Flooding situation estimation device, flooding situation estimation system, flooding situation estimation method and program

Abstract

To provide a flood state estimation device, a flood state estimation method, a system, and a program that can estimate in real time, from a posted image, a flood state of an area including a photographed place.SOLUTION: A flood state is estimated by using a flood state estimation device 1 having: an image collection unit 11 that collects posted flood images through the Internet; a flood level specification unit 12 that specifies a flood level in the posted flood images by using a learning model that is built with a plurality of data sets as training data which are formed of flood images including a water surface and reference objects to be a reference of flood depth estimation and flood level data of the flood images; a photographed place specification unit 13 that specifies a photographed place of the posted flood images; a related data acquisition unit 14 that acquires photographed places of the posted flood images, and topographic data and precipitation data of the periphery of the place; and a flood state estimation unit 15 that estimates a flood state of an arbitrary area including the photographed places of the posted flood images from the flood level of the posted flood images and the topographic data and precipitation data.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to an apparatus, a system, a method, and a program for estimating the inundation status of a specific area by using posted images publicly available on the Internet.

Background Art

[0002] In recent years, flood disasters caused by typhoons and heavy rain have occurred frequently. From the viewpoints of prompt disaster response and minimizing damage, it has been required to grasp the inundation status at an early stage. In this regard, images posted on social media have attracted attention as information sources for knowing the latest situation of disasters.

[0003] Conventionally, techniques for estimating the inundation depth from various images have been proposed (see Patent Document 1, Non-Patent Documents 1 and 2). In Patent Document 1, an inundation depth estimation apparatus for estimating the inundation depth using satellite images has been proposed. Further, in Non-Patent Documents 1 and 2, techniques for estimating the inundation depth from images taken of the state of inundation collected from the Internet are disclosed.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Non-Patent Documents

[0005]

Non-Patent Document 1

Non-Patent Document 2

[0006] However, the aforementioned conventional technologies have the following problems. The device described in Patent Document 1 employs a method of performing simulations and comparing the results with satellite images, which requires performing simulations beforehand, which involve difficult and uncertain information, and also requires acquiring simulation skills in advance. Furthermore, the device described in Patent Document 1 has the problem that it takes time to make decisions because it compares simulation results with satellite images, and it is not possible to estimate the flooding situation in real time.

[0007] On the other hand, the technology described in Non-Patent Document 1 has the problem that, because it makes judgments based on people, it cannot detect the flood depth if there are no people in images such as photographs of flooded areas. Furthermore, the technology described in Non-Patent Document 2 uses a method of creating a map from multiple images that include the flood boundary, so it has the problem that if multiple images showing the flood boundary cannot be collected, the flood area and flood depth cannot be mapped.

[0008] Therefore, the present invention aims to provide a flood situation estimation device, a flood situation estimation system, a flood situation estimation method, and a program that can estimate the flood situation of an area including the shooting location in real time from posted images. [Means for solving the problem]

[0009] The flooding situation estimation device according to the present invention includes: an image collection unit that collects posted flooding images taken during flooding via the internet; a flooding level identification unit that identifies the flooding level of the posted flooding images using a learning model constructed with multiple datasets consisting of flooding images including the water surface and reference objects that serve as a basis for estimating the flooding depth, and flooding level data of the said flooding images as training data; a shooting location identification unit that identifies the shooting location of the posted flooding images; a related data acquisition unit that acquires topographic data and precipitation data of the shooting location of the posted flooding images identified by the shooting location identification unit and its surrounding area; and a flooding situation estimation unit that estimates the flooding situation of an arbitrary area including the shooting location of the posted flooding images from the flooding level of the posted flooding images identified by the flooding level identification unit and the topographic data and precipitation data acquired by the related data acquisition unit. The flooding situation estimation unit estimates the current flooding situation from previously posted flooding images, or predicts future flooding situations from currently posted flooding images. Furthermore, the flooding situation estimation device of the present invention may also include a map generation unit that generates a flooding estimation map showing the flooded area and flooded level of the area including the location where the posted flooded image was taken, as estimated by the flooding situation estimation unit, on a two-dimensional or three-dimensional map.

[0010] The flooding situation estimation system according to the present invention comprises an image collection device that collects posted flooding images taken during flooding via the Internet, and a flooding situation estimation device that estimates the flooding situation of an arbitrary area including the location where the posted flooding images were taken. The flooding situation estimation device includes a flooding level identification unit that identifies the flooding level of the posted flooding images using a learning model constructed with multiple datasets consisting of flooding images including the water surface and reference objects that serve as a basis for estimating the flooding depth, and flooding level data of the said flooding images as training data; a shooting location identification unit that identifies the location where the posted flooding images were taken; a related data acquisition unit that acquires topographic data and precipitation data of the shooting location of the posted flooding images identified by the shooting location identification unit and its surrounding area; and a flooding situation estimation unit that estimates the flooding situation of an arbitrary area including the location where the posted flooding images were taken from the flooding level of the posted flooding images identified by the flooding level identification unit and the topographic data and precipitation data acquired by the related data acquisition unit. The flooding situation estimation device estimates the current flooding situation from previously posted flooding images, or predicts future flooding situations from currently posted flooding images. Furthermore, the flooding estimation device may also be equipped with a map generation unit that generates a flooding estimation map showing the flooded area and flooding level of any area, including the location where the posted flooding image was taken, on a two-dimensional or three-dimensional map. In that case, the flooding situation estimation system of the present invention may further include a distribution device for distributing the flooding estimation map created by the map generation unit to an external party.

[0011] The flooding situation estimation method according to the present invention comprises: an image collection step of collecting submitted flooding images taken during flooding via the internet using one or more flooding situation estimation devices; a flooding level identification step of identifying the flooding level of the submitted flooding images using a learning model constructed with multiple datasets consisting of flooding images including the water surface and reference objects that serve as a basis for estimating the flooding depth, and flooding level data of the said flooding images as training data; a shooting location identification step of identifying the shooting location of the submitted flooding images; a related data acquisition step of acquiring topographic data and precipitation data of the shooting location of the submitted flooding images identified in the shooting location identification step and its surrounding area; and a flooding situation estimation step of estimating the flooding situation of an arbitrary area including the shooting location of the submitted flooding images from the flooding level of the submitted flooding images identified in the flooding level identification step and the topographic data and precipitation data acquired in the related data acquisition step.

[0012] The flooding situation estimation program according to the present invention causes a computer to perform the following functions: an image collection function that collects flooding images posted via the internet showing the state of flooding; a flooding level identification function that identifies the flooding level of the posted flooding images using a learning model constructed with multiple datasets consisting of flooding images including the water surface and reference objects that serve as a basis for estimating the flooding depth, and flooding level data of the said flooding images as training data; a shooting location identification function that identifies the shooting location of the posted flooding images; a related data acquisition function that acquires topographic data and precipitation data of the shooting location of the posted flooding images identified by the shooting location identification function and its surrounding area; and a flooding situation estimation function that estimates the flooding situation of an arbitrary area including the shooting location of the posted flooding images from the flooding level of the posted flooding images identified by the flooding level identification function and the topographic data and precipitation data acquired by the related data acquisition function. [Effects of the Invention]

[0013] According to the present invention, it is possible to estimate the flooding situation in an area, including the location where a photo was taken, in real time from images posted on social media. [Brief explanation of the drawing]

[0014] [Figure 1] It is a block diagram showing a configuration example of the flood situation estimation device according to the first embodiment of the present embodiment. [Figure 2] It is a flowchart showing a method of estimating the flood situation using the flood situation estimation device 1 shown in FIG. 1. [Figure 3] It is a diagram showing an example of teacher data when constructing the learning model used in the flood level identification step S2b. [Figure 4] It is a diagram showing another example of teacher data when constructing the learning model used in the flood level identification step S2b. [Figure 5] It is a block diagram showing a configuration example of the flood situation estimation device according to the modified example of the first embodiment of the present embodiment. [Figure 6] A is a two-dimensional map showing the flooded area and the flood level, and B is a three-dimensional map showing the flooded area and the flood depth. [Figure 7] It is a flowchart showing a method of creating a flood estimation map using the flood situation estimation device 10 shown in FIG. 5. [Figure 8] It is a block diagram showing a configuration example of the flood situation estimation system according to the second embodiment of the present embodiment.

Embodiments for Carrying Out the Invention

[0015] Hereinafter, embodiments for carrying out the present invention will be described in detail with reference to the accompanying drawings. Note that the present invention is not limited to the embodiments described below.

[0016] (First Embodiment) First, the flood situation estimation device according to the first embodiment of the present invention will be described. The flood situation estimation device of the present embodiment estimates the flood situation of an area including the shooting location from an image taken of the state during flooding posted on social media (hereinafter referred to as a posted flood image).

[0017] <Device Configuration> Figure 1 is a block diagram showing an example configuration of the flooding situation estimation device of this embodiment. As shown in Figure 1, the flooding situation estimation device 1 of this embodiment includes, for example, an image acquisition unit 11, a flooding level identification unit 12, a shooting location identification unit 13, a related data acquisition unit 14, and a flooding situation estimation unit 15.

[0018] [Image Collection Section 11] The image collection unit 11 collects one or more submitted flood images in real time via the internet. The submitted flood images collected here include not only still images but also videos. Furthermore, if the area (prefecture, region, etc.) or river to be monitored for flooding is limited, the submitted flood images to be collected may be narrowed down in advance by place name or river name. This reduces the time required for collecting and classifying submitted images.

[0019] [Flood level identification section 12] The flood level identification unit 12 identifies the flood level of the flood images submitted by the image collection unit 11. In identifying the flood level of the submitted flood images, the flood level identification unit 12 uses a learning model constructed using multiple datasets as training data, each dataset consisting of flood images including the water surface and reference objects that serve as a basis for flood depth estimation, and the flood level data of those flood images.

[0020] In this way, by training a model with reference objects for estimating flood levels and depths, along with actual flood levels or depths, a learning model that can easily output flood levels can be obtained, making it possible to accurately identify the flood levels in submitted flood images. Reference objects can be anything readily available, such as utility poles, antennas, streetlights, buildings and their fences or walls, bridges, and signs. In addition to the man-made objects mentioned above, terrain features such as slopes and cliffs can also be used as reference objects. In recent years, map applications have been equipped with a Street View function that allows users to view images of areas shown on the map, making it easy to view images of the terrain of any given area. By using these terrain images as training data, it is possible to create sets of images and flood levels.

[0021] [Location Identification Section 13] The location identification unit 13 identifies the location where the submitted flood images were taken. Methods for identifying the location of a submitted image include using location information attached to the image or location information of the submission.

[0022] However, many images posted on social media do not include location information of where they were taken, and even when location information is included, the location where the image was taken and the location where it was posted do not always match. In such cases, a method can be applied to identify the location where a posted image was taken by analyzing information about the poster, past posts by the same poster, and accompanying images, as described in Japanese Patent Publication No. 6496952.

[0023] Furthermore, while the flooding situation estimation device 1 of this embodiment can estimate the flooding situation of an area including the shooting location from a single posted flooding image, if there is a discrepancy between the shooting location of the posted flooding image identified by the shooting location identification unit 13, the accuracy of the estimated flooding situation may decrease. In such cases, it is preferable to use a method for identifying the shooting location of a posted image by utilizing information obtained from signs included in the posted image, as described in Japanese Patent Application Publication No. 2022-051069. This allows for pinpoint identification of the shooting location of the posted flooding image, enabling accurate estimation of the flooding situation of an area including the shooting location from a single posted flooding image.

[0024] [Related data acquisition unit 14] The related data acquisition unit 14 acquires topographic data and precipitation data for the location and surrounding area of ​​the flood-related posted image identified by the shooting location identification unit 13. The source of the topographic data and precipitation data is not particularly limited, but for example, topographic data can be obtained from map information services provided on the internet, topographic and elevation databases, or storage devices that have this information stored in advance. Precipitation data can be obtained from weather information services of the Japan Meteorological Agency or weather companies.

[0025] [Flooding Situation Estimation Unit 15] The flooding situation estimation unit 15 estimates the flooding situation of the area including the shooting location from the flooding level of the submitted flooded image identified by the flooding level identification unit 12, and the topographic data and precipitation data acquired by the related data acquisition unit 14. The flooding situation estimation unit 15 calculates the current and future flooding depth of the shooting location from the flooding level of the submitted flooded image and the precipitation data, and calculates the flooding depth of the surrounding area (display area) of the shooting location from these values ​​and topographic data such as elevation.

[0026] <Operation> Next, the operation of the flooding situation estimation device 1 described above, that is, the method of estimating the flooding situation using the flooding situation estimation device 1 of the first embodiment of the present invention, will be explained. Figure 2 is a flowchart of the flooding situation estimation method of this embodiment. As shown in Figure 2, in the flooding situation estimation method of this embodiment, the flooding situation estimation device 1 performs the following steps: an image collection step S1 for collecting submitted flooding images, a flooding level identification step S2a for identifying the flooding level of the submitted flooding images, a shooting location identification step S2b for identifying the shooting location of the submitted flooding images, a related data acquisition step S3 for acquiring topographic data and precipitation data for the shooting location and its surroundings of the submitted flooding images, and a flooding situation estimation step S4 for estimating the flooding situation of the area including the shooting location.

[0027] [S1: Image acquisition process] In the image acquisition step S1, one or more submitted flood images are collected in real time via the internet. The flood situation estimation method of this embodiment may use two or more submitted flood images, but it is possible to estimate the flood situation of an area including the location where the image was taken from a single submitted flood image. This image acquisition step S1 is performed, for example, in the image acquisition unit 11 of the flood situation estimation device 1.

[0028] Here, the method for selectively collecting images depicting flooding from various posted images on social media is not particularly limited, but for example, a method can be applied in which multiple image determination units equipped with machine learning functions, as described in Japanese Patent Publication No. 6779641, are arranged in a tree structure, and determinations are made based on different training data for each determination unit.

[0029] The image classification method described in Japanese Patent Publication No. 6779641 is an image classification method that classifies target images using a machine learning method, comprising: a first determination step in which the target image is input, two or more training data constructed based on two or more types of image data belonging to a first category, and features extracted from the entire target image, and a determination of which type of image in the first category the target image is based solely on the determination rate; and a second determination step in which the target image is input, two or more training data constructed based on two or more types of image data belonging to a second category which is a lower concept than the first category, and features extracted from the entire target image, The method involves a second determination step of calculating the agreement rate with each training data for the image to be determined, and determining which type of image in the second category the image to be determined belongs to based solely on the agreement rate; a third determination step of calculating the agreement rate with each training data for the image to be determined from two or more training data constructed based on two or more types of image data belonging to a third category which is a lower concept than the first category, and features extracted from the entire image to be determined, and determining which type of image in the third category the image to be determined belongs to based solely on the agreement rate, and then performing the second or third determination step based on the determination result of the first determination step.

[0030] By using this method, it is possible to accurately classify and collect the desired flood images from a wide variety of posted images. In this embodiment of the flood situation estimation method, the range of posted images collected is not particularly limited, but if it is known in advance that flooding has occurred in a particular river, the image collection step S1 may be limited to collecting flood images from the river basin of that river. Also, if a dedicated surveillance camera is installed in the river where the flooding occurred, images taken by that camera may be used.

[0031] [S2a: Flood level determination process] The flood level determination step S2a uses a learning model constructed using multiple datasets, each consisting of a flood image including the water surface and a reference object used as a basis for flood depth estimation, and flood level data from said flood image, as training data, to determine the flood level of the submitted flood image collected in the image acquisition step S1. This flood level determination step S2a is performed, for example, in the flood level determination unit 12 of the flood condition estimation device 1.

[0032] Figures 3 and 4 show examples of training data used when constructing a learning model. When constructing training data used to determine the flood level in submitted flood images, for example, for the flood image shown in Figure 3, the post office sign and car are used as reference objects, and the degree of flooding of these reference objects is labeled as "flood level 50cm" to create a dataset. Similarly, for the flood image shown in Figure 4, the house and torii gate are used as reference objects, and the degree of flooding of these is labeled as "flood level 2m" to create a dataset. The learning model used in the flood situation estimation method of this embodiment is constructed using, for example, 1000 or more such training datasets.

[0033] Furthermore, by using the aforementioned learning model, it becomes possible to accurately identify the flood level in submitted flood images. In addition, in the flood situation estimation method of this embodiment, if necessary, the submitted flood images and their flood depths identified in the flood level identification step S2a can be reflected in the training data to update the learning model. Alternatively, when constructing the learning model, specific flood depths may be labeled instead of flood levels for training.

[0034] [S2b: Location identification process] In the location identification process S2b, the location where the submitted flood images were taken is identified down to the town name and street address level. If the location cannot be identified down to the town name level, the process may be stopped at the city / town level or the image may be changed to another image. This location identification process S2b is performed, for example, in the location identification unit 13 of the flood situation estimation device 1.

[0035] One method for identifying the location where a flooded image posted online was taken is to use location information attached to the image or the location information of the post. However, many images posted on social media do not have location information attached to them, and even when location information is attached, the location where the image was taken and the location where it was posted may not match. In such cases, a method can be applied to identify the location where the posted image was taken by analyzing information about the poster, past posts and accompanying images from the same poster, as described in Japanese Patent Publication No. 6496952.

[0036] The data processing method described in Japanese Patent Publication No. 6496952 is a method for processing posted data accompanied by images relating to any event collected via the Internet, using one or more data processing devices, wherein the data processing device performs a poster information analysis step of extracting place names from poster information of the posted data; a past post analysis step of extracting place names from past posted data by the poster of the posted data, concurrently with the poster information analysis step; and an image analysis step of extracting textual information relating to place names from images attached to the posted data, concurrently with the poster information analysis step and the past post analysis step. The analysis results are compared, and if the place name extracted in the poster information analysis step is the same as the place name extracted in the past post analysis step, or if a place name is extracted in only one of the poster information analysis step or the past post analysis step, the location where the event occurred is identified from that place name. If the place name extracted in the poster information analysis step is different from the place name extracted in the past post analysis step, the location where the event occurred is identified from the place name extracted in the poster information analysis step. If the location where the event occurred cannot be identified from the analysis results in the poster information analysis step and the past post analysis step, the location where the event occurred is identified from the place name extracted in the image analysis step.

[0037] Furthermore, while the flooding situation estimation method of this embodiment can estimate the flooding situation of an area including the location where a flooded image was taken from a single posted flooded image, if there is a discrepancy in the location where the identified posted flooded image was taken, the accuracy of the estimated flooding situation may decrease. In such cases, it is preferable to use a method for identifying the location where a posted image was taken by utilizing information obtained from signs included in the posted image, as described in Japanese Patent Application Publication No. 2022-051069.

[0038] The image processing method described in Japanese Patent Publication No. 2022-051069 is a method for identifying the shooting location of a posted image collected via the Internet using one or more image processing devices, and comprises a store name extraction step of recognizing the characters on a sign included in the posted image and extracting the store name, a location information acquisition step of obtaining location information of stores with the names extracted in the store name extraction step from a map database, and a shooting location identification step of performing cluster analysis on the location information of two or more stores with different store names and identifying the shooting location of the posted image from the results.

[0039] By using this method, it is possible to pinpoint the exact location where a flooded image was taken, making it possible to accurately estimate the flooding situation in the area including the location from a single flooded image.

[0040] [S3: Related Data Acquisition Process] In the related data acquisition process S3, topographic data and precipitation data are acquired for the location where the flooded images were taken and their surroundings (for example, an area within a 10km radius from the shooting location), which were identified in the shooting location identification process S2b. At this time, the range for acquiring topographic data such as elevation can be set appropriately according to the area for estimating the flood situation.

[0041] Topographic data can be obtained from map information services and topographic / elevation databases provided on the internet, such as the basic map information from the Geospatial Information Authority of Japan. Alternatively, elevation data for the entire country can be stored in a memory device beforehand, and information for the required area can be retrieved from that device. Precipitation data, on the other hand, can be obtained from weather information services provided by the Japan Meteorological Agency and weather companies.

[0042] [S4: Flood situation estimation process] In the flooding situation estimation process S4, the flooding situation of the area including the shooting location is estimated from the flooding level of the submitted flooding image identified in the flooding level identification process S2a, and the topographic data and precipitation data acquired in the related data acquisition process S3. Since water flows from higher to lower areas and always tries to become horizontal, if the elevation and flooding level (or flooding depth) of the shooting location of the submitted flooding image are known, it is possible to estimate the flooding level (or flooding depth) in the surrounding area.

[0043] For example, if the elevation of the shooting location is 100m and the flood level is 1m, then at an elevation of 99m, the flood level should be 2m. In this case, by setting the water level to 101m, the flood depth of the surrounding area can be calculated from the topographic data. On the other hand, precipitation data is mainly used for prediction. For example, if heavy rain continues to fall in the area including the shooting location, the flooded area will expand and the flood depth will increase, but if the precipitation decreases, the flooded area will shrink and the flood depth will also decrease.

[0044] Therefore, by comparing the precipitation at the time of determining the flood level with the predicted precipitation provided by the Japan Meteorological Agency and other sources, it is possible to determine whether the flooded area will expand or contract, and whether the flood depth will increase or decrease, thereby predicting changes in the flood situation. Furthermore, if there is a time difference between when the submitted flood images were taken and when the flood level was determined, comparing the precipitation at the time of taking the images with the current precipitation makes it possible to estimate the current flood situation with greater accuracy.

[0045] <Program> Each of the aforementioned steps can be carried out by creating a computer program to realize the functions of each part of the flooding situation estimation device 1 and implementing it on one or more computers.

[0046] In other words, the flooding situation estimation method of this embodiment can be implemented by running a program on a computer or flooding situation estimation device that executes: (1) an image collection function that collects one or more posted flooding images taken during flooding via the internet; (2) a flooding level identification function that identifies the flooding level of a posted flooding image using a learning model constructed with multiple datasets consisting of flooding images including the water surface and reference objects that serve as a basis for estimating the flooding depth, and flooding level data of said flooding images as training data; (3) a shooting location identification function that identifies the shooting location of the posted flooding image; (4) a related data acquisition function that acquires topographic data and precipitation data of the shooting location of the posted flooding image identified by the shooting location identification function and its surrounding area; and (5) a flooding situation estimation function that estimates the flooding situation of an arbitrary area including the shooting location of the posted flooding image from the flooding level of the posted flooding image identified by the flooding level identification function and the topographic data and precipitation data acquired by the related data acquisition function.

[0047] Furthermore, the aforementioned functions do not need to be included in a single program; they can be executed by creating separate programs for each function and linking them together. In that case, each program can be divided and implemented on two or more computers or devices.

[0048] As described in detail above, the flooding situation estimation device of this embodiment uses a learning model constructed with multiple datasets consisting of flooding images including the water surface and reference objects that serve as a basis for estimating flooding depth, and flooding level data of said flooding images, to identify the flooding level or flooding depth of a submitted flooding image, and also calculates the current flooding level or flooding depth of the surrounding area from the elevation and precipitation. Therefore, it is possible to estimate the flooding situation of an area including the shooting location in real time from a single submitted flooding image.

[0049] (Modified version of the first embodiment) Next, a flood situation estimation device according to a modified version of the first embodiment of the present invention will be described. The flood situation estimation device of the present invention can also display the flood situation of any area, including the location where the posted flood images estimated by the flood situation estimation unit were taken, on a two-dimensional or three-dimensional map.

[0050] <Device configuration> Figure 5 is a block diagram showing an example configuration of the flooding situation estimation device of this modified example. In Figure 5, the same reference numerals are used for components that are the same as those of the flooding situation estimation device 1 shown in Figure 1, and detailed explanations are omitted. As shown in Figure 5, the flooding situation estimation device 10 of this modified example is equipped with a map generation unit 16 in addition to the image acquisition unit 11, flooding level identification unit 12, shooting location identification unit 13, related data acquisition unit 14, and flooding situation estimation unit 15 mentioned above.

[0051] [Map generation unit 16] The map generation unit 16 generates a two-dimensional or three-dimensional map showing the "flood area" and "flood level or flood depth" for the location where the submitted flood images were taken and the surrounding area, based on the estimation results from the flood situation estimation unit 15. Figure 6A is a two-dimensional map showing the flood area and flood level, and Figure 6B is a three-dimensional map showing the flood area and flood depth. For example, in the two-dimensional map shown in Figure 6A, the flood area and the flood level in each area are shown by color-coding according to the flood level.

[0052] Furthermore, in the case of the three-dimensional map shown in Figure 6B, the flooded area is indicated by coloring, and the flood depth is indicated by the intensity of the color. For example, if the elevation of the shooting location is 100m and the flood level is 1m, then in the three-dimensional map shown in Figure 6B, the water surface should be colored so that it is located at an elevation of 101m.

[0053] <Operation> Next, the operation of the flooding situation estimation device 10 of this modified example, that is, the method of estimating the flooding situation using the flooding situation estimation device 10, will be explained. Figure 7 is a flowchart of the flooding situation estimation method of this modified example. In Figure 7, the same reference numerals are used for the same steps as in the flooding situation estimation method shown in Figure 2, and detailed explanations are omitted.

[0054] As shown in Figure 7, in this modified example of the flooding situation estimation method, the flooding situation estimation device 10 performs the following steps: an image collection step S1 for collecting submitted flooding images; a flooding level identification step S2a for identifying the flooding level of the submitted flooding images; a shooting location identification step S2b for identifying the shooting location of the submitted flooding images; a related data acquisition step S3 for acquiring topographic data and precipitation data for the shooting location and its surroundings; a flooding situation estimation step S4 for estimating the flooding situation of the area including the shooting location; and a map generation step S5 for generating a two-dimensional or three-dimensional map showing the flooded area and flooding depth.

[0055] [S5: Map generation process] In the map generation process S5, a flood estimation map is generated that reflects the flooding situation of the location where the submitted flood images were taken and the surrounding area, as estimated in the flooding situation estimation process S4. Specifically, a two-dimensional or three-dimensional map of the area including the location where the submitted flood images were taken is generated, showing the "flooded area" and the "flood level or flood depth." The method for showing the "flooded area" and the "flood level or flood depth" is not particularly limited, but for example, the flooded area can be colored, and the color can be differentiated or the shade can be changed according to the flood level or flood depth.

[0056] <Program> Each of the aforementioned steps can be carried out by creating a computer program to realize the functions of each part of the flooding situation estimation device 10 and implementing it on one or more computers.

[0057] In other words, the flooding situation estimation method of this modified example includes: (1) an image collection function that collects one or more submitted flooding images taken during flooding via the internet; (2) a flooding level identification function that identifies the flooding level of a submitted flooding image using a learning model constructed with multiple datasets consisting of flooding images including the water surface and reference objects that serve as a basis for estimating the flooding depth, and the flooding level data of said flooding images as training data; (3) a shooting location identification function that identifies the shooting location of the submitted flooding image; and (4) the topographic data of the shooting location of the submitted flooding image identified by the shooting location identification function and its surrounding area. This can be implemented by running a program on a computer or flood situation estimation device that performs the following: (5) a related data acquisition function to acquire precipitation data; (6) a flood situation estimation function that estimates the flood situation of an arbitrary area including the location where the posted flood image was taken, based on the flood level of the posted flood image identified by the flood level identification function, and the topographic data and precipitation data acquired by the related data acquisition function; and (7) a map generation function that generates a two-dimensional or three-dimensional map showing the flood extent and flood depth of the area including the location where the posted flood image was taken, as estimated by the flood situation estimation function.

[0058] Furthermore, the aforementioned functions do not need to be included in a single program; they can be executed by creating separate programs for each function and linking them together. In that case, each program can be divided and implemented on two or more computers or devices.

[0059] As detailed above, the flood situation estimation device of this modified version is equipped with a map generation function, and can automatically create and provide a flood estimation map showing the estimated flood situation on a two-dimensional or three-dimensional map in near real time. Furthermore, if the flood situation estimation unit predicts future flood situations, it can also create and provide a flood estimation map based on those predictions.

[0060] Furthermore, the flood estimation device of this modified version may monitor the amount of rainfall and the predicted rainfall at the shooting location or the location with the heaviest rainfall, estimate the flood situation based on these values, and recreate the flood estimation map. For example, if the same or greater amount of rainfall is expected, the flood level at the shooting location is changed from 1m to 1.5m to estimate the flood situation, and the result is reflected on the map. This creates a flood estimation map with an expanded flood area. Conversely, if less rainfall is expected, the flood level at the shooting location is changed from 1m to 50cm to estimate the flood situation, and the result is reflected on the map. This creates a flood estimation map with a reduced flood area.

[0061] The flood inundation estimation maps showing the expansion or contraction of the flooded area, as described above, may be created according to the amount of rainfall or the predicted amount of rainfall, or they may be created at the same time as the current flood inundation estimation map. In that case, three maps will be created and provided: the current flood inundation estimation map, the flood inundation estimation map for when the flooded area expands, and the flood inundation estimation map for when the flooded area contracts. These flood inundation estimation maps may also be presented as videos, in which case the process of the flooding expanding or contracting can be shown. The configuration and effects of this modified example other than those described above are the same as those of the first embodiment described above.

[0062] (Second embodiment) Next, a flood situation estimation system according to a second embodiment of the present invention will be described. Figure 8 is a block diagram showing an example of the configuration of the flood situation estimation system of this embodiment. In Figure 8, the same reference numerals are used for components that are the same as those of the flood situation estimation device 1 shown in Figure 1 and the flood situation estimation device 10 shown in Figure 5, and detailed explanations are omitted. As shown in Figure 8, the flood situation estimation system 2 of this embodiment is provided with an image acquisition device 21 in addition to the flood situation estimation device 20, and a distribution device 22 is provided to distribute the flood estimation map externally as needed.

[0063] [Image acquisition device 21] The image acquisition device 21 collects one or more submitted flood images via the internet, and its configuration and operation are the same as the image acquisition unit in the flood situation estimation device of the first embodiment and its modified form described above.

[0064] [Flooding Situation Estimation Device 20] The flooding situation estimation device 20 estimates the flooding situation of any area, including the location where the submitted flooding images collected by the image acquisition device 21 were taken, and includes a flooding level identification unit 12, a shooting location identification unit 13, a related data acquisition unit 14, a flooding situation estimation unit 15, and a map generation unit 16. Here, the flooding level identification unit 12 identifies the flooding level of the submitted flooding image using a learning model constructed with multiple datasets consisting of flooding images including the water surface and reference objects that serve as a basis for estimating the flooding depth, and flooding level data of said flooding images, as training data.

[0065] Furthermore, the shooting location identification unit 13 identifies the shooting location of the submitted flood images, and the related data acquisition unit 15 acquires topographic data and precipitation data for the shooting location of the submitted flood images and its surrounding area, as identified by the shooting location identification unit 13. The flood situation estimation unit 15 then estimates the flood situation for any area, including the shooting location of the submitted flood images, based on the flood level of the submitted flood images identified by the flood level identification unit 12 and the topographic data and precipitation data acquired by the related data acquisition unit 14.

[0066] The flooding situation estimation device 20 may be equipped with a map generation unit 16 that generates a two-dimensional or three-dimensional map (flooding estimation map) showing the extent and depth of flooding in any area, including the location where the submitted flooding image was taken. In that case, the flooding estimation map created by the map generation unit 16 can also be distributed externally by the distribution device 22 described later.

[0067] This flooding situation estimation device 20 can estimate the current flooding situation from previously posted flooding images, and it can also predict future flooding situations from currently posted flooding images.

[0068] [Distribution device 22] The distribution device 22 distributes a two-dimensional or three-dimensional map (flood estimation map) to external users, which shows the extent and depth of flooding in any area including the location where the submitted flood images were taken, as created by the map generation unit 16 of the flooding situation estimation device 20. The distribution device 22 is provided as needed.

[0069] In this embodiment of the flooding situation estimation system, similar to the flooding situation estimation device of the first embodiment and its modified form described above, a learning model constructed using multiple datasets consisting of flooding images including the water surface and reference objects that serve as a basis for estimating flooding depth, and flooding level data of said flooding images, is used to identify the flooding level or flooding depth of a submitted flooding image, and to calculate the current flooding level or flooding depth of the surrounding area from the elevation and precipitation. Therefore, the flooding situation of an area including the shooting location can be estimated in real time from a single submitted flooding image.

[0070] Furthermore, the flooding estimation system of this embodiment can automatically create and distribute flooding estimation maps showing the estimated flooding situation on a two-dimensional or three-dimensional map, as well as flooding estimation maps showing future flooding situations. Note that the configuration, operation, and effects of this embodiment other than those described above are the same as those of the first embodiment and its modified versions described above. [Explanation of Symbols]

[0071] 1, 10, 20 Flood Situation Estimation Device 2. Flood Situation Estimation System 11 Image Collection Department 12. Flood level determination section 13. Identifying the filming location 14 Related Data Acquisition Unit 15. Flood Situation Estimation Section 16 Map generation section 21 Image acquisition device 22 Distribution device

Claims

1. The image collection unit collects images of flooding posted via the internet, and An inundation level identification unit identifies the inundation level of the submitted inundation image using a learning model constructed with training data consisting of multiple datasets, each dataset comprising an inundation image including the water surface and a reference object that serves as a basis for estimating the inundation depth, and inundation level data of the said inundation image. A location identification unit that identifies the location where the aforementioned flooded image was taken, A related data acquisition unit acquires the location where the posted flood image was taken and the surrounding area, as identified by the aforementioned location identification unit, and related data acquisition unit. A flood situation estimation unit estimates the flood situation of an arbitrary area, including the location where the posted flood image was taken, based on the flood level of the posted flood image identified by the flood level identification unit and the topographic data and precipitation data acquired by the related data acquisition unit. A flooding situation estimation device having the following features.

2. The flooding situation estimation device according to claim 1, wherein the flooding situation estimation unit estimates the current flooding situation from previously posted flooding images, or predicts future flooding situations from currently posted flooding images.

3. The flooding situation estimation device according to claim 1, further comprising a map generation unit that generates a flooding estimation map showing the flooded area and flooding level in a two-dimensional or three-dimensional map of the area including the location where the aforementioned flooded image was taken.

4. An image collection device that collects images of flooding posted via the internet, The system includes a flooding situation estimation device that estimates the flooding situation in any area, including the location where the aforementioned flooded image was taken, The aforementioned flooding situation estimation device is An inundation level identification unit identifies the inundation level of the submitted inundation image using a learning model constructed with training data consisting of multiple datasets, each dataset comprising an inundation image including the water surface and a reference object that serves as a basis for estimating the inundation depth, and inundation level data of the said inundation image. A location identification unit that identifies the location where the aforementioned flooded image was taken, A related data acquisition unit acquires the location where the posted flood image was taken and the surrounding area, as identified by the aforementioned location identification unit, and related data acquisition unit. A flood situation estimation unit estimates the flood situation of an arbitrary area, including the location where the posted flood image was taken, based on the flood level of the posted flood image identified by the flood level identification unit and the topographic data and precipitation data acquired by the related data acquisition unit. A flood situation estimation system equipped with the following features.

5. The flooding situation estimation system according to claim 4, wherein the flooding situation estimation device estimates the current flooding situation from previously posted flooding images, or predicts future flooding situations from currently posted flooding images.

6. The flooding situation estimation system according to claim 4 or 5, further comprising a map generation unit that generates a flooding estimation map showing the flooded area and flooded level of an arbitrary area including the location where the posted flooded image was taken, on a two-dimensional or three-dimensional map.

7. The inundation situation estimation system according to claim 6, further comprising a distribution device for externally distributing the inundation estimation map created by the map generation unit.

8. One or more flooding situation estimation devices, The image collection process involves collecting images of flooding posted online, showing the situation during the flooding. A flood level identification step involves identifying the flood level of the submitted flood image using a learning model constructed with training data consisting of multiple datasets comprising flood images including the water surface and reference objects that serve as a basis for estimating flood depth, and flood level data of said flood images, The process of identifying the location where the aforementioned flooded image was taken, A related data acquisition step involves acquiring the location where the posted flood images were taken and the surrounding area, as identified in the aforementioned location identification step, as well as related data acquisition step. A flood situation estimation step is performed to estimate the flood situation of an arbitrary area, including the location where the posted flood image was taken, based on the flood level of the posted flood image identified in the flood level identification step and the topographic data and precipitation data acquired in the related data acquisition step. A method for estimating flood conditions.

9. On the computer, An image collection function that collects images of flooding posted via the internet, A flood level identification function that identifies the flood level of the submitted flood image using a learning model constructed with training data consisting of multiple datasets, each dataset comprising flood images including the water surface and reference objects that serve as a basis for estimating flood depth, and flood level data from said flood images, The aforementioned flooded image posting has a location identification function to identify the location where the image was taken, A related data acquisition function that acquires the location where the posted flood image was taken and the surrounding topographic data and precipitation data, which are identified by the aforementioned location identification function, A flood situation estimation function estimates the flood situation of any area, including the location where the posted flood image was taken, based on the flood level of the posted flood image identified by the flood level identification function and the topographic data and precipitation data acquired by the related data acquisition function. A program that executes the command.

Images