A method for evaluating landslide disaster risk by using bipolar fuzzy set

Abstract

The present application relates to a kind of method for judging landslide disaster risk using bipolar fuzzy set, comprising the following steps: step 1), data collection includes two parts;First part: collect landslide point data;Second part: collect Sentinel-1A radar image data, precise orbit ephemeris data, Google satellite image data, DEM data, slope, slope and curvature data, fault data, river system data, geomorphic type data, road network data, seismic intensity data;Step 2) determination of landslide disaster risk evaluation factor;Step 3) establish landslide disaster risk evaluation model based on bipolar fuzzy set.The result of the present application using bipolar fuzzy set evaluation method is basically consistent with the result of BP neural network evaluation.By taking Zhaotong City, Yunnan Province, Luding County as an example, it is finally concluded that the evaluation results of 10 of the 12 townships are relatively close, reaching 83%, which shows that the method has strong applicability;Thus it can be popularized in other areas.

Description

Technical Field

[0001] This invention belongs to the technical field of hazard assessment methods, and in particular relates to a method for assessing the hazard of landslide disasters using bipolar fuzzy sets. Background Technology

[0002] In recent years, with the rapid development of computer technology and geographic information technology, landslide hazard assessment has gradually become a research hotspot.

[0003] Numerous scholars have conducted research on geological hazard risk assessment at different scales. Currently, there are many methods for geological hazard risk assessment, which can be mainly divided into three categories: empirical models, statistical analysis models, and machine learning models.

[0004] ① Empirical Models: Empirical models essentially involve deeply mining the relationships between data points based on a large amount of historical landslide data. They assign values ​​to various factors influencing landslide occurrence according to certain standards, based on the degree of contribution. ② Statistical Analysis Models: Statistical analysis models statistically analyze the impact of various factors on landslide hazard. Landslide mechanisms are complex and require consideration of multiple factors. There may be a special correlation between relevant influencing factors and the final evaluation indicators or predicted values, often expressed as mathematical formulas or computational models. ③ Machine Learning Models: Machine learning is formed through the interdisciplinary integration of probability theory, statistics, and complex algorithms in the field of artificial intelligence. Computers automatically improve algorithms through large samples or optimize their program performance based on data and past experience, using inductive and synthetic methods to approximate the relationships between objective physical quantities. Common models include decision tree models, support vector machine models, and artificial neural network models.

[0005] The above three methods have demonstrated their effectiveness in geological hazard assessment. However, different types of landslides exhibit significant differences in their formation mechanisms and failure modes. Landslide precipitation mechanisms are complex, controlled by the combined effects of multiple internal and external factors. Hazard assessment involves numerous factors, and the degree of influence of each factor on landslide induction varies, making it difficult to accurately determine their contribution values. Furthermore, the influence of landslide hazard factors on the deformation and fracturing process of the landslide's internal structure exhibits nonlinear and unstable characteristics with changes in the geographical environment. Incorporating these factors into the model may lead to problems such as slow model operation, model overcomplication, and model overfitting, affecting the model's evaluation results and rendering it unsuitable for landslide hazard assessment in areas with complex topography. Therefore, a feasible and effective method needs to be proposed to address the problem of uncertain contribution values ​​of assessment factors in the landslide hazard assessment system. Summary of the Invention

[0006] The present invention aims to address the aforementioned problems and deficiencies by providing a method for assessing the risk of landslide disasters using bipolar fuzzy sets.

[0007] The present invention is implemented using the following technical solution.

[0008] A method for assessing landslide hazard risk using bipolar fuzzy sets, the method of the present invention includes the following steps:

[0009] Step 1), data collection includes two parts; Part 1: collecting landslide point data; Part 2: collecting Sentinel-1A radar imagery data, precise orbit determination ephemeris data, Google satellite imagery data, DEM data, slope, aspect and curvature data, fault data, river system data, landform type data, road network data, and seismic intensity data.

[0010] Step 2) Determination of landslide hazard assessment factors;

[0011] Step 3) Establish a landslide hazard assessment model based on bipolar fuzzy sets.

[0012] Furthermore, step 2) of the present invention, determining the landslide hazard assessment factors, includes:

[0013] The landslide hazard assessment factors in the study area include: altitude, slope, distance to road, distance to river, distance to fault, slope aspect and plane curvature, and seismic intensity index.

[0014] Using the frequency ratio FR )Analyze the correlation between landslide distribution and landslide condition factors. F FR for FR The value, F FR Defined as the ratio of landslide occurrence to the total area of ​​the study area;

[0015] if F FR If the value is 1, then the landslide is considered to be generally correlated with the condition factors. F FR A value greater than 1 indicates a high correlation. F FR If the correlation is less than 1, it is considered low. F FR The calculation formula (1) is as follows:

[0016] (1)

[0017] In the above formula, f ij It is a condition factor j Classificationi Medium landslide density; f It is the landslide density in the entire study area; P ij * It is a condition factor j Classification i Number of medium-sized landslides; P ij It is a condition factor j Classification i The area in; P It is the total area of ​​the study area; P * This represents the total number of landslides that occurred in the study area.

[0018] Further, step 2) of the present invention, determining the landslide hazard assessment factor, includes: a deformation rate factor, the steps of which are: based on SBAS-InSAR technology, 103 Sentinel-1A ascending-orbit slant-range single-look complex images after registration and cropping are selected; according to the optimal principle of temporal baseline and vertical baseline, the temporal baseline threshold and spatial baseline are set; to suppress speckle noise, the MinimumCost Flow unwrapping method and the Goldstein filtering method are used for interferometric processing to obtain the registered combined interferometric phase:

[0019] (2)

[0020] In the formula: Phase generated by slant-range deformation, Indicates terrain phase, Phase caused by atmospheric delay Phase caused by coherent noise; the interference phase is expressed as the average phase velocity between two time phases. v and time t The product of these factors is used to select GCP points without redundant terrain fringes and phase jumps, and those far from the deformation region, for orbit refinement and re-flattening, to obtain the first... i Phase value of amplitude interferogram ;

[0021] As shown in equations (3) and (4):

[0022] (3)

[0023] (4)

[0024] The integral of each temporal velocity between the master and slave images is rewritten as follows: m × n matrix B Thus, we obtain matrix equation (5).

[0025] (5)

[0026] Because small baseline set difference interferometry employs a multi-master image strategy, the matrix... B Since rank deficiency is easily generated, the least squares method and singular value matrix decomposition are used to perform deformation inversion. Then, the atmospheric phase is estimated and removed to obtain the time series deformation information of the study area. After geocoding the time series information, the deformation rate in the LOS direction of the study area is obtained.

[0027] Furthermore, the data processing procedure for the deformation rate results described in this invention includes:

[0028] (1) Generation of connecting lines;

[0029] (2) Interference with the workflow;

[0030] This process performs interferometric processing on all paired interferometric image pairs, including coherence generation, flattening, filtering, and phase unwrapping. Ultimately, all data pairs are registered onto the super master image to prepare for the subsequent first and second inversions.

[0031] (3) First inversion

[0032] The purpose of the first inversion is to use the first inversion model to calculate the remaining height and displacement velocity, which is used to process complex interferograms; the first inversion identifies a certain number of permanent scattering points, and then processes the pixels around them.

[0033] (4) Second inversion

[0034] The purpose of the second inversion is to use the linear model results from the first inversion to estimate the atmospheric phase components, then remove the atmospheric phase components, and fit the data to obtain the final displacement velocity.

[0035] (5) Geographic coding

[0036] (6) InSAR data processing and analysis

[0037] Based on the InSAR surface deformation monitoring results, mathematical statistical theory and other methods are used to analyze the spatiotemporal evolution of surface deformation in the monitoring area and obtain the deformation rate results of geological disasters in the monitoring area.

[0038] Furthermore, step 3) of the present invention includes extending the domain of the fuzzy set from [0,1] x [0,1] to [0,1] x [1,0]; let... X It is a finite domain of discourse, a bipolar fuzzy set B The following form (6):

[0039] (6)

[0040] Positive membership degree Represents element x Regarding bipolar fuzzy sets B Satisfaction with a certain attribute, negative membership degree Represents element x Regarding bipolar fuzzy sets B The degree of satisfaction of the opposite property of this attribute. , There are two mappings;

[0041] Suppose there is a set of alternative solutions There is also a set of attributes ;set up It is the weight of the attribute, for example It is an attribute The weights, and

[0042] , (7)

[0043] Assumption It is a bipolar fuzzy set decision matrix. μ ij Indicates positive membership degree. v ij Indicates negative membership degree. The membership degree and negative membership degree are used by decision-makers to evaluate a given alternative. μ i Satisfy a certain attribute p j The degree of.

[0044] Furthermore, this invention employs the analytic hierarchy process (AHP) to determine the weights of each evaluation factor, as detailed below:

[0045] (1) Establish a hierarchical structure

[0046] First, the problem is decomposed into three layers: the goal layer, the criteria layer, and the solution layer, establishing a hierarchical model. Then, the characteristics of the five selected indicators are analyzed.

[0047] (2) Construct the judgment matrix

[0048] The analytic hierarchy process (AHP) determines factor weights by decomposing each element into three levels (upper, middle, and lower) according to different attributes from top to bottom, and establishing a hierarchical evaluation model based on the stratification. Factors within the same level of two or more levels with subordinate relationships are compared pairwise to establish a series of judgment matrices in the following form:

[0049] (8)

[0050] The standard for measurement adopts the 1-9 scale method; the eigenvector of the evaluation matrix is ​​calculated, and after normalization, it becomes the weight vector of the system. :

[0051] (9).

[0052] Furthermore, step 3) of this invention includes using a bipolar fuzzy set evaluation method to assess landslide hazard, expressing uncertainty based on a bipolar membership function; according to the definition of the bipolar fuzzy set evaluation method, the process of solving the decision problem is as follows:

[0053] Step 1: Use the following operators to process the information of each alternative in R;

[0054] (11)

[0055] For the alternative solutions in the decision matrix R v The information in section 2 is processed, and the attribute weights are all p.

[0056] (12)

[0057] Step 2: Calculate the score function for each alternative.

[0058] (13)

[0059] Step 3: Sort the score functions of each candidate solution. If two candidate solutions have the same score function, calculate the accuracy function of these two candidate solutions. ,

[0060] (14)

[0061] Those with higher precision functions are ranked higher.

[0062] Step 4: Rank the landslide risks.

[0063] The beneficial effects of this invention are as follows: it identifies more suitable landslide hazard factors and employs the analytic hierarchy process (AHP) to determine the weights of each evaluation factor; the evaluation results using the bipolar fuzzy set evaluation method are largely consistent with those using the backpropagation (BP) neural network. Taking Ludian County, Zhaotong City, Yunnan Province as an example, the evaluation results of 10 out of 12 townships were found to be relatively similar, reaching 83%, indicating that the method has strong applicability and can therefore be promoted in other regions.

[0064] The present invention will be further explained below with reference to the accompanying drawings and specific embodiments. Attached Figure Description

[0065] Figure 1 This is a distribution map of landslide points in Ludian County, as described in an embodiment of the present invention.

[0066] Figure 2 This is a diagram of landslide susceptibility evaluation factors according to the present invention;

[0067] Figure 3 This is a deformation rate diagram of Ludian County according to an embodiment of the present invention;

[0068] Figure 4 This is a map showing the distribution of townships in Ludian County according to an embodiment of the present invention;

[0069] Figure 5 This is a deformation rate diagram of various townships in Ludian County according to an embodiment of the present invention;

[0070] Figure 6 This is a hierarchical structure diagram for evaluating the present invention;

[0071] Figure 7 This is a comparison chart of the scores of various townships in an embodiment of the present invention;

[0072] Figure 8 This is a diagram showing the landslide hazard assessment results of a bipolar fuzzy set according to an embodiment of the present invention.

[0073] Figure 9 This is a graph showing the landslide susceptibility prediction results using a BP neural network according to an embodiment of the present invention. Detailed Implementation

[0074] A method for assessing landslide hazard risk using bipolar fuzzy sets includes the following steps (this invention takes Ludian County, Zhaotong City, Yunnan Province as an example):

[0075] 1. Data Collection

[0076] 1.1 Landslide Point Distribution Data

[0077] Ludian County has relatively lush vegetation. Landslide characteristics vary depending on the scale and development stage of the landslides, mainly falling into three categories: ① Landslides with clear boundaries often exhibit a distinct armchair-like shape; the landslide walls are mostly steep slopes and lack vegetation. ② Landslides in the creep stage show obvious tensile cracks in the middle and rear sections, with the leading edge primarily exhibiting collapse damage. ③ Landslides with weak surface deformation have less clear image features, requiring assessment based on vegetation cover, topographic damage, and overall geomorphological conditions. A total of 458 landslide points were marked using remote sensing imagery. Figure 1 As shown.

[0078] Other data

[0079] Other data includes: ① Sentinel-1A radar imagery (from January 2019 to June 2022), available for free download from the European Space Agency (ESA) and accessible online; ② Precise orbit determination ephemeris, used to improve satellite orbit accuracy, also available online; ③ Google satellite imagery, obtained from the Bigemap map downloader; ④ A 30-meter resolution digital elevation model (DEM), used to extract elevation, slope, aspect, curvature of the study area and remove topographic phase effects from Sentinel-1A data processing, obtained from the Japan Aerospace Exploration Agency (JAXA); ⑤ Faults, river systems, landform types, road networks, and seismic intensity, used as landslide hazard assessment factors. Fault and river system data were obtained from the National Geomatics Center of China; landform type data from the National Geographic Information Center; road network data from the Bigemap map downloader; and seismic intensity data from the China Earthquake Administration. Detailed data descriptions are provided in Table 1.

[0080]

[0081] Selection of landslide hazard assessment factors and assessment units

[0082] 2.1 Selection of conventional evaluation factors

[0083] The risk of geological hazards mainly stems from three aspects. First, there are inherent factors of the slope itself, including its material composition, morphology, structure, topography, geological structure, and hydrogeological conditions. These factors determine the ease or difficulty of slope instability. Another important factor is evidence of slope instability, which mainly reflects the history of landslides in the study area and the current deformation state of the slope. Second, there are triggering factors, including earthquakes, rainfall, and human activities. The presence of these factors increases the likelihood of slope instability. Third, there are the number and intensity of landslides; the scale, velocity, and sliding distance of landslides all reflect their destructive power. These three aspects comprehensively reflect the magnitude of landslide risk. The selection of geological hazard risk assessment factors needs to be combined with the actual situation of the study area, summarizing and analyzing the disaster-inducing background and mechanisms, and identifying the factors most relevant to the occurrence of geological hazards as assessment factors. Only in this way can the assessment results be most accurate and valuable. Based on previous research on the landslide formation mechanism in the study area, eight traditional landslide factors were considered: elevation, slope, distance to road, distance to river, distance to fault, slope aspect and plane curvature, and seismic intensity index, as landslide hazard assessment factors for the study area. Figure 2 As shown.

[0084] Using the frequency ratio FR)Analyze the correlation between landslide distribution and landslide condition factors. F FR for FR The value, F FR Defined as the ratio of landslide occurrence to the total area of ​​the study area. If F FR If the value is 1, then the landslide is considered to be generally correlated with the condition factors. F FR A value greater than 1 indicates a high correlation. F FR If the correlation is less than 1, it is considered low. F FR The calculation formula (1) is as follows:

[0085] (1)

[0086] In the above formula, f ij It is a condition factor j Classification i Medium landslide density; f It is the landslide density in the entire study area; P ij * It is a condition factor j Classification i Number of medium-sized landslides; P ij It is a condition factor j Classification i The area in; P It is the total area of ​​the study area; P * This represents the total number of landslides that occurred in the study area.

[0087]

[0088]

[0089] Analysis of Table 2 shows that, for the slope factor, the FR (Front-to-Slope Ratio) for slopes greater than 62° is greater than 1.32, indicating a high correlation between landslide distribution and landslide condition factors. For the aspect and curvature factors, the FR for each direction is close to 1, indicating a low correlation between landslide distribution and landslide condition factors. For the elevation factor, the FR is 2.18 between 1312m and 1755m, indicating a high correlation between landslide distribution and landslide condition factors. In areas close to rivers, roads, faults, and earthquake intensity, the FR is greater than 1, indicating a high correlation between landslide distribution and factors.

[0090] The percentage of landslides with FR > 1 is shown in Table 3. The factors that have a greater impact on landslide distribution are: seismic intensity > distance from river > altitude > distance from road > distance from fault > aspect > curvature > slope.

[0091]

[0092] Deformation rate factor

[0093] Based on SBAS-InSAR technology, 103 Sentinel-1A ascending-orbit slant-range single-look complex (SLC) images from January 9, 2019 to June 22, 2022, after registration and cropping, were selected. Following the principle of optimal temporal and vertical baselines, the image from February 14, 2019, was used as the super master image. A temporal baseline threshold of 36 days and a spatial baseline of 45% of the critical baseline threshold were set, generating a total of 318 interferometric pairs. To suppress speckle noise, the multilook ratio was set to 1:4. The Minimum Cost Flow unwrapping method and Goldstein filtering method were used for interferometric processing to obtain the registered combined interferometric phase.

[0094] (2)

[0095] In the formula: Phase generated by slant-range deformation, Indicates terrain phase, Phase caused by atmospheric delay Phase caused by coherent noise. The interference phase is expressed as the average phase velocity between two time phases. v and time t The product of these factors is used to select GCP points without redundant terrain fringes and phase jumps, and those far from the deformation region, for orbit refinement and re-flattening, to obtain the first... i Phase value of amplitude interferogram As shown in equations (3) and (4):

[0096] (3)

[0097] (4)

[0098] The integral of each temporal velocity between the master and slave images is rewritten as follows: m × n matrix B Thus, we obtain matrix equation (5).

[0099] (5)

[0100] Because small baseline set difference interferometry employs a multi-master image strategy, the matrix...B Because rank deficiency is easily generated, least squares method and singular value matrix decomposition are used for deformation inversion. Then, atmospheric phase is estimated and removed to obtain time-series deformation information of the study area. After geocoding the time-series information, the deformation rate in the LOS (Line of Sight, LOS) direction of the study area is obtained. The data processing process includes:

[0101] (1) Generation of connecting lines;

[0102] (2) Interference with the workflow;

[0103] This process performs interferometric processing on all paired interferometric image pairs, including coherence generation, flattening, filtering, and phase unwrapping. Ultimately, all data pairs are registered onto the super master image to prepare for the subsequent first and second inversions.

[0104] (3) First inversion

[0105] The purpose of the first inversion is to calculate the remaining height and displacement velocity using the initial inversion model, which is used to process complex interferograms. The first inversion identifies a certain number of permanent scattering points, and then processes the pixels around them.

[0106] (4) Second inversion

[0107] The purpose of the second inversion is to use the linear model results from the first inversion to estimate the atmospheric phase component, then remove the atmospheric phase component, and fit the result to obtain the final displacement velocity.

[0108] (5) Geographic coding

[0109] (6) InSAR data processing and analysis

[0110] Based on the InSAR surface deformation monitoring results, mathematical statistical theory and other methods are used to analyze the spatiotemporal evolution of surface deformation in the monitoring area and obtain the deformation rate results of geological disasters in the monitoring area.

[0111] Correlation analysis among evaluation factors

[0112] Multicollinearity analysis is a statistical evaluation method used in linear regression analysis to assess whether there is a strong linear correlation among independent variables. Generally, the variance inflation factor (VIF) and tolerance (TOL) are used to evaluate multicollinearity among factors. If TOL is less than 0.5 and VIF is greater than 2, it indicates strong multicollinearity among the factors; otherwise, there is no multicollinearity problem. The statistical results are shown in Table 4.

[0113]

[0114] As can be seen from the table, the variance inflation of these five factors is between 0.598 and 0.876, and the tolerance is between 1.141 and 1.673, indicating that there is no multicollinearity problem.

[0115] Landslide Hazard Assessment Unit

[0116] Currently, in geological hazard risk assessment, the common evaluation units are of the following four types: (1) regular grid, where the grid can be squares, rectangles, or other shapes of the same size and shape; (2) irregular primitives, which are irregular primitives automatically generated by overlaying various image factor layers; (3) using a single landslide or other geological hazard unit; and (4) administrative units, which directly use administrative units that are smaller than the evaluation units. Regardless of which evaluation unit is used, existing mathematical models will treat each evaluation unit as an indivisible potential landslide or unstable slope. However, there is currently no good method for dividing natural hazard units. This study uses the townships of Ludian County as evaluation units to conduct landslide hazard risk assessment, such as Figure 4 As shown.

[0117] Landslide Hazard Assessment Model Based on Bipolar Fuzzy Sets

[0118] The biggest difference between bipolar fuzzy sets and traditional fuzzy sets is that the domain of the fuzzy set is extended from [0,1] x [0,1] to [0,1] x [1,0]. Let... X It is a finite domain of discourse, a bipolar fuzzy set B It manifests in the following form (6):

[0119] (6)

[0120] Positive membership degree Represents element x Regarding bipolar fuzzy sets B Satisfaction with a certain attribute, negative membership degree Represents element x Regarding bipolar fuzzy sets B The degree of satisfaction of the opposite property of this attribute. , There are two mappings.

[0121] Suppose there is a set of alternative solutions There is also a set of attributes .set up It is the weight of the attribute, for example It is an attribute The weights, and

[0122] , (7)

[0123] Assumption It is a bipolar fuzzy set decision matrix (a table below provides a brief explanation). μ ij Indicates positive membership degree. v ij Indicates negative membership degree. The membership degree and negative membership degree are used by decision-makers to evaluate a given alternative. μ i Satisfy a certain attribute p j The degree of.

[0124] Statistical Analysis of Landslide Disaster Evaluation Factors

[0125] Evaluation factors were selected based on three aspects of the sources of geological hazard risk: first, the landslide itself; second, the triggering factors; and third, the number of landslides. These three aspects can comprehensively reflect the magnitude of landslide risk. Seismic intensity, fault, slope, number of landslides, and deformation rate were selected as factors. To assign membership values ​​to each factor, factor analysis was performed on each unit. For statistical convenience, P1 represents the landslide point, P2 represents the fault, P3 represents the slope, P4 represents the seismic intensity, and P5 represents the deformation rate. Township names were represented by V1-V12. The statistical analysis results are shown in Table 5.

[0126]

[0127] As can be seen from Table 5, Longtoushan Town has the most landslides, while Ciyuan Hui Township has the fewest. Lehong Town and Suoshan Town have a higher proportion of slopes greater than 20°, while Taoyuan Hui Township has the fewest. Almost all towns have faults, and the only town without faults is Xinjie Town. Longtoushan Town has the highest seismic intensity.

[0128] Based on SBAS-InSAR technology, 103 Sentinel-1A ascending-orbit slant-range single-look complex (SLC) images, registered and cropped from January 9, 2019 to June 22, 2022, were selected. Following the principle of optimal temporal and vertical baselines, the image from February 14, 2019, was chosen as the super master image. The deformation rate of each township was analyzed, and the deformation rates are as follows: Figure 5 As shown:

[0129] right Figure 5 The proportion of deformation values ​​in townships was statistically analyzed, and the results are shown in Table 6.

[0130]

[0131] Combination Figure 5 As can be seen from Table 6, the deformation values ​​of most townships are concentrated between -100mm and 100mm, while some areas in Lehong Township have deformation values ​​between -300mm and 400mm.

[0132] The weights of each landslide factor were determined using the analytic hierarchy process.

[0133] The Analytic Hierarchy Process (AHP) is a multi-factor decision analysis method combining quantitative and qualitative approaches, proposed by American mathematician TLSatty in the 1970s. This method divides the problem into different levels, constructs judgment matrices between adjacent levels, finds the largest eigenvalue and corresponding eigenvector of the matrix, and then performs a weighted summation of the hierarchical ranking to obtain the weight values ​​of the evaluation factors relative to the target level. The advantages of AHP lie in its ability to build judgment matrices through pairwise comparisons of evaluation factors. The final result comprehensively considers the interrelationships between evaluation factors, rather than the results of a single or few factors acting independently. This better reflects the results of multiple factors acting together in reality. Furthermore, AHP is easy to understand and operate, and specialized AHP software is available, improving work efficiency in practical work. It is currently widely used in geological hazard risk assessment.

[0134] Based on the disaster-prone conditions and disaster-causing patterns summarized from remote sensing interpretation and field investigations, landslide data, faults, slope, seismic intensity, and deformation rate were selected as evaluation factors. The analytic hierarchy process (AHP) was used to determine the weights of each evaluation factor; the specific steps are as follows:

[0135] (1) Establish a hierarchical structure

[0136] First, the problem is decomposed into three layers: the objective layer, the criterion layer, and the solution layer, establishing a hierarchical structure model. After analyzing the characteristics of the five selected indicators, their hierarchical relationship is as follows: Figure 6 As shown.

[0137] (2) Construct the judgment matrix

[0138] The analytic hierarchy process (AHP) determines factor weights by decomposing each element into three levels (upper, middle, and lower) according to different attributes from top to bottom, and establishing a hierarchical evaluation model based on the stratification. Factors within the same level of two levels with subordinate relationships are compared pairwise, establishing a series of judgment matrices in the following form:

[0139] (8)

[0140] The standard for its measurement adopts the 1-9 scale method, and the specific meaning is shown in Table 7.

[0141]

[0142]

[0143] The eigenvectors of the evaluation matrix are calculated and then normalized to obtain the weight vectors of the system. :

[0144] (9)

[0145] To improve the reliability and accuracy of the judgment matrix sorting and avoid interference from other factors, the consistency of the judgment matrix should be checked.

[0146] (10)

[0147] At that time, To determine if the matrix has good consistency.

[0148]

[0149]

[0150] From Tables 9 and 10, we can see that the deformation rate has the largest weight (0.418), while the slope has the smallest weight (0.263). The consistency test of these results is passed.

[0151] Landslide hazard assessment results

[0152] The bipolar fuzzy set evaluation method was used to assess landslide hazard, using bipolar membership functions to express uncertainty. For the five main factors in the study area—existing landslide data, faults, slope, seismic intensity, and deformation rate—the fuzzy membership degree was calculated for each factor using an expert scoring method to evaluate the risk of small-scale regional geological hazards. The membership degree values ​​of the evaluated factors are shown in Table 11.

[0153]

[0154] According to the definition of the bipolar fuzzy set evaluation method, the process of solving decision problems is as follows:

[0155] Step 1: Use the following operators to process the information of each alternative in R.

[0156] (11)

[0157] For the alternative solutions in the decision matrix R v The information in section 2 is processed, and the attribute weights are all p.

[0158] (12)

[0159] Step 2: Calculate the score function for each alternative.

[0160] (13)

[0161] Step 3: Sort the score functions of each candidate solution. If two candidate solutions have the same score function, calculate the accuracy function of these two candidate solutions. ,

[0162] (14)

[0163] The functions with higher precision are ranked higher.

[0164] Step 4: Rank the landslide risks.

[0165] The evaluation scores for each township are shown in Table 12.

[0166]

[0167] The landslide risk is ranked as follows: Longtoushan Town (v7) > Lehong Town (v9) > Suoshan Town (v5) > Xiaozhai Town (v2) > Shuimo Town (v6) > Huodehong Town (v11) > Jiangdi Town (v10) > Xinjie Town (v1) > Longshu Town (v8) > Wenping Town (v3) > Taoyuan Hui Township (v4) > Ciyuan Hui Township (v12). Based on the evaluation results, Longtoushan Town, Lehong Town, and Suoshan Town have the highest landslide risk, indicating a relatively high risk coefficient for these towns. Wenping Town, Taoyuan Hui Township, and Ciyuan Hui Township have the lowest risk levels. Figure 8 As shown.

[0168] Backpropagation (BP) neural networks can learn and store a large number of input-output pattern mappings. The BP model consists of an input layer, hidden layers, and an output layer. Its core principle is to implicitly establish a functional relationship between input samples and expected values ​​using the neural network. Under training stimulation from the input training sample set, the network weights are continuously changed to adjust the mapping relationship between training samples and output results. The squared network error is used as the objective function, and the minimum value of the objective function is calculated using gradient descent. BP neural networks possess arbitrarily complex pattern classification capabilities, excellent multidimensional function mapping capabilities, strong nonlinear mapping capabilities, and a flexible network structure.

[0169] The Back Propagation (BP) algorithm was selected, utilizing DEM, slope, aspect, and curvature data, potential landslide points, and removing the influence of topographic phase on surface deformation. Rainfall, faults, stratigraphic lithology, river systems, land use, Normalized Difference Vegetation Index (NDVI), topography, road network, and seismic intensity were used as evaluation factors for the model. A natural discontinuity grading method combined with expert participation was employed to classify susceptibility into five levels: extremely low susceptibility, low susceptibility, moderate susceptibility, high susceptibility, and extremely high susceptibility. Figure 9 As shown.

[0170] The landslide susceptibility assessment results and landslide point distribution patterns calculated by the above model show the following main characteristics in the zoning results:

[0171] (1) The high to extremely high landslide susceptibility areas are mainly distributed along the boundary line within the territory, from the northwest corner to the southeast. Landslides are more developed in the area north of the Niulan River, especially concentrated in Longtoushan Town, Lehong Town and Xinjie Town.

[0172] (2) The high to very low landslide-prone areas are mainly distributed in areas such as Ciyuan Hui Township, Taoyuan Hui Township, and Wenping Town.

[0173] In summary, the results of the evaluation using the bipolar fuzzy set method are basically consistent with those of the BP neural network method. Longtoushan Town and Lehong Town have the highest landslide disaster risk, indicating that the landslide risk coefficient of these towns is relatively large. Wenping Town, Taoyuan Hui Township, and Ciyuan Hui Township have the lowest risk levels. The largest differences are found in Xinjie Town and Xiaozhai Town. Among the 12 towns, the evaluation results of 10 towns are relatively close, reaching 83%, indicating that the method has strong applicability.

[0174] The above descriptions are merely some specific embodiments of the present invention. Commonly known details or common knowledge in the solutions are not described in detail here (including but not limited to abbreviations, acronyms, and units conventionally used in the art). It should be noted that the above embodiments do not limit the present invention in any way. For those skilled in the art, any technical solutions obtained by equivalent substitution or equivalent transformation fall within the protection scope of the present invention. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A method for assessing landslide hazard risk using bipolar fuzzy sets, characterized in that, The method includes the following steps: Step 1), data collection includes two parts; Part 1: collecting landslide point data; Part 2: collecting Sentinel-1A radar imagery data, precise orbit determination ephemeris data, Google satellite imagery data, DEM data, slope, aspect and curvature data, fault data, river system data, landform type data, road network data, and seismic intensity data. Step 2) Determination of landslide hazard assessment factors; Step 3) Establish a landslide hazard assessment model based on bipolar fuzzy sets; Step 2) Determination of landslide hazard assessment factors, including: The landslide hazard assessment factors in the study area include: altitude, slope, distance to road, distance to river, distance to fault, slope aspect and plane curvature, and seismic intensity index. The correlation between landslide distribution and landslide condition factors was analyzed using the frequency ratio (FR). F FR for FR The value, F FR Defined as the ratio of landslide occurrence to the total area of ​​the study area; if F FR If the value is 1, then the landslide is considered to be generally correlated with the condition factors. F FR A value greater than 1 indicates a high correlation. F FR If the correlation is less than 1, it is considered low. F FR The calculation formula (1) is as follows: （1） In the above formula, f ij It is a condition factor j Classification i Medium landslide density; f It is the landslide density in the entire study area; P ij * It is a condition factor j Classification i Number of medium-sized landslides; P ij It is a condition factor j Classification i The area in; P It is the total area of ​​the study area; P * This represents the total number of landslides that occurred in the study area; Step 3) includes expanding the domain of the fuzzy set from [0,1] x [0,1] to [0,1] x [1,0]; let... X It is a finite domain of discourse, a bipolar fuzzy set B The following form (6): （6） Positive membership degree Represents element x Regarding bipolar fuzzy sets B Satisfaction with a certain attribute, negative membership degree Represents element x Regarding bipolar fuzzy sets B The degree of satisfaction of the opposite property of this attribute. , There are two mappings; Suppose there is a set of alternative solutions There is also a set of attributes ;set up It is the weight of the attribute, for example It is an attribute p A weight of 1, and （7） Assumption It is a bipolar fuzzy set decision matrix. μ ij Indicates positive membership degree. v ij Indicates negative membership degree. The membership degree and negative membership degree are used by decision-makers to evaluate a given alternative. μ i Satisfy a certain attribute p j The degree; Step 3) includes using the bipolar fuzzy set evaluation method to assess landslide hazard, expressing uncertainty based on the bipolar membership function; according to the definition of the bipolar fuzzy set evaluation method, the process of solving the decision problem is as follows: Step 1: Process the information of each candidate solution in R using operators, and process the information of the candidate solutions in the decision matrix R. The attribute weights are all p. Step 2: Calculate the score function for each alternative. s ( r i ()( i =1,2…12); Step 3: Sort the score functions of each candidate solution. If two candidate solutions have the same score function, calculate the accuracy function of these two candidate solutions. A ( r i Furthermore, those with higher precision functions are ranked higher. Step 4: Rank the landslide risks.

2. The method for assessing landslide hazard risk using bipolar fuzzy sets according to claim 1, characterized in that, Step 2) Determination of landslide hazard assessment factors, including: deformation rate factor. The steps are as follows: Based on SBAS-InSAR technology, 103 Sentinel-1A ascending-orbit slant-range single-look complex images after registration and cropping are selected. According to the principle of optimal temporal and vertical baselines, the temporal baseline threshold and spatial baseline are set. To suppress speckle noise, the Minimum CostFlow unwrapping method and Goldstein filtering method are used for interferometric processing to obtain the registered combined interferometric phase. （2） In the formula: Phase generated by slant-range deformation, Indicates terrain phase, Phase caused by atmospheric delay Phase caused by coherent noise; the interference phase is expressed as the average phase velocity between two time phases. v and time t The product of these factors is used to select GCP points without redundant terrain fringes and phase jumps, and those far from the deformation region, for orbit refinement and re-flattening, to obtain the first... i Phase value of amplitude interferogram ; As shown in equations (3) and (4): （3） （4） The integral of each temporal velocity between the master and slave images is rewritten as follows: m × n matrix B Thus, we obtain matrix equation (5). （5） Because small baseline set difference interferometry employs a multi-master image strategy, the matrix... B Since rank deficiency is easily generated, the least squares method and singular value matrix decomposition are used to perform deformation inversion. Then, the atmospheric phase is estimated and removed to obtain the time series deformation information of the study area. After geocoding the time series information, the deformation rate in the LOS direction of the study area is obtained.

3. The method for evaluating landslide hazard risk using bipolar fuzzy sets according to claim 2, characterized in that, The data processing procedure for the deformation rate results includes: (1) Generation of connecting lines; (2) Interference with the workflow; The process involves interferometry processing of all paired interferometric images, including coherence generation, flattening, filtering, and phase unwrapping. Ultimately, all data pairs are registered onto the super master image in preparation for the subsequent first and second inversions. (3) First inversion The purpose of the first inversion is to use the first inversion model to calculate the remaining height and displacement velocity, which is used to process complex interferograms; the first inversion identifies a certain number of permanent scattering points, and then processes the pixels around them. (4) Second inversion The purpose of the second inversion is to use the linear model results from the first inversion to estimate the atmospheric phase components, then remove the atmospheric phase components, and fit the data to obtain the final displacement velocity. (5) Geographic coding (6) InSAR data processing and analysis Based on the InSAR surface deformation monitoring results, mathematical statistical theory was used to analyze the spatiotemporal evolution of surface deformation in the monitoring area and obtain the deformation rate results of geological disasters in the monitoring area.

4. The method for evaluating landslide hazard risk using bipolar fuzzy sets according to claim 2, characterized in that, The analytic hierarchy process (AHP) was used to determine the weights of each evaluation factor. The specific steps are as follows: (1) Establish a hierarchical structure First, the problem is decomposed into three layers: the goal layer, the criteria layer, and the solution layer, establishing a hierarchical model. Then, the characteristics of the five selected indicators are analyzed. (2) Construct the judgment matrix The analytic hierarchy process (AHP) determines factor weights by decomposing each element into three levels (upper, middle, and lower) according to different attributes from top to bottom, and establishing a hierarchical evaluation model based on the stratification. Factors within the same level of two or more levels with subordinate relationships are compared pairwise to establish a series of judgment matrices in the following form: （8） Its measurement standard adopts the 1-9 scale method; The eigenvectors of the evaluation matrix are calculated and then normalized to obtain the weight vectors of the system. : （9）。 5. The method for evaluating landslide hazard risk using bipolar fuzzy sets according to claim 1, characterized in that, The operator is: （11） The alternative solutions in the decision matrix R v When processing information from point 2, the attribute weights are all p. （12） The scoring function s ( r i ) for: （13） The precision function A ( r i )for, （14）。

Images