Prediction device

Abstract

This prediction device (100) comprises: a prediction data generation unit (110) that generates prediction topography data that indicates the topography of the area around each of a plurality of structures located in a target area on the basis of location information for each of the plurality of structures and elevation information that indicates the elevation of the target area; and a prediction unit (140) that performs disaster prediction for each of the plurality of structures located in the target area by inputting the generated prediction topography data into a prediction model built by learning of training data that, for each of a plurality of structures located in a disaster area, associates training topography data that indicates the topography of the area around the structure and whether the structure has been damaged by a disaster.

Description

Prediction Device

[0001] The present disclosure relates to a prediction device.

[0002] As a system for predicting damage caused by wind and floods to ground-based structures such as utility poles, a system (RAMP-T: Risk Assessment and Management system for Power lifeline for Typhoon) has been developed that calculates wind speed for typhoon wind damage, predicts damage to structures based on the calculated wind speed, and outputs the results (see Non-Patent Document 1). A system (RAMP-Er: Risk Assessment and Management system for Power lifeline for Earthquake) has also been developed that predicts damage caused by earthquakes, and the prediction results of these systems can be displayed together on a map (see Non-Patent Document 2). A system has also been developed that predicts the number of buildings damaged by typhoons, heavy rain, and earthquakes and the damage rate for each city, ward, town, and village, and displays the results on a map (see Non-Patent Document 3).

[0003] On the other hand, there is no system that can predict damage from rainfall-induced landslides at the structural level, such as utility poles. The Japan Meteorological Agency provides a risk map called "Dossa Kikikuru." Landslide warning information is issued based on the soil moisture content predicted using a tank model, which uses a perforated tank to model the flow of rain through the soil, and the accumulated rainfall over a 60-minute period. Here, the parameters of the tank model are standardized nationwide, and regional topographical information is not taken into account (see Non-Patent Document 4).

[0004] The National Institute for Land and Infrastructure Management has created a proposed probability map that calculates the probability of landslides based on topography and predisposing factors. This probability map calculates the probability of landslides occurring in three stages based on three types of maps: landslide warning areas, deep-seated collapse estimated frequency maps, and landslide topography distribution maps (see Non-Patent Document 5).

[0005] Sugimoto et al., "Introduction of a Typhoon Damage Estimation System and Weather Simulation," IEEJ Journal Vol. 138 No. 3 pp. 141-144, 2018. T. Tadokoro "Automatic Generation of Input Data for Distribution System Simulation Programs," 2018 IEEE Innovative Smart Grid Technologies - Asia (ISGT Asia), IEEE, 2018. Real-time damage prediction website cmap, [Retrieved November 29, 2023], Internet<URL: https: / / aioinissaydowa.co.jp / corporate / service / cmap / > Japan Meteorological Agency, Landslide Warning Information / Dossan Kikikuru (Heavy Rain Warning (Landslide) Risk Distribution), [Searched November 29, 2023], Internet<URL:https: / / www.jma.go.jp / jma / kishou / know / bosai / doshakeikai.html> Matsuda et al., "Considerations on Estimating the Risk of Landslides Nationwide Using Topographical and Geological Thematic Maps," National Institute for Land and Infrastructure Management Document No. 1120, 2020, [Retrieved November 29, 2023], Internet<URL: http: / / www.nilim.go.jp / lab / bcg / siryou / tnn / tnn1120.htm>

[0006] As mentioned above, research is underway on methods for predicting damage to structures caused by typhoons or earthquakes and the probability of landslides. However, there has been insufficient research on methods for predicting damage to multiple structures caused by rainfall-related disasters.

[0007] The purpose of the present disclosure, made in consideration of the above-mentioned problems, is to provide a prediction device that can predict damage to each of multiple structures due to a disaster caused by rainfall.

[0008] In order to solve the above problem, the prediction device disclosed herein is a prediction device that predicts damage to each of a plurality of structures located in a target area due to a disaster caused by rainfall, and includes a prediction data generation unit that generates prediction terrain data indicating the terrain of the surrounding area of ​​each of the plurality of structures based on location information of each of the plurality of structures and elevation information indicating the elevation of the target area, and a prediction unit that inputs the generated prediction terrain data into a prediction model constructed by learning learning data that corresponds, for each of a plurality of structures located in a disaster-stricken area where a disaster caused by rainfall has occurred, training terrain data indicating the terrain of the surrounding area of ​​the structure and whether or not the structure has been damaged by the disaster, and makes a damage prediction for each of the plurality of structures located in the target area.

[0009] According to the prediction device of the present disclosure, it is possible to predict damage to each of a plurality of structures due to a disaster caused by rainfall.

[0010] 1 is a diagram illustrating an example configuration of a prediction device according to an embodiment of the present disclosure. FIG. 1 is a diagram illustrating an example configuration of position information illustrated in FIG. 1. FIG. 2 is a diagram illustrating an example configuration of elevation information illustrated in FIG. 1. FIG. 2 is a flowchart illustrating an example of a method for acquiring position information and elevation information. FIG. 3 is a diagram for explaining a method for acquiring position information and elevation information. FIG. 4 is a diagram for explaining linking of position information and elevation information by the calculation data generation unit illustrated in FIG. 1. FIG. 5 is a diagram illustrating an example of setting of a surrounding area by the calculation data generation unit illustrated in FIG. 1. FIG. 6 is a diagram illustrating another example of setting of a surrounding area by the calculation data generation unit illustrated in FIG. 1. FIG. 7 is a diagram illustrating yet another example of setting of a surrounding area by the calculation data generation unit illustrated in FIG. 1. FIG. 8 is a diagram illustrating yet another example of setting of a surrounding area by the calculation data generation unit illustrated in FIG. 1. FIG. 9 is a diagram illustrating yet another example of setting of a surrounding area by the calculation data generation unit illustrated in FIG. 1. FIG. 2 is a diagram for explaining generation of prediction terrain data by the prediction data generating unit shown in FIG. 1. FIG. 3 is a diagram for explaining generation of prediction terrain data by the prediction data generating unit shown in FIG. 1. FIG. 4 is a diagram for explaining generation of prediction terrain data by the prediction data generating unit shown in FIG. 1. FIG. 5 is a diagram showing an example of the relationship between the elevation of a structure and the damage rate of the structure. Maximum slope angle a ｍａｘ 1 is a diagram showing an example of the relationship between the minimum tilt angle a and the damage rate of a structure. ｍｉｎ Fig. 2 is a diagram showing an example of the relationship between the magnitude of a damage to a structure and the magnitude of a damage rate of a structure. Fig. 3 is a flowchart showing an example of the operation of the prediction device shown in Fig. 1. Fig. 4 is a diagram showing an example of the configuration of a computer that functions as the prediction device shown in Fig. 1.

[0011] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings.

[0012] 1 is a diagram illustrating an example configuration of a prediction device 100 according to an embodiment of the present disclosure. The prediction device 100 according to the present disclosure predicts damage to each of a plurality of structures located in a target area due to a disaster caused by rainfall. The structures to be predicted for damage include, for example, utility poles, steel towers, signal poles, or base stations installed on the ground, but are not limited to these, and may be any structures installed in a dispersed manner in the target area.

[0013] As shown in FIG. 1 , the prediction device 100 according to this embodiment includes a prediction data generation unit 110 , a learning data generation unit 120 , a model construction unit 130 , a prediction unit 140 , and an output unit 150 .

[0014] The prediction data generation unit 110 generates prediction terrain data used by the prediction unit 140, described below, to predict damage to each of multiple structures located in a target area due to a disaster caused by rainfall. The prediction terrain data is data indicating the terrain of the surrounding area of ​​each of multiple structures located in the target area. The prediction data generation unit 110 generates prediction terrain data for structures of the same type. As described above, structures may be various types of equipment such as utility poles, steel towers, signal poles, or base stations, and the prediction data generation unit 110 generates prediction terrain data for structures of the same type. Therefore, for example, if the structure for which damage prediction is performed is a utility pole, the prediction data generation unit 110 generates prediction terrain data for the utility pole located in the target area. The following description uses an example in which the structure is a utility pole.

[0015] 1 , location information and elevation information are input to the calculation data generation unit 110. Based on the input location information and elevation information, the calculation data generation unit 110 generates prediction topography data that indicates the topography of the surrounding areas of each of a plurality of structures located in the target area.

[0016] The location information input to the prediction data generating unit 110 is information indicating the location (for example, latitude and longitude) of each of a plurality of structures arranged in the target area, as shown in FIG. 2A.

[0017] The altitude information input to the calculation data generation unit 110 is information indicating the altitude of the target area. Specifically, the altitude information input to the calculation data generation unit 110 is information indicating the altitude of each of multiple points in the target area, as shown in Fig. 2B. The points whose altitudes are indicated by the altitude information do not need to coincide with the positions of the structures.

[0018] The altitude information may be obtained, for example, from publicly available map information (mesh information). The altitude information may also be an actual measurement value obtained by surveying, such as laser measurement. The altitude information may also be obtained by analyzing images captured by a camera and depth images captured by a ToF (Time of Flight) camera. A method for acquiring location information and altitude information using captured images and depth images will be described below.

[0019] FIG. 3A is a flowchart illustrating an example of a method for obtaining location information and altitude information.

[0020] First, a photographed image and a depth image are acquired by photographing a predetermined range including a structure and the ground surface around it using a photographing terminal whose photographing position and photographing direction are known, as shown in FIG. 3B (step S101).

[0021] Next, using the depth image, the distance l from the camera to the structure is calculated. １ Since the photographing position and photographing direction of the photographing terminal are known, the distance l from the photographing terminal to the structure is obtained (step S102). １ By obtaining this, the location of the structure can be identified (the location information of the structure can be obtained).

[0022] In addition, the ground surface (ground, cliff, etc.) is identified from the captured image, and the distance l from the capturing device to the identified ground surface is calculated using the depth image. ２ Since the photographing position and photographing direction of the photographing terminal are known, the distance l from the photographing terminal to the ground surface is obtained (step S103). ２ By obtaining this, the position of the earth's surface can be identified.

[0023] Next, the elevation difference ΔE between the structure and the ground surface is calculated from the captured image (step S104), thereby obtaining elevation information of the ground surface.

[0024] It is also possible to photograph a single structure from multiple directions and obtain position information and altitude information from the photographed images and depth images photographed from each direction.

[0025] Referring back to FIG. 1 , the calculation data generation unit 110 associates the input location information with the elevation information. Specifically, as shown in FIG. 4 , the calculation data generation unit 110 associates the position of each of a plurality of structures arranged in the target area with the elevation of the structure (the elevation E of the structure). The association between the position of each of a plurality of structures and the elevation of the structure may be performed using, for example, a GIS (Geographic Information System). The calculation data generation unit 110 may use, for example, the elevation of the center position of the structure as the elevation of the structure. Furthermore, for example, if the structure has a large area and spans multiple elevations, the calculation data generation unit 110 may use, for example, the average value of the multiple elevations as the elevation of the structure, or may determine a single representative point and use the elevation of the representative point as the elevation of the structure.

[0026] Next, the calculation data generation unit 110 sets a surrounding area, which is an area within a certain range from each of a plurality of structures arranged in the target area, and acquires the elevation of the set surrounding area. Figures 5A to 5G are diagrams showing examples of setting surrounding areas by the calculation data generation unit 110. In Figures 5A to 5G, the target area is assumed to be divided into multiple meshes of predetermined shapes such as rectangles, polygons, and circles (rectangles in Figures 5A to 5G).

[0027] For example, as shown in FIG. 5A , the calculation data generation unit 110 may set the mesh containing the center of the structure and adjacent meshes (meshes adjacent in the up, down, left, right, upper left, lower left, upper right, and lower right directions) as the surrounding area.

[0028] Furthermore, for example, as shown in FIG. 5B, when a structure spans multiple meshes, the calculation data generating unit 110 may set meshes adjacent to the mesh containing the structure as the surrounding area.

[0029] Furthermore, the calculation data generating unit 110 may set, as shown in FIG. 5C, for example, meshes that exist within a certain distance (for example, several meshes) from the center of the structure as the surrounding area.

[0030] Furthermore, for example, as shown in FIG. 5D , when a structure spans multiple meshes, the prediction data generation unit 110 may set the meshes that exist within a certain distance (e.g., several meshes) from the center of gravity of the structure as the surrounding area.

[0031] Furthermore, for example, when a structure spans multiple meshes, as shown in FIG. 5E, the prediction data generation unit 110 may set meshes that exist within a certain distance (e.g., several meshes) from any end of the structure as the surrounding area.

[0032] Furthermore, the calculation data generating unit 110 may set, as the surrounding area, meshes that exist within a range determined according to a certain policy from the center of the structure, as shown in FIG. 5F, for example.

[0033] Furthermore, for example, as shown in FIG. 5G, when a structure spans multiple meshes, the prediction data generation unit 110 may set the meshes that exist within a range determined in accordance with a certain policy from the structure as the surrounding area.

[0034] After acquiring the elevation of the surrounding area, the prediction data generating unit 110 calculates the maximum elevation (surrounding maximum elevation E ｍａｘ ) and minimum elevation (surrounding minimum elevation E ｍｉｎ The estimation data generating unit 110 associates the elevation E of the structure with the maximum elevation E in the surrounding area. ｍａｘ If E is higher than E ｍａｘ ), the elevation E of the structure is the maximum elevation E of the surrounding area ｍａｘ Furthermore, if the elevation of the structure is lower than the minimum elevation of the surrounding area (E<E ｍｉｎ ), the elevation E of the structure is the minimum elevation E ｍｉｎ may be set to

[0035] Next, the prediction data generating unit 110 calculates the elevation E of the structure, the maximum elevation E of the surrounding area, ｍａｘ , minimum surrounding elevation E ｍｉｎ , and the structure, the surrounding maximum elevation E ｍａｘ The point with the highest elevation and the surrounding minimum elevation E ｍｉｎ Based on the distance between points (minimum elevation points) having the above-mentioned features, predicted terrain data is generated that indicates the terrain of the area surrounding each of the multiple structures.

[0036] The prediction data generating unit 110 may use, for example, the surrounding maximum elevation E ｍａｘ and the minimum surrounding elevation E ｍｉｎ The calculation data generating unit 110 may calculate the difference between the maximum slope angle a , which is the slope angle between the structure and the maximum elevation point, as the calculation topographical data. ｍａｘ = (tan －１ (E ｍａｘ -E) / d ｍａｘ ), and the minimum slope angle a, which is the slope angle between the structure and the minimum elevation point ｍｉｎ = (tan －１ (E-E ｍｉｎ ) / d ｍｉｎ ) can be calculated. ｍａｘ is the distance between the structure and the highest point, and d ｍｉｎ is the distance between the structure and the point of minimum elevation. The prediction data generating unit 110 may also use, for example, the slope angles a between the structure and each of the n points in the surrounding area as the prediction topographical data. １ ~a ｎ The calculation data generating unit 110 may calculate, for example, the slope angle a between the structure and each of n points in the surrounding area as the prediction topographical data. １ ~a ｎ The mean or standard deviation of may be calculated.

[0037] The generation of prediction terrain data by the prediction data generator 110 will be described in more detail. Hereinafter, the elevation information is assumed to be information indicating the elevation of each rectangular mesh. In this case, for example, as shown in FIG. 7A , the coordinates of the vertices of each mesh (polygon) are associated with the elevation within the mesh (e.g., average value, representative value, etc.).

[0038] Based on the position information and elevation information, the calculation data generation unit 110 associates each of the plurality of structures with the elevation of the mesh nearest to the structure, as shown in FIG. 7B. Next, the calculation data generation unit 110 sets a surrounding area for each of the plurality of structures. In the following, as shown in FIG. 7C, the mesh M including the structure is ０ and the adjacent meshes M in the up / down, left / right, upper left, lower left, upper right and lower right directions. １ ~M ８ is set as the surrounding area. ０ Elevation) is E ０ and mesh M １ ~M ８ The altitude of １ ~E ８ Let us assume that:

[0039] As shown in FIG. 7D, the prediction data generating unit 110 calculates the position of the structure, the elevation of the structure, and the maximum elevation E ｍａｘ and the surrounding minimum elevation E ｍｉｎ (Mesh M １ ~M ８ Altitude E １ ~E ８ In the following, the elevation of the structure is determined as E ０ and the maximum surrounding elevation E ｍａｘ is E ４ and the minimum surrounding elevation E ｍｉｎ is E ６ Let us assume that:

[0040] The prediction data generating unit 110 may calculate, for example, the maximum surrounding elevation E ４ and the minimum surrounding elevation E ６ The difference between (= E ４ -E ６ ) as the prediction topographical data. ０ and the mesh M with the highest altitude ４ The maximum tilt angle a between ４ (=tan －１ (E ４ -E ０) / R) as prediction topographical data. ０ and the mesh M with the lowest elevation ６ The minimum tilt angle a between ６ (=tan －１ (E ０ -E ６ ) / R) is calculated as prediction topographical data, where R is the distance between the centers of adjacent meshes. ０ and Mesh M １ ~M ８ The inclination angle a between them １ ~a ８ The average value or standard deviation of the above is calculated as the topographical data for prediction.

[0041] In this way, the prediction data generator 110 generates prediction terrain data indicating the topography of the surrounding area of ​​each of the multiple structures (e.g., the slope of the area around the structure) based on the location information of each of the multiple structures located in the target area and the elevation information indicating the elevation of the target area. The susceptibility of a structure to damage from a rainfall-related disaster (e.g., a landslide) tends to depend on the topography of the surrounding area of ​​the structure (e.g., the slope of the area around the structure). Therefore, by predicting damage to a structure using prediction terrain data indicating the topography of the surrounding area of ​​the structure (e.g., the slope of the area around the structure), it is possible to improve the accuracy of the prediction. Note that in this embodiment, the prediction terrain data is an example indicating the slope of the area around the structure, but the prediction terrain data and the learning terrain data described later are not limited to this example, as long as they are data indicating topographical features that affect whether or not a structure will be damaged by a rainfall-related disaster.

[0042] 1 , the prediction data generating unit 110 outputs the generated prediction terrain data to the prediction unit 140. In addition to the prediction terrain data, the prediction data generating unit 110 may output variables such as the number of years since the installation of the structure to the prediction unit 140.

[0043] The training data generation unit 120 receives input of location information, elevation information, and damage information. The location information input to the training data generation unit 120 is information indicating the location (e.g., latitude and longitude) of each of multiple structures located in a disaster-stricken area where a rainfall-induced disaster occurred in the past. The elevation information input to the training data generation unit 120 is information indicating the elevation of the disaster-stricken area. Specifically, the elevation information input to the training data generation unit 120 is information indicating the elevation of each of multiple points in the disaster-stricken area. The damage information is information indicating whether each of multiple structures located in the disaster-stricken area suffered damage when a disaster occurred in the disaster-stricken area. The damage information is, for example, information indicating whether the structure suffered damage due to the disaster. The damage information may also be, for example, information indicating the degree of damage to the structure due to the disaster (e.g., breakage, tilt, no damage, etc.).

[0044] It should be noted that the structure whose position is indicated by the position information input to the learning data generation unit 120 and the structure whose position is indicated by the position information input to the prediction data generation unit 110 do not need to be the same structure, but they do need to be the same type of structure. Therefore, if the structure whose position is indicated by the position information input to the prediction data generation unit 110 is a utility pole, the structure whose position is indicated by the position information input to the learning data generation unit 120 also needs to be a utility pole.

[0045] Furthermore, the altitude information input to the learning data generation unit 120 and the altitude information input to the prediction data generation unit 110 need to have the same resolution. For example, if the altitude information input to the prediction data generation unit 110 is information indicating the altitude for each 10-meter square mesh, the altitude information input to the learning data generation unit 120 also needs to be information indicating the altitude for each 10-meter square mesh. However, as long as the resolutions are the same, the altitude information input to the learning data generation unit 120 and the altitude information input to the prediction data generation unit 110 may be obtained using different methods.

[0046] The learning data generation unit 120 generates learning terrain data indicating the terrain of the surrounding area of ​​each of the plurality of structures located in the disaster-stricken area based on the input location information and elevation information. That is, the learning data generation unit 120 generates learning terrain data indicating the terrain of the surrounding area of ​​each of the plurality of structures based on the location information of each of the plurality of structures located in the disaster-stricken area and elevation information indicating the elevation of the disaster-stricken area. The learning data generation unit 120 generates learning terrain data in the same manner as the prediction data generation unit 110. That is, the learning data generation unit 120 and the prediction data generation unit 110 use the same method for setting the surrounding area of ​​a structure. Furthermore, the learning data generation unit 120 and the prediction data generation unit 110 set the elevation E of the structure, the maximum surrounding elevation E, and the maximum surrounding elevation E. ｍａｘ and the surrounding minimum elevation E ｍｉｎ Topographical data calculated based on the above is the same type of data.

[0047] The training data generation unit 120 generates training data that associates the generated training terrain data with the presence or absence of damage to each of multiple structures located in a disaster-stricken area due to a disaster that occurred in the disaster-stricken area. Note that the past rainfall-related disasters used to generate the training data are not localized disasters, but disasters in which damage is expected to be relatively uniform over a certain area, such as disasters in which the 24-hour accumulated rainfall was within a predetermined range over a certain area or disasters that are judged to have a certain level of danger or higher (e.g., alert level 4) in the Japan Meteorological Agency's landslide warning system.

[0048] The learning data generation unit 120 outputs the generated learning data to the model construction unit 130 .

[0049] The model construction unit 130 constructs a prediction model that predicts damage to structures due to rainfall-induced disasters based on the predictive terrain data for the area surrounding the structures, by learning from the training data generated by the training data generation unit 120. The model construction unit 130 constructs the prediction model by machine learning, for example, using the training terrain data as an explanatory variable and the presence or absence of damage to each of multiple structures located in the disaster-stricken area as a target variable. As a machine learning method, a supervised learning method such as a gradient boosting decision tree or a random forest can be used.

[0050] The model construction unit 130 may construct a prediction model by, for example, a statistical method. As a statistical method, a prediction model is constructed by a method using a function (model formula) represented by a linear or sigmoid curve. For example, the model construction unit 130 constructs a prediction model by using the elevation E of a structure (electric pole) shown in FIG. 8A . ０ and the function f (E ０ ), the maximum tilt angle a ｍａｘ and the function f(a ｍａｘ ) and the minimum tilt angle a ｍｉｎ and the function f(a ｍｉｎ ) to build a predictive model.

[0051] 1 , the prediction unit 140 inputs the prediction topographical data generated by the prediction data generation unit 110 into the prediction model constructed by the model construction unit 130, and performs damage prediction for each of the multiple structures located in the target area. The prediction unit 140 outputs the prediction results to the output unit 150.

[0052] The output unit 150 outputs the prediction result of the prediction unit 140 by displaying it on a display device, etc. For example, the output unit 150 outputs the prediction result of the prediction unit 140 by superimposing it on a map of the target area.

[0053] Next, the operation of the prediction device 100 according to this embodiment will be described.

[0054] FIG. 9 is a flowchart showing an example of the operation of the prediction device 100 according to this embodiment, and is a diagram for explaining the prediction method executed by the prediction device 100 according to this embodiment.

[0055] The learning data generation unit 120 generates learning terrain data indicating the terrain of the area surrounding each of the plurality of structures based on the position information of each of the plurality of structures located in the disaster-stricken area and the elevation information indicating the elevation of the disaster-stricken area.The learning data generation unit 120 then generates learning data that associates the generated learning terrain data with the presence or absence of damage caused by the disaster to each of the plurality of structures located in the disaster-stricken area (step S11).

[0056] The model construction unit 130 constructs a prediction model by learning from the generated training data (step S12).

[0057] The prediction data generation unit 110 generates prediction terrain data indicating the terrain of the surrounding areas of each of the multiple structures based on the location information of each of the multiple structures placed in the target area and the elevation information indicating the elevation of the target area (step S13).

[0058] The prediction unit 140 inputs the generated prediction terrain data into the prediction model and performs damage prediction for each of the multiple structures located in the target area (step S14). Here, the prediction model is a model constructed by learning from training data in which, for each of the multiple structures located in the disaster-stricken area where a rainfall-induced disaster has occurred, training terrain data indicating the terrain of the area surrounding each of the multiple structures is associated with whether or not the structure has been damaged by the disaster.

[0059] As described above, the prediction device 100 according to this embodiment includes a prediction data generation unit 110 and a prediction unit 140. The prediction data generation unit 110 generates prediction terrain data indicating the terrain of the surrounding areas of each of the multiple structures located in the target area, based on position information for each of the multiple structures located in the target area and elevation information indicating the elevation of the target area. The prediction unit 140 inputs the generated prediction terrain data into a prediction model constructed by learning training data that associates, for each of the multiple structures located in the disaster-stricken area, training terrain data indicating the terrain of the surrounding areas of the structure with whether or not the structure has been damaged by the disaster, and performs damage prediction for each of the multiple structures located in the target area.

[0060] By using the prediction model constructed by learning the above-mentioned training data to make damage predictions, it is possible to predict damage to each of multiple structures due to disasters caused by rainfall.

[0061] In the present embodiment, the prediction device 100 has been described as having a configuration for constructing a prediction model (the learning data generation unit 120 and the model construction unit 130), but this is not limiting. The prediction device 100 may not have the learning data generation unit 120 and the model construction unit 130, and a device (model construction device) having a configuration for constructing a prediction model may be provided separately from the prediction device 100.

[0062] The prediction device 100 described above can be realized by a computer 10 shown in FIG. 10. A program for causing the computer 10 to function as the prediction device 100 may be provided. The program may be stored in a storage medium or provided via a network. FIG. 10 is a block diagram showing a schematic configuration of a computer 10 functioning as the prediction device 100. The computer 10 may be a general-purpose computer, a dedicated computer, a workstation, a PC (Personal Computer), an electronic notepad, or the like. The program instructions may be program code, code segments, or the like for executing necessary tasks.

[0063] 10, a computer 10 includes a processor 11, a ROM (Read Only Memory) 12, a RAM (Random Access Memory) 13, a storage 14, an input unit 15, a display unit 16, and a communication interface (I / F) 17. Each component is communicably connected to one another via a bus 19. The processor 11 is specifically a CPU (Central Processing Unit), an MPU (Micro Processing Unit), a GPU (Graphics Processing Unit), a DSP (Digital Signal Processor), a SoC (System on a Chip), or the like, and may be configured with multiple processors of the same or different types.

[0064] The processor 11 is a control unit that controls each component and performs various arithmetic operations. That is, the processor 11 reads a program from the ROM 12 or the storage 14 and executes the program using the RAM 13 as a work area. The processor 11 controls each component and performs various arithmetic operations in accordance with the program stored in the ROM 12 or the storage 14. In this embodiment, the ROM 12 or the storage 14 stores a program for causing the computer 10 to operate as the prediction device 100 according to the present disclosure. The program is read and executed by the processor 11 to realize each component of the prediction device 100, i.e., the prediction data generation unit 110, the training data generation unit 120, the model construction unit 130, the prediction unit 140, and the output unit 150.

[0065] The program may be provided in a form stored on a non-transitory storage medium such as a CD-ROM (Compact Disk Read Only Memory), a DVD-ROM (Digital Versatile Disk Read Only Memory), a USB (Universal Serial Bus) memory, etc. The program may also be provided in a form downloaded from an external device via a network.

[0066] The ROM 12 stores various programs and various data. The RAM 13 temporarily stores programs or data as a working area. The storage 14 is configured with an HDD (Hard Disk Drive) or an SSD (Solid State Drive) and stores various programs including the operating system and various data.

[0067] The input unit 15 includes a pointing device such as a mouse and a keyboard, and is used to input various types of information.

[0068] The display unit 16 is, for example, a liquid crystal display, and displays various types of information. The display unit 16 may employ a touch panel system and function as the input unit 15. The display unit 16 displays, for example, the prediction results of the prediction unit 140.

[0069] The communication interface 17 is an interface for communicating with other devices, for example, an interface for a LAN.

[0070] The following additional notes are provided regarding the above-described embodiments.

[0071] [Supplementary Item 1] A prediction device that predicts damage to each of a plurality of structures located in a target area due to a disaster caused by rainfall, comprising a control unit, wherein the control unit generates prediction terrain data indicating the terrain of the surrounding area of ​​each of the plurality of structures based on position information of each of the plurality of structures and elevation information indicating the elevation of the target area, and is configured to input the generated prediction terrain data into a prediction model constructed by learning learning data that associates, for each of a plurality of structures located in a disaster-stricken area where a disaster caused by rainfall has occurred, training terrain data indicating the terrain of the surrounding area of ​​the structure with whether or not the structure has been damaged by the disaster, thereby predicting damage to each of the plurality of structures located in the target area.

[0072] [Supplementary Item 2] In the prediction device described in Supplementary Item 1, the control unit generates the training terrain data based on position information of each of a plurality of structures located in the disaster-stricken area and elevation information indicating the elevation of the disaster-stricken area, generates the training data that associates the generated training terrain data with whether or not each of the plurality of structures located in the disaster-stricken area has been damaged by the disaster, and constructs the prediction model by learning the generated training data.

[0073] [Supplementary Item 3] In the prediction device described in Supplementary Item 1 or 2, the prediction terrain data is data indicating the slope of the surrounding area of ​​each of a plurality of structures located in the target area, and the learning terrain data is data indicating the slope of the surrounding area of ​​each of a plurality of structures located in the disaster-prone area.

[0074] [Supplementary Item 4] A prediction method executed by a prediction device that predicts damage to each of a plurality of structures located in a target area due to a disaster caused by rainfall, comprising: a step of generating prediction terrain data indicating the terrain of the surrounding area of ​​each of the plurality of structures based on position information of each of the plurality of structures and elevation information indicating the elevation of the target area; and a step of inputting the generated prediction terrain data into a prediction model constructed by learning learning data in which, for each of a plurality of structures located in a disaster-stricken area where a disaster caused by rainfall has occurred, learning terrain data indicating the terrain of the surrounding area of ​​the structure is associated with whether or not the structure has been damaged by the disaster, thereby predicting damage to each of the plurality of structures located in the target area.

[0075] [Supplementary Item 5] A non-transitory storage medium storing a program executable by a computer, the non-transitory storage medium storing the program causing the computer to operate as the prediction device according to any one of Supplementary Items 1 to 3.

[0076] Although the above-described embodiments have been described as typical examples, it will be apparent to those skilled in the art that many modifications and substitutions can be made within the spirit and scope of the present disclosure. Therefore, the present invention should not be interpreted as being limited by the above-described embodiments, and various modifications and alterations are possible without departing from the scope of the claims. For example, multiple building blocks shown in the block diagrams of the embodiments can be combined into one, or one building block can be divided.

[0077] REFERENCE SIGNS LIST 10 Computer 11 Processor 12 ROM 13 RAM 14 Storage 15 Input unit 16 Display unit 17 Communication I / F 19 Bus 100 Prediction device 110 Prediction data generation unit 120 Learning data generation unit 130 Model construction unit 140 Prediction unit 150 Output unit

Claims

1. A prediction device for predicting the damage of each of a plurality of structures arranged in a target area due to disasters caused by rainfall, comprising: a prediction data generation unit that generates prediction terrain data indicating the terrain of the peripheral area of each of the plurality of structures based on the position information of each of the plurality of structures and the elevation information indicating the elevation of the target area; and a prediction unit that inputs the generated prediction terrain data into a prediction model constructed by learning learning data in which learning terrain data indicating the terrain of the peripheral area of each of the plurality of structures arranged in a disaster occurrence area where a disaster caused by rainfall has occurred is associated with the presence or absence of damage to the structure due to the disaster, and predicts the disaster damage of each of the plurality of structures arranged in the target area.

2. The prediction device according to claim 1, further comprising: a learning data generation unit that generates the learning data by generating the learning terrain data based on the position information of each of the plurality of structures arranged in the disaster occurrence area and the elevation information indicating the elevation of the disaster occurrence area, and associating the generated learning terrain data with the presence or absence of damage to each of the plurality of structures arranged in the disaster occurrence area due to the disaster; and a model construction unit that constructs the prediction model by learning the generated learning data.

3. The prediction device according to claim 1, wherein the prediction terrain data is data indicating the slope of the peripheral area of each of the plurality of structures arranged in the target area, and the learning terrain data is data indicating the slope of the peripheral area of each of the plurality of structures arranged in the disaster occurrence area.

Images