Information processing device and program

Abstract

We utilize flood information obtained from flood sensors, social media images, aerial photographs, satellite images, etc., to calculate highly accurate flood map data. [Solution] An information processing device for creating an inundation map of an evaluation target area including a river, comprising: a HAND model acquisition means that acquires a HAND model in which the HAND ground height, which is the relative height difference with the nearest drainage point, is the value of a grid cell, and rotates the HAND model so that the X-axis direction is the direction of river flow; and an inundation map creation means that creates an inundation map based on the HAND model and dynamic sensing information including information on inundation, wherein the inundation map creation means comprises a measured HAND water surface height calculation means, an inundation map creation segment discrimination means, a measured HAND water surface height maximum value calculation means, an inundation map creation interpolation function calculation means, and an inundation map calculation means.

Description

[Technical Field]

[0001] The present invention relates to an information processing device that enables the calculation of highly accurate flood maps for use by administrative agencies, residents, etc., for disaster response and evacuation decisions, based on a wide variety of dynamic sensing information such as flood sensors, SNS (Social Networking Service) images (still images), aerial photographs (still images), and satellite images (SAR: Synthetic Aperture Radar images) when flooding occurs due to external flooding caused by river levee breaches, etc., and internal flooding caused by insufficient drainage capacity of sewers, etc., and to a program for causing a computer to function as the above-mentioned information processing device. [Background technology]

[0002] Patent Document 1 discloses a technique for estimating the flooded area based on flood information from a flood sensor, i.e., actual flood information.

[0003] Patent Document 2 discloses technology relating to a flood simulation device, a flood simulation method, and a program capable of simulating flood conditions with high accuracy.

[0004] Patent Document 3 discloses a technique for appropriately estimating the distribution of flood depths within a flooded area, using flood depths and flooded areas obtained from on-site surveys as input.

[0005] Non-patent document 1 discloses a technique for creating a topographic model called HAND (Height Above Nearest Drainage) based on a topographic model called DEM (Digital Elevation Model). Here, HAND is a topographic model in which the relative elevation difference with respect to the nearest drainage point, such as a river, is used as the value of each grid cell.

[0006] Non-patent document 2 discloses a technique for calculating static flood inundation area map data by drawing contour lines based on a topographic model called HAND.

[0007] Non-patent document 3 discloses a technique for calculating a curve called a water-level-flow curve based on a topographic model called HAND and Manning's formula, calculating the water level based on the water-level-flow curve and the flood discharge during a flood, and then calculating inundation map data based on the water level. It also discloses a technique for calculating division points that divide a river into equal intervals in the direction of flow, using the areas where water collects at each division point (catchment areas) as segments, and calculating inundation map data using water-level-flow curves for each segment. [Prior art documents] [Patent Documents]

[0008] [Patent Document 1] Japanese Patent Publication No. 2024-030214 [Patent Document 2] Japanese Patent Publication No. 2023-096271 [Patent Document 3] Japanese Patent Publication No. 2008-310036 [Non-patent literature]

[0009] [Non-Patent Document 1] Camilo Daleles Renno, Antonio Donato Nobre, Luz Adriana Cuartas, Joao Vianei Soares, Martin G.Hodnett, Javier Tomasella, Maarten J.Waterloo, HAND, a new terrain descriptor using SRTM-DEM:Mapping terra-firme rainforest environments in Amazonia, Remote Sensing of Environment, Netherlands, Elsevier, 2008 / 3 / 18 [Non-Patent Document 2] Antonio Donato Nobre, Luz Adriana Cuartas, Marcos Rodrigo Momo, Dirceu Luis Severo, Adilson Pinheiro, Carlos Afonso Nobre, HAND contour: a new proxy predictor of inundation extent, Hydrological Processes, USA, Wiley, 2015 [Non-Patent Document 3] Xing Zheng, David R. Maidment, David G. Tarboton, Yan Y. Liu, Paola Passalacqua, GeoFlood: Large-Scale Flood Inundation Mapping Based on High-Resolution Terrain Analysis, Water Resources Research, USA, Wiley, 2018 / 12 / 13, pp.10013-10033 [Overview of the Initiative] [Problems that the invention aims to solve]

[0010] Flood sensors only detect the presence or absence of flooding at the sensor's installation height. For example, if a flood sensor is installed 50 cm above ground level, it cannot detect flood depths less than 50 cm. Even if the flood depth is 100 cm, it can only detect limited information, such as that the flood depth is 50 cm or more. When one or more flood sensors are installed on a structure such as a utility pole, it is easy to compare the relative magnitudes of the water levels detected by each sensor. However, when one or more flood sensors are dispersed throughout a given area, there is a problem in that it is not possible to compare the relative magnitudes of the water levels detected by each sensor due to the ground slope in the direction of river flow.

[0011] Furthermore, if one or more flood sensors are installed in a designated area, it is conceivable to calculate the maximum water level based on the water level detected by each sensor, and to use that maximum water level as the representative water level for the entire designated area. In such cases, because there is a ground slope in the direction of river flow, the water level near the upstream area of ​​the designated area is generally used, which has led to the problem of overestimating the water level near the middle and downstream areas of the designated area.

[0012] Thus, conventionally, when one or more flood sensors were installed in a designated area, it was not possible to calculate flood map data with high accuracy because there was a ground slope in the direction of river flow.

[0013] On the other hand, if one or more flood sensors are installed in a given area, it is also conceivable to calculate an interpolation function that passes through the installation location (latitude, longitude) of each flood sensor and the water level detected by each flood sensor, and then use the water level represented by this interpolation function as the water level of the given area. In such cases, for example, there was a problem in that the water level and flood depth were underestimated in low-lying areas where flood depth tends to be deep, and near flood sensors installed at a low height.

[0014] Furthermore, if one or more flood sensors are installed in a designated area, it is conceivable to divide the designated area into multiple segments and, for example, calculate the maximum water level for each segment based on the water level detected by one or more flood sensors within each segment, and use this as the representative water level for each segment. In such a case, although there is a ground slope in the direction of river flow, its effect is limited to each segment, so an improvement in the accuracy of the calculated water level can be expected. However, there was a trade-off problem between the number of flood sensors within a segment and the number of segments.

[0015] Thus, conventionally, when one or more flood sensors are installed in a given area, even if the water level is calculated using an interpolation function, or even if the given area is divided into multiple segments, it has been impossible to calculate flood map data with high accuracy because there is a ground gradient in the direction of river flow, and there is a trade-off problem between the number of flood sensors in a segment and the number of segment divisions.

[0016] On the other hand, the installation and maintenance of flood sensors are costly, making it difficult to install a large number of them in a given area. In such cases, to compensate for the shortage of flood sensors, it is conceivable to use SNS images (still images) posted to SNS during flooding, aerial photographs (still images) taken during flooding, and satellite images (SAR images) observed during flooding. While this can compensate for the shortage of flood sensors, there is a problem that SNS images (still images), aerial photographs (still images), and satellite images (SAR images) generally do not contain information about water level. For this reason, conventionally, even when integrating sensing information such as flood sensors, SNS images (still images), aerial photographs (still images), and satellite images (SAR images) as input data, it was not possible to calculate flood map data with high accuracy. Furthermore, while data from flood sensors and SNS images (still images) can be received at short intervals, such as every 10 minutes, aerial photograph data can only be received during the daytime and after the weather has improved to some extent. For example, with satellite imagery (SAR) data, the ALOS-2 satellite passes over Japan twice a day, which presented a problem of a large time lag between the occurrence of flooding and the reception of the data.

[0017] On the other hand, the technology described in Patent Document 1 was able to estimate the flooded area by calculating the flood depth based on flood information detected by a flood sensor and comparing it with location information at adjacent locations. However, while it is possible to estimate the flooded area in a small area by comparing it only with adjacent locations, there were problems in estimating the flooded area in a wide area, such as an entire watershed, based on flood information detected by a small number of flood sensors, because the terrain has repeated undulations.

[0018] Furthermore, the technology described in Patent Document 2 allowed for the calculation of the drainage ditch's gradient by modifying Manning's formula based on information about the drainage ditch, such as its location, depth, average flow velocity, roughness coefficient, and wetted perimeter length. However, there was a problem in accurately determining the roughness coefficient required for using Manning's formula. There was also a problem in accurately determining the average flow velocity of the drainage ditch during floods. Moreover, there was a problem in that the technology for changing the ground surface height based on the drainage ditch's gradient was not sufficiently disclosed.

[0019] Furthermore, the technology described in Patent Document 3 uses inundation depth and inundation area obtained from on-site surveys as input to calculate the inundation level of multiple sections including the contour line of the inundation area by adding a predetermined inundation depth to the ground elevation of each section. The distribution of inundation depth in the inundation area can then be calculated by subtracting the ground elevation from the inundation level calculated by spatial interpolation. However, accurately investigating inundation depth during floods is difficult, and inundation sensors only detect the presence or absence of inundation at the sensor installation height. Social media images, aerial photographs, and satellite images do not contain information about water level or inundation depth.

[0020] Furthermore, the technology described in Non-Patent Document 1 allowed for the creation of a topographic model called HAND based on a topographic model called DEM. However, simply using the relative elevation difference with the nearest drainage point, such as a river, as the value for each grid cell did not allow for the calculation of flood map data, although it did remove the ground gradient in the direction of river flow.

[0021] In addition, the technology described in Non-Patent Document 2 was able to calculate static flood inundation area data by drawing contour lines based on a topographic model called HAND. However, because it did not use on-site observation data such as inundation sensors, it could not calculate inundation map data that changes moment by moment during an actual flood. Furthermore, because the predetermined area was not divided into segments, there was a problem in that, for example, if no inundation occurred near the upstream but inundation occurred near the midstream or downstream, the inundation depth near the upstream would be overestimated.

[0022] Furthermore, in the technology described in Non-Patent Document 3, a curve called a water-level-flow curve was calculated based on a topographic model called HAND and Manning's formula. Based on this water-level-flow curve and the flood discharge during a flood, the water level was calculated, and inundation map data was calculated based on this water level. In addition, division points were calculated to divide the river at equal intervals in the direction of flow, and the areas where water collects at each division point (catchment areas) were used as segments. Inundation map data using water-level-flow curves was calculated for each segment. However, there was a problem in that it was difficult to accurately observe the flood discharge during a flood. There was also a problem in that it was difficult to accurately determine the roughness coefficient necessary to obtain the water-level-flow curve. Moreover, although division points were calculated to divide the river at equal intervals in the direction of flow, and the areas where water collects at each division point (catchment areas) were divided as segments, a method for optimally adjusting these segments was not disclosed, leaving room for improvement in terms of accuracy.

[0023] This invention was made in view of the above problems, and its main purpose is to calculate highly accurate flood map data based on flood information obtainable from flood sensors, SNS images, aerial photographs, satellite images, etc. [Means for solving the problem]

[0024] To achieve this objective, the present invention provides an information processing device for creating an inundation map of an evaluation target area including a river, comprising: a HAND model acquisition means for acquiring a HAND model in which the HAND ground height, which is the relative elevation difference with the nearest drainage point, is the value of a grid cell, and rotating it so that the X-axis direction is the direction of river flow; and an inundation map creation means for creating an inundation map based on the HAND model and dynamic sensing information including information on inundation, wherein the inundation map creation means includes a measured HAND water surface height calculation means for calculating the measured HAND water surface height based on the HAND ground height and the dynamic sensing information for each acquisition position of the dynamic sensing information, and divides the evaluation target area in a direction intersecting the direction of river flow to set up a plurality of segments. The system is characterized by comprising: a segment discrimination means for creating a flood map; a measured HAND water level maximum value calculation means for obtaining the maximum value of the measured HAND water level for each segment; an interpolation function calculation means for creating a flood map, which projects the maximum value of the measured HAND water level onto a projection plane with the river flow direction as the horizontal axis and the measured HAND water level as the vertical axis, and calculates an interpolation function; and a flood map calculation means that associates the calculated value obtained from the interpolation function with each grid cell in the HAND model that is continuous in the X-axis direction, and associates the same value with adjacent grid cells in the Y-axis direction, and then calculates the flood depth for each grid cell by subtracting the HAND ground level from the calculated value, determines that grid cells with a flood depth greater than 0 are flooded, and creates a flood map.

[0025] The information processing device of the present invention may also include, in the flood diagram creation means, a flood diagram correction means that corrects the flood diagram so that the grid cells of the flood diagram corresponding to locations where flooding is determined to be present in the dynamic sensing information are determined to be flooded.

[0026] The information processing device of the present invention may also be configured such that the dynamic sensing information is information obtained from at least one of a flood sensor, aerial photographs, satellite images, or SNS posts, and when information indicating flooding is obtained based on aerial photographs, satellite images, and SNS posts, the HAND ground height at the acquisition location is set to the measured HAND water level.

[0027] The information processing device of the present invention may also include an abnormal value detection means in the flood diagram creation means for detecting abnormal values ​​contained in the flood sensor for each of the multiple segments.

[0028] The information processing device of the present invention further comprises a virtual flood map creation means for creating a virtual flood map of the area to be evaluated based on the HAND model and existing flood information, the virtual flood map creation means for creating virtual dynamic sensing information by arbitrarily placing a plurality of virtual flood sensors in the flood range acquired from the existing flood information, a virtual HAND water level calculation means for setting the HAND ground height at the placement position of the virtual flood sensors as the virtual HAND water level, and a virtual HAND water level maximum value means for dividing the area to be evaluated in a direction intersecting the flow direction of the river to set a plurality of segments and acquiring the maximum value of the virtual HAND water level for each segment. The system may also include: a calculation means; an interpolation function calculation means for creating a virtual flood map that projects the maximum value of the virtual HAND water level onto a projection plane with the river flow direction as the horizontal axis and the virtual HAND water level as the vertical axis, and calculates an interpolation function; and a virtual flood map calculation means that associates each of the grid cells in the HAND model that are continuous in the X-axis direction with a calculated value obtained from the interpolation function obtained by the interpolation function calculation means for creating a virtual flood map, and associates the same value with adjacent grid cells in the Y-axis direction, and then calculates a virtual flood depth for each grid cell by subtracting the HAND ground level from the calculated value, determines that grid cells with a virtual flood depth greater than 0 are flooded, and creates a virtual flood map.

[0029] The information processing device of the present invention may include existing flood inundation prediction information or past flood performance information in the existing flood inundation information.

[0030] The information processing device of the present invention further comprises an evaluation index value average calculation means, wherein the virtual flood map creation means further comprises an evaluation index value average calculation means, which compares each grid cell with the existing flood information for each of the multiple virtual flood maps created by arbitrarily changing the number of segment divisions, the number of virtual flood sensors, and the placement positions of the virtual flood sensors, obtains an F value which is the harmonic mean of the precision and recall rates related to the presence or absence of flooding, and calculates the average F value for each number of segment divisions of the virtual flood maps with the same virtual sensor density, and the virtual sensor density may be the number of virtual flood sensors relative to the area of ​​the flooded area in the evaluation target region obtained from the existing flood information.

[0031] The information processing device of the present invention includes a flood map creation segment discrimination means that calculates an interpolation function based on three-dimensional scatter data consisting of the number of segment divisions, the virtual sensor density, and the average F value, and sets a plurality of segments in the evaluation target area with the number of segment divisions that maximizes the average F value calculated based on the interpolation function and the measured sensor density, wherein the measured sensor density may be the total number of acquisition locations of the dynamic sensing information relative to the area of ​​the flood-prone area in the evaluation target area.

[0032] The present invention can also be implemented by a program that causes a computer to function as the information processing device described above.

[0033] According to the information processing device and program of the present invention, the measured HAND water surface height is calculated based on the HAND ground height and the dynamic sensing information for each location where dynamic sensing information is acquired. This makes it possible to compare the relative magnitudes of the water surface heights at the locations where dynamic sensing information is acquired using the measured HAND water surface height, which is free from the ground gradient in the direction of river flow. Therefore, even when using information acquired from multiple flood sensors that are dispersed from the upstream to the downstream side of a river and have different installation heights (installation height from the ground surface) as dynamic sensing information, the evaluation target area can be divided into multiple segments in a direction intersecting the river's flow direction, and the maximum value of the measured HAND water surface height can be calculated for each segment. This maximum value can then be obtained as the water surface height that corresponds to reality for each segment. Furthermore, by calculating an interpolation function for the maximum value acquired for each segment and calculating the flood depth of the evaluation target area based on this interpolation function, it becomes possible to calculate continuous flood map data with high accuracy, even when the flooding situation differs between the upstream and downstream areas of a river. Furthermore, by setting the number of segments based on the measured sensor density of the area to be evaluated, it becomes possible to calculate more accurate flood map data.

[0034] Furthermore, as dynamic sensing information, flood-related information obtained from flood sensors, SNS images (still images), aerial photographs (still images), and satellite images (SAR images) can be used. In this way, even when flood-related information has various conditions, such as being point data or area data, it is possible to integrate these and reflect them in the flood map data. Moreover, although there is a time lag between the occurrence of flooding and the acquisition of data from flood sensors, SNS images, aerial photographs, and satellite images, by adding this data each time it is acquired and updating the information, it becomes possible to capture the moment-by-moment changes in the flood situation in the evaluation area as data for the flood map. [Effects of the Invention]

[0035] According to the present invention, by employing flood-related information obtainable from flood sensors, SNS images (still images), aerial photographs (still images), satellite images (SAR images), etc., as dynamic sensing information, it becomes possible to calculate highly accurate flood map data for evaluation target areas, including rivers. [Brief explanation of the drawing]

[0036] [Figure 1] This is a schematic diagram illustrating the challenges of creating flood maps using DEM (Data Element Method). [Figure 2] This is a schematic diagram illustrating the effectiveness of creating flood maps using HAND. [Figure 3] This is a schematic diagram illustrating how the measured sensor density changes over time. [Figure 4] This flowchart shows an example of creating a HAND model. [Figure 5] This figure shows an example of the system configuration of an information processing device. [Figure 6] This figure shows an example of the hardware configuration of a terminal device. [Figure 7] This figure shows an example of the hardware configuration of a server device. [Figure 8] This figure shows an example of the functional configuration of an information processing device. [Figure 9] This figure shows an example of the functional configuration of the HAND model acquisition means. [Figure 10] This figure shows another example of the functional configuration of the HAND model acquisition means. [Figure 11] This is a schematic diagram illustrating the function of rotating the image coordinate space. [Figure 12] This figure shows an example of dynamic sensing information related to a water ingress sensor. [Figure 13] This figure shows an example of dynamic sensing information related to social networking services (SNS). [Figure 14] This figure shows an example of dynamic sensing information related to aerial photographs. [Figure 15] This figure shows an example of dynamic sensing information related to satellite imagery. [Figure 16]This is a schematic diagram illustrating the interpolation function calculation function using radial basis functions. [Figure 17] This is a schematic diagram illustrating an example of the segmentation function. [Figure 18] This is a schematic diagram illustrating another example of the segmentation function. [Figure 19] This is a schematic diagram illustrating the function of projecting the measured HAND water level onto a cross-section. [Figure 20] This figure shows an example of the functional configuration of a flood inundation map calculation device. [Figure 21] This figure shows another example of the functional configuration of the flood map calculation means. [Figure 22] This is a schematic diagram explaining the flood inundation map calculation function. [Figure 23] This figure shows an example of the functional configuration of an existing flood information acquisition method. [Figure 24] This figure shows another example of the functional configuration of existing flood information acquisition methods. [Figure 25] This figure shows an example of the average value of an evaluation index. [Figure 26] This flowchart shows an example of the process for creating a flood diagram of a server device. [Figure 27] This flowchart shows an example of the process for creating a virtual flood diagram for a server device. [Figure 28] This figure shows an example of a flood map calculated by an information processing device. [Modes for carrying out the invention]

[0037] This invention employs flood-related information obtained from flood sensors, social media posts, aerial photographs, satellite images, etc., as dynamic sensing information, and utilizes a topographic model called the HAND (Hand-On-the-Right Distance) model to calculate flood map data for an evaluation target area (hereinafter referred to as a predetermined area) including rivers, as illustrated in Figure 28.

[0038] While details of HAND will be explained later, it is a topographic model calculated using the "Basic Map Information 5m Mesh Digital Elevation Model" (Geospatial Information Authority of Japan) to cover the entire country in approximately 5m mesh units. It is a widely used indicator for evaluating areas where water tends to accumulate locally (areas where water does not easily drain downstream).

[0039] An embodiment of this disclosure will be described below with reference to Figures 1 to 27. Note that the figures are examples, and this disclosure is not limited to those shown in the figures. For example, the system configuration of the illustrated information processing device is an example, and this disclosure is not limited thereto. Furthermore, in the description of the drawings, the same elements are denoted by the same reference numerals, and redundant explanations are omitted.

[0040] <<<Background Technology and Definitions of Terms>>> Figure 1 is a schematic diagram illustrating the challenges of creating flood maps using DEM (Digital Elevation Model). Assume that flood sensors S1 are installed in a predetermined area. In Figure 1, the installation locations of flood sensors S1 are shown as black circles, rivers as solid lines, and contour lines in the DEM as dashed lines. The right side of the figure is upstream, and the left side is downstream. Let's consider creating a flood map assuming that flood sensors S1 have detected flooding. Based on the DEM, since there is a ground slope in the direction of river flow, if we assume that the flooded surface is a still water surface with no water movement, the shaded area will be the flooded area. As an example, if flood sensors S1 are installed at a ground elevation of 103m, the flood depth near the coast will be an unrealistic value of 103m. On the other hand, if we assume that the flooded surface is a dynamic surface with moving water, it becomes necessary to solve partial differential equations such as the two-dimensional advection-diffusion equation, making it difficult to create flood maps at short time intervals, such as every 10 minutes.

[0041] Figure 2 is a schematic diagram illustrating the effect of creating flood maps using HAND. Assume that flood sensors S2 are installed in a predetermined area. In Figure 2, the installation locations of flood sensors S2 are shown as black circles, rivers as solid lines, and contour lines in the HAND model as dashed lines. The right side of the figure is upstream, and the left side is downstream. Let's consider creating a flood map assuming that flood sensors S2 have detected flooding. Based on the HAND model, the ground elevation represented by the HAND model at the river's location is 0m, and since there is a ground slope perpendicular to the river's flow direction, the shaded area represents the flooded area. Thus, because the ground slope in the river's flow direction is taken into account in the HAND model, an approximate solution to the flood map can be calculated without having to solve partial differential equations such as the two-dimensional advection-diffusion equation.

[0042] Dynamic sensing information Figure 3 is a schematic diagram illustrating how the measured sensor density calculated from dynamic sensing information changes over time. For example, information regarding the presence or absence of flooding detected by flooding sensors 10 installed on structures such as utility poles, information regarding the presence or absence of flooding determined from SNS images (still images) posted to SNS 11, information regarding the extent of flooding determined from aerial photographs (still images) acquired by an aircraft 12, and information regarding the extent of flooding determined from satellite images (SAR images) acquired by an artificial satellite 13 are all named "dynamic sensing information."

[0043] ≪Measured sensor density, virtual sensor density≫ In the evaluation area, the value calculated by (N1+N2+N3+N4) / A1 is named "measured sensor density". Here, N1 is the number of inundation sensors installed within the flood inundation area specified in the flood inundation area map in the given region, N2 is the number of posts on social media about inundation within the flood inundation area in the given region, N3 is the number of grid cells within the flood inundation area and within the range of aerial photography in the given region, N4 is the number of grid cells within the flood inundation area and within the observation range of satellite images in the given region, and A1 is the area of ​​the flood inundation area in the given region. In addition, the value calculated by N / A2 in the given region is named "virtual sensor density". Here, N is the number of virtual inundation sensors installed within the inundation area of ​​existing inundation information in the given region, and A2 is the area of ​​the inundation range of existing inundation information in the given region. The flood inundation area map will be described later.

[0044] ≪HAND Ground Elevation≫ The HAND model is a type of terrain model. The HAND model consists of grid cells arranged in rows and columns, and the value of each grid cell is named "HAND Ground Elevation." "Ground Elevation" represents the relative height of the target location from sea level. On the other hand, "HAND Ground Elevation" represents the relative elevation difference of the target location to the nearest drainage point, such as a river.

[0045] ≪Installation height of flood sensor≫ Unless otherwise specified, the term "flood sensor installation height" as used herein refers to the height from the ground to the sensor part of the flood sensor at the location where the flood sensor is installed.

[0046] <Hand water level, virtual hand water level> In the case of a flood sensor, the value obtained by adding "HAND Ground Height" and "Flood Sensor Installation Height" is named "HAND Water Level." In the case of SNS, the value of "HAND Ground Height" is named "HAND Water Level." In the case of aerial photographs, the value of "HAND Ground Height" is named "HAND Water Level." In the case of satellite images, the value of "HAND Ground Height" is named "HAND Water Level." "Water Level" represents the relative height of the target water surface from the sea level. On the other hand, "HAND Water Level" represents the relative height difference of the target water surface to the nearest drainage point, such as a river. In the case of a virtual flood sensor, the value of "HAND Ground Height" is named "Virtual HAND Water Level."

[0047] ≪Segment≫ A given area is divided into equally spaced sections perpendicular to the average direction of river flow (transverse direction), and these sections are named "segments." Furthermore, a given area is divided into equally spaced sections perpendicular to the direction of river flow (transverse direction), and then further divided longitudinally along the centerline of the river, and these sections are also named "segments."

[0048] ≪Measured HAND water level maximum value, virtual HAND water level maximum value≫ Assume that multiple flood sensors are installed in a certain segment. Assume that multiple of these flood sensors are detecting flooding. The maximum value of the actual HAND water level measured by the flood sensors that are detecting flooding will be named "Maximum Actual HAND Water Level". Also, assume that multiple virtual flood sensors are installed in a certain segment. Assume that multiple of these virtual flood sensors are detecting flooding. The maximum value of the virtual HAND water level measured by the virtual flood sensors that are detecting flooding will be named "Maximum Virtual HAND Water Level". The "Maximum Actual HAND Water Level" is not limited to flood sensors. The maximum value of the actual HAND water level calculated from flood sensors, SNS images (still images), aerial photographs (still images), and satellite images (SAR images) will also be named "Maximum Actual HAND Water Level".

[0049] ≪Interpolation function for creating flood maps, interpolation function for creating virtual flood maps≫ The interpolation function calculated using a radial basis function based on the measured maximum HAND water level for each segment is named the "interpolation function for flood diagram creation." Similarly, the interpolation function calculated using a radial basis function based on the virtual maximum HAND water level for each segment is named the "interpolation function for virtual flood diagram creation." Both the "interpolation function for flood diagram creation" and the "interpolation function for virtual flood diagram creation" are expressed as linear combinations of radial basis functions.

[0050] ≪Measured HAND water level for flood map creation, virtual HAND water level for flood map creation≫ In the image coordinate space (row number, column number) of the HAND model, the value calculated by the substitution formula from the right side to the left side of the equation "Measured HAND water level for flood map creation = interpolation function for flood map creation" is named "Measured HAND water level for flood map creation". In addition, in the image coordinate space (row number, column number) of the HAND model, the value calculated by the substitution formula from the right side to the left side of the equation "Virtual HAND water level for flood map creation = virtual interpolation function for flood map creation" is named "Virtual HAND water level for flood map creation".

[0051] ≪Flood Inundation Area Map≫ Unless otherwise indicated, the term "flood inundation area map" as used herein refers to flood inundation areas designated by the Ministry of Land, Infrastructure, Transport and Tourism and prefectural governments. A flood inundation area map designates areas where inundation is expected in the event of river flooding due to the maximum conceivable rainfall, and includes the expected area, expected water depth, and duration of inundation. In the context of this disclosure, the terms "inundation map," "flood inundation area map," or "inundation area map" may be used rather than "flood inundation area map." However, a person skilled in the art will understand that the terms "flood inundation map," "flood inundation area map," or "inundation area map" can be replaced with "flood inundation area map" where applicable.

[0052] ≪Existing Flood Information≫ Existing flood information may be the flood inundation area map of the maximum assumed scale published by the Geospatial Information Authority of Japan, the flood inundation area map of the planned scale published by the Geospatial Information Authority of Japan, the flood inundation map created from past flood records, or the flood inundation map calculated by solving partial differential equations such as the two-dimensional advection-diffusion equation.

[0053] ≪Average of Evaluation Metric Values≫ Unless otherwise specified, the term "mean evaluation index value" as used herein refers to TP (True Positive), FP (False Positive), TN (True Negative), FN (False Negative), precision, recall, F-score, and mean F-score. TP, FP, TN, FN, precision, recall, F-score, and mean F-score will be discussed later.

[0054] ≪Amount of water≫ Unless otherwise specified, "water volume" as used in this specification refers to the volume of water inundated by river flooding during a flood. Water volume is calculated as the sum of inundation depth (m) × grid cell width (m) × grid cell height (m).

[0055] ≪Water Inundation Sensor≫ Unless otherwise specified, the term "flood sensor" as used herein refers to a device installed on a structure such as a utility pole that, upon detecting flooding, transmits sensing information, such as the presence or absence of flooding, to a server device via a communication network. Known detection methods include radio wave type, pressure type, contact type, and float type. Known communication methods include using a mobile phone network and using low-power radio and relay devices that enable communication over short distances of several tens of meters. The main focus of one embodiment of this disclosure is an implementation method for creating a flood map based on sensing information detected by a sensor such as a flood sensor; therefore, a detailed explanation of the component configuration, operating principle, and function of the flood sensor is omitted.

[0056] <<<Publicly Known Technique: An Example of HAND Model Calculation>>> The HAND model can be calculated using publicly available computer programs based on DEM data published by the Geospatial Information Authority of Japan. One example of a publicly available computer program is TauDEM (Terrain Analysis Using Digital Elevation Model), which can be obtained from "https: / / hydrology.usu.edu / taudem / taudem5 / ". Here, an example of calculating the HAND model using TauDEM will be explained below, referring to the flowchart in Figure 4.

[0057] Based on user instructions, the process of removing depressions from the TauDEM (ST1) is initiated. Here, a depression is a grid cell that is surrounded by grid cells with higher ground elevations within the grid cells that make up the DEM data. In this step ST1, the server's computing unit reads the DEM data stored in the storage device. Next, the server's computing unit determines whether each grid cell is a depression or not based on the read DEM data. Next, for each grid cell, the server's computing unit increases the ground elevation of the grid cells determined to be depressions to remove them. Next, the server's computing unit stores the DEM data with the depressions removed as a GeoTIFF file in the storage device. After that, the process proceeds to step ST2.

[0058] In step ST2, the server's calculation unit performs the calculation of the eight flow directions included in the TauDEM. In step ST2, the server's calculation unit calculates the difference in ground elevation between each grid cell and the eight adjacent grid cells based on the DEM data with depressions removed calculated in step ST1. Next, the server's calculation unit determines the flow direction for each grid cell based on the difference in ground elevation between it and the eight adjacent grid cells, and stores a GeoTIFF file in the storage device with the grid cell values ​​set to East=1, Northeast=2, North=3, Northwest=4, West=5, Southwest=6, South=7, and Southeast=8. After that, the process proceeds to step ST3.

[0059] In step ST3, the server's calculation unit performs the calculation of the infinitely directional flow included in the TauDEM. In step ST3, based on the depression-removed DEM data calculated in step ST1, the server's calculation unit calculates the direction of the maximum gradient for each grid cell, not limited to the eight adjacent grid cells. Next, for each grid cell, the server's calculation unit stores a GeoTIFF file in the storage device, where the grid cell values ​​are consecutive floating-point values ​​from 0 to 2pi, starting from east and moving counterclockwise, based on the direction of the maximum gradient of the grid cell. After that, the process proceeds to step ST4.

[0060] In step ST4, the server's calculation unit performs the calculation of the upstream drainage area and flow path included in TauDEM. In step ST4, the server's calculation unit calculates the upstream drainage area for each grid cell based on the flow direction data for the eight directions calculated in step ST2. It also determines whether each grid cell is a flow path or not based on the flow direction data for the eight directions calculated in step ST2. Next, the server's calculation unit stores a GeoTIFF file containing the value of the upstream drainage area for the grid cells determined to be flow paths in the storage device. After that, the process proceeds to step ST5.

[0061] In step ST5, the server's calculation unit performs the calculation of rivers included in TauDEM. In this step ST5, the server's calculation unit determines whether each grid cell is a river or not, based on the upstream drainage area data and the flow path data calculated in step ST4. Here, grid cells whose upstream drainage area is above a predetermined threshold and which were determined to be a flow path in step ST4 are determined to be rivers. Next, the server's calculation unit stores a GeoTIFF file in the storage device, with the value of grid cells determined to be rivers set to "1" and the value of grid cells determined not to be rivers set to "0". After that, the process proceeds to step ST6.

[0062] In step ST6, the server's calculation unit performs the calculation of the relative height to the nearest drainage point included in the TauDEM. In step ST6, the server's calculation unit calculates the difference between the ground elevation of the nearest drainage point and the ground elevation of each grid cell for each grid cell, based on the DEM data with depressions removed calculated in step ST1, the infinite flow direction data calculated in step ST3, and the river data calculated in step ST5. Next, the server's calculation unit stores a GeoTIFF file containing the difference between the ground elevation of the nearest drainage point and the ground elevation of each grid cell in the storage device. After that, the calculation process for the HAND model is completed.

[0063] <<An example of the system configuration of Information Processing Device 1>> Figure 5 is a diagram showing an example of the system configuration of an information processing device according to one embodiment of the present disclosure. As shown in Figure 5, the information processing device 1 comprises a flood sensor 10, an SNS 11, an aircraft 12, an artificial satellite 13, a terminal device 20 used by disaster prevention officials of the national or local government or by residents, a server device 30 that receives information on the presence or absence of flooding detected by sensors such as the flood sensor and creates a flood map, and a communication network 90 that connects these in a communicative manner.

[0064] <<An example of the hardware configuration of terminal device 20 and server device 30>> Figure 6 shows an example of the hardware configuration of a terminal device 20 according to one embodiment of the present disclosure. The terminal device 20 is a computer such as a smartphone, tablet, or personal computer. The terminal device 20 is mainly composed of an arithmetic unit 2001, a memory 2002, a storage device 2003, a communication unit 2004, a display unit 2005, and an input unit 2006. The terminal device 20 may also include a web browser.

[0065] Figure 7 shows an example of the hardware configuration of a server device 30 according to one embodiment of the present disclosure. The server device 30 is, for example, a general-purpose server computer such as a rack-mount server, or a virtual server computer such as a cloud server. The server device 30 is mainly composed of an arithmetic unit 3001, a memory 3002, a storage device 3003, a communication unit 3004, a display unit 3005, and an input unit 3006. The server device 30 may also include a web server.

[0066] The arithmetic unit 2001 consists of a processor such as a CPU. The arithmetic unit 2001 receives response information from the communication unit 2004, such as HTML (Hyper Text Markup Language), CSS (Cascading Style Sheets), JavaScript (registered trademark) such as leaflet.js, and images of flood maps, and displays the HTML processed by the browser, images of flood maps, etc., on the display unit 2005.

[0067] Memory 2002 consists of storage devices such as RAM (Random Access Memory). Memory 2002 stores the OS, computer programs such as browsers, HTML, CSS, JavaScript (registered trademark) such as leaflet.js, and images of flood maps received from the server device 30. leaflet.js is JavaScript (registered trademark) for displaying maps in a web browser. leaflet.js is publicly available from "https: / / leafletjs.com / ". To display maps in a web browser, open-source JavaScript (registered trademark) such as Open Layers can also be used instead of leaflet.js. Open Layers is publicly available from "https: / / openlayers.org / ".

[0068] The storage device 2003 consists of a hard disk, an SSD (Solid State Drive), etc. The storage device 2003 stores computer programs such as the OS and browser, HTML, CSS, JavaScript (registered trademark) such as leaflet.js, and images of flood maps received from the server device 30.

[0069] The display unit 2005 consists of a liquid crystal display panel, an organic EL panel, etc. The display unit 2005 displays a browser, as well as HTML processed by the browser, images of flood maps, etc.

[0070] The input unit 2006 consists of a keyboard, mouse, etc. The input unit 2006 accepts user operations such as scrolling, swiping, clicking on zoom icons, zoom icons, etc., and keyboard input.

[0071] The communication unit 2004 receives response information such as HTML, CSS, JavaScript (registered trademark) including leaflet.js, and images of flood maps transmitted from the server device 30. The communication unit 2004 also transmits request information such as URLs to the server device 30 based on the control of the calculation unit 2001.

[0072] The arithmetic unit 3001 is composed of a processor such as a CPU. The processor executes computer programs stored in memory, etc., and the processor and memory work together to realize functions such as an OS, web server, database, and flood map creation.

[0073] Memory 3002 consists of storage devices such as RAM. Memory 3002 stores computer programs such as the OS, web server, and database, as well as computer programs for creating flood maps.

[0074] The storage device 3003 is composed of a hard disk, SSD, etc. The storage device 3003 stores computer programs such as the OS, web server, and database, a computer program for creating flood maps, a HAND model, a flood inundation area map, a flood map, a database data file, and files such as HTML, CSS, and JavaScript (registered trademark) such as leaflet.js. The flood map calculated by the flood map creation means 36 is stored in the directory of the flood map storage means 44 included in the storage device 3003, for example, " / 17 / 114460 / ", with a file name such as "52432.png". Here, "17", "114460", and "52432" are the values ​​of zoom level Z, tile coordinate X, and tile coordinate Y in the XYZ method. This XYZ method is a known technology used to efficiently distribute map information in communication between a web server and a web browser.

[0075] The communication unit 3004 receives request information such as a URL transmitted from the terminal device 20. The calculation unit 3001 receives the request information such as a URL from the communication unit 3004 and, according to the URL etc. contained in the request information, reads a file such as "52432.png" from a directory of the flood diagram storage means 44 contained in the storage device 3003, for example, " / 17 / 114460 / ", and transmits response information including an image of the flood diagram to the terminal device 20 via the communication unit 3004. Here, "17", "114460", and "52432" are the values ​​of zoom level Z, tile coordinate X, and tile coordinate Y in the XYZ system.

[0076] The display unit 3005 is composed of a liquid crystal display panel, etc.

[0077] The input unit 3006 consists of a keyboard, mouse, etc. The input unit 3006 accepts scrolling operations, icon click operations, and key input operations from the user.

[0078] <<An example of the functional configuration of server device 30>> As shown in Figure 8, the server device 30 includes a sensing information reception discrimination means 32, a common processing means 34, a flood diagram creation means 36, a virtual flood diagram creation means 38, a dynamic sensing information storage means 40, an evaluation index value storage means 42, a flood diagram storage means 44, and a flood diagram transmission control means 46.

[0079] As shown in Figure 8, the sensing information reception discrimination means 32 includes a flood sensor reception control means 320, an SNS discrimination means 322, an aerial photograph discrimination means 324, and a satellite image discrimination means 326.

[0080] As shown in Figure 8, the common processing means 34 includes a HAND model acquisition means 340.

[0081] As shown in Figure 8, the flood diagram creation means 36 includes a dynamic sensing information acquisition means 360, a measured HAND water level calculation means 362, a flood diagram creation segment discrimination means 364, a measured HAND water level maximum value calculation means 366, a flood diagram creation interpolation function calculation means 368, and a flood diagram calculation means 370.

[0082] As shown in Figure 8, the flood map creation means 36 may also be configured to include, in addition to the dynamic sensing information acquisition means 360, the measured HAND water level calculation means 362, the flood map creation segment discrimination means 364, the measured HAND water level maximum value calculation means 366, the flood map creation interpolation function calculation means 368, and the flood map calculation means 370, a flood map correction means 372 and an anomaly value detection means 374. By including the flood map correction means 372, the accuracy of the flood map can be further improved. Furthermore, by including the anomaly value detection means 374, the accuracy of the flood map can be further improved.

[0083] As shown in Figure 8, the virtual flood map creation means 38 includes an existing flood information acquisition means 380, a virtual flood map creation condition setting means 382, ​​a virtual dynamic sensing information generation means 384, a virtual HAND water level calculation means 386, a virtual HAND water level maximum value calculation means 388, an interpolation function calculation means for virtual flood map creation 390, a virtual flood map calculation means 392, and an evaluation index value average calculation means 394.

[0084] The sensing information reception discrimination means 32, common processing means 34, flood diagram creation means 36, virtual flood diagram creation means 38, dynamic sensing information storage means 40, evaluation index value storage means 42, flood diagram storage means 44, flood diagram transmission control means 46, and each of the elements included therein are realized through the collaborative execution of a program by the calculation unit 3001, memory 3002, storage device 3003, and communication unit 3004.

[0085] <<An example of detailed functions of server device 30>> <<Water immersion sensor reception control means 320 of sensing information reception discrimination means 32>> The flood sensor receiving control means 320 has the function of receiving information regarding the presence or absence of flooding from the flood sensor 10 via the communication network 90. ​​It also has the function of storing the received information regarding the presence or absence of flooding in the dynamic sensing information storage means 40. Here, the information regarding the presence or absence of flooding includes the flood sensor ID that identifies the flood sensor, the observation date and time, the location of the flood sensor (latitude, longitude), the height at which the flood sensor is installed, the height of the HAND ground, and whether or not there is flooding.

[0086] ≪SNS discrimination means 322 of sensing information reception discrimination means 32≫ The SNS identification means 322 has the function of receiving SNS information from server devices such as cloud servers managed by organizations operating SNS via the communication network 90. ​​It also has the function of determining whether or not there is flooding from the received SNS information. Furthermore, it has the function of storing the determined information regarding the presence or absence of flooding in the dynamic sensing information storage means 40. Here, SNS information refers to the SNS ID that identifies the SNS post, the posting date and time, the posting location (latitude, longitude), and image information (still image) attached to the SNS post. The information regarding the presence or absence of flooding refers to the SNS ID that identifies the SNS post, the posting date and time, the posting location (latitude, longitude), image information (still image) attached to the SNS post, the HAND ground level, and whether or not there is flooding.

[0087] A publicly known method for determining whether or not there is flooding from SNS information is semantic segmentation technology using neural networks such as U-NET. The neural network is trained as follows: Image information (still images) attached to SNS posts and information about whether or not there is flooding created by humans are used as training data. The image information (still images) attached to SNS posts are used as input data, and the sum of the squared errors is calculated, for example, between the output data of the neural network and the information about whether or not there is flooding created by humans. Based on the sum of the squared errors, the weights and bias values ​​of the neural network are optimized using a method known as backpropagation. Whether or not there is flooding is determined as follows: Image information (still images) attached to SNS posts are input to the trained neural network. The data output from the trained neural network is used as the determination of whether or not there is flooding.

[0088] As described above, by including information regarding the presence or absence of flooding, which can be determined from SNS posts, in the dynamic sensing information, the actual sensor density increases. This has the effect of enabling information processing equipment that employs flooding sensors and SNS images (still images) to calculate highly accurate flooding map data.

[0089] <Aerial photograph discrimination means 324 of sensing information reception discrimination means 32> The aerial photograph discrimination means 324 has the function of receiving aerial photograph information from a server device, such as a cloud server managed by the organization that takes aerial photographs, via the communication network 90. ​​It also has the function of discriminating information regarding the flooded area from the received aerial photograph information. Furthermore, it has the function of storing the information regarding the flooded area that has been discriminated in the dynamic sensing information storage means 40. Here, aerial photograph information refers to the aerial photograph ID that identifies the aerial photograph, the date and time of shooting, the shooting location (latitude, longitude), and the image information (still image) that is captured. The information regarding the flooded area refers to the identification ID that identifies the polygon of the flooded area, the aerial photograph ID that identifies the aerial photograph, the date and time of shooting, the shooting location (latitude, longitude), the image information (still image) that is captured, and the flooded area (latitude, longitude of the vertices of the outer boundary).

[0090] A publicly known method for determining information about flood areas from aerial photographs is semantic segmentation, which uses neural networks such as U-NET. The neural network is trained as follows: Aerial photographs taken during flooding and human-created flood areas (with predetermined values ​​inside, for example, polygons filled with 1) are used as training data. The aerial photographs taken during flooding are used as input data, and the sum of the squared errors is calculated between the output data of the neural network and the human-created flood areas (with predetermined values ​​inside, for example, polygons filled with 1). Based on the sum of squared errors, the weights and biases of the neural network are optimized using a method known as backpropagation. The flood areas (with predetermined values ​​inside, for example, polygons filled with 1) are determined as follows: Aerial photographs taken during flooding are input to the trained neural network. The data output from the trained neural network is used to determine flood area 1 (with predetermined values ​​inside, for example, polygons filled with 1).

[0091] Next, flooded area 2 (a polygon defined by the latitude and longitude of the vertices of the outer boundary) is calculated from flooded area 1 (a polygon whose interior is filled with a predetermined value, for example, a polygon filled with 1), and this flooded area 2 (a polygon defined by the latitude and longitude of the vertices of the outer boundary) is used as information about the flooded area. Here, edge detection techniques are publicly known as methods for calculating polygons defined by the vertices of the outer boundary from a polygon whose interior is filled with a predetermined value (for example, an image composed of pixels). To calculate flooded area 2 (a polygon defined by the latitude and longitude of the vertices of the outer boundary) from flooded area 1 (a polygon whose interior is filled with a predetermined value, for example, a polygon filled with 1), a computer program such as OpenCV can also be used. OpenCV is publicly available from "https: / / opencv.org / ".

[0092] As described above, by including information on the flooded area determined from aerial photographs (still images) in the dynamic sensing information, the actual sensor density increases, resulting in the effect of being able to calculate highly accurate flood map data even in an information processing device that employs flood sensors, SNS images (still images), and aerial photographs (still images).

[0093] <<Satellite image discrimination means 326 of sensing information reception discrimination means 32>> The satellite image discrimination means 326 has the function of receiving satellite image information from a server device, such as a cloud server managed by an organization operating artificial satellites, via the communication network 90. ​​It also has the function of discriminating information regarding the flooded area from the received satellite image information. Furthermore, it has the function of storing the discriminated information regarding the flooded area in the dynamic sensing information storage means 40. Here, satellite image information refers to the satellite image ID that identifies the satellite image, the date and time of observation, the observation location (latitude, longitude), and the observed satellite image (SAR image). The information regarding the flooded area refers to the identification ID that identifies the polygon of the flooded area, the satellite image ID that identifies the satellite image, the date and time of observation, the observation location (latitude, longitude), the observed satellite image (SAR image), and the flooded area (latitude, longitude of the vertices of the outer boundary).

[0094] As a method for determining information about the extent of flooding from satellite imagery, semantic segmentation techniques using neural networks such as U-NET are publicly known. The neural network is trained as follows: Satellite images (SAR images) observed during flooding and human-created flood areas (polygons filled with predetermined values, for example, 1s) are used as training data. Using the satellite images (SAR images) observed during flooding as input data, the sum of the squared errors is calculated, for example, between the output data of the neural network and the human-created flood areas (polygons filled with predetermined values, for example, 1s). Based on the sum of squared errors, the weights and biases of the neural network are optimized using a method known as backpropagation. The flood areas (polygons filled with predetermined values, for example, 1s) are determined as follows: Satellite images (SAR images) observed during flooding are input into the trained neural network. The data output from the trained neural network is classified as a flooded area 1 (with a predetermined value inside, for example, a polygon filled with 1).

[0095] Next, flooded area 2 (a polygon defined by the latitude and longitude of the vertices of the outer edge) is calculated from flooded area 1 (a polygon whose interior is filled with a predetermined value, for example, a polygon filled with 1), and this flooded area 2 (a polygon defined by the latitude and longitude of the vertices of the outer edge) is used as information about the flooded area. Here, edge detection techniques are publicly known as methods for calculating polygons defined by the vertices of the outer edge from a polygon whose interior is filled with a predetermined value (for example, an image composed of pixels). Computer programs such as OpenCV can also be used to calculate flooded area 2 (a polygon defined by the latitude and longitude of the vertices of the outer edge) from flooded area 1 (a polygon whose interior is filled with a predetermined value, for example, a polygon filled with 1).

[0096] As described above, by including information on the flooded area determined from satellite images (SAR images) in the dynamic sensing information, the density of measured sensors increases, which has the effect of enabling the calculation of highly accurate flood map data even in information processing devices that employ flood sensors, SNS images (still images), aerial photographs (still images), and satellite images (SAR images).

[0097] ≪Common processing means 34 HAND model acquisition means 340≫ [HAND model acquisition function, memory saving function] As shown in Figure 9, the HAND model acquisition means 340 has the function of acquiring the HAND model stored in the storage device 3003 and the river centerline stored in the storage device 3003. It also has the function of storing the acquired HAND model and river centerline in the memory 3002. Here, the HAND model is raster data composed of grid cells arranged in a grid of rows and columns.

[0098] The HAND model may be stored in storage device 3003 as a GeoTIFF file, or metadata such as the Coordinate Reference System (CRS) and Bounding Box, along with the HAND ground elevation of the grid cells, may be stored based on a predetermined database management method. Here, the Coordinate Reference System is, for example, EPSG:6668-JGD2011, which represents a method for projecting positions on the spherical Earth onto a plane. The Bounding Box refers to the longitude of the western end, eastern end, northern end, and southern end of the GeoTIFF file. A computer program such as rasterio can be used to read GeoTIFF files. rasterio is publicly available from "https: / / rasterio.readthedocs.io / en / stable / ".

[0099] [Image coordinate space rotation function] As shown in Figure 10, the HAND model acquisition means 340 may also have a function to rotate the image coordinate space (row number, column number) of the HAND model. As shown in Figure 11, when the direction of flow of the river is inclined from the north-south or east-west direction, or when the river meanders and the average direction of flow is inclined from the north-south or east-west direction, it is preferable to rotate the image coordinate space (row number, column number) of the HAND model. In the following description, it will be assumed that by rotating the image coordinate space (row number, column number) of the HAND model, the X-axis direction of the HAND model coincides with the average direction of flow of the river. Also, it will be assumed that the Y-axis direction of the HAND model coincides with the direction perpendicular to the average direction of flow of the river.

[0100] The following describes the details of the function for rotating the image coordinate space. When rotating the image coordinate space (row number, column number) clockwise, the row number after rotation is calculated by rounding the result of (row number before rotation × cos(rotation angle) - column number before rotation × sin(rotation angle)) to the nearest integer. The column number after rotation is calculated by rounding the result of (row number before rotation × sin(rotation angle) + column number before rotation × cos(rotation angle)) to the nearest integer. Here, the rotation angle is assumed to be a positive real number. If the rotation of the image coordinate space (row number, column number) of the HAND model results in grid cells where no HAND ground elevation exists, known techniques such as nearest neighbor interpolation, bilinear interpolation, and bicubic interpolation are used for interpolation.

[0101] <<Dynamic sensing information acquisition means 360 for flood map creation means 36>> The dynamic sensing information acquisition means 360 has the function of acquiring dynamic sensing information stored in the dynamic sensing information storage means 40. It also has the function of storing the acquired dynamic sensing information in the memory 3002. Furthermore, it has the function of calculating the measured sensor density based on the dynamic sensing information and storing the measured sensor density in the memory 3002. The dynamic sensing information may be stored in the dynamic sensing information storage means 40 based on a predetermined database management method, or it may be stored in the storage device 3003 as a comma-separated file.

[0102] Figure 12 shows an example of a database table when the dynamic sensing information is related to flood sensors. Figure 13 shows an example of a database table when the dynamic sensing information is related to social networking services (SNS). Figure 14 shows an example of a database table when the dynamic sensing information is related to aerial photographs. Figure 15 shows an example of a database table when the dynamic sensing information is related to satellite images.

[0103] The following describes the details of the function for calculating the measured sensor density. The measured sensor density is calculated using the following equation (1). JPEG2026046222000002.jpg12170

[0104] Here, D is the measured sensor density (number of sensors / m²). 2 N1 is the number of flood sensors installed within the flood-prone area in a given region. N2 is the number of posts on social media about flooding within the flood-prone area in a given region. N3 is the number of grid cells within the flood-prone area and within the range of aerial photography in a given region. N4 is the number of grid cells within the flood-prone area and within the observation range of satellite images in a given region. A is the area (m²) of the flood-prone area in a given region. 2 )

[0105] <<Measurement HAND water level calculation means 362 of flood diagram creation means 36>> The measured HAND water level calculation means 362 has the function of calculating the measured HAND water level based on the HAND model stored in memory 3002 and the dynamic sensing information stored in memory 3002. It also has the function of storing the calculated measured HAND water level in memory 3002. The details of the function for calculating the measured HAND water level are described below.

[0106] 1) When the dynamic sensing information is related to the water ingress sensor. The measured HAND water level (i, j) is obtained by adding the height of the flood sensor installation to the HAND ground level (i, j) at the flood sensor installation location. Here, i and j are the row and column numbers (row and column numbers) in the image coordinate space (row and column numbers) of the HAND model. The HAND ground level (i, j) is the value of a grid cell in the HAND model, and its value is stored in the image coordinate space (row and column numbers). Furthermore, since the value of the flood sensor installation location is stored in geographic coordinate space (latitude and longitude), the latitude and longitude of the flood sensor installation location are converted to row number i and column number j in the image coordinate space (row and column numbers) to calculate the measured HAND water level (i, j).

[0107] 2) When the dynamic sensing information is related to social networking services. The HAND ground elevation (i, j) of the SNS posting location is taken as the measured HAND water level elevation (i, j). Here, since the SNS posting location is stored in geographic coordinate space (latitude, longitude), the latitude and longitude of the SNS posting location are converted to row number i and column number j in the image coordinate space (row number, column number) to calculate the measured HAND water level elevation (i, j).

[0108] 3) When the dynamic sensing information is related to aerial photographs. The HAND ground elevation (i, j) within the flooded area identified from aerial photographs is defined as the measured HAND water level elevation (i, j). Here, since the presence or absence of flooding is stored in geographic coordinate space (latitude, longitude) as identified from aerial photographs, the latitude and longitude within the flooded area are converted to row number i and column number j in the image coordinate space (row number, column number) to calculate the measured HAND water level elevation (i, j).

[0109] 4) When the dynamic sensing information is related to satellite imagery. The HAND ground elevation (i, j) within the flooded area determined from satellite imagery is defined as the measured HAND water level elevation (i, j). Here, since the flooded area determined from satellite imagery is stored in geographic coordinate space (latitude, longitude), the latitude and longitude within the flooded area are converted to row number i and column number j in the image coordinate space (row number, column number) to calculate the measured HAND water level elevation (i, j).

[0110] The processes described in 1) to 4) above are repeated for each item in the list of dynamic sensing information to calculate the measured HAND water level (i, j). Here, the reason why the HAND ground level (i, j) in 2) to 4) is used as the measured HAND water level (i, j) is that, generally, the presence or absence of flooding determined from image information (still images) posted on social media, the extent of flooding determined from aerial photographs (still images), and the extent of flooding determined from satellite images (SAR images) do not include information about the depth of flooding.

[0111] Furthermore, the conversion from latitude and longitude in geographic coordinate space to row number i and column number j in image coordinate space is calculated based on the longitude of the western end, the longitude of the eastern end, the latitude of the northern end, the latitude of the southern end, the number of grid cells in the width direction and height direction of the HAND model, the longitude per grid cell width, and the latitude per grid cell height. Row number i and column number j, with the northwest corner (upper left corner) as the origin, are calculated by the following equation (2). JPEG2026046222000003.jpg22170

[0112] Here, i and j are the row number, column number, and lat in the image coordinate space (row number, column number) of the HAND model. north This is the latitude of the northernmost point, lon west is the longitude of the westernmost point, lat is the latitude, lon is the longitude, lat cell This is latitude per grid cell height, lon cellThis represents the longitude per grid cell width. Also, since the right-hand side of equation (2) is a real value, it is rounded to the nearest integer to calculate the row number i and column number j. Here, the latitude per grid cell height is calculated using the formula (latitude of the northernmost point - latitude of the southernmost point) / (number of grid cells in the height direction) from the latitude of the northernmost point, the latitude of the southernmost point, and the number of grid cells in the height direction. Similarly, the longitude per grid cell width is calculated using the formula (longitude of the easternmost point - longitude of the westernmost point) / (number of grid cells in the width direction) from the longitude of the westernmost point, the longitude of the easternmost point, and the number of grid cells in the width direction.

[0113] As described above, since the measured HAND water level is calculated using the HAND ground elevation, the ground gradient in the direction of river flow is removed. This has the effect of allowing comparison of the relative magnitudes of the measured HAND water levels even when one or more flood sensors are installed in a given area.

[0114] <<Flood Map Creation Means 364, Segment Discrimination Means for Flood Map Creation of Flood Map Creation Means 36>> The flood map creation segment determination means 364 has the function of determining the number of segments to divide into for flood map creation based on the measured sensor density stored in the memory 3002 and the average evaluation index value stored in the evaluation index value storage means 42. It also has the function of determining the necessity of dividing the area into segments longitudinally along the river's centerline based on the average evaluation index value. Furthermore, it has the function of calculating segments obtained by dividing a predetermined area perpendicular to the average flow direction of the river (cross-sectional direction) based on the number of segments to divide into for flood map creation. In addition to this, if necessary, it has the function of calculating segments obtained by further dividing the area longitudinally along the river's centerline based on the necessity of dividing the area into segments longitudinally along the river's centerline.

[0115] [Function to determine the number of segments to divide into for flood inundation map creation] The following describes in detail the function for determining the number of segments to divide into for creating the flood map. 1) Assuming there are N average values ​​for evaluation metrics, (X1, Y1, Z1), ..., (XN , Y N , Z N ) is represented by a sequence of points. Here, X1, …, X N is the virtual sensor density, Y1, …, Y N is the number of segment divisions for creating the virtual flooding map, Z1, …, Z N is the average F value.

[0116] 2) Next, using the radial basis function, a smooth interpolation function passing through the vicinity of each point of (X1, Y1, Z1), …, (X N , Y N , Z N ) is obtained. As shown in the following equation (3), the interpolation function is approximated by a weighted linear combination of radial basis functions. JPEG2026046222000004.jpg15170

[0117] Here, z(x) is the interpolation function, λ j is the weight of the radial basis function centered at the point x j , φ is the radial basis function, ||x - x j || is the distance from the point x to the point x j . Also, x, x j are vector representations of the position coordinates, and in the Cartesian coordinate system (orthogonal coordinate system, Cartesian coordinate system), they are (X, Y), (X j , Y j ), respectively. Also, well-known functions such as the Gaussian function, multiquadric function, and inverse multiquadric function are used as the radial basis function. In one embodiment of the present disclosure, as shown in the following equation (4), a two-dimensional Gaussian function is used, but other two-dimensional radial basis functions may also be used. JPEG2026046222000005.jpg1 / 5170

[0118] Here, ε is a parameter related to the shape of the radial basis function. As the value increases, the shape becomes sharp, and as the value decreases, the shape becomes smooth. Assuming that the values of the point sequence (X1, Y1, Z1), …, (X N , Y N , Z N ) are given, for each point (X i , Yi ) from other points (X j , Y j Next, as shown in Figure 16, calculate the value of the radial basis function for the distance to point x. i The value of the interpolation function at is approximated by equation (5) below. JPEG2026046222000006.jpg16170

[0119] Next, based on equation (5), point x1, point x2, ..., point x N The approximate value of the interpolation function at x is approximated by the following equation (6). Equation (6) below is given for points x1, x2, ..., x N This describes the approximate values ​​of the interpolation function in a given context using matrices and vectors. JPEG2026046222000007.jpg26170

[0120] The following equation (7) is equation (6) written in terms of vector Z, matrix Φ, and vector Λ. JPEG2026046222000008.jpg12170

[0121] Next, the matrix Φ is decomposed into an N x N orthogonal matrix Q and an N x N upper triangular matrix R using a known technique called QR decomposition. Equation (7) shows the inverse matrix Φ from the left. -1 Multiplying by and swapping the left and right sides, the weight vector Λ of the weighted linear combination of radial basis functions is expressed by equation (8) below. Here, due to the symmetry of the radial basis functions, the matrix Φ is a real symmetric matrix. A real symmetric matrix always has the characteristic of being QR decomposable into an N x N orthogonal matrix Q and an N x N upper triangular matrix R. Furthermore, an N x N upper triangular matrix R has the characteristic that its inverse matrix can be calculated if all diagonal elements are non-zero. JPEG2026046222000009.jpg15170

[0122] Here, Q T Q is the transpose of the N x N orthogonal matrix Q. In equation (8), Q is the inverse matrix of the N x N orthogonal matrix Q. -1 This utilizes the characteristic that it is the transpose matrix of Q.

[0123] 3) Next, substitute the measured sensor density into the variable X of the interpolation function 1 calculated in 2) (equation (3) above) to calculate interpolation function 2, where the number of segment divisions for creating the virtual flood map is the variable Y and the average F value is the variable Z.

[0124] 4) Next, using the interpolation function 2 calculated in 3), the number of segments for creating the virtual flood map that maximizes the average F value is calculated, and this value is used as the number of segments for creating the flood map. Here, the interpolation function using radial basis functions can also be calculated using a computer program such as scikit-learn. scikit-learn is publicly available from "https: / / scikit-learn.org / stable / ".

[0125] In calculating the interpolation function, the position coordinates (X1, Y1), ..., (X N , Y N Although the example uses radial basis functions centered at the nodes of the grid, the calculation of the interpolation function is not limited to this. The interpolation function can also be found using the least squares method with radial basis functions centered at the nodes of the grid, or using Gaussian process regression, or using gradient descent with a radial basis function network.

[0126] [A function to determine the necessity of dividing the river into segments in the longitudinal direction along its centerline.] The following describes in detail the function that determines the necessity of dividing the river into segments in the longitudinal direction along its centerline. F divide ≥F notdivide In this case, the segment is identified by dividing it longitudinally along the centerline of the river. Also, F divide <F notdivide In this case, it is determined that the river is not divided into segments in the longitudinal direction along the centerline. Here, F divide This refers to the average F value when the river is divided into segments longitudinally along its centerline, based on the measured sensor density and the number of segments used for creating flood maps. notdivideThis refers to the average F-value when the measured sensor density and the number of segments for creating flood maps are not divided into segments in the longitudinal direction along the centerline of the river.

[0127] [Cross-sectional division function] The following describes in detail the function of dividing a given area into equal intervals perpendicular to the average flow direction of the river (cross-sectional direction). As shown in Figure 17, the column numbers of the points where the lines dividing the given area perpendicular to the average flow direction of the river (cross-sectional direction) intersect the X-axis are J1, J2, ..., J N-1 Let's assume J1, J2, ..., J N-1 These are calculated by rounding (X2-X1) / N, 2×(X2-X1) / N, ..., and (N-1)×(X2-X1) / N to the nearest integer. Here, X1 is the minimum column number of the HAND model, X2 is the maximum column number of the HAND model, and N is the number of segments for creating the flood map. In this case, segment 1 will consist of grid cells with column numbers greater than or equal to X1 and less than or equal to J1. Similarly, segment 2 will consist of grid cells with column numbers greater than J1 and less than or equal to J2. Similarly, segment N will consist of grid cells with column numbers greater than J1 and less than or equal to J2. N-1 It will be larger and consist of grid cells up to X2 in size.

[0128] [Vertical and Cross-sectional Segmentation Function] The following describes in detail the function of dividing a given area into equal intervals perpendicular to the average flow direction of the river (cross-sectional direction), and further dividing it longitudinally along the river's centerline. As shown in Figure 18, the column numbers of the points where the lines dividing the given area perpendicular to the average flow direction of the river (cross-sectional direction) intersect the X-axis are J1, J2, ..., J N-1 Let's assume J1, J2, ..., J N-1These are calculated by rounding (X2-X1) / N, 2×(X2-X1) / N, ..., and (N-1)×(X2-X1) / N to the nearest integer. Here, X1 is the minimum column number of the HAND model, X2 is the maximum column number of the HAND model, and N is the number of segments for creating the flood map. Next, the map is divided longitudinally along the centerline of the river. In this case, segment 1 will consist of grid cells with column numbers between X1 and J1, and row numbers above the centerline of the river. Similarly, segment N will consist of grid cells with column numbers between J N-1 Larger than X2, the row numbers will consist of grid cells above the river's centerline. Similarly, segment N+1 will consist of grid cells below the river's centerline, with column numbers between X1 and J1. Similarly, segment 2N will consist of grid cells below the river's centerline, with column numbers between J N-1 For larger sizes, up to X2, the row numbers will consist of grid cells below the centerline of the river.

[0129] As described above, the optimal number of segment divisions is automatically determined by using the measured sensor density data and the average evaluation index data to calculate the number of segment divisions that maximizes the average F value. Furthermore, by using the average evaluation index data to determine the necessity of dividing a given area longitudinally along the river's centerline, it is possible to prevent unnecessary segment divisions that do not contribute to improving accuracy, thus preventing an unnecessary decrease in the number of flood sensors per segment.

[0130] <<Method 366 for calculating the maximum measured water level using the flood diagram creation method 36>> The measured HAND water level maximum value calculation means 366 has the function of calculating the measured HAND water level maximum value based on the segments stored in the memory 3002 and the measured HAND water level stored in the memory 3002. Furthermore, it has the function of storing the calculated measured HAND water level maximum value in the memory 3002.

[0131] [HAND measured water level maximum value calculation function (when divided in the transverse direction)] The following describes in detail the function for calculating the maximum measured HAND water level when the river is segmented perpendicular to the average flow direction (transverse direction). As an example, let's assume that flood sensors 1 to 5 are installed in segment 1. The row and column numbers in the image coordinate space (row number, column number) for the installation locations of flood sensors 1 to 5 are (1, 1), (1, 5), (1, 10), (3, 5), and (3, 10). Here, the number on the left in parentheses is the row number, and the number on the right in parentheses is the column number. Also, let's assume that the measured HAND water levels for flood sensors 1 to 5 are 1.0m, 1.5m, 2.0m, 2.5m, and 3.0m, respectively. In this case, the maximum measured HAND water level in segment 1 is 3.0m, and the row and column numbers in its image coordinate space (row number, column number) are (3, 10). Furthermore, if a water inundation sensor is not installed in a segment, the maximum measured HAND water level in that segment will not be calculated. Next, for segments 2 through N, the maximum measured HAND water level and its corresponding row and column numbers in the image coordinate space (row number, column number) will be calculated in the same manner.

[0132] [HAND measured water level maximum value calculation function (when divided in both the transverse and longitudinal directions)] The following describes in detail the function for calculating the maximum measured HAND water level when segmenting the river longitudinally along the river's centerline, in addition to segmenting perpendicular to the average flow direction of the river. As an example, let's assume that flood sensors 1-3 are installed in segment 1. The row and column numbers in the image coordinate space (row number, column number) for the installation locations of flood sensors 1-3 are (1, 1), (1, 5), and (1, 10). Here, the number on the left in parentheses is the row number, and the number on the right in parentheses is the column number. Also, let's assume that the measured HAND water levels for flood sensors 1-3 are 1.0m, 1.5m, and 2.0m, respectively. In this case, the maximum measured HAND water level in segment 1 is 2.0m, and the row and column numbers in its image coordinate space (row number, column number) are (1, 10). Furthermore, if a water inundation sensor is not installed in a segment, the maximum measured HAND water level in that segment will not be calculated. Next, for segments 2 to 2N, the maximum measured HAND water level and its corresponding row and column numbers in the image coordinate space (row number, column number) will be calculated in the same manner.

[0133] <<Inundation diagram creation means 368, interpolation function calculation means for inundation diagram creation means 36>> The flood map creation interpolation function calculation means 368 has the function of calculating a flood map creation interpolation function based on the measured maximum HAND water level stored in memory 3002, and the row number and column number in the image coordinate space (row number, column number). It also has the function of storing the calculated flood map creation interpolation function in memory 3002.

[0134] [Interpolation function calculation function for creating flood maps (when divided in the cross-sectional direction)] The following describes the details of the function for calculating the interpolation function for creating flood maps. Here, we assume that a given area is divided into segments perpendicular to the average flow direction of the river (transverse direction). Furthermore, we assume that there are N segments, and that the maximum measured HAND water level is calculated for each segment.

[0135] 1) The row number, column number, and the measured maximum HAND water level in the image coordinate space (row number, column number) are (X1, Y1, M1), ..., (X N , Y N M N Let it be represented by a sequence of points X1, ..., X N This represents the column number, Y1, ..., Y in the image coordinate space (row number, column number) of the HAND model. N This represents the row number, M1, ..., M in the image coordinate space (row number, column number) of the HAND model. N This represents the maximum measured HAND water level in each segment.

[0136] 2) Next, as shown in Figure 19, (X1, Y1, M1), ..., (X N , Y N M N The sequence of points ) is projected onto the projection plane for creating the flood map, (X1, M1), ..., (X N M N The sequence of points in the image is calculated. Here, the projection plane for creating the flood map is a plane where the column number in the image coordinate space (row number, column number) of the HAND model is on the horizontal axis and the maximum measured HAND water level is on the vertical axis.

[0137] 3) Next, using radial basis functions, (X1, M1), ..., (X N M N An interpolation function for creating a smooth flood map is obtained that passes through the neighborhood of each point in the map. The calculation of the interpolation function using the radial basis function is performed using the method described in the flood map creation segment discrimination means 364.

[0138] [Interpolation function calculation function for creating flood maps (when the map is divided in both the transverse and longitudinal directions)] The function of the interpolation function calculation means 368 for creating flood maps to calculate the interpolation function for creating flood maps may be such that, in addition to dividing the river into segments perpendicular to the average flow direction (transverse direction), the river is divided into segments longitudinally along the centerline of the river. Here, it is assumed that there are 2N segments, and that the maximum measured HAND water level is calculated for each segment.

[0139] 1) In the image coordinate space (row number, column number) of the maximum measured HAND water surface height, the row number, the column number, and the maximum measured HAND water surface height are (X1, Y1, M1), …, (X N , Y N , M N ), (X N+1 , Y N+1 , M N+1 ), …, (X 2N , Y 2N , M 2N ), and are represented by a point sequence. Here, X1, …, X N are the column numbers in the image coordinate space (row number, column number) of the HAND model above the center line of the river, and X N+1 , …, X 2N are the column numbers in the image coordinate space (row number, column number) of the HAND model below the center line of the river. Y1, …, Y N are the row numbers in the image coordinate space (row number, column number) of the HAND model above the center line of the river, and Y N+1 , …, Y 2N are the row numbers in the image coordinate space (row number, column number) of the HAND model below the center line of the river. M1, …, M N are the maximum measured HAND water surface heights in each segment above the center line of the river, and M N+1 , …, M 2N are the maximum measured HAND water surface heights in each segment below the center line of the river.

[0140] 2) Next, using the radial basis function, a smooth interpolation function passing through the vicinity of each point of (X1, Y1, M1), …, (X N , Y N , M N ), (X N+1 , Y N+1 , M N+1 ), …, (X 2N , Y 2N , M 2N ) is obtained. The calculation of the interpolation function using the radial basis function uses the method described in the segment discrimination means 364 for flood inundation map creation.

[0141] As described above, in each segment, the maximum value of the measured HAND water level is calculated, and an interpolation function for creating the flood map is calculated using an interpolation technique based on a radial basis function. This has the effect of preventing the water level and flood depth from being underestimated near flood sensors where the installation height of the flood sensors is small.

[0142] ≪Flood map calculation means 370 of flood map creation means 36≫ As shown in Figure 20, the flood map calculation means 370 has the function of calculating a flood map based on an interpolation function for creating a flood map stored in memory 3002 and a HAND model stored in memory 3002. It also has the function of storing the calculated flood map in the flood map storage means 44. Furthermore, it has the function of calculating a tile image suitable for use in a web application based on the calculated flood map and storing it in the storage device 3003. In addition to these, as shown in Figure 21, the flood map calculation means 370 may also have the function of rotating the image coordinate space (row number, column number) of the flood map, the function of calculating the amount of water inundated based on the calculated flood map, and the function of masking unnecessary areas including river channels from the calculated flood map based on a mask diagram stored in the storage device 3003.

[0143] [Flood Inundation Map Calculation Function (when divided in the cross-sectional direction)] The following describes the details of the function for calculating the flood map. Here, we assume that a given area is divided into segments perpendicular to the average flow direction of the river (cross-sectional direction). In this case, as shown in Figure 22, the interpolation function for creating the flood map is defined on a projection plane for creating the flood map, where the column number in the image coordinate space (row number, column number) of the HAND model is on the horizontal axis and the maximum measured HAND water level is on the vertical axis.

[0144] 1) Based on the interpolation function (j) for creating the flood map, the measured HAND water level (i, j) for creating the flood map is calculated. The measured HAND water level (i, j) for creating the flood map is calculated using the substitution equation from the right side to the left side of the equation: measured HAND water level (i, j) = interpolation function (j) for creating the flood map. Here, i and j are the row number and column number in the image coordinate space (row number, column number) of the HAND model. Also, as is clear from the substitution equation, the measured HAND water level (i, j) for creating the flood map depends only on the column number j and is independent of the row number i.

[0145] 2) Next, the flood depth 1(i, j) is calculated based on the measured HAND water level (i, j) for flood map creation calculated in 1). The flood depth 1(i, j) is calculated using the formula: measured HAND water level (i, j) - HAND ground level (i, j). Here, the HAND ground level (i, j) is the value of the grid cell of the HAND model at row number i and column number j in the image coordinate space (row number, column number).

[0146] 3) Next, calculate the flood depth (i, j) based on the flood depth 1(i, j) calculated in 2). If the flood depth 1(i, j) is 0m or less, it is determined that there is no flooding, and the flood depth (i, j) is set to 0m. If the flood depth 1(i, j) is greater than 0m, it is determined that there is flooding, and the flood depth 1(i, j) is set to the flood depth (i, j).

[0147] 4) Next, the inundation depth (i, j) calculated in 3) is stored in the inundation map storage means 44 as a GeoTIFF file. Alternatively, the inundation depth (i, j) and metadata such as a coordinate reference system and bounding boxes may be stored based on a predetermined database management method. Here, the coordinate reference system is, for example, EPSG:6668-JGD2011, which represents a method of projecting a position on the spherical Earth onto a plane. The bounding boxes are the longitude of the western end, the longitude of the eastern end, the latitude of the northern end, and the latitude of the southern end of the GeoTIFF file. A computer program such as rasterio can be used to write the GeoTIFF file.

[0148] [Flood Inundation Map Calculation Function (when divided into longitudinal and transverse directions)] The following describes in detail the function of the flood map calculation means 370 for calculating flood maps, specifically in the case where a predetermined area is divided not only perpendicularly (cross-sectionally) to the average flow direction of the river, but also longitudinally along the centerline of the river.

[0149] 1) Based on the interpolation function (i, j) for creating the flood map, the measured HAND water level (i, j) for creating the flood map is calculated. The measured HAND water level (i, j) for creating the flood map is calculated using the substitution equation from the right side to the left side of the equation: measured HAND water level (i, j) for creating the flood map = interpolation function (i, j) for creating the flood map. Here, i and j are the row number and column number in the image coordinate space (row number, column number) of the HAND model.

[0150] 2) Next, the flood depth 1(i, j) is calculated based on the measured HAND water level (i, j) for flood map creation calculated in 1). The flood depth 1(i, j) is calculated using the formula: measured HAND water level (i, j) - HAND ground level (i, j). Here, the HAND ground level (i, j) is the value of the grid cell of the HAND model at row number i and column number j in the image coordinate space (row number, column number).

[0151] 3) Next, calculate the flood depth (i, j) based on the flood depth 1(i, j) calculated in 2). If the flood depth 1(i, j) is 0m or less, it is determined that there is no flooding, and the flood depth (i, j) is set to 0m. If the flood depth 1(i, j) is greater than 0m, it is determined that there is flooding, and the flood depth 1(i, j) is set to the flood depth (i, j).

[0152] 4) Next, the inundation depth (i, j) calculated in 3) is stored in the inundation map storage means 44 as a GeoTIFF file. Alternatively, the inundation depth (i, j) and metadata such as the coordinate reference system and bounding box may be stored based on a predetermined database management method.

[0153] [Tile Image Calculation Function] The following details the function that calculates tile images suitable for use in web applications based on the generated flood map.

[0154] 1) The flood map stored in the flood map storage means 44 is read, and metadata such as the coordinate reference system and bounding box, along with the flood depth (i, j) data, are stored in the memory 3002.

[0155] 2) Next, a predetermined zoom level, predetermined tile coordinates X and Y are set to convert the flood depth (i, j) data into XYZ tile format data. Here, XYZ tile format data refers to image data of small squares obtained by dividing the map into a grid. The zoom level represents the scale of the map; when the zoom level is 0, the entire world is represented by one tile. When the zoom level is 1, the map is divided into four tiles. Furthermore, tile coordinates X refers to the position of the tile in the east-west direction (horizontal direction). Also, tile coordinates Y refers to the position of the tile in the north-south direction (vertical direction).

[0156] 3) Next, the XYZ tile format data converted in step 2) is stored in the storage device 3003 as a PNG file with a width of 256 pixels and a height of 256 pixels. At this time, the directory is set to, for example, " / 17 / 114460 / " and the name of the PNG file is set to "52432.png" according to a predetermined zoom level and predetermined tile coordinates. Here, "17", "114460", and "52432" are the values ​​of the zoom level Z, tile coordinate X, and tile coordinate Y in the XYZ format. This XYZ method is a well-known technology used to efficiently distribute map information in communication between a web server and a web browser. Computer programs such as rasterio can be used to read GeoTIFF format files. Computer programs such as riotiler can be used to convert to XYZ tile format data. riotiler is publicly available from "https: / / cogeotiff.github.io / rio-tiler / ".

[0157] [Image coordinate space rotation function] In the description of the HAND model acquisition means 340, as shown in Figure 11, it was explained that when the direction of river flow is inclined from the north-south or east-west direction, or when the river meanders and the average direction of flow is inclined from the north-south or east-west direction, it is preferable to rotate the image coordinate space (row number, column number) of the HAND model. In such cases, it is preferable to rotate the image coordinate space (row number, column number) of the inundation depth (i, j) in the opposite direction to return it to the original image coordinate space (row number, column number).

[0158] The following describes in detail the function that rotates the image coordinate space (row number, column number) of the flood depth (i, j) in the opposite direction to the rotation used when acquiring the HAND model.

[0159] 1) When the image coordinate space (row number, column number) is rotated counterclockwise, the row number after rotation is calculated by rounding the result of (row number before rotation × cos(-rotation angle) - column number before rotation × sin(-rotation angle)) to the nearest integer. The column number after rotation is calculated by rounding the result of (row number before rotation × sin(-rotation angle) + column number before rotation × cos(-rotation angle)) to the nearest integer. Here, the rotation angle is assumed to be a positive real number. If the rotation of the image coordinate space (row number, column number) of the flood depth (i, j) results in grid cells where no flood depth exists, interpolation is performed using known techniques such as nearest neighbor interpolation, bilinear interpolation, or bicubic interpolation.

[0160] 2) Next, the values ​​of the inundation depth (i, j) rotated in 1) are stored in the inundation map storage means 44 as a GeoTIFF file. Alternatively, the inundation depth (i, j) and metadata such as the coordinate reference system and bounding box may be stored based on a predetermined database management method.

[0161] [Water level calculation function] The following describes the details of the function for calculating the water level. The water level is calculated using the following equation (9). JPEG2026046222000010.jpg17170

[0162] Here, V is the water volume (m³ 3 ), Dij is the inundation depth (m) at the position of row number i and column number j in the inundation diagram, W is the width (m) of the grid cell, H is the height (m) of the grid cell, N1 is the row number of the grid cell, and N2 is the column number of the grid cell.

[0163] As described above, by calculating water volume data in addition to flood map data, the national and local governments can efficiently deploy drainage pump trucks, which has a beneficial effect.

[0164] [Mask function] The following describes in detail the function for masking unnecessary areas, including river channels, from the flood map. The masked map is assumed to be stored in storage device 3003 beforehand as a GeoTIFF file.

[0165] 1) The flood diagram stored in the flood diagram storage means 44 and the mask diagram stored in the storage device 3003 are read and stored in the memory 3002. Here, the flood diagram is raster data composed of grid cells arranged in rows and columns, and the inside of the flooded area is filled with flood depth (i, j). The mask diagram is raster data composed of grid cells arranged in rows and columns, and the area to be masked is filled with 1, for example.

[0166] 2) Next, for each row and column number of the grid cells in the flood diagram, as an example, if the value of the grid cell in the mask diagram is 1, the value of the grid cell in the flood diagram will be set to 0m, and the resulting flood diagram will be masked.

[0167] 3) Next, the masked flood map calculated in step 2) is stored in the flood map storage means 44 as a GeoTIFF file.

[0168] <<Flood map correction means 372 of flood map creation means 36>> The flood diagram correction means 372 has the function of correcting the flood diagram based on the dynamic sensing information stored in the memory 3002 and the flood diagram stored in the flood diagram storage means 44. The details of the function for correcting the flood diagram are described below.

[0169] 1) The flood map stored in the flood map storage means 44 is read, and metadata such as the coordinate reference system and bounding box, along with the flood depth (i, j) data, are stored in the memory 3002.

[0170] 2) Next, if the dynamic sensing information is related to flood sensors, the latitude and longitude of the flood sensor installation location included in the dynamic sensing information are converted to row number i and column number j in the image coordinate space (row number, column number). If the dynamic sensing information is related to social media, the latitude and longitude of the social media posting location are converted to row number i and column number j in the image coordinate space (row number, column number). If the dynamic sensing information is related to aerial photographs, the latitude and longitude within the flooded area are converted to row number i and column number j in the image coordinate space (row number, column number). If the dynamic sensing information is related to satellite images, the latitude and longitude within the flooded area are converted to row number i and column number j in the image coordinate space (row number, column number).

[0171] 3) Next, if the dynamic sensing information is related to a flood sensor, determine whether the flood depth (i, j) at the row number i and column number j of the flood sensor installation location calculated in 2) is 0m or less, and whether the presence or absence of flooding included in the dynamic sensing information is not inconsistent with the presence or absence of flooding.

[0172] 4) Next, if the dynamic sensing information is related to SNS, determine whether the flood depth (i, j) at the row number i and column number j of the SNS posting location calculated in 2) is 0m or less, and whether the presence or absence of flooding included in the dynamic sensing information is not inconsistent with the presence or absence of flooding.

[0173] 5) Next, if the dynamic sensing information is related to aerial photographs, determine whether the flood depth (i, j) at the position of row number i and column number j within the flooded area calculated in 2) is 0m or less, and whether the presence or absence of flooding included in the dynamic sensing information is not inconsistent with the presence or absence of flooding.

[0174] 6) Next, if the dynamic sensing information is related to satellite images, determine whether the flood depth (i, j) at the position of row number i and column number j within the flooded area calculated in 2) is 0m or less, and whether the presence or absence of flooding included in the dynamic sensing information is not inconsistent with the presence or absence of flooding.

[0175] 7) Next, if the results of the determinations in 3) to 6) are inconsistent, determine whether the row number i and column number j at that location are inside or outside the flood area of ​​the flood map. The determination of whether the row number i and column number j at that location are inside or outside the flood area of ​​the flood map is as follows: If there is a grid cell with a flood depth (i, j) value greater than 0m within a predetermined distance from that location, it is determined to be inside the flood area. If there is no grid cell with a flood depth (i, j) value greater than 0m within a predetermined distance from that location, it is determined to be outside the flood area.

[0176] 8) Next, if in step 7) the row number i and column number j of the location in question are determined to be outside the flooded area of ​​the flood map, the following are added to the list of maximum measured HAND water level values: the latitude, longitude, and measured HAND water level of the location where the flood sensor was installed; the latitude, longitude, and measured HAND water level of the location where the SNS post was made; the corresponding latitude, longitude, and measured HAND water level within the flooded area determined from aerial photographs; and the corresponding latitude, longitude, and measured HAND water level within the flooded area determined from satellite images.

[0177] 9) Next, repeat steps 2) to 8) until there are no more row numbers i and column numbers j whose judgment results from steps 3) to 6) are inconsistent.

[0178] 10) Next, based on the measured maximum HAND water level calculated in steps 2) to 9), and the row number i and column number j in the image coordinate space (row number, column number), an interpolation function for creating a flood map is calculated in the same manner as described for the interpolation function calculation means 368 for creating a flood map.

[0179] 11) Next, based on the interpolation function for creating the flood map calculated in 10) and the HAND model stored in memory 3002, a flood map is calculated in the same manner as described for the flood map calculation means 370, and this corrected flood map is stored in the flood map storage means 44.

[0180] As described above, even if the dynamic sensing information indicates flooding, it becomes possible to use data that shows no flooding in the flood map data. Therefore, even information processing devices that employ flood sensors, SNS images (still images), aerial photographs (still images), and satellite images (SAR images) can calculate highly accurate flood map data.

[0181] ≪Abnormal Value Detection Means 374 of Flood Diagram Creation Means 36≫ The abnormal value detection means 374 has the function of detecting abnormal values ​​based on the measured HAND water level stored in the memory 3002. It also has the function of storing the detected abnormal values ​​in the memory 3002. Here, an abnormal value refers to a situation in the flood sensor where, due to causes such as a malfunction of the flood sensor or the adhesion of mud to the flood sensor, information indicating flooding is not transmitted even though there is flooding at the installation height of the flood sensor, or information indicating flooding is transmitted even though there is no flooding. The details of the function for detecting abnormal values ​​are described below. Here, it is assumed that a predetermined area is divided into segments perpendicular to the average flow direction of the river (transverse direction). In addition to the perpendicular (transverse direction) division, it is also possible that the area is divided into segments longitudinally along the centerline of the river.

[0182] 1) In a certain segment, water ingress sensors S1, S2, ..., S N Assume that it is installed. Also, the measured HAND water level of the flood sensor is H1, H2, ..., H N For example, the water ingress sensors S1, S2, ..., S N Assume the numbers are sorted in ascending order of the measured HAND water level. Furthermore, the reliability of the water ingress sensors is assigned as W1, W2, ..., W N Let's assume that W1, W2, ..., W N This value ranges from 0 to 1, with 0 representing the lowest reliability and 1 representing the highest reliability. Here, as an example, the reliability of all flood sensors is set to 1.0. Alternatively, the reliability of a flood sensor may be calculated from information such as whether it transmitted flood information in the past when there was clearly no flooding.

[0183] 2) Next, as an example, assume that the water immersion sensors S1, S2, …, S K indicate no water immersion, the water immersion sensors S K+1 …, S L indicate water immersion, and the water immersion sensors S L+1 …, S N indicate no water immersion. Even though S1, S2, …, S K with a lower measured HAND water surface height indicate no water immersion, S K+1 …, S L with a higher measured HAND water surface height than these indicate water immersion, and assume that S1, S2, …, S K are abnormal values, or either S K+1 …, S L is an abnormal value.

[0184] 3) Next, for the water immersion sensors S1, S2, …, S K add up the values of the reliabilities W1, W2, …, W K and set this value as A1. Also, for the water immersion sensors S K+1 …, S L add up the values of the reliabilities W K+1 …, W L and set this value as A2.

[0185] 4) Next, compare A1 and A2. If A1 is greater than A2, determine that the information indicating no water immersion for the water immersion sensors S1, S2, …, S K is a normal value, and the information indicating water immersion for the water immersion sensors S K+1 …, S L is an abnormal value. If A1 is less than A2, determine that the information indicating no water immersion for the water immersion sensors S1, S2, …, S K is an abnormal value, and the information indicating water immersion for the water immersion sensors S K+1 …, S L is a normal value. Further, if A1 and A2 are the same value, determine that the information of S1, S2, …, S K S K+1 …, S L is a normal value.

[0186] 5) Next, repeat the above processes 1) to 4) for the number of segments to detect abnormal values.

[0187] As described above, in the water immersion sensor, abnormal values such as the presence of water immersion despite the absence of water immersion can be excluded due to phenomena such as mud adhesion. Therefore, in the information processing apparatus that employs the water immersion sensor, SNS images (still images), aerial photographs (still images), and satellite images (SAR images), the effect of being able to calculate highly accurate data of the water immersion map is achieved.

[0188] ≪Existing water immersion information acquisition means 380 of virtual water immersion map creation means 38≫ [Existing water immersion information acquisition function, memory storage function] As shown in FIG. 23, the existing water immersion information acquisition means 380 has a function of acquiring the existing water immersion information stored in the storage device 3003. It also has a function of storing the acquired existing water immersion information in the memory 3002. The existing water immersion information may be stored in the storage device 3003 as a file in GeoTIFF format, or may be stored based on a predetermined database management method with metadata such as a coordinate reference system and a bounding box, and the water depth of the grid cell.

[0189] Here, the coordinate reference system, as an example, is EPSG:6668-JGD2011, which indicates a method of projecting positions on the spherical Earth onto a plane. The bounding box refers to the longitude of the west end, the longitude of the east end, the latitude of the north end, and the latitude of the south end of a file in GeoTIFF format. For reading a file in GeoTIFF format, a computer program such as rasterio can be used.

[0190] [Image coordinate space rotation function] As shown in FIG. 24, the existing flood information acquisition means 380 may further have a function of rotating the image coordinate space (row number, column number) of the existing flood information. As shown in FIG. 11, when the flow direction of the river is inclined from the north-south direction or the east-west direction, or when the river meanders and flows with the average flow direction inclined from the north-south direction or the east-west direction, it is preferable to rotate the image coordinate space (row number, column number) of the existing flood information. In the following description, it is assumed that the X-axis direction of the existing flood information coincides with the average flow direction of the river by rotating the image coordinate space (row number, column number) of the existing flood information. Also, it is assumed that the Y-axis direction of the existing flood information coincides with the direction perpendicular to the average flow direction of the river. The rotation of the image coordinate space uses the method described for the HAND model acquisition means 340.

[0191] ≪Virtual Flood Map Creation Means 38's Virtual Flood Map Creation Condition Setting Means 382≫ The virtual flood map creation condition setting means 382 has a function of setting a list of the number of segment divisions for creating a virtual flood map. It also has a function of setting a list of the number of virtual flood sensors. Furthermore, it has a function of setting the number of repetitions of the random arrangement of virtual flood sensors. Here, the list of the number of segment divisions for creating a virtual flood map is, for example, a list of positive integers such as 2, 3, 4, 5, 6, 7, 8, 9, 10. Also, the list of the number of virtual flood sensors is, for example, a list of positive integers such as 10, 20, 30, 40, 50, 60, 70, 80, 90, 100. Furthermore, the number of repetitions of the random arrangement of virtual flood sensors is, for example, a positive integer such as 10. The virtual flood map creation condition setting means 382 has a function of storing these lists of the number of segment divisions for creating a virtual flood map, the list of the number of virtual flood sensors, and the number of repetitions of the random arrangement of virtual flood sensors in the memory 3002.

[0192] ≪Virtual Flood Map Creation Means 38's Virtual Dynamic Sensing Information Generation Means 384≫ The virtual dynamic sensing information generation means 384 has a function to calculate virtual dynamic sensing information based on existing flood information stored in memory 3002, a list of virtual flood sensors, and the number of repetitions of the random arrangement of virtual flood sensors. It also has a function to store the calculated virtual dynamic sensing information in memory 3002. Furthermore, it has a function to calculate the virtual sensor density based on the virtual dynamic sensing information and store the virtual sensor density in memory 3002. The details of the function for calculating virtual dynamic sensing information are described below.

[0193] 1) Obtain the westernmost longitude, easternmost longitude, northernmost latitude, and southernmost latitude of the GeoTIFF file from the boundary box metadata included in the existing flood information. Define the area enclosed by these westernmost longitude, easternmost longitude, northernmost latitude, and southernmost latitude as the designated area.

[0194] 2) Next, within the range of the longitude of the westernmost, easternmost, northernmost, and southernmost points of the designated region, a random number for latitude and a random number for longitude are calculated.

[0195] 3) Next, the random numbers of latitude and longitude calculated in 2) are converted to row number i and column number j in the image coordinate space (row number, column number). Here, equation (2) explained in the measured HAND water level calculation means 362 is used for the conversion to row number i and column number j.

[0196] 4) Next, determine whether the position of row number i and column number j calculated in 3) is included within the inundation range of the existing inundation information. Here, if the inundation depth (i, j) at the position of row number i and column number j is greater than 0m, then it is assumed that the position of row number i and column number j is included within the inundation range of the existing inundation information. Also, if the inundation depth (i, j) at the position of row number i and column number j is 0m or less, then it is assumed that the position of row number i and column number j is not included within the inundation range of the existing inundation information.

[0197] 5) Next, if the row number i and column number j calculated in 3) are within the inundation range of the existing inundation information, the random latitude number, the random longitude number, and the identification ID calculated in 2) are added to the list of virtual dynamic sensing information. Here, the identification ID should be set to a value that does not overlap with other identification IDs such as sequence numbers.

[0198] 6) Next, the processes described in 2) to 5) above are repeated until the number of items in the list of virtual dynamic sensing information reaches the number of virtual flood sensors selected from the list of virtual flood sensors, thereby calculating the virtual dynamic sensing information.

[0199] The following describes the details of the function for calculating virtual sensor density. The virtual sensor density is calculated using the following equation (10). JPEG2026046222000011.jpg12170

[0200] Here, D is the virtual sensor density (number of sensors / m²). 2 ) is the number of virtual flood sensors installed within the flood range of existing flood information in a given area. A is the area (m²) of the flood range of existing flood information in a given area. 2 ) is. For example, in the case of a flood inundation area map of the maximum expected scale published by the Geospatial Information Authority of Japan, N is the number of virtual flood sensors installed within the flood inundation area in a given region, and A is the area of ​​the flood inundation area in a given region (m²). 2 )

[0201] ≪Virtual HAND water level calculation means 386 of virtual flood map creation means 38≫ The virtual HAND water level calculation means 386 has the function of calculating the virtual HAND water level based on the HAND model stored in the memory 3002 and the virtual dynamic sensing information stored in the memory 3002. It also has the function of storing the calculated virtual HAND water level in the memory 3002.

[0202] The following describes the details of the function for calculating the virtual HAND water level. The HAND ground level (i, j) at the virtual flood sensor installation location is defined as the virtual HAND water level (i, j). Here, since the values ​​of the virtual flood sensor installation location are stored in geographic coordinate space (latitude, longitude), the latitude and longitude of the virtual flood sensor installation location are converted to row number i and column number j in image coordinate space (row number, column number) to calculate the virtual HAND water level (i, j). The above process is repeated for the number of items in the virtual dynamic sensing information list to calculate the virtual HAND water level (i, j).

[0203] ≪Virtual HAND water level maximum value calculation means 388 of virtual flood map creation means 38≫ The virtual HAND water level maximum value calculation means 388 has the function of dividing a predetermined area into segments based on a list of the number of segments for creating a virtual flood map stored in memory 3002 and a HAND model stored in memory 3002. It also has the function of calculating the maximum virtual HAND water level based on the segment to be calculated and the virtual HAND water level stored in memory 3002. Furthermore, it has the function of storing the calculated maximum virtual HAND water level in memory 3002.

[0204] [Cross-sectional division function] The following describes in detail the function of dividing a given area into equal intervals perpendicular to the average flow direction of the river (cross-sectional direction). As shown in Figure 17, the column numbers of the points where the lines dividing the given area perpendicular to the average flow direction of the river (cross-sectional direction) intersect the X-axis are J1, J2, ..., J N-1 Let's assume J1, J2, ..., J N-1These are calculated by rounding (X2-X1) / N, 2×(X2-X1) / N, ..., and (N-1)×(X2-X1) / N to the nearest integer. Here, X1 is the minimum column number of the HAND model, X2 is the maximum column number of the HAND model, and N is the number of segments for creating a virtual flood map selected from the list of segments for creating a virtual flood map. In this case, segment 1 will consist of grid cells with column numbers greater than or equal to X1 and less than or equal to J1. Similarly, segment 2 will consist of grid cells with column numbers greater than J1 and less than or equal to J2. Similarly, segment N will consist of grid cells with column numbers greater than J1 and less than or equal to J2. N-1 It will be larger and consist of grid cells up to X2 in size.

[0205] [Vertical and Cross-sectional Segmentation Function] The function of the virtual HAND water level maximum value calculation means 388 for dividing a predetermined area into segments may be a function that divides the area at equal intervals perpendicular to the average flow direction of the river (transverse direction), as well as dividing it longitudinally along the centerline of the river.

[0206] The following describes in detail the function of dividing a given area into equal intervals perpendicular to the average flow direction of the river (cross-sectional direction), and further dividing it longitudinally along the river's centerline. As shown in Figure 18, the column numbers of the points where the lines dividing the given area perpendicular to the average flow direction of the river (cross-sectional direction) intersect the X-axis are J1, J2, ..., J N-1 Let's assume J1, J2, ..., J N-1 These are calculated by rounding (X2-X1) / N, 2×(X2-X1) / N, ..., and (N-1)×(X2-X1) / N to the nearest integer. Here, X1 is the minimum column number of the HAND model, X2 is the maximum column number of the HAND model, and N is the number of segments for creating a virtual flood map selected from the list of segments for creating a virtual flood map. Next, the map is divided longitudinally along the centerline of the river. In this case, segment 1 will consist of grid cells with column numbers greater than or equal to X1 and less than or equal to J1, and row numbers above the centerline of the river. Similarly, segment N will consist of grid cells with column numbers J N-1Larger than X2, the row numbers will consist of grid cells above the river's centerline. Similarly, segment N+1 will consist of grid cells below the river's centerline, with column numbers between X1 and J1. Similarly, segment 2N will consist of grid cells below the river's centerline, with column numbers between J N-1 For larger sizes, up to X2, the row numbers will consist of grid cells below the centerline of the river.

[0207] [Function to calculate the maximum virtual HAND water level (when divided in the transverse direction)] The following describes the details of the function for calculating the maximum virtual HAND water level. As an example, let's assume that virtual flood sensors 1 to 5 are installed in segment 1. The row and column numbers in the image coordinate space (row number, column number) for the installation locations of virtual flood sensors 1 to 5 are (1, 1), (1, 5), (1, 10), (3, 5), and (3, 10). Here, the number on the left in parentheses is the row number, and the number on the right in parentheses is the column number. Also, let's assume that the virtual HAND water levels for virtual flood sensors 1 to 5 are 1.0m, 1.5m, 2.0m, 2.5m, and 3.0m, respectively. In this case, the maximum virtual HAND water level in segment 1 is 3.0m, and the row and column numbers in its image coordinate space (row number, column number) are (3, 10). If no virtual flood sensors are installed in a segment, the maximum virtual HAND water level in that segment is not calculated. Next, for segments 2 through N, we similarly calculate the maximum virtual HAND water level and its corresponding row and column numbers in its image coordinate space (row number, column number).

[0208] [Function to calculate the maximum virtual HAND water level (when the water level is divided in both the transverse and longitudinal directions)] The details of the function for calculating the maximum virtual HAND water surface height are described below. As an example, it is assumed that virtual immersion sensors 1 to 3 are installed in segment 1. The row number and column number in the image coordinate space (row number, column number) of the installation locations of virtual immersion sensors 1 to 3 are assumed to be (1, 1), (1, 5), and (1, 10). Here, the left number within the parentheses is the row number, and the right number within the parentheses is the column number. Also, the virtual HAND water surface heights of virtual immersion sensors 1 to 3 are assumed to be 1.0 m, 1.5 m, and 2.0 m, respectively. In this case, the maximum virtual HAND water surface height in segment 1 is 2.0 m, and the row number and column number in its image coordinate space (row number, column number) are (1, 10). Also, when no virtual immersion sensor is installed in a segment, the maximum virtual HAND water surface height in that segment is not calculated. Next, for segments 2 to 2N, similarly, the maximum virtual HAND water surface height and the row number and column number in its image coordinate space (row number, column number) are calculated.

[0209] ≪Virtual immersion map creation interpolation function calculation means 390 of virtual immersion map creation means 38≫ The virtual immersion map creation interpolation function calculation means 390 has a function of calculating an interpolation function for virtual immersion map creation based on the maximum virtual HAND water surface height stored in the memory 3002, the row number and column number in its image coordinate space (row number, column number). Also, it has a function of storing the calculated interpolation function for virtual immersion map creation in the memory 3002. The details of the function for calculating the interpolation function for virtual immersion map creation are described below. Here, it is assumed that a predetermined area is divided into segments in a direction perpendicular (transverse direction) to the average flow direction of the river. Also, it is assumed that there are N segments, and the maximum virtual HAND water surface height has been calculated for each segment.

[0210] [Function for calculating interpolation function for virtual immersion map creation (when divided in the transverse direction)] 1) The row number, column number, and maximum virtual HAND water surface height in the image coordinate space (row number, column number) of the maximum virtual HAND water surface height are (X1, Y1, M1), …, (X N , Y N , MN Let it be represented by a sequence of points X1, ..., X N This represents the column number, Y1, ..., Y in the image coordinate space (row number, column number) of the HAND model. N This represents the row number, M1, ..., M in the image coordinate space (row number, column number) of the HAND model. N This represents the maximum virtual HAND water level in each segment.

[0211] 2) Next, (X1, Y1, M1), ..., (X N , Y N M N The sequence of points ) is projected onto the projection plane for creating the virtual flood map, (X1, M1), ..., (X N M N The sequence of points in the image is calculated. Here, the projection plane for creating the virtual flood map is a plane where the column number in the image coordinate space (row number, column number) of the HAND model is the horizontal axis and the maximum value of the virtual HAND water level is the vertical axis.

[0212] 3) Next, using radial basis functions, (X1, M1), ..., (X N M N An interpolation function for creating a smooth virtual flood map that passes through the neighborhood of each point in the map is obtained. The calculation of the interpolation function using the radial basis function is performed using the method described in the flood map creation segment discrimination means 364.

[0213] [Function to calculate interpolation functions for creating virtual flood maps (when the map is divided in both the transverse and longitudinal directions)] The function of the interpolation function calculation means 390 for creating a virtual flood map to calculate the interpolation function for creating a virtual flood map may be such that, in addition to dividing the river into segments perpendicular to the average flow direction (transverse direction), the river is divided into segments longitudinally along the centerline. Here, it is assumed that there are 2N segments, and that the maximum virtual HAND water level is calculated for each segment.

[0214] 1) The row number, column number, and maximum virtual HAND water level in the image coordinate space (column number, row number) are (X1, Y1, M1), ..., (X N , YN M N ), (X N+1 , Y N+1 M N+1 ), …, (X 2N , Y 2N M 2N Let it be represented by a sequence of points X1, ..., X N This is above the river centerline, and the column number in the image coordinate space (row number, column number) of the HAND model, X N+1 , ..., X 2N Below the river centerline, the column number, Y1, ..., Y in the image coordinate space (row number, column number) of the HAND model. N This is above the river centerline, and the row number and Y are in the image coordinate space (row number, column number) of the HAND model. N+1 , ..., Y 2N Below the river centerline, the row number, M1, ..., M in the image coordinate space (row number, column number) of the HAND model. N This is the maximum virtual HAND water level in each segment above the river centerline, M N+1 ... M 2N This is defined as the maximum virtual HAND water level in each segment below the river centerline.

[0215] 2) Next, using radial basis functions, (X1, Y1, M1), ..., (X N , Y N M N ), (X N+1 , Y N+1 M N+1 ), …, (X 2N , Y 2N M 2N An interpolation function for creating a smooth virtual flood map that passes through the neighborhood of each point in the map is obtained. The calculation of the interpolation function using the radial basis function is performed using the method described in the flood map creation segment discrimination means 364.

[0216] ≪Virtual flood map calculation means 392 of virtual flood map creation means 38≫ The virtual flood map calculation means 392 has the function of calculating a virtual flood map based on an interpolation function for creating a virtual flood map stored in memory 3002 and a HAND model stored in memory 3002. It also has the function of storing the calculated virtual flood map in memory 3002. Furthermore, it may have the function of rotating the image coordinate space (row number, column number) of the virtual flood map. The details of the function for calculating the virtual flood map are described below. Here, it is assumed that a predetermined area is divided into segments perpendicular to the average flow direction of the river (cross-sectional direction). In this case, the interpolation function for creating a virtual flood map is defined as a projection plane for creating a virtual flood map where the column number in the image coordinate space (row number, column number) of the HAND model is the horizontal axis and the maximum value of the virtual HAND water level is the vertical axis.

[0217] [Virtual flood map calculation function (when divided in the cross-sectional direction)] 1) Based on the interpolation function (j) for creating the virtual flood map, the virtual HAND water level (i, j) for creating the flood map is calculated. The virtual HAND water level (i, j) for creating the flood map is calculated using the substitution equation from the right side to the left side of the equation: virtual HAND water level (i, j) = interpolation function (j) for creating the virtual flood map. Here, i and j are the row number and column number in the image coordinate space (row number, column number) of the HAND model. Also, as is clear from the substitution equation, the virtual HAND water level (i, j) for creating the flood map depends only on the column number j and is independent of the row number i.

[0218] 2) Next, the virtual inundation depth 1(i, j) is calculated based on the virtual HAND water level (i, j) calculated in 1). The virtual inundation depth 1(i, j) is calculated using the formula: virtual HAND water level (i, j) - HAND ground level (i, j). Here, the HAND ground level (i, j) is the value of the grid cell of the HAND model at row number i and column number j in the image coordinate space (row number, column number).

[0219] 3) Next, calculate the virtual flood depth (i, j) based on the virtual flood depth 1(i, j) calculated in 2). If the virtual flood depth 1(i, j) is 0m or less, it is determined that there is no flooding, and the virtual flood depth (i, j) is set to 0m. If the virtual flood depth 1(i, j) is greater than 0m, it is determined that there is flooding, and the virtual flood depth 1(i, j) is set to the virtual flood depth (i, j).

[0220] 4) Next, the virtual flood depth (i, j) calculated in step 3) is stored in memory 3002.

[0221] [Virtual flood map calculation function (when divided in both the transverse and longitudinal directions)] The following describes in detail the function of the virtual flood map calculation means 392 for calculating a virtual flood map, specifically in the case where a predetermined area is divided not only perpendicular (cross-sectional) to the average flow direction of the river, but also longitudinally along the centerline of the river.

[0222] 1) Based on the interpolation function (i, j) for creating the virtual flood map, the virtual HAND water level (i, j) for creating the flood map is calculated. The virtual HAND water level (i, j) for creating the flood map is calculated using the substitution equation from the right side to the left side of the equation: virtual HAND water level (i, j) = interpolation function (i, j) for creating the virtual flood map. Here, i and j are the row number and column number in the image coordinate space (row number, column number) of the HAND model.

[0223] 2) Next, the virtual inundation depth 1(i, j) is calculated based on the virtual HAND water level (i, j) calculated in 1). The virtual inundation depth 1(i, j) is calculated using the formula: virtual HAND water level (i, j) - HAND ground level (i, j). Here, the HAND ground level (i, j) is the value of the grid cell of the HAND model at row number i and column number j in the image coordinate space (row number, column number).

[0224] 3) Next, calculate the virtual flood depth (i, j) based on the virtual flood depth 1(i, j) calculated in 2). If the virtual flood depth 1(i, j) is 0m or less, it is determined that there is no flooding, and the virtual flood depth (i, j) is set to 0m. If the virtual flood depth 1(i, j) is greater than 0m, it is determined that there is flooding, and the virtual flood depth 1(i, j) is set to the virtual flood depth (i, j).

[0225] 4) Next, the virtual flood depth (i, j) calculated in step 3) is stored in memory 3002.

[0226] [Function to rotate the image coordinate space of the virtual flood map] In the description of the HAND model acquisition means 340 and the existing flood information acquisition means 380, it was explained that, as shown in Figure 11, if the direction of river flow is inclined from the north-south or east-west direction, or if the river meanders and the average direction of flow is inclined from the north-south or east-west direction, it is preferable to rotate the image coordinate space (row number, column number) of the HAND model and the existing flood information. In such cases, it is preferable to rotate the image coordinate space (row number, column number) of the virtual flood depth (i, j) in the opposite direction to return it to the original image coordinate space (row number, column number). The method described in the flood map calculation means 370 is used to rotate the image coordinate space (row number, column number) in the opposite direction.

[0227] <<Method 394 for calculating the average evaluation index value of the virtual flood map creation method 38>> The evaluation index value average calculation means 394 has the function of calculating the evaluation index value average based on existing flood information stored in memory 3002 and a virtual flood diagram stored in memory 3002. It also has the function of storing the calculated evaluation index value average in the evaluation index value storage means 42. The details of the function for calculating the evaluation index value average are described below.

[0228] 1) TP is calculated as the number of grid cells that were determined to be flooded in the virtual flood map and that are also flooded in existing flood information. 2) FP is calculated as the number of grid cells that were determined to be inundated in the virtual inundation map but were not inundated in the existing inundation information. 3) TN is calculated as the number of grid cells that were determined to be free from flooding in the virtual flood map, and which are also free from flooding in existing flood information. 4) FN is calculated as the number of grid cells that were determined to be free from flooding in the virtual flood map, but which are actually flooded according to existing flood information. 5) The precision rate is calculated as TP / (TP+FP). 6) The recall rate is calculated as TP / (TP+FN). 7) The F-score is calculated as 2 × precision × recall / (precision + recall). 8) The average F-value is calculated by averaging the F-values ​​for the same virtual sensor density and the same number of segment divisions for creating the virtual flood map, as shown in Figure 25. Figure 25 illustrates the case where the average F-value is calculated for each case where division is made along the river centerline or not.

[0229] Next, the identification ID, virtual sensor density, number of segments for creating the virtual flood map, and the TP, FP, TN, FN, precision, recall, F-value, and average F-value calculated in 1) to 8) are stored in the evaluation index value storage means 42.

[0230] ≪Dynamic sensing information storage means 40≫ The dynamic sensing information storage means 40 has the function of storing information regarding the presence or absence of flooding received by the flooding sensor reception control means 320, information regarding the presence or absence of flooding determined by the SNS determination means 322, information regarding the flooded area determined by the aerial photograph determination means 324, and information regarding the flooded area determined by the satellite image determination means 326, based on a predetermined database management method. The dynamic sensing information storage means 40 also has the function of allowing the dynamic sensing information to be read out by the dynamic sensing information acquisition means 360.

[0231] As illustrated in Figure 12, when dynamic sensing information is related to a flood sensor, the database table stores an identification ID to identify the dynamic sensing information, a sensor ID to identify the flood sensor, the date and time of observation, the location (latitude, longitude) of the flood sensor, the height at which the flood sensor was installed, the ground level of the HAND, and whether or not there was flooding.

[0232] Furthermore, as illustrated in Figure 13, when the dynamic sensing information is related to SNS, the database table stores an identification ID to identify the dynamic sensing information, an SNS ID to identify the SNS post, the posting date and time, the posting location (latitude, longitude), image information (still image) attached to the SNS post, the HAND ground elevation, and whether or not there is flooding.

[0233] Furthermore, as illustrated in Figure 14, when the dynamic sensing information is related to aerial photographs, the database table stores an identification ID to identify the polygons of the flooded area, an aerial photograph ID to identify the aerial photograph, the date and time of shooting, the shooting location (latitude, longitude), the image information (still image) to be captured, and the latitude and longitude of the vertices of the outer boundary of the polygons of the flooded area. Since there are multiple vertices on the outer boundary of the polygons of the flooded area, the item for the flooded area (latitude, longitude of the vertices of the outer boundary) may be stored in CSV format (comma-separated format).

[0234] Furthermore, as illustrated in Figure 15, when the dynamic sensing information is related to satellite imagery, the database table stores an identification ID to identify the polygon of the flooded area, a satellite image ID to identify the satellite image, the observation date and time, the observation location (latitude, longitude), the observed satellite image (SAR image), and the latitude and longitude of the vertices of the outer boundary of the polygon of the flooded area. Since there are multiple vertices on the outer boundary of the polygon of the flooded area, the item for flooded area (latitude, longitude of the vertices of the outer boundary) may be stored in CSV format (comma-separated format).

[0235] <<Evaluation index value storage means 42>> The evaluation index value storage means 42 has the function of storing the average evaluation index value calculated by the evaluation index value average calculation means 394 based on a predetermined database management method. The evaluation index value storage means 42 also has the function of being read out by the flood map creation segment discrimination means 364. As illustrated in Figure 25, the database table of the evaluation index value average stores an identification ID for identifying the evaluation index value average, virtual sensor density, number of segment divisions for creating a virtual flood map, TP, FP, TN, FN, precision, recall, F value, and average F value.

[0236] ≪Flooding diagram storage means 44≫ The flood map storage means 44 has the function of storing the flood map calculated by the flood map calculation means 370 and the tile image in the storage device 3003. The flood map storage means 44 also has the function of reading the tile image by the flood map transmission control means 46. Here, the flood map may be stored in the storage device 3003 as a GeoTIFF file, or metadata such as the coordinate reference system and bounding boxes, and the flood depth of the grid cells may be stored based on a predetermined database management method. The tile image is stored in a directory of the storage device 3003, for example, " / 17 / 114460 / ", with a file name, for example, "52432.png". Here, "17", "114460", and "52432" are the values ​​of the zoom level Z, tile coordinate X, and tile coordinate Y in the XYZ method. This XYZ method is a known technology used to efficiently distribute map information in communication between a web server and a web browser.

[0237] ≪Flooding diagram transmission control means 46≫ The flood map transmission control means 46 has the function of receiving transmission requests from the web browser of the terminal device 20 via the communication network 90. ​​It also has the function of reading tile images stored in the flood map storage means 44 and storing them in the memory 3002 in response to the received transmission request. Furthermore, it has the function of transmitting the tile images stored in the memory 3002 to the web browser of the terminal device 20 via the communication network 90. ​​The flood map transmission control means 46 may be a computer program called a web server, or it may be a combination of a web server and a computer program that cooperates with the web server.

[0238] <<An example of a flowchart for creating a flood map>> Here, an example of the operation for creating the flood diagram in this embodiment will be described below with reference to the flowchart in Figure 26.

[0239] Based on instructions from a program that runs periodically, such as cron, the arithmetic unit 3001 of the server device 30 starts the process of creating a flood map. Here, cron is a daemon program in UNIX® and Linux® that starts a specific program at a specified date and time.

[0240] Subsequently, the arithmetic unit 3001 of the server device 30 executes the process of acquiring the HAND model (ST11). In this step ST11, the arithmetic unit 3001 of the server device 30 acquires the HAND model stored in the storage device 3003. As a result, the HAND model is stored in the memory 3002.

[0241] Furthermore, as shown in Figure 12, if the direction of river flow is inclined from the north-south or east-west direction, or if the river meanders and the average direction of flow is inclined from the north-south or east-west direction, the image coordinate space (row number, column number) of the HAND model may be rotated. Then, proceed to step ST12.

[0242] In step ST12, the arithmetic unit 3001 of the server device 30 executes the process of acquiring dynamic sensing information. In this step ST12, the arithmetic unit 3001 of the server device 30 acquires the dynamic sensing information stored in the dynamic sensing information storage means 40. The arithmetic unit 3001 of the server device 30 also calculates the measured sensor density based on the dynamic sensing information. As a result, the dynamic sensing information is stored in the memory 3002. The measured sensor density is also stored in the memory 3002. After that, the process proceeds to step ST13.

[0243] In step ST13, the calculation unit 3001 of the server device 30 performs the calculation of the measured HAND water level. In step ST13, if the dynamic sensing information is related to the flood sensor, the calculation unit 3001 of the server device 30 adds the height of the flood sensor to the HAND ground level at the location where the flood sensor is installed to obtain the measured HAND water level. If the dynamic sensing information is related to SNS, the calculation unit 3001 of the server device 30 uses the HAND ground level at the location of the SNS post as the measured HAND water level. If the dynamic sensing information is related to aerial photographs, the calculation unit 3001 of the server device 30 uses the HAND ground level within the flooded area determined from the aerial photographs as the measured HAND water level. If the dynamic sensing information is related to satellite images, the calculation unit 3001 of the server device 30 uses the HAND ground level within the flooded area determined from the satellite images as the measured HAND water level. As a result, the measured HAND water level is stored in the memory 3002. Then, proceed to step ST14.

[0244] In step ST14, the arithmetic unit 3001 of the server device 30 performs a determination of the end of the dynamic sensing information list. In step ST14, if there is only one dynamic sensing information item (YES in ST14), the process proceeds to step ST15. In step ST14, if there are multiple dynamic sensing information items, and it is the last dynamic sensing information item (YES in ST14), the process proceeds to step ST15. In step ST14, if there are multiple dynamic sensing information items, and it is not the last dynamic sensing information item (NO in ST14), the process returns to step ST13, and steps ST13 to ST14 are repeated until the last dynamic sensing information item is found.

[0245] In step ST15, the calculation unit 3001 of the server device 30 performs the process of calculating the number of segments for creating the flood map. In this step ST15, the calculation unit 3001 of the server device 30 calculates the number of segments for creating the flood map based on the measured sensor density calculated in step ST12 and the average evaluation index value stored in the evaluation index value storage means 42. As a result, the number of segments for creating the flood map is stored in the memory 3002. After that, the process proceeds to step ST16.

[0246] In step ST16, the calculation unit 3001 of the server device 30 executes the segment list setting process. In this step ST16, the calculation unit 3001 of the server device 30 divides a predetermined area into multiple segments and sets the segment list based on the HAND model acquired in step ST11 and the number of segments for flood map creation calculated in step ST15. As a result, the segment list is stored in memory 3002. After that, the process proceeds to step ST17.

[0247] In step ST17, the calculation unit 3001 of the server device 30 performs the process of calculating the maximum measured HAND water level. In this step ST17, the calculation unit 3001 of the server device 30 calculates the maximum measured HAND water level for a segment selected from the segment list, based on the segment list set in step ST16 and the measured HAND water level calculated in steps ST13 to ST14, and adds it to the list of maximum measured HAND water level values. As a result, the list of maximum measured HAND water level values ​​is stored in the memory 3002. After that, the process proceeds to step ST18.

[0248] In step ST18, the arithmetic unit 3001 of the server device 30 performs a determination of the end of the segment list. If it is the last segment in the segment list in step ST18 (YES in ST18), the process proceeds to step ST19. If it is not the last segment in the segment list in step ST18 (NO in ST18), the process returns to step ST17, and steps ST17 to ST18 are repeated until it becomes the last segment in the segment list.

[0249] In step ST19, the calculation unit 3001 of the server device 30 performs the calculation of an interpolation function for creating a flood map. In this step ST19, the calculation unit 3001 of the server device 30 calculates an interpolation function for creating a flood map based on the list of measured HAND water level maximum values ​​calculated in steps ST17 to ST18, and the row number and column number in the image coordinate space (row number, column number). As a result, the interpolation function for creating a flood map is stored in memory 3002. After that, the process proceeds to step ST20.

[0250] In step ST20, the calculation unit 3001 of the server device 30 performs the process of calculating the flood map. In step ST20, the calculation unit 3001 of the server device 30 calculates the flood map based on the HAND model acquired in step ST11 and the interpolation function for creating the flood map calculated in step ST19. The calculation unit 3001 of the server device 30 also calculates tile images suitable for use in a web application based on the flood map calculated in step ST20. As a result, the flood map is stored in the flood map storage means 44. The tile images are also stored in the storage device 3003.

[0251] After that, the process of creating the flood map is completed.

[0252] <<An example of a flowchart for creating a virtual flood map>> Next, an example of the operation for creating a virtual flood map in this embodiment will be described below with reference to the flowchart in Figure 27.

[0253] Based on instructions from the user via the input unit 3006 of the server device 30, the calculation unit 3001 of the server device 30 starts the process of creating a virtual flood diagram.

[0254] Subsequently, the arithmetic unit 3001 of the server device 30 executes the process of acquiring the HAND model (ST31). In this step ST31, the arithmetic unit 3001 of the server device 30 acquires the HAND model stored in the storage device 3003. As a result, the HAND model is stored in the memory 3002.

[0255] Furthermore, as shown in Figure 12, if the direction of river flow is inclined from the north-south or east-west direction, or if the river meanders and the average direction of flow is inclined from the north-south or east-west direction, the image coordinate space (row number, column number) of the HAND model may be rotated. Then, proceed to step ST32.

[0256] In step ST32, the arithmetic unit 3001 of the server device 30 performs the process of acquiring existing flood information. In this step ST32, the arithmetic unit 3001 of the server device 30 acquires the existing flood information stored in the storage device 3003. As a result, the existing flood information is stored in the memory 3002.

[0257] Furthermore, as shown in Figure 12, if the direction of river flow is inclined from the north-south or east-west direction, or if the river meanders and the average direction of flow is inclined from the north-south or east-west direction, the image coordinate space (row number, column number) of the existing flood information may be rotated. After that, proceed to step ST33.

[0258] In step ST33, the arithmetic unit 3001 of the server device 30 executes the process of setting the virtual flood map creation condition list. In this step ST33, the server device sets a list of the number of segments to be divided into for creating the virtual flood map, a list of the number of virtual flood sensors, and the number of repetitions for the random arrangement of virtual flood sensors. Here, the list of the number of segments to be divided into for creating the virtual flood map is, for example, a list of positive integers such as 2, 3, 4, 5, 6, 7, 8, 9, and 10. The list of the number of virtual flood sensors is, for example, a list of positive integers such as 10, 20, 30, 40, 50, 60, 70, 80, 90, and 100. Furthermore, the number of repetitions for the random arrangement of virtual flood sensors is, for example, a positive integer such as 10. As a result, the list of the number of segments to be divided into for creating the virtual flood map, the list of virtual flood sensors, and the number of repetitions for the random arrangement of virtual flood sensors are stored in memory 3002. Then, proceed to step ST34.

[0259] In step ST34, the arithmetic unit 3001 of the server device 30 executes the process of generating virtual dynamic sensing information. In step ST34, the arithmetic unit 3001 of the server device 30 calculates virtual dynamic sensing information based on the existing flood information acquired in step ST32, the list of virtual flood sensors set in step ST33, random latitude numbers, and random longitude numbers. The arithmetic unit 3001 of the server device 30 also calculates the virtual sensor density based on the virtual dynamic sensing information calculated in step ST34. As a result, the virtual dynamic sensing information is stored in the memory 3002. The virtual sensor density is also stored in the memory 3002. After that, the process proceeds to step ST35.

[0260] In step ST35, the calculation unit 3001 of the server device 30 performs the calculation of the virtual HAND water level. In this step ST35, the calculation unit 3001 of the server device 30 uses the HAND ground level at the virtual flood sensor installation location as the virtual HAND water level. As a result, the virtual HAND water level is stored in memory 3002. After that, the process proceeds to step ST36.

[0261] In step ST36, if there is one virtual dynamic sensing information (YES in ST36), proceed to step ST37. In step ST36, if there are multiple virtual dynamic sensing information items, and it is the last virtual dynamic sensing information item (YES in ST36), proceed to step ST37. In step ST36, if there are multiple virtual dynamic sensing information items, and it is not the last virtual dynamic sensing information item (NO in ST36), return to step ST35, and steps ST35 to ST36 are repeated until the last virtual dynamic sensing information item is found.

[0262] In step ST37, the arithmetic unit 3001 of the server device 30 executes the segment list setting process. In this step ST37, the arithmetic unit 3001 of the server device 30 divides a predetermined area into multiple segments and sets a segment list based on the HAND model acquired in step ST31 and the number of segments for creating a virtual flood map selected from the list of number of segments for creating a virtual flood map set in step ST33. As a result, the segment list is stored in memory 3002. After that, the process proceeds to step ST38.

[0263] In step ST38, the arithmetic unit 3001 of the server device 30 performs the calculation of the maximum virtual HAND water level. In this step ST38, the arithmetic unit 3001 of the server device 30 calculates the maximum virtual HAND water level for a segment selected from the segment list, based on the segment list set in step ST37 and the virtual HAND water level calculated in steps ST35 to ST36, and adds it to the list of maximum virtual HAND water level values. As a result, the list of maximum virtual HAND water level values ​​is stored in memory 3002. After that, the process proceeds to step ST39.

[0264] In step ST39, if it is the last segment in the segment list (YES in ST39), proceed to step ST40. In step ST39, if it is not the last segment in the segment list (NO in ST39), return to step ST38, and steps ST38 to ST39 are repeated until it becomes the last segment in the segment list.

[0265] In step ST40, the arithmetic unit 3001 of the server device 30 performs the calculation of an interpolation function for creating a virtual flood map. In this step ST40, the arithmetic unit 3001 of the server device 30 calculates an interpolation function for creating a virtual flood map based on the list of maximum virtual HAND water level values ​​calculated in steps ST38 to ST39, and the row number and column number in the image coordinate space (row number, column number). As a result, the interpolation function for creating a virtual flood map is stored in memory 3002. After that, the process proceeds to step ST41.

[0266] In step ST41, the arithmetic unit 3001 of the server device 30 performs the calculation of a virtual flood diagram. In step ST41, the arithmetic unit 3001 of the server device 30 calculates a virtual flood diagram based on the HAND model acquired in step ST31 and the interpolation function for creating a virtual flood diagram calculated in step ST40. As a result, the virtual flood diagram is stored in memory 3002. After that, the process proceeds to step ST42.

[0267] In step ST42, the calculation unit 3001 of the server device 30 performs the process of calculating the average evaluation index value. In this step ST42, the calculation unit 3001 of the server device 30 calculates the average evaluation index value based on the existing flood information acquired in step ST32 and the virtual flood diagram calculated in step ST41. As a result, the average evaluation index value is stored in the evaluation index value storage means 42. After that, the process proceeds to step ST43.

[0268] In step ST43, the arithmetic unit 3001 of the server device 30 performs a termination check of the virtual flood map creation condition list. In step ST43, if the number of segments to be created for virtual flood map creation is the last in the list of number of segments to be created for virtual flood map creation, and the number of virtual flood sensors is the last in the list of number of virtual flood sensors, and the number of repetitions of the random arrangement of virtual flood sensors has been reached (YES in ST43), the virtual flood map creation operation is completed. In step ST43, if the number of segments to be created for virtual flood map creation is not the last in the list of number of segments to be created for virtual flood map creation, or if the number of virtual flood sensors is not the last in the list of number of virtual flood sensors, or if the number of repetitions of the random arrangement of virtual flood sensors has not been reached (NO in ST43), the process returns to step ST34, and steps ST34 to ST43 are repeated until the last number of segments to be created for virtual flood map creation, the last number of virtual flood sensors, and the number of repetitions of the random arrangement of virtual flood sensors are reached.

[0269] <<<Note>>> The features of the above-described embodiment are noted below.

[0270] The information processing device includes a dynamic sensing information storage means, a flood diagram storage means, a dynamic sensing information acquisition means, a HAND model acquisition means, a measured HAND water level calculation means, a segment discrimination means for creating a flood diagram, a measured HAND water level maximum value calculation means, an interpolation function calculation means for creating a flood diagram, and a flood diagram calculation means.

[0271] (a) In this configuration, the dynamic sensing information storage means stores dynamic sensing information to which data such as an identification ID for the flood sensor, sensor ID, observation date and time, location of the flood sensor (latitude, longitude), height of the flood sensor, and presence or absence of flooding are associated. (b) In this configuration, the flood map storage means stores the data of the flood map calculated by the flood map calculation means. (c) In this configuration, the dynamic sensing information acquisition means acquires dynamic sensing information from the dynamic sensing information storage means, including an identification ID for the flood sensor, a sensor ID, observation date and time, the location of the flood sensor (latitude, longitude), the height at which the flood sensor is installed, and whether or not there is flooding. (d) In this configuration, the HAND model acquisition means acquires HAND model data and river centerline data from the storage device. In this embodiment, the river centerline data is the river data from the National Land Numerical Information published by the Geospatial Information Authority of Japan. The HAND model data is calculated using the technology described in Non-Patent Document 1, based on the Fundamental Geospatial Information published by the Geospatial Information Authority of Japan. (e) In this configuration, the measured HAND water level calculation means calculates the measured HAND water level data for the flood sensor installation location (latitude, longitude) by adding the HAND ground level of the flood sensor installation location (latitude, longitude) to the flood sensor installation height of the flood sensor installation location (latitude, longitude). (f) In this configuration, the inundation map creation segment discrimination means calculates segment data obtained by dividing a predetermined area perpendicular to the average flow direction of the river (cross-sectional direction) based on data for a predetermined number of segments to be divided for inundation map creation. Furthermore, based on data on whether or not there is a division at a predetermined river centerline and data on the river centerline, if necessary, it calculates segment data obtained by further dividing the predetermined area longitudinally at the river centerline. (g) In this configuration, the measured HAND water level maximum value calculation means calculates the maximum value of the measured HAND water level in each segment based on the measured HAND water level data at the location (latitude, longitude) of the flood sensor and the segment data. (h) In this configuration, the interpolation function calculation means for creating flood maps uses interpolation techniques based on radial basis functions to calculate the data for the interpolation function for creating flood maps for each segment, based on the data of the location (latitude, longitude) of the flood sensor where the measured HAND water level was maximum and the data of the maximum measured HAND water level. Here, the interpolation function for creating flood maps is expressed as a weighted linear combination of radial basis functions. (i) In this configuration, the flood map calculation means calculates the flood depth data for each latitude and longitude by subtracting the HAND ground elevation for each latitude and longitude from the flood map creation interpolation function for each latitude and longitude, based on the data of the interpolation function for flood map creation and the HAND ground elevation data included in the HAND model. (j) In this configuration, the flood map calculation means calculates flood map data for the entire predetermined area based on the flood depth data for each latitude and longitude, and stores the said flood map data in the flood map storage means.

[0272] As described above, in this configuration, the maximum value of the measured HAND water level is calculated in each segment, and an interpolation function for creating flood maps is calculated using interpolation technology with a radial basis function. Therefore, it is possible to prevent the water level and flood depth from being underestimated in low-lying areas where flood depth tends to be deep, and near flood sensors where the installation height of the flood sensors is small. Furthermore, since the measured HAND water level is calculated using the HAND ground level, the ground gradient in the direction of river flow is removed, making it possible to compare the magnitudes of the measured HAND water levels even when one or more flood sensors are installed in a given area. As a result, this configuration improves the accuracy of the water level and flood depth at each latitude and longitude in the geographic coordinate space in a given area, and enables the calculation of highly accurate flood map data even in information processing devices that employ flood sensors.

[0273] 2) In the configuration of 1) above, it is preferable that the information processing device of the present invention further comprises: an evaluation index value storage means; an existing flood information acquisition means; a virtual flood diagram creation condition setting means; a virtual dynamic sensing information generation means; a virtual HAND water level calculation means; a virtual HAND water level maximum value calculation means; an interpolation function calculation means for creating a virtual flood diagram; a virtual flood diagram calculation means; and an evaluation index value average calculation means.

[0274] (a) In this configuration, the evaluation index value storage means stores data of the average evaluation index value associated with an identification ID for the evaluation index value, virtual sensor density, number of segments for creating a virtual flood map, presence or absence of division at the river centerline, TP, FP, TN, FN, precision, recall, F value, and average F value. (b) In this configuration, the means for acquiring existing flood information acquires existing flood information from a storage device. Here, the existing flood information may be data from the maximum-scale flood inundation area map published by the Geospatial Information Authority of Japan, or data from the planned-scale flood inundation area map published by the Geospatial Information Authority of Japan, or data from inundation maps created from past flood records, or data from inundation maps calculated by solving partial differential equations such as the two-dimensional advection-diffusion equation. (c) In this configuration, the virtual flood map creation condition setting means sets a list of the number of segments to be divided for creating the virtual flood map. The virtual flood map creation condition setting means also sets a list of the number of virtual flood sensors. Furthermore, the virtual flood map creation condition setting means sets the number of repetitions for the random arrangement of virtual flood sensors. In addition to these, it sets a list of whether or not to divide along the river centerline. (d) In this configuration, the virtual dynamic sensing information generation means calculates virtual dynamic sensing information based on existing flood information and a list of virtual flood sensors. Here, virtual dynamic sensing information is data that associates an identification ID to identify the virtual dynamic sensing information, the location of the virtual flood sensor (latitude, longitude), and whether or not there is flooding. The location of the virtual flood sensor (latitude, longitude) is calculated using random numbers calculated by the server device's calculation unit for each virtual flood sensor. Furthermore, the location of the virtual flood sensor (latitude, longitude) is calculated so that it is included in the flood range of the existing flood information. (e) In this configuration, the virtual dynamic sensing information generation means calculates virtual sensor density data based on the virtual dynamic sensing information and the area data of the flood-prone area. (f) In this configuration, the virtual HAND water level calculation means uses the data of the virtual flood sensor installation location (latitude, longitude) included in the virtual dynamic sensing information and the HAND ground level data included in the HAND model to set the HAND ground level of the virtual flood sensor installation location (latitude, longitude) as the data of the virtual HAND water level of the virtual flood sensor installation location (latitude, longitude). (g) In this configuration, the virtual HAND water level maximum value calculation means calculates segment data for a predetermined area divided perpendicular to the average flow direction of the river (cross-sectional direction) based on the HAND model data and the list of the number of segments for creating a virtual flood map. In addition, based on the list of whether or not division is made along the river centerline, if necessary, it calculates segment data for further division in the longitudinal direction along the river centerline. (h) In this configuration, the virtual HAND water level maximum value calculation means calculates the maximum value of the virtual HAND water level in each segment based on the virtual HAND water level data of the virtual flood sensor installation location (latitude, longitude) and the segment data. (i) In this configuration, the interpolation function calculation means for creating a virtual flood map calculates the data for the interpolation function for creating a virtual flood map based on the data of the virtual flood sensor installation location (latitude, longitude) where the virtual HAND water level was maximized and the data of the maximum virtual HAND water level for each segment, using interpolation techniques based on radial basis functions. Here, the interpolation function for creating a virtual flood map is expressed as a weighted linear combination of radial basis functions. (j) In this configuration, the virtual flood map calculation means calculates the virtual flood depth data for each latitude and longitude by subtracting the HAND ground elevation for each latitude and longitude from the virtual flood map creation interpolation function for each latitude and longitude, based on the data of the interpolation function for creating the virtual flood map and the HAND ground elevation data included in the HAND model. (k) In this configuration, the virtual flood map calculation means calculates virtual flood map data for the entire predetermined area based on the virtual flood depth data for each latitude and longitude. (l) In this configuration, the evaluation index value average calculation means calculates TP, FP, TN, FN, precision, recall, F value, and average F value data based on existing flood information and virtual flood map data, associates these with data on identification ID, virtual sensor density, number of segments for creating virtual flood map, and whether or not there is division at the river centerline, and stores them in the evaluation index value storage means. (m) If the list of the number of segment divisions for creating the virtual flood map, the list of the number of virtual flood sensors, the number of repetitions of the random placement of virtual flood sensors, and the list of whether or not divisions are made along the river centerline are not the last values, return to (d) and repeat the calculation of virtual dynamic sensing information, virtual sensor density data, virtual HAND water level data, segment data, virtual HAND water level maximum data, data for the interpolation function for creating the virtual flood map, data for virtual flood depth, data for the virtual flood map, and data for the average evaluation index values. (n) In this configuration, the dynamic sensing information acquisition means calculates measured sensor density data based on the dynamic sensing information and the area of ​​the flood-prone zone in a predetermined region. (o) In this configuration, the segment determination means for creating flood maps determines the number of segments for creating flood maps that maximizes the average F value, based on the measured sensor density data and the average evaluation index value data. Furthermore, based on the average evaluation index value data, it determines the necessity of dividing a predetermined area in a direction perpendicular to the average flow direction of the river (transverse direction), as well as dividing it along the river centerline. The necessity of dividing along the river centerline is F divide ≥F notdivide In this case, it is determined to be divided. Also, F divide <F notdivide In this case, it is determined that it will not be divided. Here, F divide This refers to the average F value when the number of segments used for creating the flood map is divided along the river centerline. notdivide This refers to the average F-value when there is no segmentation along the river centerline in the number of segments used to create the flood map. (p) In this configuration, the segment determination means for creating flood maps calculates segment data obtained by dividing a predetermined area perpendicular to the average flow direction of the river (cross-sectional direction) based on the data of the number of segments to be divided for creating flood maps as determined by determination. Furthermore, based on the necessity of division along the river centerline as determined by determination and the data of the river centerline, if necessary, it calculates segment data obtained by further dividing the predetermined area longitudinally along the river centerline.

[0275] As described above, in this configuration, the number of segments for flood map creation that maximizes the average F value is calculated using the measured sensor density data and the average evaluation index value data, so the optimal number of segments for flood map creation is automatically determined. In addition, the necessity of dividing a given area longitudinally along the centerline of the river is determined using the average evaluation index value data, so it is possible to prevent unnecessary reductions in the number of flood sensors per segment due to unnecessary segment divisions that do not contribute to accuracy improvement. As a result, in this configuration, the accuracy of water surface height and flood depth at each latitude and longitude of the geospatial coordinates is improved in a given area, and the information processing device employing flood sensors can calculate highly accurate flood map data.

[0276] 3) In the configuration described in 2) above, it is preferable that the dynamic sensing information storage means further stores data related to SNS. Here, data related to SNS refers to the SNS ID that identifies the SNS post, the date and time of posting, the location of posting (latitude, longitude), image information (still image) attached at the time of SNS posting, and data on whether or not there is flooding.

[0277] As described above, this configuration has a higher density of actual sensors compared to configuration 2) above, and therefore, even an information processing device that employs flood sensors and SNS images (still images) can calculate highly accurate flood map data.

[0278] 4) In the configuration described in 3) above, it is preferable that the dynamic sensing information storage means further stores data relating to aerial photographs. Here, data relating to aerial photographs refers to an identification ID for identifying polygons of the flooded area, an aerial photograph ID for identifying the aerial photograph, the date and time of shooting, the shooting location (latitude, longitude), the image information to be captured (still image), and the flooded area (latitude, longitude of the vertices of the outer boundary).

[0279] As described above, this configuration has a higher density of actual sensors compared to configuration 3) above, and therefore, even an information processing device that employs flood sensors, SNS images (still images), and aerial photographs (still images) can calculate highly accurate flood map data.

[0280] 5) In the configuration of 4) above, it is preferable that the dynamic sensing information storage means further stores data relating to satellite images. Here, the data relating to satellite images includes an identification ID for identifying polygons of the flooded area, a satellite image ID for identifying the satellite image, observation date and time, observation location (latitude, longitude), observed satellite image (SAR image), and flooded area (latitude, longitude of the vertices of the outer perimeter).

[0281] As described above, this configuration has a higher density of actual sensors compared to the configuration in 4) above, and therefore, even in an information processing device that employs flood sensors, SNS images (still images), aerial photographs (still images), and satellite images (SAR images), it is possible to calculate highly accurate flood map data.

[0282] 6) In the configuration described in 5) above, it is preferable that the information processing device of the present invention further includes a means for correcting the flood diagram.

[0283] (a) In this configuration, the flood map correction means identifies data where there is flooding in the dynamic sensing information, but there is no flooding in the flood map data. (b) In this configuration, the flood map correction means, based on the data of the discrimination result, if necessary, adds the data of the location (latitude, longitude) of the flood sensor and the data of the measured HAND water level at the location (latitude, longitude) of the flood sensor to the list of the maximum measured HAND water level, calculates the corrected flood map data, and stores the corrected flood map data in the flood map storage means.

[0284] As described above, this configuration allows the use of data where flooding is present in the dynamic sensing information but not in the flood map data. Therefore, even in information processing devices that employ flood sensors, SNS images (still images), aerial photographs (still images), and satellite images (SAR images), it is possible to calculate highly accurate flood map data.

[0285] 7) In the configuration described in 6) above, it is preferable that the information processing device of the present invention further includes an abnormal value detection means.

[0286] (a) In this configuration, the abnormal value detection means determines that the data from one of the flood sensors may be abnormal if, in a certain segment, a flood sensor with a smaller measured HAND water level shows no flooding, while a flood sensor with a larger measured HAND water level shows flooding. (b) In this configuration, the abnormal value detection means calculates the number of flood sensors that may have abnormal values ​​and determines which flood sensor's data is abnormal by majority vote. (c) In this configuration, the abnormal value detection means removes the data from the measured HAND water level data based on the discrimination result data if necessary, calculates the corrected flood map data, and stores the corrected flood map data in the flood map storage means.

[0287] As described above, in this configuration, the flood sensor can eliminate abnormal values ​​such as mud adhesion, which would otherwise cause flooding to occur even when there is no flooding. Therefore, an information processing device that employs a flood sensor, SNS images (still images), aerial photographs (still images), and satellite images (SAR images) can calculate highly accurate flood map data. [Explanation of Symbols]

[0288] 1…Information Processing Device 10…Water ingress sensor 11…SNS 12...Aircraft 13…Artificial satellite 20…Terminal device 30…Server equipment 32... Sensing information reception determination means 34... Common processing means 36…Methods for creating flood maps 38…Method for creating a virtual flood map 40... Dynamic sensing information storage means 42...Means for storing evaluation index values 44...Method for storing flood diagrams 46... Flooding diagram transmission control means 90…Communication Network 320...Water ingress sensor receiving control means 322…SNS discrimination method 324... Aerial photograph identification method 326... Satellite image discrimination method 340…Method for acquiring HAND models 360...Means for acquiring dynamic sensing information 362...Measured HAND water level calculation method 364... Segment identification means for creating flood maps 366...Measurement method for calculating the maximum value of water level using HAND 368...Method for calculating interpolation function for creating flood maps 370...Method for calculating flood maps 372...Method for correcting flood maps 374... Abnormal Value Detection Means 380…Existing methods for obtaining flood information 382…Method for setting conditions for creating a virtual flood map 384... Virtual Dynamic Sensing Information Generation Means 386…Virtual HAND water level calculation method 388…Method for calculating the maximum value of virtual HAND water level 390...Method for calculating interpolation function for creating virtual flood diagrams 392...Method for calculating virtual flood map 394...Method for calculating the average of evaluation index values 2001...Arithmetic section 2002…Memory 2003…Storage device 2004... Communications Department 2005…Display section 2006... Input section 3001...Arithmetic section 3002...Memory 3003...Storage device 3004... Communications Department 3005...Display section 3006...Input section

Claims

1. An information processing device that creates a flood map of an area to be evaluated, including rivers, A means for acquiring a HAND model, in which the HAND ground elevation, which is the relative elevation difference with the nearest drainage point, is used as the value of the grid cell, and the HAND model is rotated so that the X-axis direction is the direction of river flow. The system includes a flood map creation means that creates a flood map based on the aforementioned HAND model and dynamic sensing information including information about flooding, The means for creating the flood map is, A measured HAND water level calculation means calculates the measured HAND water level based on the HAND ground height and the dynamic sensing information for each location where the dynamic sensing information is acquired. A segment determination means for creating a flood map, which divides the evaluation target area in a direction intersecting the flow direction of the river and sets up multiple segments, For each of the aforementioned segments, a means for calculating the maximum value of the measured HAND water level is provided, An interpolation function calculation means for creating a flood diagram calculates an interpolation function by projecting the maximum value of the measured HAND water level onto a projection plane with the river flow direction as the horizontal axis and the measured HAND water level as the vertical axis, and calculating the interpolation function. An information processing device comprising: an inundation map calculation means that associates a calculated value obtained from the interpolation function with each grid cell in the HAND model that is continuous in the X-axis direction, and associates the same calculated value with adjacent grid cells in the Y-axis direction; calculates the inundation depth for each grid cell by subtracting the HAND ground height from the calculated value; determines that grid cells with an inundation depth greater than 0 are inundated; and creates an inundation map.

2. In the information processing apparatus according to claim 1, The information processing device is characterized in that the flood map creation means includes a flood map correction means for correcting the flood map so that the grid cells of the flood map corresponding to locations where flooding is determined to be present based on the dynamic sensing information are determined to be flooded.

3. In the information processing apparatus according to claim 1, The aforementioned dynamic sensing information is information obtained from at least one of the following: flood sensors, aerial photographs, satellite images, or SNS posts. An information processing device characterized in that, when information indicating flooding is obtained based on aerial photographs, satellite images, and SNS posts, the HAND ground elevation at the acquisition location is set to the measured HAND water level.

4. In the information processing apparatus described in claim 3, The information processing apparatus is characterized in that the flood diagram creation means includes an abnormal value detection means for detecting abnormal values ​​contained in the flood sensor for each of the multiple segments.

5. In the information processing apparatus according to claim 1, The system further comprises a virtual flood map creation means that creates a virtual flood map of the area to be evaluated based on the aforementioned HAND model and existing flood information. The means for creating the virtual flood map is, A virtual dynamic sensing information generation means creates virtual dynamic sensing information by arbitrarily placing multiple virtual flood sensors in the flooded area acquired from the aforementioned existing flood information, A virtual HAND water level calculation means that sets the HAND ground level at the placement location of the virtual flood sensor as the virtual HAND water level, The evaluation target area is divided into multiple segments in a direction intersecting the flow direction of the river, and a virtual HAND water level maximum value calculation means is provided to obtain the maximum value of the virtual HAND water level for each segment. An interpolation function calculation means for creating a virtual flood map, which projects the maximum value of the virtual HAND water level onto a projection plane with the river flow direction as the horizontal axis and the virtual HAND water level as the vertical axis, and calculates an interpolation function, An information processing device comprising: a virtual flood map calculation means that associates each of the grid cells in the HAND model that are continuous in the X-axis direction with a calculated value obtained from an interpolation function acquired by the virtual flood map creation interpolation function calculation means, and associates the same calculated value with adjacent grid cells in the Y-axis direction; calculates a virtual flood depth for each grid cell by subtracting the HAND ground height from the calculated value; determines that grid cells with a virtual flood depth greater than 0 are flooded; and creates a virtual flood map.

6. In the information processing apparatus described in claim 5, An information processing device characterized in that the aforementioned existing flood information includes existing flood inundation prediction information or past flood performance information.

7. In the information processing apparatus described in claim 5, The virtual flood map creation means further comprises an evaluation index value average calculation means, The means for calculating the average of the evaluation index values ​​is: For each of the multiple virtual flood diagrams created by arbitrarily changing the number of segments, the number of virtual flood sensors, and the placement positions of the virtual flood sensors, each grid cell is compared with the existing flood information to obtain the F-value, which is the harmonic mean of the precision and recall rates related to the presence or absence of flooding. The average F value of the virtual flood diagrams with the same virtual sensor density is calculated for each segment division. The aforementioned virtual sensor density is An information processing device characterized by the number of virtual flood sensors relative to the area of ​​the flooded area in the evaluation target region, obtained from the existing flooding information.

8. In the information processing apparatus according to claim 7, The inundation map creation segment discrimination means provided in the inundation map creation means, An interpolation function is calculated based on the three-dimensional scatter data consisting of the number of segment divisions, the virtual sensor density, and the average F value. Multiple segments are set in the evaluation target area, with the number of segments being such that the average F value calculated based on the interpolation function and the measured sensor density is maximized. The information processing device is characterized in that the measured sensor density is the total number of acquisition locations of the dynamic sensing information relative to the area of ​​the flood-prone area in the evaluation target region.

9. A program for causing a computer to function as an information processing device according to any one of claims 1 to 8.

Images