results for each type of landslide susceptibility assessment model are output, and then weighted and fused using corresponding weighting coefficients to obtain the comprehensive landslide susceptibility assessment result.

[0106]

[0107] in, To comprehensively assess the landslide susceptibility results, y i,k =f k (X i ) represents the prediction result output by the landslide susceptibility assessment model of the k-th category based on the feature vector of the input slope i to be predicted.

[0108] In one embodiment, a landslide susceptibility map of the region can be drawn based on the comprehensive landslide susceptibility assessment results of each slope in the region. Based on the output comprehensive landslide susceptibility probability of each slope, the risk interval to which the slope belongs is determined (extremely low risk, low risk, medium risk, high risk, extremely high risk; the landslide susceptibility probability range for each risk interval can be preset). By color-coding the slope units according to the risk intervals, the landslide susceptibility map of the region can be generated. Figure 3 As shown.

[0109] This application also proposes a landslide susceptibility assessment system based on second-order environmental similarity constraints, including:

[0110] The feature extraction module is used to determine the landslide occurrence area and collect landslide disaster-causing characteristic factors; this module is used to implement the function of step S100 in the above embodiment, and will not be described in detail here.

[0111] The slope determination module is used to determine the slope objects in the area through DEM elevation, extract slope features by the type of landslide disaster characteristic factors, and define the slope objects that intersect with the landslide area as landslide bodies, and the rest as non-landslide bodies; this module is used to implement the function of step S200 in the above embodiment, and will not be described in detail here.

[0112] The second-order environmental similarity constraint module is used to divide slope features into macro-environmental features and local environmental features. It calculates the first-order environmental similarity between landslides through macro-environmental features, determines a set of landslides with similar macro-environment through first-order environmental similarity, and calculates the second-order environmental similarity within the set of landslides through local environmental features. This module is used to implement the function of step S300 in the above embodiment, and will not be described in detail here.

[0113] The landslide classification database construction module is used to classify landslides based on second-order environmental similarity, randomly extract an equal number of non-landslides according to the number of landslides in each class, and construct an intra-class susceptibility database for each class based on the landslide data and the extracted non-landslide data. This module is used to implement the function of step S400 in the above embodiment, and will not be described again here.

[0114] The landslide susceptibility assessment module is used to construct landslide susceptibility assessment models based on deep learning models according to the in-class susceptibility databases of each category, and to assess the landslide susceptibility of the region through the landslide susceptibility assessment models of each category; this module is used to implement the function of step S500 in the above embodiments, and will not be described again here.

[0115] This application also proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the landslide susceptibility assessment method based on second-order environmental similarity constraints as described in any embodiment of the present invention.

[0116] This application also proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the landslide susceptibility assessment method based on second-order environmental similarity constraints as described in any embodiment of the present invention.

[0117] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent the existence of A alone, A and B simultaneously, or B alone. A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of singular or plural items. For example, at least one of a, b, and c can represent: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, and c can be single or multiple.

[0118] Those skilled in the art will recognize that the units and algorithm steps described in the embodiments disclosed herein can be implemented using electronic hardware, computer software, or a combination of electronic hardware and software. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0119] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0120] In the several embodiments provided in this application, any function, if implemented as a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0121] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A landslide susceptibility assessment method based on second-order environmental similarity constraints, characterized in that, Includes the following steps: Determine the landslide area and collect landslide-causing characteristic factors; The slope objects in the area are determined by DEM elevation, and the slope features are extracted by the type of landslide disaster characteristic factors. The slope objects that intersect with the landslide area are defined as landslide bodies, and those that do not are defined as non-landslide bodies. The slope features are divided into macro-environmental features and local environmental features. The first-order environmental similarity between landslides is calculated through macro-environmental features. The first-order environmental similarity determines the set of landslides with similar macro-environment. The second-order environmental similarity is calculated within the set of landslides through local environmental features. Landslides are classified based on second-order environmental similarity. An equal number of non-landslides are randomly selected based on the number of landslides in each class. A class-specific susceptibility database is constructed based on the landslide data and the selected non-landslide data for each class. Landslide susceptibility assessment models based on deep learning models were constructed according to the intra-category susceptibility databases for each category, and the landslide susceptibility of the region was assessed through the landslide susceptibility assessment models for each category.

2. The landslide susceptibility assessment method based on second-order environmental similarity constraints according to claim 1, characterized in that, The steps of determining the slope objects in the region using DEM elevation and extracting the hierarchical features of the slope according to the type of landslide-causing characteristic factors include: The DEM elevation data is preprocessed, and the slope and aspect of each grid cell are calculated after preprocessing. The grid cell block with the largest slope and aspect within the preset tolerance range is connected as a slope body. The raster values ​​of landslide-causing characteristic factors in the slope are extracted according to whether the landslide-causing characteristic factors are discrete or continuous. For discrete factors, the maximum frequency in the slope area is taken, and for continuous factors, the mean value in the slope area is taken.

3. The landslide susceptibility assessment method based on second-order environmental similarity constraints according to claim 2, characterized in that, The process of identifying the slope object also includes the following steps: A hydrological constraint algorithm is introduced to calculate the water flow direction of each grid cell; A flow direction network is constructed based on the flow direction of each grid cell, and the watershed boundary is extracted based on the flow direction network; Hydrological zones are determined based on watershed boundaries, and slope objects are restricted to the largest connected grid cell block within the same hydrological zone.

4. The landslide susceptibility assessment method based on second-order environmental similarity constraints according to claim 1, characterized in that: The slope characteristics include slope, aspect, topographic humidity index, rainfall, topographic relief, elevation, stratum lithology, distance from stratum fissures, stratum fissure density, distance from rivers, river density, vegetation normalization index, land type, distance from roads, and road density. Among them, rainfall, topographic relief, elevation, stratum lithology, distance from rivers, river density, land type, distance from roads, and road density are classified as macro-environmental characteristics; Slope, aspect, topographic humidity index, distance from ground fissures, density of ground fissures, and vegetation normalization index are classified as local environmental characteristics.

5. The landslide susceptibility assessment method based on second-order environmental similarity constraints according to claim 1, characterized in that, The steps of calculating the first-order environmental similarity between landslide bodies through macroscopic environmental characteristics, determining a set of landslide bodies with similar macroscopic environments through the first-order environmental similarity, and calculating the second-order environmental similarity within this set of landslide bodies through local environmental characteristics include: For each pair of landslides, the first-order environmental similarity is calculated based on their macro-environmental feature vectors. The landslide pairs with the highest first-order environmental similarity values ​​are retained as the set of landslides with similar macro-environment. Second-order environmental similarity is calculated by considering the local environmental characteristics of each pair of landslides in a set of landslides with similar macroscopic environments.

6. The landslide susceptibility assessment method based on second-order environmental similarity constraints according to claim 5, characterized in that, The steps for classifying landslides based on second-order environmental similarity include: Landslide pairs with the highest second-order environmental similarity values ​​(top M%) were classified. For the preserved landslide pairs, hierarchical clustering was used for classification, and the number of clusters was determined by the Gap statistic.

7. The landslide susceptibility assessment method based on second-order environmental similarity constraints according to claim 1, characterized in that, The steps for assessing regional landslide susceptibility using various landslide susceptibility assessment models include: Obtain the feature vector of the slope to be predicted, and calculate the similarity between the slope to be predicted and various types of landslide cluster centers using the feature vector of the slope to be predicted and the feature vector of the cluster centers of various types of landslides. The weighting coefficients of the landslide susceptibility assessment models for each category are calculated based on the similarity between the slope to be predicted and various types of landslides. The landslide susceptibility assessment results of the slope to be predicted are output through various types of landslide susceptibility assessment models, and then weighted and fused using corresponding weighting coefficients to obtain a comprehensive landslide susceptibility assessment result.

8. A landslide susceptibility assessment system based on second-order environmental similarity constraints, characterized in that, include: The feature extraction module is used to determine the landslide occurrence area and collect landslide-causing characteristic factors; The slope determination module is used to determine the slope objects in the area through DEM elevation, extract slope features by the type of landslide disaster characteristic factors, and define the slope objects that intersect with the landslide area as landslide bodies, and the rest as non-landslide bodies. The second-order environmental similarity constraint module is used to divide slope features into macro-environmental features and local environmental features. It calculates the first-order environmental similarity between landslides based on macro-environmental features, determines the set of landslides with similar macro-environment based on first-order environmental similarity, and calculates the second-order environmental similarity within the set of landslides based on local environmental features. The landslide classification database construction module is used to classify landslides based on second-order environmental similarity. It randomly selects an equal number of non-landslides based on the number of landslides in each class and constructs an intra-class susceptibility database for each class based on the landslide data and the selected non-landslide data for each class. The landslide susceptibility assessment module is used to construct landslide susceptibility assessment models based on deep learning models according to the in-class susceptibility databases of each category, and to conduct landslide susceptibility assessment of the region through the landslide susceptibility assessment models of each category.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the landslide susceptibility assessment method based on second-order environmental similarity constraints as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the landslide susceptibility assessment method based on second-order environmental similarity constraints as described in any one of claims 1 to 7.

Images