Method and device for evaluating loss scale of disaster loss area

Abstract

In an embodiment of the present specification, a working method of a server for evaluating a loss scale of a disaster loss area can be provided. At this time, the working method of the server can include a step of acquiring at least one or more first disaster images, a step of deriving a loss area from the at least one or more first disaster images, respectively, and acquiring loss area-related information by tagging based on the derived loss area, and a step of learning a first learning model using the at least one or more first disaster images and the loss area-related information.

Description

Technical Field

[0001] The embodiments relate to a method and apparatus for assessing the scale of losses in a disaster-affected area. In particular, the embodiments relate to a method and apparatus for identifying a disaster-affected area using features of the surrounding area, and assessing the scale of losses and compensation for the identified area. Background Art

[0002] When natural disasters occur, significant human resources are required to identify and accurately assess the losses. Furthermore, it can be difficult to gather sufficient information using human resources alone to accurately assess and calculate the scale of losses.

[0003] In consideration of the above-mentioned problems, some technologies have been developed to obtain disaster loss information by using drones or satellites to obtain aerial images of disaster-damaged areas and analyzing them. However, even when using drones or satellites to obtain aerial images of disaster-damaged areas, identifying disaster losses solely through images is still subject to certain limitations. In particular, disaster loss images, as aerial images, cannot reflect detailed information related to disaster losses, but can only reflect the general situation related to disaster losses. Therefore, they are also subject to certain limitations when it comes to specific disaster loss information. In consideration of the above-mentioned problems, the following will analyze the characteristics of disaster loss trends and even disaster-surrounding areas through aerial images, and then analyze additional information related to the characteristics of disaster-surrounding areas based on artificial intelligence (AI), thereby explaining a method for identifying disaster loss areas.

[0004] In addition, after deriving the disaster loss area, it may be necessary to determine the loss scale and compensation scale while taking into account the characteristics of the loss area. This will be explained below.

[0005] Prior art literature

[0006] Patent Literature

[0007] (Patent Document 1) Korean Registered Patent No. 10-2203135 Summary of the Invention

[0008] This specification relates to a method and apparatus for identifying disaster-damaged areas using features of disaster-prone areas.

[0009] This specification can provide a method and apparatus for extracting feature information from a disaster-prone area and simultaneously using the extracted feature information and surrounding loss area information to identify the disaster-prone area.

[0010] This specification can provide a method and device for learning based on disaster loss area information and feature information extracted from the disaster surrounding area to identify the disaster loss area.

[0011] This specification relates to a method and apparatus for outputting the scope and type of disaster damage areas by inputting an image of a disaster damage area.

[0012] This specification relates to a method and apparatus for evaluating the scale of loss and compensation in a disaster-damaged area while taking into account the characteristics of the disaster-damaged area.

[0013] This specification relates to a method and apparatus for evaluating the scale of compensation for a disaster-damaged area by taking into account characteristics of the disaster-damaged area, additional data, and related images.

[0014] The problems to be solved by this specification are not limited to the problems described above, but can be expanded to various items that can be derived from the embodiments of the present invention described below.

[0015] In one embodiment of the present disclosure, a server operating method for assessing the scale of damage in a disaster-damaged area may be provided. The server operating method may include: obtaining at least one first disaster image; deriving damage areas from the at least one first disaster image, and obtaining information related to the damage areas by labeling the derived damage areas; learning a first learning model using the at least one first disaster image and the information related to the damage areas; and assessing the scale of damage information for the damage areas in the disaster images based on the first learning model.

[0016] Furthermore, in one embodiment of the present specification, a server for assessing the scale of losses in disaster-damaged areas may be provided. The server may include: a transceiver for communicating with an external device; and a processor for controlling the transceiver. The processor may acquire at least one first disaster image, derive damage areas from each of the at least one first disaster image, obtain information related to the damage areas through labeling based on the derived damage areas, learn a first learning model using the at least one first disaster image and the damage area information, and then assess the scale of losses in the damage areas in the disaster images based on the first learning model.

[0017] In addition, the following items can also be applied to a server for evaluating the scale of loss in a disaster damage area and a method for operating the server.

[0018] In addition, in one embodiment of the present specification, it may also include: a step of acquiring a second disaster image from an external device; a step of deriving a loss area and a surrounding area from the second disaster image, and acquiring a plurality of disaster-related information by labeling based on the derived loss area and the surrounding area; a step of assigning weighted values ​​to the plurality of disaster-related information acquired respectively; a step of inputting the second disaster image and the plurality of disaster-related information into the learned first learning model; and a step of outputting disaster loss area confirmation information and disaster loss type information based on the first learning model.

[0019] In addition, in one embodiment of the present specification, the second disaster image can be input into a second learning model, and the second learning model can deduce the loss area and surrounding area of ​​the second disaster image, and provide multiple disaster-related information as output information through labeling based on the derived second loss area and surrounding area.

[0020] In addition, in one embodiment of the present specification, it may also include: a step of obtaining information on the use of a disaster loss area; a step of confirming the area of ​​the disaster loss area, the type of the disaster loss area, and the use of the disaster loss area based on the disaster loss area confirmation information, the disaster loss type information, and the disaster loss area use information; a step of deriving characteristic information of the disaster loss area based on the confirmed area of ​​the disaster loss area, the type of the disaster loss area, and the use of the disaster loss area; and a step of confirming the loss scale information based on the derived characteristic information of the disaster loss area.

[0021] In addition, in one embodiment of the present specification, it may also include: a step of obtaining at least one third disaster image related to the disaster loss area and at least one external data related to the disaster loss area; the loss scale information can be confirmed in a manner that further reflects at least one third disaster image and at least one external data.

[0022] In addition, in one embodiment of the present specification, weighted values ​​may be assigned to disaster loss area characteristic information, at least one third disaster image, and at least one external data, respectively, and the loss scale information may be confirmed in a manner reflecting the assigned weighted values.

[0023] In addition, in one embodiment of the present specification, disaster loss area characteristic information, at least one third disaster image and at least one external data can be provided as input data of the third learning model, and the third learning model can derive loss scale information as output data based on the input data.

[0024] Furthermore, in one embodiment of the present specification, the loss scale information may be provided as feedback information to the third learning model, and the third learning model may be updated based on the loss scale information.

[0025] In addition, in one embodiment of the present specification, when providing external data as input data of the third learning model, it can be provided as input data of the third learning model after assigning a weighted value to the external data, and the server can obtain the external data from the disaster statistics database.

[0026] In addition, in one embodiment of the present specification, when the server obtains external data from a disaster statistics database, the disaster type can be confirmed based on the at least one first disaster image, and at least one piece of information corresponding to the disaster type can be extracted from the disaster statistics database as external data. The at least one piece of information may include at least one of loss scale information, loss frequency information, loss amount information, death toll information, loss cause information, and fault-related information.

[0027] This specification can provide a method for identifying disaster loss areas using characteristics of disaster surrounding areas.

[0028] This specification can extract feature information from the disaster surrounding area and use the extracted feature information and surrounding loss area information to identify the disaster loss area.

[0029] This manual can learn based on disaster loss area information and feature information extracted from the disaster surrounding area to identify the disaster loss area.

[0030] This specification can output the scope of the disaster damage area and the disaster loss type by inputting a disaster damage area image.

[0031] This specification can evaluate the scale of losses and the scale of compensation in a disaster-damaged area while taking into account the characteristics of the disaster-damaged area.

[0032] This specification allows for the assessment of the scale of compensation for disaster-damaged areas by taking into account the characteristics of the disaster-damaged areas, additional data, and related images.

[0033] The problems to be solved by this specification are not limited to the problems described above, but can be expanded to various items that can be derived from the embodiments of the present invention described below. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 This is a schematic diagram illustrating an example of a working environment of a system applicable to one embodiment of this specification.

[0035] Figure 2 This is a block diagram for explaining the internal structure of the computing device 200 in one embodiment of this specification.

[0036] Figure 3a This is a schematic diagram illustrating a method for obtaining disaster images in one embodiment of this specification.

[0037] Figure 3b This is a schematic diagram illustrating a method for obtaining disaster images in one embodiment of this specification.

[0038] Figure 3c This is a schematic diagram illustrating a method for obtaining disaster images in one embodiment of this specification.

[0039] Figure 4 This is a schematic diagram illustrating a method for labeling disaster images in one embodiment of this specification.

[0040] Figure 5a This is a schematic diagram illustrating a method for identifying a disaster damage area and a disaster surrounding area in one embodiment of this specification.

[0041] Figure 5b This is a schematic diagram illustrating a method for identifying a disaster damage area and a disaster surrounding area in one embodiment of this specification.

[0042] Figure 5c This is a schematic diagram illustrating a method for identifying a disaster damage area and a disaster surrounding area in one embodiment of this specification.

[0043] Figure 6 This is a schematic diagram illustrating a method for extracting features of a disaster surrounding area in one embodiment of this specification.

[0044] Figure 7a This is a schematic diagram illustrating a method for determining the scale of loss based on the disaster loss area in one embodiment of this specification.

[0045] Figure 7b This is a schematic diagram illustrating a method for determining the scale of loss based on the disaster loss area in one embodiment of this specification.

[0046] Figure 8a This is a schematic diagram illustrating a method for determining the scale of loss based on the disaster loss area in one embodiment of this specification.

[0047] Figure 8b This is a schematic diagram illustrating a method for determining the scale of loss based on the disaster loss area in one embodiment of this specification.

[0048] Figure 8c This is a schematic diagram illustrating a method for determining the scale of loss based on the disaster loss area in one embodiment of this specification.

[0049] Figure 9a This is a schematic diagram illustrating a method for determining the scale of loss based on the disaster loss area in one embodiment of this specification.

[0050] Figure 9b This is a schematic diagram illustrating a method for determining the scale of loss based on the disaster loss area in one embodiment of this specification.

[0051] Figure 9c This is a schematic diagram illustrating a method for determining the scale of loss based on the disaster loss area in one embodiment of this specification.

[0052] Figure 10a This is a schematic diagram illustrating a method for constructing a learning model based on disaster images in one embodiment of this specification.

[0053] Figure 10b This is a schematic diagram illustrating a method for constructing a learning model based on disaster images in one embodiment of this specification.

[0054] Figure 11a This is a schematic diagram illustrating a method for determining the scale of disaster losses and the scale of compensation based on the disaster loss area in one embodiment of this specification.

[0055] Figure 11b This is a schematic diagram illustrating a method for determining the scale of disaster losses and the scale of compensation based on the disaster loss area in one embodiment of this specification.

[0056] Figure 12 This is a sequence diagram illustrating a method for identifying a disaster damage area in one embodiment of this specification. DETAILED DESCRIPTION

[0057] In the process of describing the embodiments of this specification, if it is determined that the detailed description of well-known structures or functions may make the main points of the embodiments of this specification unclear, the relevant detailed description will be omitted. In addition, in the drawings, parts that are not related to the description of the embodiments of this specification are omitted, and similar parts are assigned similar figure numbers.

[0058] In the embodiments of this specification, when a component is described as being "connected," "coupled," or "in contact with" another component, this includes not only a direct connection relationship but also an indirect connection relationship where another component exists between the two. In addition, when a component is described as "including" or "having" other components, unless otherwise expressly stated, this does not exclude other components but rather means that other components may be included.

[0059] In the embodiments of this specification, terms such as first and second are used only to distinguish one component from other components and, unless otherwise explicitly stated, do not limit the order or importance of the components. Therefore, within the scope of the embodiments of this specification, the first component in an embodiment may also be referred to as the second component in other embodiments, and similarly, the second component in an embodiment may also be referred to as the first component in other embodiments.

[0060] In the embodiments of this specification, the distinction between components is intended only to clarify the characteristics of each component and does not necessarily imply that the components must be separated. Specifically, multiple components can be integrated to form a single hardware or software unit, and a single component can be dispersed across multiple hardware or software units. Therefore, even if not specifically mentioned, such integrated or dispersed embodiments are included within the scope of the embodiments of this specification.

[0061] In this specification, the term "network" may include both wired and wireless networks. In this case, the term "network" may refer to a communication network that enables data exchange between devices and systems, and between devices, and is not limited to a specific network.

[0062] The embodiments described in this specification may be all hardware, part hardware and part software, or all software. In this specification, terms such as "unit", "device" or "system" refer to hardware, a combination of hardware and software, or a computer-related entity such as software. For example, in this specification, a unit, module, device or system may be a running process, process, object, executable file, thread of execution, program and / or computer, but is not limited to this. For example, an application running in a computer and both sides of the computer may be equivalent to a unit, module, device or system in this specification.

[0063] In addition, in this specification, a device can be not only a mobile device such as a smartphone, tablet computer (PC), wearable device, and head-mounted display (HMD), but also a fixed device such as a personal computer (PC) or a home appliance equipped with a display function. In addition, as an example, a device can also be an in-vehicle instrument or an Internet of Things (IoT) device. That is, in this specification, a device can refer to an instrument that can run application software functions and is not limited to a specific type. In the following, for the convenience of explanation, the instrument that runs application software is referred to as a device.

[0064] In this specification, the communication mode of the network is not subject to special limitations, and the connections between the various components may not use the same network mode. The network may include not only communication modes using communication networks (for example, mobile communication networks, wired Internet, wireless Internet, broadcast networks, and Wei Xingwang, etc.), but also short-range wireless communications between devices. For example, the network may include all communication methods that can establish interconnection between objects, and is not limited to wired communication, wireless communication, 3G, 4G, 5G or other methods.For example, the wired and / or network may refer to a network based on a local area network (LAN), a metropolitan area network (MAN), a global system for mobile communications (GSM), an enhanced data global system for mobile communications environment (EDGE), a high speed downlink packet access (HSDPA), a wideband code division multiple access (W-CDMA), a code division multiple access (CDMA), a time division multiple access (TDMA), a Bluetooth, a Zigbee, a wireless local area network (Wi-Fi), a voice over Internet protocol (VoIP), a long term evolution technology upgraded version (LTE Advanced), IEEE802.16m, a wireless metropolitan area network upgraded version (Wireless MAN-Advanced), a high speed packet access (HSPA+), a third generation partnership project long term evolution technology (3GPP Long Term Evolution (LTE)), a mobile global interoperability for microwave access (Mobile World Interoperability for Microwave Access) A communication network using one or more communication methods selected from the group consisting of WiMAX (IEEE 802.16e), Ultra-Wideband Mobile Technology (UMB, formerly EV-DO Rev.C), Flash-Orthogonal Frequency Division Multiplexing (Flash-OFDM), iBurst and MBWA (IEEE 802.20) systems, High-Performance Metropolitan Area Network Specification (HIPERMAN), Beam-Division Multiple Access (BDMA), World Interoperability for Microwave Access (Wi-MAX), and ultrasonic application communication, but is not limited thereto.

[0065] The components described in the various embodiments are not necessarily essential components; some of them may be optional components. Therefore, embodiments consisting of a partial collection of the components described in the embodiments are also included within the scope of the embodiments of this specification. Furthermore, embodiments that include additional components in addition to the components described in the various embodiments are also included within the scope of the embodiments of this specification.

[0066] Next, embodiments of the present specification will be described in detail with reference to the accompanying drawings.

[0067] Figure 1 This is a diagram illustrating an example of a working environment of a system applicable to one embodiment of this specification. Figure 1 , the user device 110 and one or more servers 120, 130, 140 are connected via a network 1. Figure 1 This is just an example for illustrating the invention. The number of user devices or servers is not limited to Figure 1 The contents shown in the figure.

[0068] The user device 110 may be a fixed terminal or a mobile terminal implemented by a computer system. The user device 110 may be, for example, a smart phone, a mobile phone, a navigation system, a computer, a laptop computer, a digital broadcast terminal, a personal digital assistant (PDA), a portable multimedia player (PMP), a tablet computer (PC), a game console, a wearable device, an Internet of Things (IoT) device, a virtual reality (VR) device, and an augmented reality (AR) device. As an embodiment, the user device 110 in the embodiment may refer to one of various physical computer systems that can essentially communicate with other servers 120-140 via the network 1 by wireless or wired communication.

[0069] Each server can be implemented using a computer device or multiple computer devices that communicate with the user device 110 through the network 1 and provide instructions, codes, files, content, services, etc. For example, the server can be a system that provides services to the connected user devices 110 respectively through the network 1. As a more specific example, the server can provide the user device 110 with services required by the corresponding application software (for example, information provision, etc.) through a computer program, i.e., application software, installed and running in the user device 110. As another example, the server can provide corresponding services by distributing files for installing and running the application software to the user device 110 and receiving user input information. As an example, each of the servers described below can be Figure 1 The server can be a subject that communicates with at least one of other devices, drones, satellites, and aircraft through the network and obtains data. In addition, the terminal that receives the disaster loss identification information can be Figure 1 As another example, drones, satellites, and airplanes can also Figure 1 The network in the embodiment is used as the basis for communicating with other terminals or servers via the network. That is, the subjects can communicate with each other via the network and work in a data exchange manner, which is not limited to the above-mentioned embodiment.

[0070] Figure 2 This is a block diagram for explaining the internal structure of a computing device 200 in one embodiment of this specification. The computing device 200 may be applicable to the above content. Figure 1 In the one or more user devices 110 - 1 and 110 - 2 or servers 120 to 140 described above, each device and server may have the same or similar internal configuration by adding or excluding some components.

[0071] See Figure 2The computing device 200 may include a memory 210, a processor 220, a communication module 230, and a transceiver 240. The memory 210 is a non-volatile computer-readable storage medium and may include permanent mass storage devices such as random access memory (RAM), read-only memory (ROM), a disk drive, a solid state drive (SSD), and flash memory. Among them, permanent mass storage devices such as read-only memory (ROM), solid state drive (SSD), flash memory, and disk drive are separate permanent storage devices different from the memory 210 and may also be included in the device or server. In addition, the memory 210 may store an operating system and at least one program code (for example, a browser installed and running in the user device 110, etc., or code required by application software installed in the user device 110, etc. to provide a specific service). The software components can be loaded from a separate computer-readable storage medium separate from the memory 210. The separate computer-readable storage medium may include computer-readable storage media such as a floppy disk, a hard disk, a magnetic tape, a digital versatile disk (DVD) / a CD-ROM drive, and a memory card.

[0072] In another embodiment, software components may be loaded into the memory 210 instead of the computer-readable storage medium via the communication module 230. For example, at least one program may be loaded into the memory 210 based on a computer program (e.g., the application software) installed using a file provided by a developer or a file distribution system (e.g., the server) that distributes the application software installation file via the network 1.

[0073] The processor 220 can process the instructions of the computer program by performing basic arithmetic, logical, and input / output operations. The instructions can be provided to the processor 220 by the memory 210 or the communication module 230. For example, the processor 220 can execute the received instructions according to the program code stored in the storage device such as the memory 210.

[0074] The communication module 230 can provide a function for the user device 110 and the servers 120-140 to communicate with each other through the network 1, and can also provide a function for the user device 110 and / or the servers 120-140 to communicate with other electronic devices respectively.

[0075] The transceiver unit 240 may be a component for providing an interface with external input / output devices (not shown). For example, external input devices may include devices such as a keyboard, a mouse, a microphone, and a camera, while external output devices may include devices such as a display, a speaker, and a haptic feedback device.

[0076] As another example, the transceiver 240 may also be a component for providing an interface with a device such as a touch screen that integrates the functions required for input and output.

[0077] In addition, in another embodiment, the computing device 200 may also include a plurality of Figure 2 For example, when the computing device 200 is applied to the user device 110, it may include at least a part of the input and output devices as described above, or may also include other components such as a transceiver, a global positioning system (GPS) module, a camera, various sensors, and a database. As a more specific example, when the user device is a smart phone, it may also include various components commonly included in a smart phone, such as an acceleration sensor or a gyroscope sensor, a camera module, various physical buttons, buttons using a touch screen, input and output ports, and a vibrator for emitting vibrations.

[0078] As an example, the terminals, drones, satellites and aircraft mentioned below may be Figure 2 That is, the terminals, drones, satellites and aircraft described below may be based on Figure 1 The main body that connects and communicates with the server through the network between them.

[0079] Among them, each main body can be equipped with Figure 2 The computing device that is composed and operated by the basic memory, processor, communication module, transceiver and other components is not limited to a specific computing device.

[0080] As an example, in the event of a disaster (e.g., flood, fire, earthquake, waterlogging, and landslide, etc.), at least one of a drone, satellite, and airplane can be used to obtain images or videos related to the disaster area. That is, at least one of the drone, satellite, and airplane can take pictures of the disaster area and transmit the captured images to a server. The server can confirm the disaster loss area and the type of disaster loss based on the received images. In this case, the server can be equipped with a disaster identification learning model and can use the received images (or videos, hereinafter referred to as images) as input to the disaster identification learning model, and as output of the disaster identification learning model, can provide disaster loss area confirmation information and disaster loss type information. As an example, the server can provide the disaster loss area confirmation information and disaster loss type information to a terminal or device, and the terminal or device can use the received information. In addition, the server can evaluate the loss scale information of the loss area from the disaster image based on the learning model.

[0081] However, as an example, ensuring that the disaster identification learning model can accurately identify disaster damage areas and types of damage may require extensive data training. However, because disasters do not occur all the time, insufficient data related to disaster images may hinder learning. Furthermore, as an example, when learning based on the disaster identification learning model, it is necessary to incorporate features related to the disaster-damaged area and surrounding areas to improve the accuracy of damage area identification.

[0082] Considering the above-mentioned issues, the following describes an advanced method of a disaster identification learning model that inputs an aerial image as a disaster image and outputs the extent of the disaster area and the type of disaster loss.

[0083] As an example, the disaster image may be an aerial image. Figure 3a Server 310 can obtain aerial images of the disaster area from drone 320. For example, a drone is a small aircraft that can move freely in the air, and aerial images can be obtained from it as disaster images. Alternatively, server 310 can be a device equipped with a disaster identification learning model and capable of identifying disaster-damaged areas, but this is not limited to a specific form.

[0084] As another example, see Figure 3b , the server 310 can obtain aerial images related to the disaster area from the satellite 330. As another example, see Figure 3c , the aerial image may be an image obtained by the aircraft 330. As an example, the server 310 may Figures 3a to 3c The server 310 may obtain a disaster image from at least one of the following, and is not limited to a specific form. Furthermore, the server 310 may obtain multiple disaster images through multiple devices, and may use multiple disaster images. Furthermore, as an example, the server 310 may obtain disaster images in other forms from other devices. As an example, the server 310 may obtain, as an aerial image, images in other forms that capture the same area as the damage area captured by the disaster image. The images in other forms may be close-up images of the damage area, map images, terrain images, or other images, and are not limited to a specific form.

[0085] That is, the server 310 can Figures 3a to 3c Aerial images are obtained as disaster images based on the present invention, and images of other forms may be obtained through other devices or databases. The present invention is not limited to the above-mentioned embodiments.

[0086] The disaster identification learning model of the server may take at least one disaster image obtained in the manner described above as input, and derive and provide disaster loss area confirmation information and disaster loss type information as output values.

[0087] However, as mentioned above, because disasters don't always occur, there might not be enough standard data for the disaster recognition learning model to learn from. Furthermore, as mentioned above, only by incorporating information about domestic terrain and other features into the disaster recognition learning model can it provide accurate output values ​​related to disaster images.

[0088] Considering the above issues, the disaster identification learning model can be trained in advance based on multiple existing disaster images. In other words, the disaster identification learning model can use existing disaster images as standard data to build a learning model for identifying disaster damage areas.

[0089] However, as an example, using only disaster images as input for a disaster recognition learning model may not yield sufficient standard data, and its advanced learning may be limited. Considering these issues, disaster images can be labeled beforehand, and the labeled information used as input for the disaster recognition learning model. As another example, information about the disaster damage area and surrounding areas can be extracted from the disaster images, and this information can be used as input for the disaster recognition learning model.

[0090] As an example, see Figure 4 The labeling of disaster images can be performed on the disaster loss area of ​​the disaster image. That is, the disaster loss area can be derived from the disaster image, and multiple loss-related information can be generated through labeling based on the derived loss area.

[0091] As an example, see Figure 4 The multiple loss-related information generated by labeling disaster images may include at least one of loss area information, loss type information, peripheral feature information, loss diffusion information, meteorological information at the time of the disaster, and other information.

[0092] As an example, as described above, disaster images can be images related to a disaster that has already ended. That is, the loss-related information obtained through labeling can be records related to images of a disaster that has already ended. The server can obtain the above-mentioned records related to disaster images and label each disaster image based on them. The labeled disaster images and the labeled information can each be used as input for advanced learning of the disaster recognition learning model described above.

[0093] At this point, the disaster recognition learning model can use multiple disaster images and loss-related information obtained through labeling as input for learning. During the learning process, the disaster recognition learning model can match the disaster images with the loss-related information obtained through labeling and assign weighted values ​​to each loss-related information. As a more specific example, the disaster recognition learning model can identify loss-related information for disaster images occurring in the summer, i.e., at temperatures above 30 degrees Celsius, through labeling. Based on this information, the disaster recognition learning model can derive commonalities and differences related to each disaster. Next, the disaster recognition learning model can learn by matching the derived commonalities and differences with each disaster image.

[0094] As an example, the server can use an information inference learning model to derive damage areas from existing disaster images and generate damage-related information by performing labeling. In other words, a learning model different from the disaster identification learning model can be used. The information inference learning model can take existing disaster images as input and output damage-related information by deriving damage areas and labeling them.

[0095] Next, the server can use the damage area and damage-related information derived by the information inference learning model, along with the disaster image, as input information for the disaster identification learning model. Based on this input information, the disaster identification learning model can provide disaster damage area confirmation information and disaster loss type information as output information.

[0096] As another example, as described above, the damaged area and the surrounding area of ​​the damaged area can be derived from the disaster image, and based on this, information can be obtained through labeling to perform advanced learning of the disaster identification learning model. As an example, the damaged area and the surrounding area can be derived from the disaster image used for learning the disaster identification learning model. In this case, the damage type information (e.g., flooding, landslide) and the feature information of the surrounding area of ​​the damaged area (e.g., hills, forests, pine trees, valleys, bridges, rice fields, ginseng fields) can be derived through labeling, and the corresponding information can be used as learning data.

[0097] More specifically, Figures 5a to 5c This is a schematic diagram illustrating a method for identifying a disaster-damaged area and a disaster-surrounding area in one embodiment of this specification. Figure 5a as well as Figure 5b , the loss area 510 and the surrounding area 520 can be derived from the disaster image. As an example, the server can deduce the loss area 510 and the surrounding areas 520 and 530 by using the above-mentioned information derivation learning model. As a specific example, in Figure 5a as well as Figure 5b In this example, the damaged area 510 may be a flooded area, and the disaster loss type information may be flooding. However, this is merely an example; other disaster losses are also possible. In this case, as an example, the server can label the damaged area 510 and surrounding areas 520 and 530 to derive their respective feature information. Surrounding areas 520 and 530 may comprise multiple areas, and each of these areas 520 and 530 may include its own feature information. For example, the first surrounding area 520 may be a river near a flooded area, and feature information related to the first surrounding area 520 can be derived through labeling. The feature information related to the first surrounding area 520 may be information generated for the damaged area 510 with respect to flooding. In other words, the feature information can be derived by considering the correlation between flooding and the river. Furthermore, the second surrounding area 520 may be an adjacent road, and the feature information related to the second surrounding area 520 may also be information generated with respect to flooding. Based on the above, feature information related not only to the damaged area 510 but also to surrounding areas 520 and 530 can be included in the learning data.

[0098] As an example, the feature information related to the river, i.e., the first surrounding area 520, can include at least one of the following as learning data: the river's morphology, region, and direction. That is, the information related to the first surrounding area 520 can be included as learning data. Furthermore, the feature information related to the road beside the river, i.e., the second surrounding area 530, can include information comparing the height of the second surrounding information 530 with the loss area 510, terrain morphology, paving status, and other information as learning data. That is, the learning data can include the derived feature information related to the loss area 510 as information related to the loss area 410, as well as information related to the surrounding areas 520 and 530 of the loss area 510.

[0099] As a more specific example, see Figure 5c In addition to the first and second surrounding areas 520 and 530, information from the third, fourth, and fifth surrounding areas 540 and 560 can also be used as learning data. The third surrounding area 540 can be buildings surrounding the damaged area 510, and information related to buildings and waterlogging losses can be used as feature information for the third surrounding area 540. As an example, the feature information for the third surrounding area 540 can be information about building losses at the time of the waterlogging damage in the damaged area 510. On the other hand, the fourth surrounding area 550 can be farmland, and information related to farmland and waterlogging can be used as feature information. As an example, the feature information can include at least one of the following: farmland morphology, the type of crops grown on the farmland, terrain elevation, and comparison information with the damaged area.

[0100] Furthermore, the fifth surrounding area 560 may be a plastic greenhouse and may include feature information associated with the loss area 510. That is, the learning data may further include information related to the surrounding areas 510, 520, 530, 540, and 550, and learning may be performed based on this information.

[0101] As another example, information related to loss regions may also be included in the learning data. For example, see Figure 6 , the feature information related to the loss area 610 and the surrounding area 620 can be derived and reflected in the learning data, which is consistent with the description in the above content. As an example, Figure 6The loss area 610 in the example can be a flooded area. In this case, disaster-related information can be added to the disaster identification learning model as loss area-related information. As an example, disaster-related information can be reflected by deriving fluid flow information 630 and soil flow information 640. As an example, fluid flow information 630 and soil flow information 640 are information related to the cause of the loss area 610, namely, flooding. They can be measured at a location or multiple locations at a predetermined distance from the loss area 610, and such information can be reflected in the learning data. In other words, not only the loss area information and surrounding area feature information, but also disaster information can be reflected in the learning data and used as a basis for learning.

[0102] exist Figure 6 In the above description, the flooded area is described, but the present invention is not limited thereto. That is, the learning data may include not only information related to the damaged area and the surrounding area, but also information related to the disaster, and is not limited to the above-mentioned embodiment.

[0103] As another example, disaster-related information may also include loss diffusion information. The loss diffusion information may be a value predetermined based on the loss area or the surrounding area. As a specific example, when the surrounding area of ​​the loss area is a flood control embankment or a forest area, the diffusion value may be set to a lower value because the possibility of loss diffusion is low. In contrast, in the case of plastic greenhouses, power plants, reservoirs, or empty homesteads, the loss diffusion value of the corresponding area may be set to a higher value because the surrounding area is more likely to suffer losses in the event of a disaster. That is, the loss diffusion information may be determined in advance based on the characteristics of each loss area or surrounding area, and the corresponding information may be reflected in the learning data as disaster-related information.

[0104] As another example, as described above, when learning information derived as learning data through a learning model, a weighted value can be assigned to each piece of information. As an example, a weighted value can be assigned based on information related to the loss, that is, the degree of impact on the loss. As a specific example, weighted values ​​can be assigned to the slope of the terrain, the shape of the river, and the depth of the river, so that the slope, the shape of the river, and the depth of the river are preferentially identified in the input data for comparison, thereby improving the accuracy of the judgment. In the manner described above, the disaster identification learning model can be learned based on at least one of the disaster image, the loss area obtained by labeling, the characteristic information of the surrounding area, and the disaster-related information.

[0105] Next, in the event of an actual disaster, the Figures 3a to 3cThe disaster recognition learning model can use the acquired disaster images as input to calculate the disaster loss area confirmation information and disaster loss type determination information as output information.

[0106] At this time, as an example, landmark information can be extracted from the disaster image used as input information for the disaster identification learning model. The landmark information, as described above, can be information related to the characteristics of the periphery of the damaged area. As a more specific example, see Figures 5a to 5c , at least one of the following information can be extracted: the shape and area of ​​the river, the direction of the river, the shape and elevation of the cultivated land to the left of the flooded area, the use of the flooded area (e.g., dry farmland, paddy field, or apple orchard), and the use of the surrounding area of ​​the flooded area (e.g., dry farmland, paddy field, or plastic greenhouse). As an example, the landmark information can be extracted based on the information inference learning model as another learning model.

[0107] Specifically, when a server receives a disaster image requiring analysis, it can extract the aforementioned damaged area, surrounding area, and other disaster-related information by applying a learning model designed to extract landmark information from the disaster image. The server can then use the disaster image and the information extracted as input to a disaster identification learning model. The model then calculates disaster damage area confirmation information and disaster damage type determination information as output, thereby improving learning accuracy.

[0108] As another example, for advanced learning to identify damaged areas, pre-disaster images related to the disaster images can be obtained. As another example, the server can obtain pre-disaster images from other databases or other devices. As another example, the server can obtain map images or other types of images related to the disaster images and use them as input information for the disaster recognition learning model to improve accuracy.

[0109] As another example, the server may identify the disaster-damaged area and the type of disaster-damaged area based on the disaster-damaged area confirmation information and the disaster-damaged type determination information. Specifically, the server may determine the area and type of the disaster-damaged area based on this information. Furthermore, the server may confirm the purpose of the disaster-damaged area and derive characteristic information about the damaged area. As one example, the purpose of the disaster-damaged area may be confirmed using the disaster image or external data. As one example, the purpose of the disaster-damaged area may be differentiated based on whether it is a vacant homestead, dry farmland, paddy field, mountainous area, or residential area. As another example, the purpose of the disaster-damaged area may be differentiated based on more specific purposes. As one example, if the purpose of the disaster-damaged area is dry land, the purpose of the damaged area may be determined by considering information about plants grown in the damaged area (e.g., peppers, apples, and ginseng). As another example, the purpose of the damaged area may be determined by considering additional related information, such as whether the damaged area is a plastic greenhouse dry land or open dry land. As another example, if the damaged area is a warehouse, the purpose of the damaged area may be determined by further considering the products stored in the warehouse. As another example, if the damaged area is a poultry house, the purpose of the damaged area can be determined by further considering the poultry housed therein. In this case, as an example, the purpose of the damaged area can be determined not only based on the disaster image, but also based on other images or related data of the damaged area. Furthermore, as an example, the purpose of the damaged area can be determined based on land records or other data, and is not limited to the above-described embodiment.

[0110] In the manner described above, the server can confirm the area, type and use of the loss area. Specifically, when the area, type and use information of the loss area have been determined, the server can derive the characteristic information of the loss area, and confirm the loss scale and compensation scale information based on this. In this specification, the loss scale and compensation scale can be understood as the amount of loss or the amount of compensation. As an example, the server can derive the characteristic information of the loss area through aerial images as a disaster image of the loss area. At this time, the server can derive the characteristic information based on the features identified from the aerial images while considering the area, type and use of the loss area. As a specific example, when the loss of the loss area can be identified from the aerial images, the server can confirm the area, type and use of the loss area based on the identified loss and derive the characteristic information. Next, the server can confirm the loss scale information based on the derived characteristic information.

[0111] As another example, if the damage type is flooding and the damaged area is a plastic greenhouse, aerial imagery may not be able to identify the damage because the greenhouse is not actually damaged. In this case, the server can use not only the aerial imagery but also information about the damaged area to derive characteristic information. As an example, the server can derive the appearance of the plastic greenhouse from the aerial imagery. However, because the appearance of the plastic greenhouse in the aerial imagery may be in a state where its damage cannot be identified, deriving characteristic information is difficult. Therefore, additional related images or external data can be used to derive characteristic information and confirm the scale of the damage.

[0112] As another example, in a case where the damage to the damaged area cannot be directly identified through aerial images, the damaged area can be a ginseng field. In this case, because the shading film top of the ginseng field is still present in the aerial image, the damage may not be recognizable from the appearance. As an example, the server can further use the characteristic information of the surrounding area in the aerial image to identify the damage and derive characteristic information. As a specific example, the case where the shading film of the ginseng field is present, but the surrounding area has all suffered waterlogging damage can be taken as an example. In this case, although the damage to the damaged area cannot be identified from the aerial image, the characteristic information of the surrounding area can be derived. The server can determine that the area including the damaged area is in a completely waterlogged state based on the characteristic information of the surrounding area, and derive characteristic information based on this. That is, the server can use not only the damaged area, but also the information obtained from the surrounding area to derive characteristic information related to the damaged area, and confirm the scale of the loss information based on this.

[0113] As another example, in the case of a facility (such as a ginseng field) with a light-shielding film installed in one direction, the server can request an image of the field's interior from the image capture device through the portion of the film where the film is open. In this case, the interior of the field can be captured by adjusting the flight position and shooting angle. Because the angle of the light-shielding film is fixed, the drone's altitude can be used to calculate the location within the field where the camera can be captured. The server can then provide the drone with the shooting position and angle, using the aforementioned calculations.

[0114] As another example, the server can further utilize external data as relevant supplementary information. For example, while the surrounding area can be determined to be completely flooded using the aforementioned method, further utilizing cadastral information, for example, can reduce the probability of this determination being completely flooded if the ginseng field is higher than the surrounding area or has drainage facilities. In other words, the server can further utilize external data to derive characteristic information, and use this information as a basis for confirming the scale of the damage.

[0115] As another example, the server can use the additional information that can be obtained through the aerial image to derive the characteristic information of the loss area. As an example, in the case of a ginseng field, information related to the degree of curvature of the light-shielding film bracket pattern can be confirmed from the aerial image. The server can confirm the degree of waterlogging based on the degree of curvature of the light-shielding film bracket pattern, and thereby derive the characteristic information of the loss area and confirm the loss scale information. That is, the server can further derive additional information from the aerial image, and after deriving the characteristic information using the additional information, confirm the loss scale information. In addition, the server can also receive and utilize external data in order to derive the loss scale information, which is the same as the description in the above content.

[0116] In this case, as a specific example, if crops are present in the damaged area, the server can obtain at least one of the following as external data: information on the time of damage, information on the time of aerial image capture, information on the area of ​​the damaged area, information on whether crops in the damaged area have been shipped out, information on the market price of crops in the damaged area, information on the cost of damaged crops, and information on the damaged crops. Based on this information, the server can derive characteristic information and confirm the scale of damage. For example, the information on the time of damage and the time of aerial image capture can be necessary to confirm whether the crops in the damaged area are pre-harvest or in the delivery period. Furthermore, for example, information on the market price or cost of damaged crops can be obtained from other databases (e.g., an agricultural cooperative server) or other devices. Furthermore, information on whether damaged crops are susceptible to influences such as temperature, wind speed, and rainfall can be considered as information on damaged crops. Furthermore, information on the scale of damage can be derived by reflecting whether the damaged crops are annual or perennial crops.

[0117] As a specific example, in the case of a ginseng field, the server can take into account the tilted nature of the light-shielding film and obtain images captured after the drone has moved to a position facing the film's opening. The server can then identify the ginseng's color as crop information within the image and compare it with database information to determine the ginseng's age. The server can also further infer characteristics of the damaged area by reflecting this information, thereby confirming the scale of the damage.

[0118] In the manner described above, the server can derive characteristic information while taking into account the area, type, and purpose of the damaged area, and thereby confirm the loss scale information. As an example, the server can also derive compensation scale information. In this case, the compensation scale information can be determined in the same manner as the loss scale information. As another example, the compensation scale information can be determined by reflecting the responsibility information of the management personnel in the loss scale information. In this case, the responsibility information can be determined based on images taken during normal times or images taken before the disaster. As another example, the responsibility information can be determined based on the degree of disaster prevention measures in place in the disaster-damaged area. As an example, the disaster prevention measures in place information can be databased information, and can be compared based on the checklist information.

[0119] As a specific example, if a roof is damaged or has been unattended for a long time, or if a dam is damaged or unattended, responsibility information can be derived taking these conditions into account. In this case, a negative weighting value can be reflected in the responsibility information, and the final compensation scale information can be derived by reflecting this value in the loss scale information. As another example, if renovations were performed or new equipment was installed before a disaster, a positive weighting value can be reflected in the responsibility information, and the final compensation scale information can be derived by reflecting this value in the loss scale information. In other words, a positive or negative weighting value can be reflected by confirming the responsibility information, and based on this, the compensation scale information can be set to be greater or less than the average value of the loss scale information, but this is not limited to the above embodiment.

[0120] As a specific example of how the server derives characteristic information from the loss area and determines the loss scale, see Figure 7a as well as Figure 7b The server can confirm the area, type and use of the disaster loss area based on the above-mentioned disaster loss area confirmation information and disaster loss type determination information. As an example, in Figure 7a as well as Figure 7bAerial images of locations 710 and 720 affected by disaster damage can be used to derive disaster damage areas 711 and 721. The server can identify the damage areas using aerial images and confirm the scale of the damage by deriving characteristic information. As an example, the disaster damage areas 711 and 721 can be identified based on aerial images, and characteristic information can be derived and the scale of the damage confirmed based on the damage information identified from the aerial images. The server can confirm the scale of the damage using only aerial images.

[0121] As another example, the server can use not only aerial images but also derive characteristic information by taking into account external data and related images and confirm the scale of the damage. Figure 7a as well as Figure 7b The server can use images captured at different angles of the disaster damage area as relevant images. The relevant images can be close-up images. Corresponding damage areas 712 and 722 can be derived from the aerial images and compared with damage areas 711 and 721 in the aerial images. Furthermore, as an example, the server can use pre-disaster images. In this case, the server can derive portions 713 and 723 corresponding to the damage areas from the pre-disaster images and compare them with damage areas 711 and 721 in the aerial images to derive characteristic information, thereby confirming the scale of the damage.

[0122] As another example, see Figures 8a to 8c The server can use images captured while taking into account the purpose of the disaster damage area as relevant images. The relevant images can be images captured directly of damaged objects (e.g., crops) in the damage areas 811, 821, and 831, from which characteristic information can be derived and the scale of the damage can be confirmed. That is, the server can confirm the damage areas 811, 821, and 831 based on aerial images of the disaster locations 810, 820, and 830, and further reflect the images captured of damaged objects in the actual damage areas as relevant images to derive characteristic information and further confirm the scale of the damage.

[0123] As another example, see Figures 9a to 9c , the server can further use external data to deduce the scale of loss. The external data can be information on a cadastral map. In addition, the external data can be information related to loss areas where similar disasters have occurred. As a specific example, in Figure 9a The information related to the area where the same type of disaster loss occurred can be further confirmed, and the loss scale information can be confirmed after the characteristic information is derived using the information.

[0124] As another example, see Figure 9b as well as Figure 9c , the server can derive characteristic information using information related to areas where disaster losses occurred and areas where disaster losses did not occur. As an example, facilities A 911 and 921 can be areas where disaster losses occurred, while facilities B 912 and 922 can be areas where disaster losses did not occur. As an example, the server can confirm the loss area and loss type of facilities A 911 and 921 in the manner described above. In addition, the server can derive characteristic information and confirm the loss scale information using image information related to facilities A 911 and 921. At this time, as an example, the server can further use image information of facilities B 912 and 922, areas where disasters did not occur. At this time, the server can derive an area corresponding to the loss area and related information from the area where no loss occurred, and derive characteristic information and confirm the loss scale information by comparing it with the loss area, which is not limited to the embodiments described above.

[0125] As another example, the server can assign weighted values ​​when incorporating external data into the disaster identification learning model. The weighted external data can also be applied to the learning model and used in deriving outputs, and is not limited to the above-described embodiments. As an example, the external data can be information obtained by the server from a network or external device, and is not limited to a specific type.

[0126] As an example, the external data may be information related to disaster images. As a more specific example, the server may obtain external data from a disaster statistics database. That is, the external data may be information obtained from a database containing statistical information related to disasters, and this information may be reflected in the learning model. As an example, the server may confirm the type of disaster based on the disaster images described above. The type of disaster may be classified based on the types of disasters recorded in the disaster statistics database. That is, the type of disaster may be determined based on the information in the disaster statistics database. Next, the server may obtain information related to a specific type of disaster as external data. As an example, the external data may include at least one of loss scale information, loss frequency information, loss amount information, death toll information, loss cause information, and fault-related information based on a specific type of disaster. In addition, the external data may also include other disaster-related information and is not limited to specific information.

[0127] Figure 10a This is a schematic diagram illustrating a method for constructing a learning model based on disaster images in one embodiment of this specification.

[0128] See Figure 10a , the server can learn the disaster identification learning model through multiple existing disaster images. Among them, in step S1011, the server can obtain multiple existing disaster images. At this time, the loss area and the surrounding area can be derived from the multiple existing disaster images. As an example, in step S1012, the server can deduce the loss area and the surrounding area from the disaster image through the information deduction learning model in the manner described above. Next, in step S1013, each feature information can be derived by labeling the derived loss area and the surrounding area. Among them, the labeling and the derivation of each feature information can also be performed based on the information deduction learning model, and are not limited to the embodiments described above. As an example, as mentioned above, because it can be an existing disaster image, the loss-related information related to the disaster can be set in advance, and the loss area and the surrounding area can be labeled and the feature information can be derived based on the corresponding information. Among them, the feature information derived by labeling is as follows. Figure 4 As shown, it may include at least one of the loss area information, loss type information, peripheral feature information, damage diffusion information, meteorological information at the time of the disaster, and date information based on each disaster image. At this time, the feature information can be used together with the disaster image to learn the disaster identification learning model. Next, in step S1014, the labeling information and the weighted values ​​related to the loss area / surrounding area can be applied. At this time, as an example, different weighted values ​​can be set for each information. As another example, the weighted value can be set while reflecting the loss diffusion information, and is not limited to the embodiment described above. Next, in step S1015, the server can use each disaster image, labeling information, and loss area / surrounding area information to learn the disaster identification learning model, thereby updating the disaster identification learning model.

[0129] In addition, see Figure 10b The server can use the updated disaster recognition learning model to analyze the actual disaster image and derive the disaster loss area confirmation information and disaster loss type information based on this. Specifically, in step S1021, the server can use Figures 3a to 3cBased on the information, aerial images are obtained as disaster images. At this point, the damaged area and surrounding areas can be derived from each disaster image. As an example, the server can derive the damaged area and surrounding areas based on a learning model that derives landmark information and obtain feature information through labeling in step S1023. Next, in step S1024, weights can be applied to each extracted information, and the weighted information can be used as input information for an updated disaster recognition learning model. Furthermore, as an example, in step S1025, the server can obtain not only the disaster image, but also at least one of pre-disaster images, images of the disaster's vicinity, map images, terrain images, and other relevant images. In this case, in step S1026, the server can use the weighted information, the disaster image, and other relevant images as input to the disaster recognition learning model in the manner described above. In step S1027, the disaster recognition learning model can use the learned database based on this input information to identify the disaster damaged area and derive relevant information. Next, the disaster identification learning model can provide disaster loss area confirmation information and disaster loss type information as output information, thereby improving the accuracy of loss area identification in disaster images.

[0130] Figure 11a This is a schematic diagram illustrating a method for determining the scale of disaster losses and the scale of compensation based on the disaster loss area in one embodiment of this specification.

[0131] See Figure 11a In step S1111, the server can Figure 10a as well as Figure 10b Based on the loss area and loss type, the server can confirm the loss area and loss type. Next, in step S1112, the server can confirm the use of the loss area. In this case, the use of the loss area can be determined based on the use information of the loss area. As an example, the use information refers to the use of the loss area, and can be based on whether it is a vacant homestead, dry farmland, paddy field, mountainous land, or residential area.

[0132] As another example, the purpose of a disaster-damaged area can be determined by considering specific usage information, taking into account the object of the damaged area's use. For example, if the disaster-damaged area is dryland, the purpose of the damaged area can be determined by considering information about plants grown in the damaged area (e.g., peppers, apples, and ginseng). For another example, the purpose of the damaged area can be determined by adding associated information, such as whether the damaged area is a plastic greenhouse or open-air dryland. For another example, if the damaged area is a warehouse, the purpose of the damaged area can be determined by further considering the products stored in the warehouse. For another example, if the damaged area is a poultry house, the purpose of the damaged area can be determined by further considering the poultry raised in the house. In step S1113, after confirming the purpose of the damaged area in the manner described above, the server can derive characteristic information of the damaged area. The characteristic information can be information related to the actual damaged object, taking into account the area, type, and purpose of the damaged area. For another example, the characteristic information of the damaged area can be identified based on the aerial imagery and can be derived using only the aerial imagery.

[0133] As another example, even if the area and type of the damaged area have been confirmed using aerial imagery, it may still be difficult to determine the characteristics of the damaged area and confirm the scale of the damage using only this information. In step S1114, considering the above-mentioned issues, the server may further consider relevant images related to the damaged area. Furthermore, in step S1115, the server may further consider external data related to the damaged area. As an example, relevant images related to the damaged area may include close-up images, images taken before the disaster, images of areas not affected by the disaster, and other relevant images. Furthermore, as an example, the external data may include at least one of topographic maps, cadastral maps, information on the time of damage, information on the time of aerial imagery capture, area of ​​the damaged area, information on whether crops have been shipped out of the damaged area, market prices for crops in the damaged area, cost information for damaged crops, information on damaged crops, and other relevant information related to the damaged area. In other words, the server can use not only aerial imagery but also relevant images and external data to derive characteristics and confirm the scale of the damage. In step S1116, the server may apply weighting values ​​to each image and external data. As an example, different weighting values ​​may be applied based on the characteristic information extracted from the image and the type of external data. Next, the server may use the information of the applied weighting values ​​to confirm the loss scale information and the compensation scale information, which is not limited to the above-described embodiment.

[0134] As another example, Figure 11b This is a diagram illustrating a method for determining the scale of disaster losses and the scale of compensation based on the disaster loss area in one embodiment of this specification. Figure 11b In step S1121, the server can Figure 10a as well as Figure 10b Based on this, the server can confirm the loss area and loss type. Next, in step S1122, the server can confirm the use of the loss area. In this case, the use of the loss area can be determined based on the use information of the loss area. As an example, the use information refers to the use of the loss area, and can be based on whether it is a vacant homestead, dry farmland, paddy field, mountainous land, or residential area.

[0135] As another example, the purpose of a disaster-damaged area can be determined based on specific usage information, taking into account the object of the damaged area's use. For example, if the damaged area is dryland, the purpose of the damaged area can be determined based on information about plants grown in the damaged area (e.g., peppers, apples, and ginseng). For another example, the purpose of the damaged area can be determined based on additional related information, such as whether the damaged area is a plastic greenhouse or open-air dryland. For another example, if the damaged area is a warehouse, the purpose of the damaged area can be determined based on the products stored in the warehouse. For another example, if the damaged area is a poultry house, the purpose of the damaged area can be determined based on the poultry raised in the house. In step S1123, after confirming the purpose of the damaged area in the manner described above, the server can derive characteristic information of the damaged area. The characteristic information can be information related to the actual damaged object, taking into account the area, type, and purpose of the damaged area. For example, the characteristic information of the damaged area can be identified and derived based on the aerial imagery.

[0136] As another example, even if the area and type of the damaged area have been confirmed using aerial imagery, it may be difficult to derive the characteristics of the damaged area and confirm the scale of the damage using this information alone. To address this issue, the server can use a learning model for assessing the scale of damage. Because it can be difficult to calculate appropriate weightings for assessing the scale of damage, a learning model can be used. As an example, the input to the learning model for assessing the scale of damage may include at least one of the characteristics of the damaged area derived in the manner described above, relevant images, and external data. In this case, the relevant images related to the damaged area may include close-up images, images taken before the disaster, images of areas not affected by the disaster, and other relevant images. Furthermore, as an example, the external data may include at least one of topographic maps, cadastral maps, information on the period of damage, information on the time of aerial imagery capture, area of ​​the damaged area, information on whether crops have been shipped out of the damaged area, market prices for crops in the damaged area, cost information for damaged crops, information on damaged crops, and other relevant information related to the damaged area.

[0137] At this point, in step S1127, the loss scale assessment learning model can derive loss scale information and compensation scale information as output values ​​based on the input information. The derived output values ​​can then be fed back into the learning model, which can then be updated. In this manner, the loss scale assessment learning model can be updated, thereby improving the accuracy of the derivation of loss scale information and compensation scale information.

[0138] Figure 12 This is a sequence diagram illustrating a method for identifying a disaster damage area in one embodiment of this specification.

[0139] See Figure 12 , a working method of a server for identifying disaster loss areas can be provided. At this time, in step S1210, the server can obtain at least one first disaster image. As an example, the first disaster image can be an existing disaster image, and the server can obtain multiple existing disaster images. Next, in step S1220, the server can derive the loss area from at least one first disaster image respectively, and obtain loss area related information by labeling based on the derived loss area. Next, in step S1230, the server can use at least one first disaster image and loss area related information to learn a first learning model. At this time, the first learning model can be the disaster identification learning model. That is, the server can use multiple existing disaster images to learn the disaster identification learning model.

[0140] At this time, the server can derive multiple surrounding areas based on the derived loss area in at least one or more first disaster images. That is, not only the loss area can be derived, but also multiple surrounding areas can be derived, and the surrounding area related information related to the multiple surrounding areas derived by labeling can be obtained respectively. Among them, the first learning model can also use the surrounding area related information related to the multiple derived surrounding areas for learning. At this time, as an example, the surrounding area related information related to the multiple derived surrounding areas can include feature information related to the multiple surrounding areas, which is the same as the description in the above content. At this time, the feature information can be information set under the premise of considering the correlation between the multiple surrounding areas and the loss area, which is the same as the description in the above content. In addition, the server can obtain the second disaster image from the external device. At this time, the external device can be combined with the above content. Figures 3a to 3c The second image may be any one of the drone, satellite, and airplane described above. Furthermore, the external device may be another device or database and is not limited to the above-described embodiment. Furthermore, the second image may be an actual disaster image to be analyzed.

[0141] At this point, the server can derive the damaged area and surrounding areas from the second disaster image, and after obtaining multiple disaster-related information through labeling based on the derived damaged area and surrounding areas, assign weighted values ​​to each of the multiple disaster-related information obtained. Next, the second disaster image and multiple disaster-related information can be provided as input to the learned first learning model, and based on the first learning model, disaster damage area confirmation information and disaster loss type information can be output, thereby ultimately identifying the damaged area.

[0142] Furthermore, as an example, the server may acquire at least one related image related to the second disaster image. This acquired image may be input into the first learning model along with the second disaster image and multiple disaster-related information. In this case, as an example, the at least one related image may include at least one of an image before the disaster, an image of the disaster's periphery, a map image, and a terrain image, as described above. Furthermore, as an example, the second disaster image may be input into a second learning model. The second learning model may derive the damage area and surrounding areas of the second disaster image and, based on the derived second damage area and surrounding areas, provide multiple disaster-related information as output information through labeling. The second learning model may be the information-derived learning model. That is, each damage area and surrounding area may be derived as landmark information based on the information-derived learning model.

[0143] Furthermore, as an example, the server may receive loss diffusivity information associated with the derived damage area and surrounding areas, and determine weighted values ​​associated with each of the plurality of acquired disaster-related information, further reflecting the loss diffusivity information. The loss diffusivity information may be numerically quantified based on the likelihood of a disaster occurring, as described above.

[0144] Furthermore, as an example, the server can obtain information on the use of the disaster-damaged area and, based on the disaster-damaged area confirmation information, the disaster-damaged area type information, and the disaster-damaged area use information, confirm the area, type, and use of the disaster-damaged area, as described above. In this case, the server can derive disaster-damaged area characteristic information based on the confirmed area, type, and use of the disaster-damaged area, and confirm the loss scale information based on the derived disaster-damaged area characteristic information. In this case, as an example, the server can further obtain at least one third disaster image related to the disaster-damaged area and at least one external data related to the disaster-damaged area.

[0145] In this case, the third disaster image, as a relevant image, can be a close-up image, an image taken before the disaster, an image of an area not affected by the disaster, or other relevant images. Furthermore, as an example, the external data can include at least one of a topographic map, a cadastral map, information about the period of damage, information about the time of aerial image capture, area of ​​the damage area, information about whether crops have been shipped out of the damage area, market prices for crops in the damage area, cost information for damaged crops, information about damaged crops, and other relevant information related to the damage area. In this case, the server can further verify the damage scale information based on the at least one third disaster image and the at least one external data, as described above. Furthermore, the server can assign weights to the disaster damage area characteristics information, the at least one third disaster image, and the at least one external data, and verify the damage scale information based on the weights. As another example, the server can provide the disaster damage area characteristics information, the at least one third disaster image, and the at least one external data as input data to a third learning model, and obtain the damage scale information as output data through the third learning model. At this time, the loss scale information can be provided as feedback information to the third learning model, and the third learning model can be updated based on the loss scale information. The third learning model can be the loss scale assessment learning model, and is not limited to the above embodiment.

[0146] At least a portion of the embodiments described above can be implemented by a computer program and stored in a computer-readable storage medium. A computer-readable storage medium storing a program for implementing the embodiment includes all types of storage devices storing computer-readable data. Examples of computer-readable storage media include read-only memory (ROM), random access memory (RAM), read-only compact disc (CD-ROM), magnetic tape, and optical data storage devices. In addition, computer-readable storage media can also be distributed to computer systems connected via a network, so that computer-readable codes are stored and executed in a distributed manner. In addition, a person skilled in the art of the present embodiment should be able to easily understand the functional programs, codes, and code segments used to implement the present embodiment.

[0147] While the foregoing description of the present specification refers to the embodiments illustrated in the accompanying drawings, this is for illustrative purposes only. A person with general knowledge of the relevant technical field should understand that the present invention is susceptible to various modifications and variations of the embodiments. However, such modifications are still within the technical protection scope of this specification. Therefore, the true technical protection scope of this specification should be defined as including other implementations and other embodiments based on the technical concepts of the appended claims, as well as content equivalent to the claims.

Claims

1. A working method, The working method of the server for evaluating the scale of loss in a disaster-damaged area includes: The step of acquiring at least one first disaster image; The steps of respectively deriving loss areas from the at least one first disaster image, and obtaining relevant information of the loss areas by labeling based on the derived loss areas; A step of learning a first learning model using the at least one first disaster image and the loss area related information; as well as, The step of evaluating the loss scale information of the loss area in the disaster image based on the first learning model, a step of acquiring a second disaster image from an external device; Derived from the second disaster image the damaged area and the surrounding area, and providing a plurality of disaster-related information by labeling the derived loss area and surrounding areas; The step of assigning weighted values ​​to the plurality of disaster-related information obtained; The step of inputting the second disaster image and the plurality of disaster related information into the learned first learning model; as well as, The step of outputting disaster loss area confirmation information and disaster loss type information based on the first learning model, Steps to obtain information on the use of disaster-damaged areas; The step of confirming the area of ​​the disaster loss area, the type of the disaster loss area, and the purpose of the disaster loss area based on the disaster loss area confirmation information, the disaster loss type information, and the disaster loss area purpose information; The step of deriving characteristic information of the disaster loss area based on the confirmed area of ​​the disaster loss area, the type of the disaster loss area, and the purpose of the disaster loss area; as well as, The step of confirming the loss scale information based on the derived disaster loss area characteristic information.

2. The working method according to claim 1, The second disaster image is input into the second learning model, The second learning model derives the loss area and the surrounding area of ​​the second disaster image, The plurality of disaster-related information are provided as output information through labeling based on the derived loss area and the surrounding area.

3. The working method according to claim 1, further comprising: A step of acquiring at least one third disaster image related to the disaster damage area and at least one external data related to the disaster damage area; The loss scale information is confirmed by further reflecting the at least one third disaster image and the at least one external data.

4. The working method according to claim 3, assigning weighted values ​​to the disaster loss area characteristic information, the at least one third disaster image, and the at least one external data respectively; The loss scale information is confirmed by reflecting the assigned weighted value.

5. The working method according to claim 3, The disaster loss area characteristic information, the at least one third disaster image and the at least one external data are provided as input data to the third learning model. The third learning model derives the loss scale information as output data based on the input data.

6. The working method according to claim 5, The loss scale information is provided as feedback information to the third learning model, The third learning model is updated based on the loss scale information.

7. The working method according to claim 5, When the external data is provided as input data to the third learning model, after assigning a weighted value to the external data, the data is provided as the input data to the third learning model. The server obtains the external data from a disaster statistics database.

8. The working method according to claim 7, When the server obtains the external data from the disaster statistics database, the disaster type is confirmed based on the at least one first disaster image. and extracting at least one piece of information corresponding to the disaster type from the disaster statistics database as the external data, The at least one piece of information includes at least one of loss scale information, loss frequency information, loss amount information, death toll information, loss cause information, and fault-related information.

9. A computer-readable recording medium having a program recorded thereon, The program executes the working method according to any one of claims 1 to 8 by combining with hardware.

10. A server, As a server for assessing the scale of losses in disaster-affected areas, it includes: The transceiver unit is used to communicate with external devices; as well as, A processor, configured to control the transceiver unit; the processor, Acquire at least one first disaster image, Derived from the at least one first disaster image, the loss area is respectively derived, and the loss area related information is obtained by labeling based on the derived loss area. Using the at least one first disaster image and the loss area related information to learn a first learning model, Based on the first learning model, the loss scale information of the loss area in the disaster image is evaluated. the processor, Acquire the second disaster image from an external device, Derived from the second disaster image the damaged area and the surrounding area, Based on the deduced loss area and surrounding areas, multiple disaster-related information is provided through labeling. Assign weighted values ​​to the plurality of disaster-related information obtained respectively, inputting the second disaster image and the plurality of disaster related information into the learned first learning model, Outputting disaster loss area confirmation information and disaster loss type information based on the first learning model, the processor, Obtain information on the use of disaster-damaged areas, Based on the disaster loss area confirmation information, the disaster loss type information and the disaster loss area use information, the area of ​​the disaster loss area, the type of the disaster loss area and the use of the disaster loss area are confirmed. Based on the confirmed area of ​​the disaster loss area, the type of the disaster loss area, and the purpose of the disaster loss area, characteristic information of the disaster loss area is derived. The loss scale information is confirmed based on the derived disaster loss area characteristic information.

11. The server according to claim 10, The second disaster image is input into the second learning model, The second learning model derives the loss area and the surrounding area of ​​the second disaster image, The plurality of disaster-related information are provided as output information through labeling based on the derived loss area and the surrounding area.

12. The server according to claim 10, the processor, acquiring at least one third disaster image related to the disaster damage area and at least one external data related to the disaster damage area, The loss scale information is confirmed by further reflecting the at least one third disaster image and the at least one external data.

13. The server according to claim 12, assigning weighted values ​​to the disaster loss area characteristic information, the at least one third disaster image, and the at least one external data respectively; The loss scale information is confirmed by reflecting the assigned weighted value.

14. The server according to claim 12, The disaster loss area characteristic information, the at least one third disaster image and the at least one external data are provided as input data to the third learning model. The third learning model derives the loss scale information as output data based on the input data.

15. The server according to claim 14, The loss scale information is provided as feedback information to the third learning model, The third learning model is updated based on the loss scale information.

16. The server according to claim 14, When the external data is provided as input data to the third learning model, after assigning a weighted value to the external data, the data is provided as the input data to the third learning model. The server obtains the external data from a disaster statistics database.

Images