Method and related equipment for assessing landslide risk in road networks

Abstract

This provides a method for assessing landslide risk in road networks. [Solution] The method includes: calculating the predicted probability of landslide occurrence in a target area using a road landslide prediction model; creating a road landslide susceptibility level map of the target area based on the predicted probability of landslide occurrence; calculating the network efficiency loss value after link disruption in the road network topology map of the target area; creating a road impact level map of the target area based on the network efficiency loss value and the original road network topology map; and determining a comprehensive road landslide risk heat map based on a nonlinear risk coupling calculation method, the road landslide susceptibility level map, and the road impact level map. Physical constraint mechanisms and complex network theory are integrated through nonlinear coupling design to realize the linkage between the geological model and road network functions, and a comprehensive road landslide risk heat map is output.

Description

[Technical Field]

[0001] This application relates to the technical field of road construction safety, and more particularly to a method for evaluating landslide risk in road networks and related equipment. [Background technology]

[0002] Mountain road networks, as vital transportation links connecting remote areas, are susceptible to frequent landslides due to slope instability caused by disasters such as earthquakes and heavy rains, leading to road closures, delays in rescue operations, and paralysis of socio-economic systems. Therefore, assessing the landslide risk of road networks under complex geological conditions is crucial for ensuring rural communication, healthcare, and regional economic systems.

[0003] In current mainstream methods for assessing landslide risk in road networks, conventional landslide susceptibility models (logistic regression, random forests, etc.) rely purely on data-driven approaches and ignore constraints imposed by geological laws (e.g., the monotonically increasing gradient-stability relationship), resulting in low reliability of predictions in areas with complex rock formations. Conventional neural networks, lacking physical equations, suffer significantly reduced generalization performance in road sections with small sample sizes. Simultaneously, much mainstream research focuses on the probability of failure of individual roads, ignoring the cascading effects of disasters on the function of road network systems. Physical information neural networks, possessing relatively high expressive power and generalization capabilities for modeling complex physical processes, are increasingly being applied to the fields of geological process modeling and prediction. At the same time, complex network theory demonstrates good adaptability in describing the topological structure of road systems and evaluating their disruption resistance and efficiency variability. [Overview of the project] [Problems that the invention aims to solve]

[0004] An object of the present application is to provide a method for evaluating landslide risks of a road network and related devices, and by combining the physical modeling of landslides with the topological evaluation of the road network, to achieve a systematic evaluation of road landslide risks, and further to achieve a scientific classification and refined management of regional road risks.

Means for Solving the Problems

[0005] To achieve the above object, the present application provides the following solutions. In a first aspect, the present application provides a method for evaluating landslide risks of a road network, including the following. Obtain the original road network topology map of the target area and each grid image within the target area, and the original road network topology map includes a plurality of nodes and links.

[0006] Input all the grid images into a trained road landslide prediction model to obtain the predicted values of landslide occurrence probabilities corresponding to each grid image. In the training process of the road landslide prediction model, the corresponding total loss function includes a data-driven term and a physical constraint term, and the physical constraint term includes a gradient monotonicity physical constraint loss term, a rainfall factor monotonicity physical constraint loss term, and a physical constraint loss term for the smoothness and continuity of elevation.

[0007] Create a road landslide susceptibility level map of the target area based on all the grid images and the predicted values of landslide occurrence probabilities corresponding to each grid image.

[0008] Calculate the first network efficiency of the original road network topology map based on the Floyd-Warshall algorithm, the population weighting mechanism, and the original road network topology map.

[0009] For each link in the original road network topology map, simulate the link interruption scenario, update the original road network topology map, obtain the updated road network topology map, and calculate the second network efficiency of the updated road network topology map based on the Floyd-Warshall algorithm, population weighting mechanism, and the updated road network topology map.

[0010] Calculate the network efficiency loss value corresponding to the interruption of each link based on the first network efficiency and the second network efficiency.

[0011] Create a road impact level map of the target area based on the network efficiency loss values corresponding to the interruption of all links in the original road network topology map and the original road network topology map.

[0012] Determine the comprehensive landslide risk heat map of the target area based on the non-linear risk coupling calculation method, the road landslide susceptibility level map, and the road impact level map. The comprehensive landslide risk heat map is used to evaluate the landslide risk of the road network.

[0013] In a second aspect, the present application provides a computer device including a memory, a processor, and a computer program stored in the memory and executable by the processor. When the processor executes the computer program, it realizes the steps in the landslide risk assessment method of the road network according to any one of the above items.

[0014] In a third aspect, the present invention provides a computer-readable storage medium in which a computer program is stored, which, when the computer program is executed by a processor, enables the implementation of the steps in the method for evaluating the landslide risk of a road network described in any one of the above paragraphs.

[0015] In a fourth aspect, the present invention provides a computer program product including a computer program, wherein when the computer program is executed by a processor, the computer program product implements the steps in the method for evaluating the landslide risk of a road network described in any one of the above paragraphs. [Effects of the Invention]

[0016] According to the specific embodiments provided herein, the present application has the following technical effects. This invention provides a method and related apparatus for evaluating landslide risk in road networks. First, a road landslide prediction model calculates a predicted landslide probability. During the training process of the road landslide prediction model, the corresponding total loss function includes a data-driven term and a physical constraint term. The physical constraint term includes a gradient monotonic physical constraint loss term, a rainfall factor monotonic physical constraint loss term, and an elevation smoothness and continuity physical constraint loss term. By considering physical mechanisms such as gradient monotonicity and rainfall constraints when predicting landslide probability, the generalization capability of the model for newly constructed roads or geologically complex areas can be significantly improved. Next, a population weighting mechanism is introduced and combined with the Floyd-Warshall algorithm and the original road network topology map to calculate the first network efficiency. For each link in the original road network topology map, link closures are simulated one by one, the road network topology map is updated, and the second network efficiency is calculated. The first and second network efficiencys are then combined to calculate the network efficiency loss value after link closure. Compared to simple path closure calculations, the introduction of the population weighting mechanism better addresses actual disaster relief needs. By comparing the changes in network efficiency before and after link closures, critical vulnerabilities in the road network can be quickly identified, and cascading effects can be predicted. This allows for accurate quantification of the impact of disasters on the reachability of medical relief. Finally, based on the nonlinear risk coupling calculation method, road landslide susceptibility level map, and road impact level map, a comprehensive road landslide risk heatmap is formed. In this way, the present invention integrates a physical information neural network and complex network theory through nonlinear coupling design, realizing the linkage between the geological model and road network functions, and outputting a comprehensive road landslide risk heatmap. This enables accurate, rapid, and intuitive systematic evaluation of road landslide risk, and further realizes the scientific classification and precise management of regional road risks. [Brief explanation of the drawing]

[0017] To more clearly explain the embodiments of this application or the prior art, the drawings necessary for describing the embodiments are briefly described below. However, the drawings described below are merely examples of embodiments of this application, and those skilled in the art can obtain other drawings based on these without requiring any creative effort. [Figure 1] Figure 1 is a diagram illustrating the application environment of a road network landslide risk assessment method in one embodiment of the present invention. [Figure 2] Figure 2 is a schematic flowchart of a method for evaluating the landslide risk of a road network according to one embodiment of the present invention. [Figure 3] Figure 3 is a schematic flowchart of the training process of a road landslide prediction model according to one embodiment of the present invention. [Figure 4] Figure 4 is a schematic diagram of the training and validation loss curves of a road landslide prediction model according to one embodiment of the present invention. [Figure 5] Figure 5 is a schematic diagram of the ROC curve result of a road landslide prediction model according to one embodiment of the present invention. [Figure 6] Figure 6 is a road landslide susceptibility level map according to one embodiment of the present invention. [Figure 7] Figure 7 shows the results of the percentage reduction in network efficiency corresponding to the top 10 links with the highest network efficiency loss values ​​when a single link is interrupted in the original road network topology map according to one embodiment of the present invention. [Figure 8] Figure 8 shows a road impact level map according to one embodiment of the present invention. [Figure 9] Figure 9 is a comprehensive road landslide risk heat map according to one embodiment of the present invention. [Figure 10] Figure 10 is a schematic diagram of the structure of a computer device according to one embodiment of the present invention. [Modes for carrying out the invention]

[0018] The technical concept of the embodiments of this application will be described clearly and completely below with reference to the drawings of the embodiments of this application. Of course, the embodiments described are only a part of the embodiments of this application, not all of them. Any other embodiments that a person skilled in the art can obtain based on the embodiments of this application without requiring inventive work are all within the scope of protection of this application.

[0019] To make the above-mentioned objectives, features, and advantages of this application clearer and easier to understand, the application will be described in more detail below with reference to the drawings, using specific embodiments.

[0020] The applicant's research has revealed that prior art lacks a methodology that effectively combines the modeling of landslide susceptibility with the efficiency analysis of road networks. This makes it difficult to realize a linked model of the physical processes of disasters and the response of road functions in actual road risk assessments, and it has become clear that road efficiency analysis is separated from the driving forces of geological mechanisms.

[0021] The landslide risk assessment method for a road network according to the embodiment of the present invention can be applied to the application environment shown in Figure 1. Terminal 102 communicates with server 104 via the network. The data storage system can store data that server 104 needs to process. The data storage system may be installed independently, integrated with server 104, or located in the cloud or on another server.Terminal 102 can transmit the original road network topology map and grid image of the target area to be processed to Server 104, and Server 104 receives the original road network topology map and grid image of the target area to be processed. Regarding the original road network topology map and grid image of the target area to be processed, Server 104 inputs all grid images into a trained road landslide prediction model, obtains a predicted landslide occurrence probability value corresponding to each grid image, creates a road landslide susceptibility level map of the target area based on all grid images and the predicted landslide occurrence probability value corresponding to each grid image, calculates the first network efficiency of the original road network topology map based on the Floyd-Warshall algorithm, population weighting mechanism and the original road network topology map, and for each link in the original road network topology map, the link The system simulates a blockage scenario, updates the original road network topology map, obtains the updated road network topology map, calculates the second network efficiency of the updated road network topology map based on the Floyd-Warshall algorithm, population weighting mechanism, and the updated road network topology map, calculates the network efficiency loss value corresponding to the blockage of each link based on the first network efficiency of the original road network topology map and the second network efficiency of the updated road network topology map, creates a road impact level map for the target area based on the network efficiency loss values ​​corresponding to the blockage of all links in the original road network topology map and the original road network topology map, and determines the overall road landslide risk heatmap for the target area based on the nonlinear risk coupling calculation method, road landslide susceptibility level map, and road impact level map. Server 104 can feed back the obtained overall road landslide risk heatmap to terminal 102.Furthermore, in some embodiments, the method for evaluating the landslide risk of a road network may be implemented by either the server 104 or the terminal 102 alone. For example, the terminal 102 may directly process the original road network topology map and grid image of the target area to be processed and obtain a comprehensive road landslide risk heatmap. Alternatively, the server 104 may obtain the original road network topology map and grid image of the target area to be processed from a data storage system, process the original road network topology map and grid image of the target area to be processed, and obtain a comprehensive road landslide risk heatmap.

[0022] Terminal 102 may be, but is not limited to, various desktop computers, laptop computers, smartphones, tablet computers, Internet of Things devices, and portable wearable devices. Internet of Things devices may include smart speakers, smart TVs, smart air conditioners, and in-car smart devices. Portable wearable devices may include smartwatches, smart bracelets, and head-mounted devices. Server 104 may be implemented as an independent server, a server cluster consisting of multiple servers, or a cloud server.

[0023] In one exemplary embodiment, as shown in Figures 2, 6, 8, and 9, a method for assessing the landslide risk of a road network is provided, which is performed by a computer device, which may be performed independently by a computer device such as a terminal or a server, or by both a terminal and a server, and in the embodiment of the present application, the case in which the method is applied to server 104 in Figure 1 is described as an example, and the method includes the following steps 201 to 208.

[0024] Step 201: Obtain the original road network topology map of the target area and each grid image within the target area, wherein the original road network topology map includes multiple nodes and links.

[0025] Step 202 All the grid images are input into the trained road landslide prediction model, and a predicted landslide probability value corresponding to each grid image is obtained, and during the training process of the road landslide prediction model, the corresponding total loss function includes a data-driven term and a physical constraint term, the physical constraint term includes a gradient monotonic physical constraint loss term, a rainfall factor monotonic physical constraint loss term, and an elevation smoothness and continuity physical constraint loss term.

[0026] Step 203 Based on all the grid images and the predicted landslide probability values ​​corresponding to each grid image, a road landslide susceptibility level map of the target area is created.

[0027] Step 204: Based on the Floyd-Warshall algorithm, population weighting mechanism, and the original road network topology map, the first network efficiency of the original road network topology map is calculated.

[0028] Step 205 For each link in the original road network topology map, a blockage scenario for the link is simulated, the original road network topology map is updated, the updated road network topology map is obtained, and the second network efficiency of the updated road network topology map is calculated based on the Floyd-Warshall algorithm, a population weighting mechanism, and the updated road network topology map.

[0029] Step 206 Based on the first network efficiency and the second network efficiency, calculate the network efficiency loss value corresponding to the disconnection of each link.

[0030] Step 207 A road impact level map for the target area is created based on the network efficiency loss values ​​corresponding to the blockage of all links in the original road network topology map, and the original road network topology map.

[0031] Step 208: Nonlinear risk coupling calculation method, based on the road landslide susceptibility level map and the road impact level map, determines a comprehensive road landslide risk heatmap for the target area, and the comprehensive road landslide risk heatmap is used to evaluate the landslide risk of the road network.

[0032] By performing steps 201 to 208 described above, the present invention provides a method for evaluating the landslide risk of a road network. First, a road landslide prediction model calculates the predicted probability of landslide occurrence. During the training process of the road landslide prediction model, the corresponding total loss function includes a data-driven term and a physical constraint term. The physical constraint term includes a gradient monotonic physical constraint loss term, a rainfall factor monotonic physical constraint loss term, and an elevation smoothness and continuity physical constraint loss term. Thus, when predicting the probability of landslide occurrence, the present invention can significantly improve the generalization capability of the model in newly constructed roads or geologically complex areas by considering physical constraint mechanisms such as gradient monotonicity and rainfall factor monotonicity. Next, a population weighting mechanism is introduced and combined with the Floyd-Warshall algorithm and the original road network topology map to calculate the first network efficiency. For each link in the original road network topology map, the blockage of the link is simulated step by step, the road network topology map is updated, and the second network efficiency is calculated. The first and second network efficiencies are then combined to calculate the network efficiency loss value after the link blockage. Compared to simple path blockage calculations, this invention, by introducing a population weighting mechanism, better addresses actual disaster relief needs. By comparing the changes in network efficiency before and after link blockage, critical vulnerabilities in the road network can be quickly identified, and cascading effects can be predicted. This allows for accurate quantification of the impact of disasters on the reachability of medical relief. Finally, based on the nonlinear risk coupling calculation method, road landslide susceptibility level map, and road impact level map, a comprehensive road landslide risk heatmap is formed. In this way, the present invention integrates physical constraint mechanisms and complex network theory through nonlinear coupling design, realizes the linkage between geological models and road network functions, and outputs a comprehensive road landslide risk heatmap. This enables accurate, rapid, and intuitive systematic evaluation of road landslide risk, and further realizes scientific classification and precise management of regional road risks.

[0033] In another exemplary embodiment of the present invention, in order to improve the stability and generalization ability of the road landslide prediction model in predicting the probability of landslide occurrence, a training multi-source factor image set and historical landslide data for a target area can be acquired, and the road landslide prediction model can be trained based on the training multi-source factor image set and historical landslide data for the target area, and the road landslide prediction model can be adjusted based on a total loss function consisting of a data-driven term and a physical constraint term until the performance of the road landslide prediction model no longer improves. As shown in Figure 3, the training process of the road landslide prediction model includes the following steps 301 to 306.

[0034] Step 301: A training multi-source factor image set and historical landslide data are acquired for the target region. The training multi-source factor image set includes multiple landslide factors and factor images corresponding to each landslide factor, and the historical landslide data includes historical landslide points and historical non-landslide points.

[0035] Step 302 The training multi-source factor image set is preprocessed to obtain a data grid group with unified spatial resolution, the data grid group including multiple grid images and a landslide factor corresponding to each grid image.

[0036] Step 303 Based on the historical landslide data, the actual probability of landslide occurrence corresponding to each grid image is determined.

[0037] Step 304: Construct a road landslide prediction model using a model with a multilayer perceptron as the core network.

[0038] Step 305: For each grid image, input the grid image into the road landslide prediction model and obtain a predicted landslide probability value corresponding to the grid image.

[0039] Step 306 Based on the predicted landslide probability, actual landslide probability, landslide factor, and total loss function corresponding to the grid image, the total loss value is calculated, and the road landslide prediction model is adjusted based on the total loss value to obtain a trained road landslide prediction model, the total loss function of the road landslide prediction model consisting of a data-driven term and a physical constraint term.

[0040] In another exemplary embodiment of the present invention, the total loss function in step 306 is as follows: JPEG0007911722000002.jpg7170JPEG0007911722000003.jpg42170JPEG0007911722000004.jpg13170JPEG0007911722000005.jpg8170In the formula, N is the total number of samples, and y i is the label of the i-th sample, and p i This represents the predicted landslide probability output from the model. The sample labels are positive and negative; positive samples indicate landslide locations where landslides have occurred in the past, and negative samples indicate non-landslide locations where landslides have not occurred in the past (usually obtained through random sampling).

[0041] The formula for calculating JPEG0007911722000006.jpg7170 is as follows: JPEG0007911722000007.jpg10170, x i indicates the position of the i-th grid image, s i This shows the gradient factor corresponding to the i-th grid image, and p(x i ) indicates the predicted landslide probability value corresponding to the i-th grid image. JPEG0007911722000008.jpg14170 shows the partial derivative with respect to the gradient factor of the landslide probability prediction value corresponding to the i-th grid image. JPEG0007911722000009.jpg8170JPEG0007911722000010.jpg11170, x i indicates the position of the i-th grid image, ri represents the rainfall factor corresponding to the i-th grid image, and p(x i ) represents the landslide probability prediction value corresponding to the i-th grid image. JPEG0007911722000011.jpg14170 represents the partial differential coefficient of the rainfall factor of the landslide probability prediction value corresponding to the i-th grid image. The calculation formula of JPEG0007911722000012.jpg7170 is as follows. In the formula of JPEG0007911722000013.jpg9170, x i represents the position of the i-th grid image, and p(x i ) represents the landslide probability prediction value corresponding to the i-th grid image. JPEG0007911722000014.jpg8170 represents the landslide probability prediction value under the action of the Laplace operator. In a two-dimensional space, the Laplace operator is defined as follows. In the formula of JPEG0007911722000015.jpg11170, (x, y) represents the coordinates corresponding to the position x of the i-th grid image i . JPEG0007911722000016.jpg8170 represents the spatial smoothness index of the probability field. It is used to constrain or evaluate the landslide probability prediction value, so that the probability distribution becomes more spatially continuous and avoids unreasonable sudden changes.

[0042] As an example, in order to further show the training process of the road landslide prediction model according to the present application, the above steps 301 to 3 step 06 are replaced by the following steps 401 to 408. Step 401: Obtain the historical landslide data of the target area, the 12 types of landslide factors in the target area, and the factor images corresponding to each landslide factor. The historical landslide data includes historical landslide points and historical non-landslide points. The 12 types of landslide factors and their data sources are shown in Table 1. <>

[0043] JPEG0007911722000017.jpg114170

[0044] Step 402: Spatial alignment, gridning, and normalization processes are performed on the factor images of the 12 types of landslide factors to construct a data grid group with unified spatial resolution. The data grid group includes the 12 types of landslide factors and the grid images corresponding to each landslide factor.

[0045] Step 403: Based on the historical landslide data, true landslide labels are set for grid images corresponding to 12 types of landslide factors, with historical landslide points as positive sample labels and historical non-landslide points as negative sample labels. Spatial datasets of landslides (positive samples) and non-landslides (negative samples) are constructed. The spatial datasets contain positive sample grid images corresponding to multiple historical landslide points and negative sample grid images corresponding to multiple historical non-landslide points. All positive and negative sample grid images are normalized to a value between 0 and 1. After processing, spatial datasets of landslides (positive samples) and non-landslides (negative samples) are obtained, and these processed spatial datasets are divided into a training set and a validation set.

[0046] Step 404 Construct a road landslide prediction model, which is a physical information neural network model. The physical information neural network model uses a multilayer perceptron (MLP) as the backbone network and includes one input layer, three hidden layers (with 256, 128, and 64 nodes respectively), and one sigmoid output layer. The activation function used is LeakyReLU to mitigate gradient vanishing. The network structure of the physical information neural network model supports an automatic differentiation mechanism, which allows for easy integration of spatial fine-grain classification constraints. The input to the road landslide prediction model is a grid image of the target area, and the output value shows the predicted probability of landslide occurrence (range [0,1]) corresponding to the current grid image.

[0047] Step 405: For each grid image in the training set, input the grid image into the road landslide prediction model and obtain the predicted landslide probability value corresponding to the grid image. Step 406: A loss value is calculated based on the predicted landslide probability value corresponding to the grid image, the true landslide label, and the total loss function, and the road landslide prediction model is adjusted based on the loss value, the total loss function of the road landslide prediction model consisting of a data-driven term and a physical constraint term.

[0048] Step 407: Construct training and validation loss curves and analyze the stability and generalization ability of the road landslide prediction model in predicting the probability of landslide occurrence.

[0049] As an example, as shown in Figure 4, to verify the stability and generalization ability of the proposed road landslide prediction model in predicting the probability of landslide occurrence, training and validation loss curves for the road landslide prediction model are constructed. The training and validation loss curves show that the loss of the road landslide prediction model decreases rapidly in the early stages of training, and in the subsequent training process, both the training loss and validation loss continuously converge. If the validation loss curve does not show a stable and clear upward trend, it is explained that the road landslide prediction model is not overfitting. In Figure 4, the gray dashed line indicates the early stopping point, which is the epoch at which the loss of the validation set is optimal. By using the Early Stopping strategy, the robustness and training efficiency of the road landslide prediction model are improved. In Figure 4, the horizontal axis (Epoch) is the number of training epochs, totaling 200 epochs. The vertical axis (Loss) is the value of the loss function, showing the difference between the model's prediction result and the true label. The black line (Total Train Loss) is the total loss in the training dataset (including the physical constraint term and the cross-entropy term), reflecting the degree to which the model fits the training data. Discontinuity line (Validation Loss): This is the loss on the validation dataset, used only to evaluate the model's generalization ability, and does not participate in weight updates. Vertical discontinuity line (Early Stopping): This is an early stopping point, indicating that the model's performance on the validation set has stopped improving, and training is terminated early to prevent overfitting.

[0050] Step 408: Compare the true landslide label with the predicted landslide probability, create an ROC curve, and calculate the Area Under the Curve (AUC) index. As shown in Figure 5, the horizontal axis of the ROC curve represents the False Positive Rate, and the vertical axis represents the True Positive Rate. By sequentially changing the threshold and creating classification performance of the road landslide prediction model at different classification thresholds, the overall discriminative ability of the road landslide prediction model under various judgment criteria can be comprehensively reflected. In Figure 5, the ROC curve of the road landslide prediction model proposed in this application shows a clear upward convex shape, and the AUC value is 0.9190, which indicates that the road landslide prediction model has excellent accuracy and robustness in predicting the probability of road landslides.

[0051] In another exemplary embodiment of the present invention, step 203 above may be replaced by steps 2031 to 2033 below in order to obtain a road landslide susceptibility level map. Step 2031 Using the natural breakpoint method, segmentation is performed on the predicted landslide probability values ​​corresponding to each grid image obtained in Step 202 above, breakpoint sections are obtained, and the predicted road landslide probability values ​​are divided into five levels: "extremely low," "low," "moderate," "high," and "extremely high."

[0052] Step 2032: Assign the corresponding level labels to each grid image based on the breakpoint interval.

[0053] Step 2033: Based on the level labels corresponding to all grid images and all grid images, a road landslide susceptibility level map is created, and the created road landslide susceptibility level map is shown in Figure 6.

[0054] Another exemplary embodiment of the present invention provides a process for determining an original road network topology map of a target area, including steps 501 to 503. Step 501: Obtain road network vector data for the target area, perform coordinate system unification, cleaning and preprocessing on the road network vector data, correct topology errors, classify and encode according to the road level, and obtain the preprocessed road network vector data.

[0055] Step 502 Based on the preprocessed road network vector data, the roads are segmented according to intersections in the road network of the target area, and the start and end nodes and attribute information of each road segment are obtained.

[0056] Step 503 Based on the preprocessed road network vector data and the start and end nodes and attribute information of each road segment, an original road network topology map is constructed, the original road network topology map includes multiple nodes and links, the nodes representing intersections in the road network of the target area, and the links indicating road segments between two adjacent nodes.

[0057] In another exemplary embodiment of the present invention, in order to accurately calculate the first network efficiency of an original road network topology map, two nodes are selected from the original road network topology map, the shortest distance between these two nodes is calculated, a first shortest distance matrix is ​​obtained until the shortest distance between any two nodes in the original road network topology map is calculated, and population data corresponding to all nodes is obtained, and the first network efficiency of the original road network topology map is calculated using a population weighting mechanism based on the shortest distance matrix and population data, in which case step 204 above is replaced by steps 2041 to 2042 below. Step 2041 The node represents an intersection in the road network of the target area, and the first operation is performed on all nodes in the original road network topology map to obtain the first shortest distance matrix between all nodes in the original road network topology map. The first operation is, Two nodes v in the aforementioned original road network topology map i and node v j Select as you like, and node v i and node v j n intermediate nodes v that exist between k The task is to determine (k∈{1,2,…,n}), Based on the Floyd-Warshall algorithm, the node v i and the node v j The shortest distance d between them ij The calculation involves and includes the following formula: JPEG0007911722000018.jpg10170In formula, JPEG0007911722000019.jpg10170 is node v under the assumption that up to k intermediate nodes are allowed. i from node v j This shows the shortest distance to node v, and when k=n, i and node v j The shortest distance d between them ij The shortest distance d is obtained, ij This is shown as follows: JPEG0007911722000020.jpg24170, l ij is node v i and node v j This shows the shortest road distance between the two points, v i =v j In this case, it indicates the distance from the node to itself, and node v i and node v j If the distance between and is infinite, then node v i and node v j This indicates that it is unreachable.

[0058] Step 2042 Calculating the first network efficiency of the original road network topology map based on the population weighting mechanism and the first shortest distance matrix, specifically, Population data for each node in the aforementioned original road network topology map P={P1,…,P i ,…,P j To obtain , ...} and Based on the population data of all nodes and the first shortest distance matrix, the population weighting mechanism is used to determine the first network efficiency E of the original road network topology map. w The calculation involves and includes the following formula: JPEG0007911722000021.jpg11170, d ij either two nodes v i and v j It shows the shortest distance between and w ij is node v i and node v j The weight of the population between P i is node v i The corresponding population data is shown, P j is node v j The corresponding population data is shown.

[0059] In another exemplary embodiment of the present invention, in order to accurately calculate the second network efficiency in a single-link failure scene of the original road network topology map and to calculate the network efficiency loss value after a single-link failure in the original road network topology map based on the change in network efficiency before and after a single-link failure, a single-link failure scene can be simulated for each link in the original road network topology map, the original road network topology map can be updated, the updated road network topology map can be obtained, and the second network efficiency of the updated road network topology map can be calculated based on the Floyd-Warshall algorithm, a population weighting mechanism and the updated road network topology map. In this case, steps 205 to 207 above can be replaced by steps 2051 to 2055 below. Step 2051: For each link in the original road network topology map, remove the link from the original road network topology map and obtain the updated road network topology map. The updated road network topology map is shown below. In the formula JPEG0007911722000022.jpg8170, G′ represents the updated road network topology map, V represents the node set of the updated road network topology map, E represents the link set, and e k -e indicates any one link in the link set. k is any one link e in the link set k Figure 7 shows the network efficiency degradation percentage results corresponding to the top 10 links with the highest network efficiency loss rates, indicating that the network was blocked.

[0060] Step 2052: Process all nodes in the updated road network topology map and obtain a second shortest distance matrix between all nodes in the updated road network topology map.

[0061] As an example, in step 2052 above, a second operation is performed on all nodes in the updated road network topology map to obtain a second shortest distance matrix between all nodes in the updated road network topology map. The second operation is similar to the first operation, the difference being that the second operation calculates the second shortest distance matrix based on the updated road network topology map.

[0062] Step 2053 Obtain population data for each node in the updated road network topology map, and, based on the population data for all nodes and the second shortest distance matrix, calculate the second network efficiency of the updated road network topology map using a population weighting mechanism. Calculate JPEG0007911722000023.jpg7170.

[0063] Step 2054 The first network efficiency E of the original road network topology map w and the second network efficiency of the updated road network topology map Based on JPEG0007911722000024.jpg7170, the network efficiency loss value corresponding to the interruption of the link is calculated. The calculation formula is as follows: JPEG0007911722000025.jpg11170In formula, C i This is the network efficiency loss value corresponding to the disruption of the aforementioned link, and indicates the relative performance loss of the road network when any one link in the original road network topology map is destroyed by a landslide event, E w This shows the first network efficiency of the original road network topology map. JPEG0007911722000026.jpg7170 shows the second network efficiency of the updated road network topology map when any one link in the original road network topology map is destroyed by a landslide event. C iA larger value indicates that the corresponding blocked link in the original road network topology map has a more severe impact on the overall road network, and therefore is more important.

[0064] Step 2055: Network efficiency loss values ​​corresponding to the blockage of all links in the original road network topology map, and a road impact level map for the target area are created based on the original road network topology map, and the created road impact level map is shown in Figure 8.

[0065] In another exemplary embodiment of the present application, the network efficiency loss value C corresponds to the blockage of all links in the original road network topology map obtained in steps 2051 to 2054. i By sorting and categorizing the data, a road network vulnerability map (not shown) can be generated. This allows for the identification of critical links and vulnerable points in the original road network topology map, thereby providing decision-making support for future disaster prevention and improvement of road disaster resilience.

[0066] In another exemplary embodiment of the present invention, in order to accurately, quickly, and intuitively systematically evaluate road landslide risk, physical constraint mechanisms and complex network theory are integrated by nonlinear coupling design, linking the geological model and road network function, and a comprehensive road landslide risk heatmap can be output. In this case, step 208 above can be replaced by steps 2081 to 2085 below. Step 2081: Construct a 2D risk mapping matrix based on nonlinear risk coupling calculations, using landslide susceptibility levels as column vectors and road impact levels as row vectors.

[0067] Step 2082 Using the Delphi method, risk levels corresponding to each combination of levels in the two-dimensional risk mapping matrix are set, and a predetermined risk matrix is ​​obtained. Each cell in the predetermined risk matrix represents the risk level in that combination scene, and the risk levels are divided into five stages: "extremely low," "low," "moderate," "high," and "extremely high," and the classification of risk levels satisfies the following basic principles. High sensitivity level × High impact level → High risk If any one of the variables is at the lowest level, the risk will not exceed a moderate level. In the case of the same level combination, the priority of sensitivity level is slightly higher than that of impact level.

[0068] The risk level mapping relationships are shown in Table 2.

[0069] JPEG0007911722000027.jpg59170

[0070] Step 2083 Grid matching is performed on the road landslide susceptibility level map and the road impact level map to ensure that the road landslide susceptibility level map and the road impact level map have the same spatial resolution, projection method and geographical range, and the values ​​of each pixel position in the road landslide susceptibility level map and the road impact level map are used as input indices (i,j) of the two-dimensional matrix.

[0071] Step 2084 Based on the predetermined risk matrix, the risk level corresponding to the input index (i,j) of the two-dimensional matrix is ​​obtained by table lookup, and the comprehensive road landslide risk map is determined.

[0072] Step 2085 The comprehensive road landslide risk map is imported into the GIS platform, and the comprehensive road landslide risk heatmap output from the GIS platform is acquired. The GIS platform processes the comprehensive road landslide risk map, sets a classification style and legend based on the risk level (from "very low" to "very high"), and uses it to form the comprehensive road landslide risk heatmap. The comprehensive road landslide risk heatmap can be expanded to any region, and by updating the landslide factors and the grid images corresponding to the landslide factors, the road network risk can be rapidly evaluated. The comprehensive road landslide risk heatmap is shown in Figure 9.

[0073] This invention further provides application scenarios for the above-described method for evaluating the landslide risk of road networks. Specifically, the method for evaluating the landslide risk of road networks according to this embodiment can be applied to landslide risk scenarios of road networks, enabling quantitative evaluation and area-specific management of the impact of landslide disasters on road networks.

[0074] In an exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be shown in Figure 10. The computer device comprises a processor, memory, input / output interfaces (I / O, Input / Output), and a communication interface. The processor, memory, and input / output interfaces are connected via a system bus, and the communication interface is connected to the system bus via the input / output interfaces. The processor of the computer device is used to provide computation and control functions. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an execution environment for the operating system and computer programs in the non-volatile storage medium. The database of the computer device is used to store video label processing data. The input / output interfaces of the computer device are used to exchange information between the processor and external devices. The communication interfaces of the computer device are used to communicate with external terminals via a network connection. When the computer program is executed by the processor, a method for assessing landslide risk of a road network is realized.

[0075] As those skilled in the art will understand, the structure shown in Figure 10 is merely a block diagram showing a part of the structure related to the solution of the present invention, and does not limit the computer device to which the solution of the present invention is applied. A specific computer device may include more or fewer components than shown, combine several components, or have a different arrangement of components. In one exemplary embodiment, a computer device is provided that includes memory and a processor, wherein a computer program is stored in the memory, and the processor executes the computer program to realize the steps in each of the above-described method embodiments.

[0076] In one exemplary embodiment, a computer-readable storage medium is provided in which a computer program is stored, and which, when the computer program is executed by a processor, realizes the steps in each of the above-described method embodiments.

[0077] In one exemplary embodiment, a computer program product is provided which includes a computer program, and when the computer program is executed by a processor, the computer program product realizes the steps in each of the above-described method embodiments.

[0078] It should be explained that all user information (including, but not limited to, user device information and user personal information) and data (including, but not limited to, analytical data, stored data, and displayed data) related to this application are information and data for which the user's consent or the full consent of each party concerned has been obtained, and the collection, use, and processing of the related data must comply with the relevant regulations.

[0079] As those skilled in the art will understand, all or part of the processes in the methods of the above embodiments may be implemented by instructing the relevant hardware with a computer program, which may be stored in a non-volatile computer-readable storage medium, and which may include the processes in each embodiment of the above methods at runtime. Any reference to memory, database or other medium used in each embodiment of the present application may include at least one of non-volatile memory and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM®), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. Rather than being a limitation, RAM may take various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0080] The databases relating to each embodiment described in this application may include at least one of relational databases and non-relational databases. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors relating to each embodiment described in this application may include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processing units, programmable logic devices, quantum computing-based data processing logic devices, and the like.

[0081] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments have been described. However, any combination of these technical features should be considered within the scope described herein, as long as they do not contradict each other.

[0082] In this specification, the principles and embodiments of the present application have been described using specific examples. The above description of examples is intended to help understand the method and core idea of ​​the present application, and at the same time, those skilled in the art can modify the specific embodiments and scope of application based on the idea of ​​the present application. In short, the contents of this specification should not be construed as limiting the present application. [Explanation of symbols]

[0083] 102 terminals 104 Storage Server

Claims

1. A computer device comprising: memory; a processor; and a computer program stored in the memory and executable by the processor, A method for assessing the landslide risk of road networks, The computer device acquires the original road network topology map of the target area and each grid image within the target area, and the original road network topology map includes multiple nodes and links. The computer device inputs all the grid images into a trained road landslide prediction model, obtains a predicted landslide probability value corresponding to each grid image, and in the training process of the road landslide prediction model, the corresponding total loss function includes a data-driven term and a physical constraint term, the physical constraint term includes a gradient monotonic physical constraint loss term, a rainfall factor monotonic physical constraint loss term, and an elevation smoothness and continuity physical constraint loss term. The computer device creates a road landslide susceptibility level map of the target area based on all the grid images and the predicted landslide probability values ​​corresponding to each grid image. The computer device calculates the first network efficiency of the original road network topology map based on the Floyd-Warshall algorithm, a population weighting mechanism, and the original road network topology map. The computer device simulates a link blockage scenario for each link in the original road network topology map, updates the original road network topology map, obtains the updated road network topology map, and calculates the second network efficiency of the updated road network topology map based on the Floyd-Warshall algorithm, population weighting mechanism, and the updated road network topology map. The computer device calculates a network efficiency loss value corresponding to the disconnection of each link based on the first network efficiency and the second network efficiency. The computer device creates a road impact level map for the target area based on the network efficiency loss values ​​corresponding to the disconnection of all links in the original road network topology map, and the original road network topology map. A method for evaluating the landslide risk of a road network, characterized in that the computer device determines a comprehensive road landslide risk heatmap of the target area based on a nonlinear risk coupling calculation method, a road landslide susceptibility level map, and a road impact level map, and the comprehensive road landslide risk heatmap is used to evaluate the landslide risk of the road network.

2. The training process for the aforementioned road landslide prediction model is as follows: The computer device acquires a training multi-source factor image set and historical landslide data for the target region, the training multi-source factor image set includes multiple landslide factors and factor images corresponding to each landslide factor, and the historical landslide data includes historical landslide points and historical non-landslide points. The computer device preprocesses the training multi-source factor image set to obtain a data grid group with unified spatial resolution, and the data grid group includes multiple grid images and landslide factors corresponding to each grid image. The computer device determines the actual probability of landslide occurrence corresponding to each grid image based on the historical landslide data, The computer device constructs the road landslide prediction model using a model with a multilayer perceptron as the core network, The computer device inputs each grid image into the road landslide prediction model and obtains a predicted landslide probability value corresponding to the grid image. A method for evaluating the landslide risk of a road network according to claim 1, characterized in that the computer device calculates a total loss value based on the predicted landslide probability value, the actual landslide probability value, the landslide factor, and the total loss function corresponding to the grid image, adjusts the road landslide prediction model based on the total loss value, obtains the trained road landslide prediction model, and the total loss function of the road landslide prediction model consists of a data-driven term and a physical constraint term.

3. The total loss function is, λ １ λ represents the gradient weighting coefficient. ２ λ represents the rainfall weighting coefficient. ３ This indicates the elevation weighting coefficient. In the formula, x ｉ indicates the position of the i-th grid image, s ｉ This shows the gradient factor corresponding to the i-th grid image, and p(x ｉ ) shows the predicted landslide probability value corresponding to the i-th grid image. This shows the partial derivative with respect to the gradient factor of the predicted landslide probability value corresponding to the i-th grid image. where \(x\) ｉ indicates the position of the \(i\)-th grid image, \(r\) ｉ indicates the rainfall factor corresponding to the \(i\)-th grid image, \(p(x\) ｉ ) indicates the predicted landslide probability value corresponding to the \(i\)-th grid image. This shows the partial derivative of the landslide probability prediction value corresponding to the i-th grid image with respect to the rainfall factor. In the formula, x ｉ indicates the position of the i-th grid image, and p(x ｉ ) shows the predicted landslide probability value corresponding to the i-th grid image. This shows the predicted landslide probability under the action of the Laplace operator. The method for evaluating the landslide risk of a road network as described in feature 1.

4. The node indicates an intersection in the road network of the target area, and the computer device calculates a first network efficiency of the original road network topology map based on the Floyd-Warshall algorithm, a population weighting mechanism, and the original road network topology map. Specifically, the above step involves: The computer device performs a first operation on all nodes in the original road network topology map to obtain a first shortest distance matrix between all nodes in the original road network topology map. The computer device includes calculating a first network efficiency of the original road network topology map based on a population weighting mechanism and the first shortest distance matrix, The first operation is, The computer device then processes two nodes v in the original road network topology map. ｉ and node v ｊ Select as you like, node v ｉ and node v ｊ n intermediate nodes v that exist between them ｋ The task is to determine (k ∈ {1, 2, ..., n}), The computer device, based on the Floyd-Warshall algorithm, processes the node v ｉ and the aforementioned node v ｊ The shortest distance d between them ｉｊ The calculation includes, and the formula is, During the ceremony, This is node v under the assumption that up to k intermediate nodes are allowed. ｉ From node v ｊ This indicates the shortest distance to node v, and when k=n, the computer device determines this. ｉ and node v ｊ The shortest distance d between them ｉｊ The shortest distance d is obtained, ｉｊ teeth, In the formula, l ｉｊ is node v ｉ and node v ｊ It shows the shortest road distance between and v ｉ = v ｊ In this case, it indicates the distance from the node to itself, and node v ｉ and node v ｊ If the distance between and is ∞, then node v ｉ and node v ｊ A method for evaluating the landslide risk of a road network according to claim 1, characterized in that it indicates that the road is unreachable.

5. The step in which the computer device determines a comprehensive road landslide risk heat map of the target area based on the nonlinear risk coupling calculation method, the road landslide susceptibility level map, and the road impact level map, specifically, The computer device constructs a two-dimensional risk mapping matrix based on nonlinear risk coupling calculations, with landslide susceptibility levels as column vectors and road impact levels as row vectors. The computer device sets risk levels corresponding to each combination of levels in the two-dimensional risk mapping matrix using the Delphi method and obtains a predetermined risk matrix. The computer device performs grid matching on the road landslide susceptibility level map and the road impact level map, and uses the values ​​of each pixel position in the road landslide susceptibility level map and the road impact level map as input indices for a two-dimensional matrix. The computer device determines a comprehensive road landslide risk map based on the predetermined risk matrix and the input index of the two-dimensional matrix. The method for evaluating the landslide risk of a road network according to claim 1, characterized in that the computer device introduces the comprehensive road landslide risk map into a GIS platform, acquires the comprehensive road landslide risk heatmap output from the GIS platform, and the GIS platform is used to process the comprehensive road landslide risk map and form the comprehensive road landslide risk heatmap.

6. The computer device calculates the first network efficiency of the original road network topology map based on the population weighting mechanism and the first shortest distance matrix, specifically, The computer device generates population data P = {P} for each node in the original road network topology map. １ , ..., P ｉ , ..., P ｊ To obtain, ...} The computer device, based on the population data of all nodes and the first shortest distance matrix, uses a population weighting mechanism to determine the first network efficiency E of the original road network topology map. ｗ The calculation includes, and the formula is, In the formula, d ｉｊ any two nodes v ｉ and v ｊ It shows the shortest distance between them, w ｉｊ is node v ｉ and node v ｊ The weight of the population between P ｉ is node v ｉ The corresponding population data is shown, P ｊ is node v ｊ The method for evaluating the landslide risk of a road network according to claim 4, characterized in that it shows corresponding population data.

7. The computer device simulates a link blockage scene for each link in the original road network topology map, updates the original road network topology map, obtains the updated road network topology map, and calculates the second network efficiency of the updated road network topology map based on the Floyd-Warshall algorithm, population weighting mechanism, and the updated road network topology map, specifically, The computer device removes each link in the original road network topology map from the original road network topology map and obtains an updated road network topology map. The computer device performs processing on all nodes in the updated road network topology map and obtains a second shortest distance matrix between all nodes in the updated road network topology map. The method for evaluating the landslide risk of a road network according to claim 1, characterized in that the computer device acquires population data for each node in the updated road network topology map, and calculates the second network efficiency of the updated road network topology map using a population weighting mechanism based on the population data for all nodes in the updated road network topology map and the second shortest distance matrix.

8. The computer device, A computer device characterized in that the processor executes the computer program to realize the method for evaluating the landslide risk of a road network according to any one of claims 1 to 7.

9. A computer-readable storage medium in which the computer program is stored, A computer-readable storage medium characterized in that, when the computer program is executed by the processor, it realizes the method for evaluating the landslide risk of a road network as described in any one of claims 1 to 7.

10. A computer program product comprising the computer program, A computer program product characterized in that, when the computer program is executed by the processor, it realizes the method for evaluating the landslide risk of a road network as described in any one of claims 1 to 7.

Images