Landslide susceptibility evaluation method and device and storage medium

Through the PU bagging algorithm and XGBoost model, the scenarios with only positive samples and unlabeled samples are processed in the landslide susceptibility evaluation, which solves the sample imbalance and noise problems, improves the accuracy and speed of landslide susceptibility prediction, and achieves more Efficient landslide proneness evaluation.

CN119066480APending Publication Date: 2024-12-03WUHAN CENT CHINA GEOLOGICAL SURVEY CENT SOUTH CHINA INNOVATION CENT FOR GEOSCIENCES

Patent Information

Application Number
CN202410908754.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-08
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The existing landslide prediction model has inaccurate prediction results in scenarios with only positive samples and unlabeled samples, and is not very accurate, mainly due to sample imbalance and reduced feature learning effect due to the influence of noise data.

Method used

The PU bagging algorithm is used to train a weak classifier from random sampling samples in the unlabeled dataset. Multiple weak classifiers are generated through the iterative sampling-modeling-training process. The final prediction model is constructed in combination with the XGBoost algorithm, and the average value of the out-of-bag samples is used for this. Forecast and divide the landslide susceptibility evaluation level through the natural breakpoint method.

Benefits of technology

It significantly improves the generalization performance and accuracy of the landslide prone prediction model, avoids the error of artificial selection of negative samples, improves the prediction speed and accuracy, overcomes the problem of sample imbalance, and reduces the impact of noise.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119066480A_ABST
    Figure CN119066480A_ABST
Patent Text Reader

Abstract

The invention relates to the field of geological disaster monitoring, and discloses a landslide susceptibility evaluation method and device and a storage medium, and the method comprises the steps: analyzing the original data of a landslide historical event and a basic environment factor, and building an evaluation factor geospatial data set through a frequency ratio method; a bagging thought is utilized, random sampling is carried out on the unmarked sample set U through a repeatable sampling Bootstrap method, the unmarked sample set U is combined with all the landslide positive sample sets P to create a training set, and a limit gradient tree is adopted as a base learning device to generate a weak classifier; taking an average value of a plurality of weak classifiers as a final prediction model through a multi-iteration process; and performing susceptibility prediction on all sample points in the evaluation factor geospatial data set by using the final prediction model to obtain a landslide susceptibility evaluation partition map. The method has the advantages that the problem that the proportion of the positive samples and the negative samples of the landslide is unbalanced is solved, and the generalization performance and accuracy of the landslide susceptibility prediction model are remarkably improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of geological disaster monitoring, and in particular to a landslide susceptibility evaluation method, equipment and storage medium. Background Art

[0002] Landslides are one of the most frequent geological disasters, often causing serious damage to natural resources, ecological environment and infrastructure, and seriously threatening people's lives and property. Against the backdrop of global climate change, extreme weather events have increased, and the originally fragile geological environment has exacerbated the risk of landslides. Landslide disaster prevention and mitigation work has become an increasingly urgent task for today's society.

[0003] Carrying out landslide susceptibility assessment is an important basic basis for landslide disaster prevention and mitigation. Landslide susceptibility refers to the possibility of landslides occurring in a certain area under certain geological and environmental conditions. It focuses on evaluating the possible locations where landslides may occur and the spatial probability of landslide occurrence.

[0004] In the process of landslide susceptibility assessment, the most critical link is to establish a suitable evaluation model. An excellent susceptibility assessment model can fully explore the mapping relationship between landslides and their basic environmental factors and construct a nonlinear function from basic environmental factors to the spatial probability of landslides.

[0005] The commonly used landslide susceptibility assessment models at home and abroad are divided into non-deterministic models and deterministic models. In the development history of susceptibility assessment models, traditional non-deterministic models based on statistical analysis have been widely used in previous studies, such as information model, evidence weight model, hierarchical analysis method, fuzzy comprehensive evaluation method, etc. With the development of data mining and artificial intelligence, scholars at home and abroad have gradually applied algorithm models in the field of machine learning to landslide susceptibility assessment, such as logistic regression (LR), decision tree (DT), support vector machine (SVM), neural network (ANN), etc.

[0006] The input of traditional binary classifiers based on machine learning usually requires two sets of samples with corresponding samples, including positive samples (landslide samples) and negative samples (non-landslide samples).

[0007] In actual landslide susceptibility prediction, imperfect situations exist. The most common situation is that there are only positive samples and no-sample samples in the training set. For ease of operation, no-sample samples are usually used as negative samples to construct a binary classifier for supervised learning, resulting in some samples being labeled as unreliable samples, which seriously affects the features learned by the algorithm and reduces the accuracy of the final algorithm. Summary of the invention

[0008] The purpose of the present invention is to propose a landslide susceptibility assessment method, device and storage medium to solve the technical problem that the prediction results of the existing landslide prediction model are inaccurate and the accuracy is not high. The present invention is applicable to specific scenarios with only positive samples and unlabeled samples.

[0009] The present invention provides a method for evaluating landslide susceptibility, comprising the following steps:

[0010] S1: Obtaining original landslide data; the original landslide data includes original data of historical landslide events and original data of basic environmental factors for landslide susceptibility assessment;

[0011] S2: Establishing geospatial dataset of basic environmental factors based on original landslide data;

[0012] S3: Extract the geospatial dataset of basic environmental factors, take the original data of historical landslide events as the positive sample set P, and take the rest of the data as the unlabeled dataset U. Use the PU bagging algorithm to randomly extract samples from the dataset U and train the weak classifier.

[0013] S4: Set the number of iterations T, repeat the sampling-modeling-training process, use the base learner to obtain T weak classifiers, and take the average of T iterations as the final prediction model;

[0014] S5: The final prediction model is used to predict each sample in all geospatial data sets to obtain the landslide susceptibility index, and the landslide susceptibility evaluation level is divided according to the natural breakpoint method and the susceptibility zoning map is drawn.

[0015] A storage medium stores instructions and data for implementing a landslide susceptibility assessment method.

[0016] A landslide susceptibility assessment device comprises: a processor and the storage medium; the processor loads and executes instructions and data in the storage medium to implement a landslide susceptibility assessment method.

[0017] The beneficial effects provided by the present invention are:

[0018] 1. It overcomes the problem of unbalanced ratio of positive and negative samples of landslides. Traditional supervised learning artificially treats unlabeled samples as negative samples, resulting in the number of negative samples being much larger than positive samples, which affects the algorithm effect of the classifier. The present invention constructs a balanced sample set training to significantly improve the generalization performance of the landslide susceptibility prediction model.

[0019] 2. There is no need to manually select negative samples, which avoids the selection error being continuously amplified during iterative training.

[0020] 3. Improve the prediction accuracy of the model. Traditional supervised learning treats unlabeled samples as negative samples, and unlabeled samples also contain a small number of positive samples, which are noisy data and seriously affect the features learned by the algorithm. The present invention uses multi-classifier combined prediction to significantly improve the accuracy of the final algorithm.

[0021] 4. By utilizing parallelization technology, the landslide susceptibility prediction speed of the present invention is faster. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a schematic flow chart of the method of the present invention;

[0023] Figure 2 The landslide susceptibility evaluation factors in the embodiment are: elevation;

[0024] Figure 3 The landslide susceptibility evaluation factors in the embodiment are: slope;

[0025] Figure 4 The landslide susceptibility evaluation factors in the embodiment are: slope direction;

[0026] Figure 5 The factors for evaluating the susceptibility of landslides in the embodiment are: curvature;

[0027] Figure 6 The landslide susceptibility evaluation factors in the embodiment are: stratum lithology;

[0028] Figure 7 The landslide susceptibility evaluation factors in the embodiment are: annual average rainfall;

[0029] Figure 8 The landslide susceptibility evaluation factor in the embodiment: NDVI;

[0030] Fig. 9 This is a heat map of the correlation analysis of basic environmental factors in the embodiment;

[0031] Fig.10 This is a graph showing the PU bagging-XGBoost landslide susceptibility evaluation results of the embodiment;

[0032] Fig.11 It is a working schematic diagram of the hardware device of the present invention. DETAILED DESCRIPTION

[0033] To make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0034] Before formally describing the present invention, the scheme of the present invention is first generally described for easy understanding.

[0035] Please refer to Figure 1 , Figure 1 It is a schematic flow diagram of the method of the present invention;

[0036] The present invention provides a landslide susceptibility evaluation method, comprising the following steps:

[0037] S1: Obtaining original landslide data; the original landslide data includes original data of historical landslide events and original data of basic environmental factors for landslide susceptibility assessment;

[0038] It should be noted that the original data of historical landslide events come from detailed geological disaster investigation information, the digital elevation model (DEM) comes from the basic data of the free and open Geographic Data Cloud, the stratigraphic lithology comes from the regional geological map, the NDVI comes from the Landsat-8 remote sensing image, and the annual average rainfall comes from the meteorological station observation data.

[0039] It should be noted that the basic environmental factors for landslide susceptibility assessment include seven factors: elevation, slope, aspect, curvature, average annual rainfall, normalized difference vegetation index (NDVI), and stratum lithology.

[0040] S2: Establishing geospatial dataset of basic environmental factors based on original landslide data;

[0041] As an embodiment, the present invention uses a landslide frequency ratio method to process the original data to establish a basic environmental factor geospatial data set. Specifically, the frequency ratio method is:

[0042]

[0043] Where: n i is the number of landslide grids in each type of factor in the i-th classification interval; N is the total number of landslide grids in the area; s i is the number of grids in the i-th classification interval of each type of factor; S is the total number of grids in the study area; FR is the frequency ratio of this type of factor;

[0044] It should be noted that, in step S2, the correlation heat map and variance inflation factor VIF are also used to eliminate the basic environmental factors for landslide susceptibility assessment whose correlation and collinearity exceed preset values.

[0045] The variance inflation factor VIF is:

[0046]

[0047] Where: is the ith independent variable x i The coefficient of determination between it and the other independent variables.

[0048] In the present invention, the heat map correlation analysis is used to eliminate factors with high correlation to improve the model efficiency and prediction accuracy; the variance inflation factor (VIF) is used to characterize the degree of collinearity between factors. When the VIF is greater than 10, it indicates that there is serious collinearity between the data and it needs to be eliminated;

[0049] S3: Extract the geospatial dataset of basic environmental factors, take the original data of historical landslide events as the positive sample set P, and take the rest of the data as the unlabeled dataset U. Use the PU bagging algorithm to randomly extract samples from the dataset U and train the weak classifier.

[0050] It should be noted that step S3 is specifically as follows:

[0051] S31. Use the repeatable sampling Bootstrap method to extract subset samples from the data set U as negative samples N, and collect all positive samples P to construct a training set. The number of negative samples N is the same as that of positive samples P.

[0052] S32, construct a weak classifier using the training set of step S31, treating the positive sample P as 1 and the negative sample N as 0;

[0053] S33. Apply the classifier to other unlabeled sample data that have not been sampled, record them as out-of-bag samples OOB, and record their scores.

[0054] S4: Set the number of iterations T, repeat the sampling-modeling-training process, use the base learner to obtain T weak classifiers, and take the average of T iterations as the final prediction model;

[0055] It should be noted that the base learner in step S4 adopts the extreme gradient boosting tree (eXtreme GradientBoosting, XGBoost).

[0056] As one of the latest algorithms, XGBoost is the trump card of ensemble learning. Currently, it is rarely used in landslide assessment research. Compared with the traditional gradient boosting tree, XGBoost is faster than other ensemble algorithms using gradient boosting, and has been considered to be an advanced estimator with super high performance in both classification and regression. The XGBoost algorithm changes to a greedy strategy for the continuous iteration of weight allocation in the classification of erroneous samples. The best training direction is the direction of gradient descent of the loss function. Taylor's second-order expansion is used to optimize the loss function. At the same time, a regularization term is added to control the complexity of the model to prevent overfitting.

[0057] XGBoost builds the optimal model by minimizing the loss function. The loss function increases the regularization term of the model complexity. The objective function is set to optimize the iterative model using the idea of ​​minimizing structural risk. The objective function of the model is:

[0058]

[0059] Where: is the model prediction value of the i-th sample in the t-th round, y i is the true value, is the prediction error of the tth round, n is the total number of samples, Ω(f k ) is the regularization term of the k-round, which represents the complexity of the k-round model and is defined as:

[0060]

[0061] Where: T is the number of leaf nodes, ‖w‖ is the modulus of the leaf node vector, γ represents the difficulty of node segmentation, and λ represents the regularization coefficient;

[0062] The Taylor second-order expansion optimization of the objective function is performed, Substitute (4) into (3):

[0063]

[0064] Where: I j ={i|q(x i )=}, represents the sample set of the jth leaf node, q(x i ) is the structure of the tree. G j and H j is a deterministic quantity. Equation (5) can be regarded as a quadratic function about the leaf node w. By minimizing (5), we can obtain the optimal parameters and the optimal loss function:

[0065]

[0066]

[0067] In step S4 of the present invention, step S3 is repeated, and the training is iterated T times to obtain the probability scores of the out-of-bag samples OOB predictions of T weak classifiers, and the average value of the out-of-bag samples OOB is taken as the final prediction value.

[0068] S5: The final prediction model is used to predict each sample in all geospatial data sets to obtain the landslide susceptibility index, and the landslide susceptibility evaluation level is divided according to the natural breakpoint method and the susceptibility zoning map is drawn.

[0069] It should be noted that after step S5, statistical indicators including positive predictive value, negative predictive value, sensitivity, specificity, accuracy, and root mean square error, as well as the area under curve (AUC) of the receiver operating characteristic (ROC) curve are also used to verify the performance and prediction accuracy of the final prediction model.

[0070] The present invention is suitable for evaluating landslide susceptibility in special scenarios with only positive samples and unlabeled samples. It uses the idea of ​​bagging, and its working principle is to create a training set by randomly combining all positive landslide samples and unlabeled samples through a repeatable sampling Bootstrap method. The XGBoost model is used as a base learner to generate multiple weak classifiers, each of which independently learns and makes predictions for out-of-bag samples OOB. These predictions are finally combined into a combined prediction, so it is better than any single classification prediction.

[0071] As an embodiment, the present invention takes District C in City B, Province A as the research area and continues to explain the detailed steps of the method of the present invention.

[0072] Investigate and collect data on historical landslide events and basic environmental factors affecting landslides in the study area. The study is located in the central part of Province A, with a total area of ​​1974.5km 2 , is the highest district in City B. The terrain slopes from north to south, with the highest point at 1210m above sea level and the lowest point at 16.2m above sea level.

[0073] The landform is dominated by low mountains and hills, with steep terrain and well-developed gullies. The main exposed strata are Proterozoic gneiss, Devonian siltstone, Carboniferous limestone dolomite, Triassic quartz sandstone, Jurassic-Cretaceous granite, granodiorite, Paleogene sandstone, conglomerate, and Quaternary clay. The complex and changeable geological and geomorphological conditions provide favorable conditions for the breeding of landslides.

[0074] On the basis of finding out the geological environment characteristics of the study area, seven landslide susceptibility evaluation factors were determined, namely elevation, slope, slope aspect, curvature, average annual rainfall, normalized difference vegetation index NDVI, and stratum lithology.

[0075] Detailed survey data of geological disasters in Conghua District to obtain the location distribution of historical landslide points; Digital Elevation Model (DEM), which comes from the free and open basic data of Geographic Data Cloud, is mainly used to analyze and extract basic topographic factors such as elevation, slope and slope aspect; 1:50,000 regional geological map of City A to obtain the distribution of strata and lithology; Landsat-8 remote sensing images to obtain the Normalized Difference Vegetation Index (NDVI) factor; Observation data of meteorological stations in the study area to obtain the multi-year average rainfall factor. According to the size of the study area and the scale of the landslide, a grid with a resolution of 30m×30m is used as the basic unit for landslide susceptibility evaluation, and the study area is divided into about 2.2 million grid units;

[0076] The natural break point method is used to divide the selected 7 basic environmental factors into 5 attribute intervals. The division results are shown in Figure 2 to Figure 8 The frequency ratio method based on statistical principles is used as a quantitative model. By calculating the ratio of the frequency of known landslide units appearing in each partition to the frequency of all grid units appearing in the corresponding partition in the study area, the contribution of the above seven basic environmental factor interval levels to the landslide susceptibility is quantitatively quantified. The frequency ratio is used as the quantitative value of each factor for the model training data set. The size of the frequency ratio determines whether the attribute interval of the environmental factor is conducive to the development of landslides.

[0077] Basic environmental factors are all related to the occurrence of landslides, but the possible correlation between factors will have an adverse effect on the prediction results and may increase the complexity and running time of the model. Therefore, before the evaluation, it is necessary to screen the quantitative factors and conduct correlation analysis to eliminate factors with high correlation and improve the model efficiency and prediction accuracy. Fig. 9 This is the correlation analysis heat map of 7 basic environmental factors. The correlation coefficients of each factor are small, with absolute values ​​ranging from 0 to 0.23, indicating that each basic environmental factor can be used independently as a characteristic variable for landslide susceptibility assessment.

[0078] Since collinearity between sample data can easily lead to a decrease in model training accuracy, the data set used to build the model must meet certain collinearity requirements. The variance inflation factor (VIF) is used to characterize the degree of collinearity between factors. When VIF is greater than 10, it indicates that there is serious collinearity between the data and it needs to be eliminated.

[0079] Referring to Table 1, the collinearity analysis results of the seven basic environmental factors of the present invention are shown. The VIF value of each basic environmental factor does not exceed 10, indicating that the collinearity of the sample data is not significant, and all basic environmental factors can participate in the susceptibility evaluation without being eliminated.

[0080] Table 1 Collinearity analysis of basic environmental factors

[0081] serial number Basic environmental factors VIF 1 slope 1.066933 2 Elevation 1.336441 3 Curvature 1.072272 4 Average annual rainfall 1.144736 5 NDVI 1.255478 6 Formation lithology 1.113084 7 Slope aspect 1.031587

[0082] The grid cells containing 365 historical landslide points are taken as the positive sample set P, and the remaining grid cells are taken as the unlabeled data set U. In the data set U, the repeatable sampling Bootstrap method is used to extract subset samples as negative samples N, and all positive samples P are combined to construct a training set. The number of negative samples N can be the same as the positive samples P. The positive sample P is regarded as 1, and the negative sample N is regarded as 0. A weak classifier is trained using the merged data set. The classifier is applied to other unlabeled sample data that have not been sampled, recorded as out-of-bag samples OOB (out of bag), and their scores are recorded. The extreme gradient boosting tree XGBoost is used as the base learner, and the iterative training is repeated T times. The average value of T weak classifiers is taken as the final prediction model;

[0083] The above landslide susceptibility evaluation method based on PU bagging algorithm was used to obtain the final prediction model. Each sample point in all data sets in the study area was predicted to obtain the landslide susceptibility index. The landslide susceptibility evaluation level was divided according to the natural breakpoint method and the susceptibility zoning map was drawn. Fig.10 The results show that most of the historical landslide sites are located in the extremely high and high susceptibility areas in the susceptibility zoning map, and the evaluation results are consistent with the actual landslide locations;

[0084] The performance and prediction accuracy of the landslide susceptibility assessment model were verified using sensitivity SST, specificity SPF, accuracy ACC, root mean square error RMSE, and the area under the line AUC of the receiver operating characteristic curve ROC. The results showed that the values ​​of sensitivity SST, specificity SPF, accuracy ACC, and root mean square error RMSE were 0.98, 0.99, 0.99, and 0.10, respectively, the AUC value on the training set was 0.999, and the AUC value on the test set was 0.998, indicating that the prediction model of the present invention has high accuracy.

[0085] See also Fig.11 , Fig.11 4 is a schematic diagram of the working of the hardware device of an embodiment of the present invention, wherein the hardware device specifically comprises: a landslide susceptibility assessment device 401, a processor 402 and a storage medium 403.

[0086] A landslide susceptibility assessment device 401: The landslide susceptibility assessment device 401 implements the landslide susceptibility assessment method.

[0087] Processor 402: The processor 402 loads and executes the instructions and data in the storage medium 403 to implement the landslide susceptibility assessment method.

[0088] Storage medium 403: The storage medium 403 stores instructions and data; the storage medium 403 is used to implement the landslide susceptibility assessment method.

[0089] The beneficial effects of the present invention are as follows: when the smart container is in operation, a face and commodity detection model is used to perform privacy processing on the detected face; by constructing a semi-automatic sample data set, the preparation speed before model training is improved, while the model detection accuracy is indirectly improved; by identifying and recommending a variety of commodities, the accuracy of commodity detection is improved.

[0090] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A landslide susceptibility assessment method, characterized in that: The method comprises the following steps: S1: Obtaining original landslide data; the original landslide data includes original data of historical landslide events and original data of basic environmental factors for landslide susceptibility assessment; S2: Establishing geospatial dataset of basic environmental factors based on original landslide data; S3: Extract the geospatial dataset of basic environmental factors, take the original data of historical landslide events as the positive sample set P, and take the rest of the data as the unlabeled dataset U. Use the PU bagging algorithm to randomly extract samples from the dataset U and train the weak classifier. S4: Set the number of iterations T, repeat the sampling-modeling-training process, use the base learner to obtain T weak classifiers, and take the average of T iterations as the final prediction model; S5: The final prediction model is used to predict each sample in all geospatial data sets to obtain the landslide susceptibility index, and the landslide susceptibility evaluation level is divided according to the natural breakpoint method and the susceptibility zoning map is drawn.

2. A landslide susceptibility assessment method according to claim 1, characterized in that: In step S1, the basic environmental factors for landslide susceptibility assessment include elevation, slope, aspect, curvature, average annual rainfall, normalized difference vegetation index and stratum lithology.

3. A landslide susceptibility assessment method according to claim 1, characterized in that: In step S2, the original data is processed based on the landslide frequency ratio method to establish a basic environmental factor geospatial data set.

4. A landslide susceptibility assessment method according to claim 1, characterized in that: In step S2, the correlation heat map and variance inflation factor (VIF) are also used to eliminate the basic environmental factors for landslide susceptibility assessment whose correlation and collinearity exceed preset values.

5. A landslide susceptibility assessment method according to claim 1, characterized in that: Step S3 is specifically as follows: S31. Use the repeatable sampling Bootstrap method to extract subset samples from the data set U as negative samples N, and collect all positive samples P to construct a training set. The number of negative samples N is the same as that of positive samples P. S32, construct a weak classifier using the training set of step S31, treating the positive sample P as 1 and the negative sample N as 0; S33. Apply the classifier to other unlabeled sample data that have not been sampled, record them as out-of-bag samples OOB, and record their scores.

6. A landslide susceptibility assessment method according to claim 1, characterized in that: In step S4, the base learner adopts an extreme gradient boosting tree.

7. A landslide susceptibility assessment method according to claim 5, characterized in that: Step S4 is specifically as follows: iterative training is performed T times to obtain the probability scores of the out-of-bag samples OOB predictions of T weak classifiers, and the average value of the out-of-bag samples OOB is taken as the final prediction value.

8. A landslide susceptibility assessment method according to claim 1, characterized in that: After step S5, the performance and prediction accuracy of the final prediction model are verified by using statistical indicators including positive predictive value, negative predictive value, sensitivity, specificity, accuracy, and root mean square error, as well as the area under the receiver operating characteristic curve.

9. A storage medium, characterized in that: The storage medium stores instructions and data for implementing a landslide susceptibility assessment method as described in any one of claims 1 to 8.

10. A landslide susceptibility assessment device, characterized in that: include: A processor and a storage medium; the processor loads and executes instructions and data in the storage medium to implement a landslide susceptibility assessment method as described in any one of claims 1 to 8.

Citation Information

Patent Citations

  • Landslide susceptibility evaluation method based on Bagging strategy

    CN114298436A

  • Landslide susceptibility evaluation method of bagged integrated gradient boosting decision tree

    CN117874710A

  • Landslide susceptibility dynamic evaluation method

    CN118296961A

Cited By

  • Power line patrol road risk assessment method and device, computer equipment, storage medium and computer program product

    CN119273163A

  • Landslide susceptibility mapping method and device and medium

    CN119918112A

  • Landslide disaster negative sample optimization method based on improved frequency ratio

    CN120123777A

  • Landslide disaster risk assessment method, equipment, medium and product

    CN120410236A

  • Landslide susceptibility ensemble learning evaluation method considering spatial heterogeneity partitioning and factor feature screening

    CN120764352A