Bedding landslide disaster risk evaluation method based on machine learning

Abstract

A bedding landslide disaster risk evaluation method based on machine learning comprises the following steps: step 1, acquiring landform image data of a certain area, identifying different types of bedding landslide hidden danger points according to development features and landform features of bedding landslide, and constructing a bedding landslide feature sample library; and 2, under the condition that the bedding landslide feature sample library is constructed, carrying out bedding landslide hidden danger identification on an unknown region by utilizing machine learning, and further obtaining bedding landslide geological disaster data of the region. The method is used for solving the problems that an existing assessment method is limited by complexity and cost of data collection, and the processing capacity for multi-factor influence is limited, so that the risk assessment precision is not high.

Description

Technical Field

[0001] The present invention relates to the field of geological disasters, and more particularly to a method for risk assessment of landslide disasters along the bedding layer based on machine learning. Background Art

[0002] The hazards of bedding landslides are multifaceted, not only affecting road traffic, river navigation and water conservancy facilities, but also destroying farmland, threatening the safety of residential areas, and causing long-term impacts on the geological and ecological environments. Therefore, it is necessary to strengthen the risk assessment of bedding landslides.

[0003] Traditional bedding landslide risk assessment methods usually rely on changes in soil physical properties. Traditional methods are often limited by the complexity and cost of data collection, as well as limited ability to handle multiple factors. With the development of society, the geological hazard risk assessment theory has now matured and has gradually been applied from academic research to specific geological hazard prevention and control work. The bedding landslide risk assessment method based on machine learning can process a large amount of complex data, improve prediction accuracy, consider the impact of multiple factors on landslide hazards, and improve the accuracy of risk assessment. Summary of the invention

[0004] The purpose of the present invention is to provide a method for risk assessment of bedding landslide disasters based on machine learning, which is used to solve the problem that existing assessment methods are limited by the complexity and cost of data collection, as well as the limited ability to handle the influence of multiple factors, resulting in low accuracy of risk assessment.

[0005] In order to solve the above problems, the technical solution of the present invention is: A method for risk assessment of landslide disaster along the layer based on machine learning includes the following steps: Step 1: Obtain topographic image data of a certain area, identify different types of bedding landslide risk points according to the development characteristics of bedding landslides and topographic and geomorphic characteristics, and build a bedding landslide feature sample library; Step 2: After building a sample library of bedding landslide characteristics, use machine learning to identify bedding landslide hazards in unknown areas, and then obtain bedding landslide geological disaster data in the area; Step 3: Extract feature samples of unknown area evaluation factors based on machine learning; Step 4: Based on the obtained evaluation factor characteristics, an evaluation index system is established, and the hazard of the layered landslide is evaluated using the random forest algorithm to obtain a hazard evaluation zoning map; Step 5: Select population density, road density, and housing distribution density as evaluation factors, use the fuzzy comprehensive evaluation algorithm to construct a vulnerability evaluation calculation formula, conduct regional vulnerability evaluation, and obtain a vulnerability evaluation zoning map; Step 6: Calculate the risk zoning map and conduct risk assessment on regional bedding landslides.

[0006] The beneficial effects of the present invention are as follows: the bedding landslide disaster risk assessment method based on machine learning can process a large amount of complex data, improve the prediction accuracy, and can consider the impact of multiple factors on landslide disasters, and can also improve the accuracy of risk assessment. According to typical topography, this method is easier to identify bedding landslides, and can also improve the assessment accuracy of landslide risk. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] The present invention will be further described below in conjunction with the accompanying drawings: Figure 1 A landslide topographic map; Figure 2 Topographic expression characteristics of landslide. DETAILED DESCRIPTION

[0008] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0009] A risk assessment method for landslide disasters along the layer based on machine learning. Step 1: Obtain high-precision topographic image data of a certain area, identify different types of bedding landslide risk points according to the development characteristics of bedding landslides and topographic and geomorphic characteristics, and build a bedding landslide feature sample library; like Figure 2 As shown in the figure, the lateral boundaries of layer-controlled bedding landslides are mostly steep cliffs, rock walls or valleys, grooves. On the topographic map, the top of the landslide is a ring-shaped closed contour line; the middle part of the landslide is equidistant and parallel contour lines; the lower and middle parts of the landslide have curved and gradually widening parallel contour lines, which represent the presence of deposits or slow deformation, and the slope becomes gentle. When the landslide has long-term slow deformation, there will generally be weak groove topographic lines on the upstream and downstream side boundaries, which will extend and intersect along the boundaries on both sides to the top of the mountain regularly, and it can be judged that this landslide is a layer-controlled landslide.

[0010] Step 2: After constructing a sample library of bedding landslide characteristics, use machine learning to identify the hidden dangers of bedding landslides in the area, and then obtain a database of bedding landslide geological hazards in the area; obtain data on bedding landslide geological hazards in the area, such as topography (elevation, slope, slope direction, surface undulation, ground roughness, surface cutting depth), geological conditions (stratigraphic age, stratigraphic lithology, geological structure), meteorology and hydrology (precipitation, groundwater level depth, temperature, humidity), human activities (land use type, vegetation index, roads) and other corresponding data.

[0011] Step 3: Extract feature samples of the evaluation factors of the region based on machine learning (random forest); Step 4: Based on the obtained evaluation factor characteristics, an evaluation index system is established, and the random forest method is used to evaluate the hazard of bedding landslides to obtain a hazard evaluation zoning map.

[0012] Step 5: Select population density, road density, and housing distribution density as evaluation factors, use the fuzzy comprehensive evaluation method to construct a vulnerability assessment calculation formula, conduct regional vulnerability assessment, and obtain a vulnerability assessment zoning map.

[0013] Step 6: Calculate the risk zoning map and conduct risk assessment on regional bedding landslides.

[0014] Based on the ArcGIS platform, with the hazard zoning map and vulnerability zoning map as the basic layers, the raster calculator overlay tool is used to perform basic layer raster overlay calculation to obtain the geological hazard risk assessment zoning map.

[0015] The random forest algorithm (RF) in step three is a bagging ensemble learning algorithm based on multiple decision trees as a basic model, and is applied to supervised learning fields such as classification and regression.

[0016] The model principle is based on the Boot-strap non-parametric sampling algorithm, which builds a variety of decision tree models by constructing different training samples and randomly selecting different input features. Random forest effectively reduces the effect of overfitting by aggregating the results and taking the average value, removes the deviation caused by random samples and input features, and effectively improves the accuracy of the risk assessment model. The model formula is:

[0017] Where F(x) is the RF model; f i (x) is a single decision tree model; Z is the output variable (indicator factor); I is the indicative function (characteristically, it can only take 0 and 1, disaster points and non-disaster points).

[0018] In the risk assessment study, the random forest model uses the Gini index or information gain to measure the importance of the evaluation factor to the development of geological hazards. The present invention uses the Gini index to calculate the importance of the evaluation factor. By calculating the reduction of the Gini index D when the evaluation factor k is split at the node GK The importance of the evaluation factor is measured by the proportion of the total average Gini reduction of all factors. The calculation formula is as follows:

[0019] Where m represents the total number of evaluation factors, n represents the number of classification trees, and t represents the number of nodes in a single class tree; D Gkhj is the Gini index reduction value of the kth evaluation factor at the jth node of the hth class tree; P k is the importance of the kth evaluation factor among all evaluation factors.

[0020] The fuzzy comprehensive evaluation method uses the fuzzy transformation principle and the maximum membership principle to consider the multi-factor influence of a thing and make a scientific evaluation method. The evaluation result is a fuzzy set. There are many factors that affect the vulnerability evaluation research and the degree of influence cannot be expressed quantitatively. Therefore, the fuzzy comprehensive evaluation method is more suitable for the vulnerability evaluation of geological hazards.

[0021] In the application process, each evaluation factor has different degrees of influence on the vulnerability assessment. Therefore, based on expert experience and risk assessment results, the vulnerability of geological hazards in the study area is divided into five levels: extremely low, low, medium, high, and extremely high, forming an evaluation set V = {v1, v2, v3, v4, v5} = {extremely high vulnerability, high vulnerability, medium vulnerability, low vulnerability, extremely low vulnerability}. The grading standards for geological hazard vulnerability assessment are shown in Table 1.

[0022] Table 1 Geological hazard vulnerability rating standards The fuzzy comprehensive evaluation algorithm in step 5 includes the following steps: Step 1: Construct membership function The present invention adopts descending (ascending) semi-trapezoidal distribution to construct a univariate linear membership function, and the mathematical model is as follows:

[0023]

[0024]

[0025]

[0026]

[0027] Where: u j(x i ) is the xth i The evaluation index data corresponds to the membership degree of the first and second level vulnerability, and the value range is [0,1]; i is the evaluation data of the i-th evaluation index; x ij (i=1,2,3; j=1,2,3,4,5) is the evaluation standard under the j-level vulnerability corresponding to the ith evaluation indicator.

[0028] Step 2: Construct evaluation matrix and weight matrix; The evaluation matrix R with a dimension of 3×5 is used to represent the fuzzy relationship between the geological hazard level factor U and the rating set V.

[0029]

[0030] Construct a weight matrix, the weight formula is:

[0031]

[0032]

[0033] According to the above three formulas, the weight matrix is:

[0034] Step 3: fuzzy synthesis operation; Each row of data in the evaluation matrix R reflects the degree of membership of the evaluated object to each level of fuzzy subset under the influence of different factors.

[55] The weight matrix w is combined with the evaluation matrix R to obtain the fuzzy evaluation matrix B, that is: B=w×R=(b1 b2 b3 b4 b5)

[0035] According to the maximum membership principle, the level where the maximum value of the fuzzy judgment matrix B is located corresponds to the level of vulnerability judgment result, that is, b j =b jmax , then the vulnerability level of the evaluation unit is j.

[0036] The contents described in the embodiments of this specification are merely an enumeration of the implementation forms of the inventive concept. The protection scope of the present invention should not be regarded as limited to the specific forms described in the embodiments. The protection scope of the present invention also extends to equivalent technical means that can be conceived by those skilled in the art based on the inventive concept.

Claims

1. A method for risk assessment of landslide disaster along the layer based on machine learning, characterized in that: The steps include: Step 1: Obtain topographic image data of a certain area, identify different types of bedding landslide risk points according to the development characteristics of bedding landslides and topographic and geomorphic characteristics, and build a bedding landslide feature sample library; Step 2: After building a sample library of bedding landslide characteristics, use machine learning to identify bedding landslide hazards in unknown areas, and then obtain bedding landslide geological disaster data in the area; Step 3: Extract feature samples of unknown area evaluation factors based on machine learning, where the machine learning is a random forest algorithm; Step 4: Based on the obtained evaluation factor characteristics, an evaluation index system is established, and the hazard of the layered landslide is evaluated using the random forest algorithm to obtain a hazard evaluation zoning map; Step 5: Select population density, road density, and housing distribution density as evaluation factors, use the fuzzy comprehensive evaluation algorithm to construct a vulnerability evaluation calculation formula, conduct regional vulnerability evaluation, and obtain a vulnerability evaluation zoning map; Step 6: Calculate the risk zoning map and conduct risk assessment on regional bedding landslides.

2. The method for risk assessment of landslide disaster along the layer based on machine learning according to claim 1, characterized in that: In step three, the random forest algorithm model includes: Where F(x) is the random forest model; f i (x) is a single decision tree model; Z is the output variable (indicator factor); I is the indicative function (characteristically only taking 0 and 1, disaster points and non-disaster points); k is the evaluation factor.

3. The method for risk assessment of landslide disaster along the layer based on machine learning according to claim 2 is characterized in that: By calculating the reduction D of the Gini index when the evaluation factor k is split at the node GK The importance of the evaluation factor is measured by the proportion of the total average Gini reduction of all factors. The calculation formula is: Where m represents the total number of evaluation factors, n represents the number of classification trees, and t represents the number of nodes in a single class tree; D Gkhj is the Gini index reduction value of the kth evaluation factor at the jth node of the hth class tree; P k is the importance of the kth evaluation factor among all evaluation factors.

4. A method for risk assessment of landslide disaster along the layer based on machine learning according to any one of claims 1 to 3, characterized in that: The fuzzy comprehensive evaluation algorithm in step 5 includes the following steps: Step 1: Construct membership function; The descending (ascending) semi-trapezoidal distribution is used to construct a univariate linear membership function. The mathematical model is as follows: Where: u j (x i ) is the xth i The evaluation index data corresponds to the membership degree of the first and second level vulnerability, and the value range is [0,1]; i is the evaluation data of the i-th evaluation index; x ij (i=1,2,3;j=1,2,3,4,5) is the evaluation standard under the j-level vulnerability corresponding to the i-th evaluation index; Step 2: Construct evaluation matrix and weight matrix; The evaluation matrix R with a dimension of 3×5 is used to represent the fuzzy relationship between the geological hazard level factor U and the rating set V. Construct a weight matrix, the weight formula is: According to the above three formulas, the weight matrix is: Step 3: fuzzy synthesis operation; Each row of data in the evaluation matrix R reflects the degree of membership of the evaluated object to each level of fuzzy subset under the influence of different factors. [55] The weight matrix w is combined with the evaluation matrix R to obtain the fuzzy evaluation matrix B, that is: B=w×R=(b1 b2 b3 b4 b5) According to the maximum membership principle, the level where the maximum value of the fuzzy judgment matrix B is located corresponds to the level of vulnerability judgment result, that is, b j =b jmax , then the vulnerability level of the evaluation unit is j.

Images