KR20240080204A - predictting method of a landslide based on unsupervised learning and an electroni device for warning of landslide supporting the same - Google Patents

Abstract

The present invention includes the steps of collecting, by a processor of a landslide warning device, a plurality of spectral images of the mountain slope by controlling a spectral camera arranged to acquire a spectral image of at least a portion of the mountain slope, the plurality of spectral images Creating a landslide prediction reference model by performing clustering model learning on the spectrum of spectral images, storing the landslide prediction reference model, and analyzing spectral images acquired at a specific time based on the landslide prediction reference model. An unsupervised learning-based landslide occurrence prediction method and a landslide warning device supporting the same can be disclosed, which includes the step of estimating whether a landslide has occurred.

Description

Unsupervised learning-based landslide occurrence prediction method and landslide warning device supporting the same {predicting method of a landslide based on unsupervised learning and an electroni device for warning of landslide supporting the same}

The present invention relates to landslide prediction, and more specifically, to a technology that can predict the occurrence of landslides through unsupervised learning-based spectral image analysis.

Landslides are very complex natural phenomena that can cause significant damage to people, property and transportation networks. Landslides generally occur frequently in mountainous areas due to steep slopes, and since Korea has large mountainous areas that account for approximately 70% of the total area, the risk of landslides occurring is high. In particular, mountainous areas composed of granite or gneiss have steep slopes, and these areas are vulnerable to landslides.

Meanwhile, in Korea, the average monthly rainfall during the rainy season is recorded at 280 to 300 mm, and these conditions can weaken the strength of the soil and increase the risk of landslides. In response, many government agencies are seeking solutions to mitigate the disastrous consequences of landslides by educating people about their serious impacts and developing appropriate planning and decision-making tools.

For example, conventionally, landslide vulnerability assessment is performed to evaluate the spatial distribution of the probability of landslides occurring in a given area based on the geographical and environmental factors of the area, and based on the assessment results, areas where landslides may occur are identified and mapped. (mapping) work has been carried out. Landslide vulnerability analysis methods include statistical analysis methods and geotechnical analysis methods. The statistical analysis method is a method of analyzing vulnerability by considering the statistical correlation between various landslide triggering factor data in a wide area and the location of landslide occurrence. In the case of statistical analysis methods, their application can be limited to areas where landslides have occurred in the past. The geotechnical analysis method is a method of analyzing stability by assuming a landslide model and considering the geometric and geotechnical characteristics of the mountain, regardless of whether a landslide has occurred, and can take into account the mechanism and process of landslide occurrence.

In the case of the conventional statistical analysis method and geotechnical analysis method described above, since they are performed based only on the analysis of environmental factors without direct monitoring of areas where landslides may occur, even if the exact time of landslide occurrence or the scale of landslide occurrence is estimated, the results are There is a problem that the reliability of is not high.

The present invention collects spectral images of areas where landslides are likely to occur, performs modeling by learning about the collected spectral images, and then performs comparative analysis on the spectral images at the current time to determine the possibility of more accurate landslide occurrence and The aim is to provide an unsupervised learning-based landslide occurrence prediction method that can support the estimation of landslide occurrence patterns and a landslide warning device that supports the same.

However, the object of the present invention is not limited to the above object, and other objects not mentioned can be clearly understood from the description below.

An unsupervised learning-based landslide warning device for achieving the above-described purpose may include a spectral camera that acquires a spectral image of at least a portion of a mountain slope, and a processor functionally connected to the spectral camera. The processor controls the spectral camera to collect a plurality of spectral images of the mountain slope, performs clustering model learning on the spectrum of the plurality of spectral images to generate a landslide prediction reference model, and the landslide prediction reference model. It is characterized in that it is set to store and estimate whether a landslide has occurred by performing analysis on the spectral image acquired at a specific time based on the landslide prediction standard model.

Specifically, the processor is configured to check weather information, collect a plurality of spectral images for each weather information, and store the plurality of spectral images together with the weather information.

Specifically, the processor is set to classify the spectral images based on weather information at the time of collecting the plurality of spectral images and generate a class for each weather information.

Specifically, the processor is set to perform artificial neural network clustering model learning based on the spectrum of spectral images divided into classes for each weather information, and generate the landslide prediction reference model including a plurality of clusters for each weather information. Do it as

Specifically, the processor detects the cluster with the highest similarity by comparing the spectrum of the spectral image acquired at the specific point in time when landslide prediction is required with the clusters included in the landslide prediction standard model, and pre-empts the detected cluster. It is characterized in that it is set to estimate whether a landslide has occurred depending on whether the spectrum of the spectral image acquired at the specific time deviates from a defined reference range.

Specifically, the processor is set to transmit a risk alert regarding the possibility of a landslide to a designated manager device or an electronic device of a designated management office when the landslide is predicted to occur.

In the unsupervised learning-based landslide occurrence prediction method according to an embodiment of the present invention, the processor of the landslide warning device controls a spectral camera arranged to acquire a spectral image of at least a portion of the mountain slope, collecting a plurality of spectral images, generating a landslide prediction reference model by performing clustering model learning on the spectra of the plurality of spectral images, storing the landslide prediction reference model, and the landslide prediction reference model. It is characterized by including a step of estimating whether a landslide has occurred by performing analysis on a spectral image acquired at a specific time based on .

Specifically, the method further includes the step of collecting weather information by the processor, wherein the step of collecting the plurality of spectral images includes collecting a plurality of spectral images for each weather information, the plurality of spectral images It may include storing the weather information together with the weather information.

Specifically, the step of generating the landslide prediction reference model includes performing artificial neural network clustering model learning based on the spectrum of spectral images divided into classes for each weather information, and predicting the landslide including a plurality of clusters for each weather information. It includes the step of generating a reference model, and the step of estimating whether a landslide has occurred includes comparing the spectrum of the spectral image acquired at the specific time for which landslide prediction is required and the clusters included in the landslide prediction reference model to determine the similarity. Detecting the highest cluster, and estimating whether a landslide has occurred based on whether the spectrum of the spectral image acquired at the specific time deviates from a predefined reference range for the detected cluster. do.

Specifically, the method further includes transmitting a risk alert regarding the possibility of a landslide occurring to a designated manager device or an electronic device of a designated management agency.

According to the present invention, the unsupervised learning-based landslide occurrence prediction method and the landslide warning device supporting the same of the present invention analyze spectral images actually taken of areas where landslides are likely to occur, and at least We can help you estimate some things more accurately.

In addition, various effects other than the effects described above may be disclosed directly or implicitly in the detailed description according to embodiments of the present invention, which will be described later.

1 is a diagram illustrating an example of a system environment related to landslide prediction based on unsupervised learning according to an embodiment of the present invention.
Figure 2 is a diagram showing an example of the configuration of a landslide warning device according to an embodiment of the present invention.
Figure 3 is a diagram showing an example of a processor configuration among the configurations of a landslide warning device according to an embodiment of the present invention.
Figure 4 is a diagram showing an example of a server device configuration according to an embodiment of the present invention.
Figure 5 is a diagram showing an example of a method of learning spectral images of a landslide warning device related to estimation of landslide occurrence according to an embodiment of the present invention.
Figure 6 is a diagram showing an example of a method of operating a landslide warning device related to estimation of landslide occurrence according to an embodiment of the present invention.

In order to make the characteristics and advantages of the problem-solving means of the present invention clearer, the present invention will be described in more detail with reference to specific embodiments of the present invention shown in the attached drawings.

However, detailed descriptions of known functions or configurations that may obscure the gist of the present invention are omitted in the following description and attached drawings. Additionally, it should be noted that the same components throughout the drawings are indicated by the same reference numerals whenever possible.

Terms and words used in the following description and drawings should not be construed as limited to their usual or dictionary meanings, and the inventor may appropriately define the concept of the term to explain his or her invention in the best way. It must be interpreted with meaning and concept consistent with the technical idea of the present invention based on the principle that it is. Therefore, the embodiments described in this specification and the configurations shown in the drawings are only one of the most preferred embodiments of the present invention, and do not represent the entire technical idea of the present invention, so at the time of filing this application, various alternatives may be used to replace them. It should be understood that equivalents and variations may exist.

In addition, terms containing ordinal numbers, such as first, second, etc., are used to describe various components, and are used only for the purpose of distinguishing one component from other components and to limit the components. Not used. For example, the second component may be referred to as the first component without departing from the scope of the present invention, and similarly, the first component may also be referred to as the second component.

Additionally, the terms used in this specification are only used to describe specific embodiments and are not intended to limit the present invention. Singular expressions include plural expressions unless the context clearly dictates otherwise. In addition, terms such as "comprise" or "have" used in the specification are intended to designate the presence of features, numbers, steps, operations, components, parts, or combinations thereof described in the specification, but are not intended to indicate the presence of one or more of the features, numbers, steps, operations, components, parts, or combinations thereof described in the specification. It should be understood that this does not exclude in advance the presence or addition of other features, numbers, steps, operations, components, parts, or combinations thereof.

Additionally, terms such as “unit,” “unit,” and “module” used in the specification refer to a unit that processes at least one function or operation, and may be implemented as hardware, software, or a combination of hardware and software. In addition, the terms "a or an", "one", "the", and similar related terms are used in the context of describing the invention (particularly in the context of the claims below) as used herein. It may be used in both singular and plural terms, unless indicated otherwise or clearly contradicted by context.

In addition to the terms described above, specific terms used in the following description are provided to aid understanding of the present invention, and the use of these specific terms may be changed to other forms without departing from the technical spirit of the present invention.

Additionally, embodiments within the scope of the present invention include computer-readable media having or transmitting computer-executable instructions or data structures stored on the computer-readable media. Such computer-readable media may be any available media that can be accessed by a general-purpose or special-purpose computer system. By way of example, such computer-readable media may include RAM, ROM, EPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage, or in the form of computer-executable instructions, computer-readable instructions or data structures. It may be used to store or transmit certain program code means, and may include, but is not limited to, a physical storage medium such as any other medium that can be accessed by a general purpose or special purpose computer system. .

The present invention described below relates to a landslide warning method. More specifically, machine learning is performed based on spectral images (or spectral data) taken of water in a landslide risk area, such as a valley or valley with a mountain slope. This is about an unsupervised learning landslide warning method that warns of landslides through a model.

Hereinafter, the system environment including the unsupervised learning-based landslide warning device of the present invention and the types and roles of each component included therein will be described.

Figure 1 is a diagram showing an example of a system environment related to unsupervised learning-based landslide prediction according to an embodiment of the present invention.

Referring to Figure 1, the landslide prediction-related system environment 10 of the present invention includes a mountain slope 50 (e.g., valley, valley water), a landslide warning device 100, a base station 20, and a server supporting landslide warning. It may include a device 200. Here, in the system environment 10, the base station 20 for forming a communication channel between the landslide warning device 100 and the server device 200 has been described as an example, but the present invention is not limited thereto. For example, the landslide warning device 100 and the server device 200 may be directly connected through a wireless communication channel or a wired cable without configuring the base station 20. In this case, the configuration of the base station 20 can be omitted from the system environment 10.

The mountain slope 50 may include a mountain slope where landslides may occur. This mountain slope 50 has a certain angle relative to the flat land, and less than a certain amount of trees grow naturally, so its shape may be changed by heavy rain or continuous monsoon rain. As an example, landslides that occur on the mountain slope 50 include slump (a phenomenon in which the mountain slope 50 covered by a thick clay sediment layer collapses due to lower erosion) and slide (soil slides on the mountain slope 50 due to long-term weathering). a burning phenomenon), flow (a phenomenon in which a highly viscous material flows down the mountain slope (50) like a river), fall (a phenomenon in which rocks, etc. fall from a cliff and roll), dopple (a phenomenon in which a rock falls from the end of a mountain slope (50) This can include debris flow (a phenomenon in which a debris flow containing rock fragments or plant remains is quickly swept down a valley or drainage ditch), etc.

The landslide warning device 100 may be arranged to acquire a spectral image of at least some points or sections (eg, a valley or valley water) of the above-described mountain slope 50. For example, the landslide warning device 100 is installed at least one of the entrance part of the mountain slope 50 (e.g., the beginning of the mountain when heading from a flat land toward the mountain top), the middle part, or the end part of the mountain slope 50. It may include a spectroscopic camera placed in the location. As described above, the type of landslide that can occur on the mountain slope 50 may vary depending on the type or shape of the soil or rock that makes up the mountain slope 50. Accordingly, the landslide warning device 100 may include spectroscopic cameras placed in various locations and locations rather than placing only one spectral camera on one mountain. As an example, the landslide warning device 100 may be arranged to obtain a spectral image of at least a portion of a mountain slope connected to a point where a village or city is located. The landslide warning device 100 activates the deployed spectral camera in response to predefined schedule information or event information according to weather or season, acquires a plurality of spectral images using the activated spectral camera, and acquires a plurality of spectral images. A landslide prediction reference model can be created by performing unsupervised learning on the spectral images. Here, the landslide prediction standard model may include clustering models divided into classes according to weather or season. The landslide warning device 100 classifies clusters using the spectra of a plurality of spectral images by class, determines which cluster the spectral image acquired at a specific time belongs to, and then determines the specific numerical value of the cluster used for landslide estimation. It is possible to estimate whether a landslide has occurred by checking whether the (e.g. average and standard deviation values) are outside the predefined standard range. That is, the landslide warning device 100 can compare and analyze the landslide prediction standard model and the spectrum of the spectral image acquired at a specific time to determine whether a landslide is likely to occur, and process warning guidance according to the results.

In this regard, the landslide warning device 100 may include a spectral camera that collects spectral images of the mountain slope 50 and a mounting structure for mounting the spectral camera. The landslide warning device 100 can estimate whether a landslide has occurred using spectral images collected using a spectral camera. Alternatively, the landslide warning device 100 according to the present invention additionally utilizes a gas detection sensor or an audio sensor in addition to the spectral image to detect gas detection or a specific roar that occurs as a prior sign when a landslide occurs, thereby increasing the reliability of the prediction of the occurrence of a landslide. can be increased. Meanwhile, the landslide warning device 100 collects information related to the mountain slope 50 (e.g., at least one of spectral images, gas detection information, and audio information that can be detected when a landslide occurs), and sends the collected information to a server device. It may be configured to transmit to 200.

In this case, the base station 20 may form a communication channel with the landslide warning device 100 and transmit information provided by the landslide warning device 100 to the server device 200. The base station 20 is disposed at a position separated from the landslide warning device 100 by a certain distance, and topographically forms a wireless communication channel of the landslide warning device 100 so as to form a communication channel with the landslide warning device 100. It can be placed at any distance or location possible. Alternatively, the base station 20 may be connected to the landslide warning device 100 by wire (eg, cable).

The server device 200 may form a communication channel with the landslide warning device 100 through (or directly) the base station 20. The server device 200 collects landslide-related information (e.g., spectral image, gas detection information, audio information) from the landslide warning device 100, and learns a landslide prediction standard model based on the collected landslide-related information. You can do it. Thereafter, the server device 200 may perform analysis on the spectral image of the landslide 50 at a specific point in time based on the learned landslide prediction standard model to estimate the possibility of a landslide occurring on the corresponding landslide 50. The server device 200 may transmit an estimated value of the possibility of a landslide occurring on the mountain slope 50 to a designated administrator terminal (or user terminal) or output it through a display device. In this operation, the server device 200 may support output of types of landslides that may occur on the mountain slope 50, response action tips depending on the type of landslide, etc.

On the other hand, when the landslide warning device 100 is designed to process learning of the landslide prediction standard model and estimation of the possibility of landslide occurrence on the landslide surface 50, the configuration of the server device 200 is omitted, and the landslide warning device ( 100) can transmit the estimated value of the possibility of a landslide occurrence, the type of landslide, action instructions in response to the occurrence of a landslide, etc., directly to a designated manager terminal or output them on a display without going through the server device 200.

As described above, the system environment 10 related to landslide prediction according to an embodiment of the present invention acquires various spectral images related to the mountain slope 50 (e.g., valley or valley water), and various acquired spectral images After creating a landslide prediction standard model based on learning, it supports estimation of landslide occurrence based on this. In this way, the system environment 10 of the present invention can estimate the occurrence of a landslide on the mountain slope 50 using only a spectral image, and if an additional sensor is present, a more accurate estimation result of the possibility of a landslide occurring can be provided. .

Figure 2 is a diagram showing an example of the configuration of a landslide warning device according to an embodiment of the present invention, and Figure 3 is a diagram showing an example of a processor configuration among the configuration of a landslide warning device according to an embodiment of the present invention.

First, referring to FIG. 2, the landslide warning device 100 according to an embodiment of the present invention includes a communication circuit 110, a spectral camera 120, a memory 130, a sensor unit 140, and a processor 150. It can be included. Here, when the landslide warning device 100 is configured to estimate the possibility of landslide occurrence using only spectral images, the configuration of the sensor unit 140 may be omitted. In addition, the landslide warning device 100 may further include a mounting structure for mounting the spectroscopic camera 120 so that the spectral camera 120 can be photographed while aiming at at least a portion of the mountain slope 50 using the spectral camera 120. You can. As another example, in FIGS. 1 and 2, the landslide warning device 100 is described as including one spectral camera, but the present invention is not limited thereto. For example, the landslide warning device 100 includes a plurality of spectral cameras, and the plurality of spectral cameras may be placed at various locations on the mountain slope 50 where landslides are likely to occur. In addition, the landslide warning device 100 includes at least one of the above-described components, such as a communication circuit 110, a spectral camera 120, a memory 130, a sensor unit 140, and a processor 150. It may further include a power source (e.g., permanent power source or battery).

The communication circuit 110 may form a communication channel with the server device 200 through the base station 20. Alternatively, the communication circuit 110 may directly form a communication channel with the server device 200 without going through the base station 20. When the server device 200 is designed to perform operations related to landslide prediction according to an embodiment of the present invention, the communication circuit 110 transmits at least one spectral image collected by the spectral camera 120 to the server device 200. Can be transmitted. Meanwhile, landslide prediction can be performed independently by the landslide warning device 100. In this case, the communication circuit 110 may transmit a landslide prediction result or a warning message according to the landslide prediction (e.g., an evacuation or fire prevention request message for a specific area) to the server device 200. Alternatively, the communication circuit 110 may transmit at least some of the landslide prediction results and warning messages to residents of a specific area or managers belonging to a specific organization in response to the control of the processor 150. The communication circuit 110 may receive an artificial neural network algorithm necessary for landslide prediction from the server device 200. The communication circuit 110 may receive weather information used to distinguish a plurality of spectral images acquired by the spectral camera 120 from a weather server. The communication circuit 110 may receive weather information from a weather server at a certain period or upon an administrator's request, and transmit the received weather information to the processor 150.

The spectral camera 120 may be arranged to capture a spectral image of at least a certain section (or one point) of the mountain slope 50 (e.g., a mountain slope, a valley, or a valley). The spectral camera 120 may capture a spectral image of at least a portion of the mountain slope 50 in response to the control of the processor 150. As an example, when a predefined event occurs, the spectral camera 120 may capture at least one spectral image related to the mountain slope 50 in response to the occurrence of the event. For example, spectroscopic camera 120 may respond to processor 150 control (or server device 200 requests) in response to a specific weather situation (e.g., clear weather, rain or snow), a specific temperature situation, or a temperature change within a specific range. Spectral images related to the mountain slope 50 can be obtained in situations where the mountain slope is above a certain speed or when strong winds blow at a certain speed or higher. Alternatively, the spectral camera 120 may capture spectral images related to the mountain slope 50 according to a predefined schedule. Alternatively, the spectral camera 120 may acquire a spectral image of the mountain slope 50 at a time when confirmation of the possibility of a landslide occurring is required.

The memory 130 may store at least one program or data necessary for operating the landslide warning device 100. As an example, the memory 130 includes a control program necessary for driving the spectral camera 120, a plurality of spectral images acquired through the spectral camera 120, weather or season information at the time the plurality of spectral images were collected, and a plurality of spectral images. At least some of the following may be stored: a landslide prediction reference model generated through spectral images, a landslide prediction result calculated through comparison with the spectral image acquired at the current time and the landslide prediction reference model, and a warning or guidance message corresponding to the landslide prediction result. You can. Additionally, the memory 130 may store an artificial neural network algorithm or a machine learning learning algorithm required for a landslide prediction standard model.

The sensor unit 140 is disposed on one side of the landslide warning device 100 (or at the point where the spectroscopic camera 120 is placed), and at the time when the spectral camera 120 acquires a spectral image, gas or Audio information can be collected. For example, the sensor unit 140 may include at least one of a gas detection sensor that detects gas erupting immediately before a landslide occurs, and an audio sensor that detects a roar that may occur just before a landslide occurs. In addition, the sensor unit 140 may further include a wind speed sensor capable of detecting changes in ambient wind speed, a temperature detection sensor capable of detecting changes in ambient temperature, and the like. If the landslide warning device 100 is configured to include a spectral camera 120, the configuration of the sensor unit 140 may be omitted.

The processor 150 may control the acquisition of at least one spectral image related to the mountain slope 50 using the spectral camera 120 in response to a predefined event or a request from the server device 200. The processor 150 may generate a landslide prediction reference model based on at least one acquired spectral image and estimate the possibility of a landslide occurrence by comparing it with the spectral image at a specific point in time.

Referring to FIG. 3, the processor 150 may include a spectral image collection unit 151, a landslide model learning unit 152, a landslide occurrence estimation unit 153, and a landslide warning processing unit 154.

The spectral image collection unit 151 may control the acquisition of spectral images using the spectral camera 120. For example, the spectral image collection unit 151 may collect a plurality of spectral images in various environments to create a landslide prediction reference model. In this regard, the spectral image collection unit 151 may connect to a weather server that provides weather to check whether predefined weather arrives. For example, when the spectral image collection unit 151 collects information about clear weather, rainy weather, snowy weather, stormy or typhoony weather, weather with a specific temperature change above a certain level, or weather with a certain amount of rainfall or snowfall, the The spectroscopic camera 120 can be controlled to collect spectral images about the mountain slope 50 in the weather. The spectral image collection unit 151 may transmit a plurality of collected spectral images related to the mountain slope 50 to the landslide model learning unit 152. In addition, the spectral image collection unit 151 may acquire a spectral image for analyzing the possibility of landslide occurrence at a specific point in time (e.g., the current point in time), and transmit the acquired spectral image at the current point in time to the landslide occurrence estimation unit 153. there is.

The landslide model learning unit 152 may generate a landslide prediction reference model using a plurality of spectral images collected by the spectral image collection unit 151. For example, the landslide model learning unit 152 may classify spectral images collected by weather, perform artificial neural network clustering modeling on the spectrum of each spectral image, and perform clustering model learning by weather. In this process, the landslide model learning unit 152 may perform clustering model learning using Variational Auto-Encoder (VAE). When model learning is completed, the landslide model learning unit 152 classifies each cluster into classes (e.g., by weather or season), and calculates the centerline value (centroid, mean of spectra) and standard deviation (standard) of each cluster. deviation of spectra) can be calculated. The landslide model learning unit 152 may build a spectral library based on information of each cluster (e.g., center line value, average value, and standard deviation value) and store it in the memory 130.

The landslide occurrence estimation unit 153 may acquire a current spectral image related to the landslide 50 using the spectral camera 120 at a time when the possibility of a landslide occurrence needs to be estimated. The landslide occurrence estimation unit 153 may analyze the spectrum of the spectral image at the current time and check which cluster the corresponding spectrum is closest to among the clusters stored in the spectral library. The landslide occurrence estimation unit 153 identifies the cluster with the highest similarity, and deviates from the predefined ideal spectrum range for the cluster with the highest similarity to the spectrum result at the current time (or the cluster with the center line with the closest distance). You can check whether As an example, the landslide occurrence estimation unit 153 may check whether the spectrum of the spectral image at the current time is outside the average + K * standard deviation range of the cluster with the nearest center line value. Here, the K value can be adjusted to 1, 1.5, 3, etc. by the landslide warning device 100 manager. The landslide occurrence estimation unit 153 may transmit whether an abnormal spectrum has occurred to the landslide warning processing unit 154.

The landslide warning processing unit 154 may receive information on whether an abnormal spectrum has occurred from the landslide occurrence estimation unit 153 and output a result accordingly. For example, if there is a possibility that a landslide may occur due to the occurrence of an abnormal spectrum, the landslide warning processing unit 154 may transmit a message predicting the occurrence of a landslide to a designated manager terminal or an electronic device of a management office in an area where damage is expected to occur due to a landslide. there is. If no abnormal spectrum occurs, the landslide warning processing unit 154 may transmit a message indicating that there is no abnormality to the designated manager terminal. Meanwhile, the landslide warning processing unit 154 suggests changing the constant K value when an abnormal spectrum does not occur, but a value within a certain range of a predefined reference value (e.g., average + 1.5 * standard deviation value) is detected. You can also output a message saying: In the case of landslides, the possibility of occurrence may gradually increase due to energy accumulation, so by proposing to further narrow the range of the reference value, the possibility of occurrence of abnormal spectra for future spectral images can be increased.

Landslides can be foreshadowed by sudden surges of water, trees shaking or falling even when there is no wind, muddy water rushing into a valley, or soil or stones collapsing or falling. Therefore, the landslide warning device 100 according to an embodiment of the present invention can minimize damage caused by landslides by quickly identifying precursor signs of a landslide and warning in advance of the occurrence of a landslide. For example, the landslide warning device 100 of the present invention applies unsupervised learning to spectral images of mountainsides (e.g., mountain slopes, valleys, valley waters, etc.) to build a landslide prediction reference model and uses this. By estimating the possibility of landslides occurring and providing warnings accordingly, it is possible to reduce human casualties, further prevent landslides, or provide a control effect that can prevent material damage resulting from landslides.

Meanwhile, when the server device 200 is designed to perform operations related to landslide prediction, the processor 150 can control the acquired spectral images to be transmitted to the server device 200.

Figure 4 is a diagram showing an example of a server device configuration according to an embodiment of the present invention. As described above, when the landslide warning device 100 is designed to perform the landslide prediction standard modeling and landslide estimation of the present invention, the configuration of the server device 200 may be omitted.

Referring to FIG. 4, the server device 200 of the present invention may include a server communication circuit 210, a server memory 230, and a server processor 250.

The server communication circuit 210 may form a communication channel with the landslide warning device 100 through the base station 20 or directly. The server communication circuit 210 may collect at least one spectral image from the landslide warning device 100 in response to a specified period or the occurrence of a predefined event. The server communication circuit 210 can transmit the landslide estimation result and the resulting warning message to a designated administrator terminal or organization.

The server memory 230 may store at least one program or data necessary for operating the server device 200. As an example, the server memory 230 may store a spectral library 231 corresponding to a landslide prediction standard model. Additionally, the server memory 230 may store at least one spectral image received from the landslide warning device 100.

The server processor 250 may control the transmission and processing of signals necessary for operating the server device 200, storage of results, transmission of results, or transmission of messages corresponding to the results. In this regard, the server processor 250 may include a data collection unit 251, a data learning unit 252, a landslide estimation unit 253, and a landslide guidance unit 254.

The data collection unit 251 may request at least one spectral image related to the mountain slope 50 from the landslide warning device 100 according to a predefined period or in response to the occurrence of a predefined event. According to one embodiment, when the data collection unit 251 needs to collect a plurality of spectral images needed to build the spectral library 231, it may request the landslide warning device 100 for spectral images at various times. For example, the data collection unit 251 checks weather information and provides spectral images in clear weather, rainy weather, snowy weather, stormy or typhoon weather, weather in which a specific temperature change is greater than a certain level, or weather in which the amount of rainfall or snowfall is greater than a certain amount. Transmission may be requested to the landslide warning device 100. The weather information can be checked from the weather server. Additionally, the data collection unit 251 may collect at least one spectral image related to the mountain slope 50 for a certain period of time prior to the time when landslide estimation is needed. And the data collection unit 251 can collect spectral images at the time when landslide estimation is needed.

The data learning unit 252 may perform clustering model learning on a plurality of spectral images collected by the data collection unit 251. In this regard, the data learning unit 252 can generate a landslide prediction reference model by performing learning on a plurality of spectral images related to the mountain slope 50 based on the artificial neural network learning algorithm stored in the server memory 230. there is. The landslide prediction reference model may include, for example, clustering models generated through a spectrum of spectral images by weather (or by season or situation).

The landslide estimation unit 253 may perform landslide estimation by analyzing spectral images at a specific point in time based on the generated landslide prediction standard model. In this process, the landslide estimation unit 253 may detect a cluster with a center line closest to the spectrum of the spectral image at the current time and analyze a comparison between the detected cluster and the spectrum of the spectral image at the current time. If the spectrum of the spectral image at the current time is outside the reference range of a specific cluster, the landslide estimation unit 253 may transmit a risk warning about the occurrence of a landslide to the landslide information unit 254.

When the landslide information unit 254 receives a risk warning about the occurrence of a landslide from the landslide estimation unit 253, it can transmit a warning message about the occurrence of a landslide to a pre-designated manager terminal. Alternatively, the landslide information unit 254 may operate an emergency siren according to predefined contents or output guidance information to an area where damage is expected to occur due to a landslide.

Figure 5 is a diagram showing an example of a method of learning spectral images of a landslide warning device related to estimation of landslide occurrence according to an embodiment of the present invention.

Referring to FIG. 5, in relation to the method of operating the landslide warning device 100 according to an embodiment of the present invention, the processor 150 of the landslide warning device 100 collects spectral images regarding the landslide risk area in step 501. You can. For example, the processor 150 may collect a plurality of spectral images related to the mountain slope 50 for a certain period of time prior to the time when an estimate of the possibility of a landslide occurring on the mountain slope 50 is needed. According to one embodiment, when precipitation is forecast to occur more than a predefined certain amount, the processor 150 generates spectroscopic images from a predefined time period (e.g., 1 hour or 10 minutes, etc.) prior to the forecast time. It can be collected. Alternatively, the processor 150 may collect spectral images for the landslide hazard area (eg, continuously) at a first predefined time period. Alternatively, while performing rainfall detection, the processor 150 may acquire spectral images in a second time period shorter than the first time period from the point in time when the amount of rainfall increases to a specified value or more. Alternatively, processor 150 may collect spectral images in various weather conditions. For example, the processor 150 operates on a mountain slope ( 50), spectroscopic images can be collected.

At step 503, processor 150 may collect weather information. For example, the processor 150 may collect weather information at the time each spectral image is collected. In this regard, the processor 150 may form a communication channel with an external server device that provides weather information through the communication circuit 110 and collect weather information from the external server device. As an example, the server device 200 described in FIG. 1 may provide weather information. The processor 150 may classify spectral images according to the collected weather information and create classes. Meanwhile, in the above description, it was explained that weather information is collected after collecting spectral images, but spectral images can be collected after collecting weather information. That is, operations 501 and 503 may be performed sequentially or simultaneously.

In step 505, the processor 150 may perform learning on the collected spectral images. For example, the processor 150 may learn an artificial neural network-based clustering model based on the spectrum of spectral images for each class. At this time, Variational Auto-Encoder (VAE), etc. can be used for clustering modeling.

In step 507, the processor 150 can check whether learning is complete. The end of the learning may include, for example, a case where clustering modeling for a preset number or more of spectral images is completed. After performing learning on a predefined number of spectral images or more, the processor 150 may additionally collect spectral images for each weather at a certain period and perform additional clustering modeling for each weather using the collected spectral images. If learning is not completed, the processor 150 may branch to step 501 and re-perform the following operations.

When learning is completed, the processor 150 can build a spectral library in step 509. In this process, the processor 150 calculates the centroid (mean of spectra) of each cluster (by weather or season) and the standard deviation of the cluster. And a spectral library can be built based on the information of each cluster (centroid, etc.).

Figure 6 is a diagram showing an example of a method of operating a landslide warning device related to estimation of landslide occurrence according to an embodiment of the present invention.

Referring to FIG. 6, the processor 150 of the landslide warning device 100 may collect a spectral image at the current time in step 601. The current time may be a time when it is necessary to estimate whether a landslide has occurred. For example, the current time is when it rained a certain amount or more, when it rained for a certain amount of time, when an earthquake occurred, when strong winds of a certain speed or more loosened, when a roar of a certain size or more occurred, and when a certain predefined type of gas was detected. It may include at least one of the viewpoints. Alternatively, the current time may include at least one of a predefined schedule for detecting the occurrence of a landslide or a request time from an operator or manager of the landslide warning device 100.

After collecting the spectral image at the current time, in step 603, the processor 150 may compare and confirm a previously stored cluster based on the spectrum of the current spectral image. In this regard, the landslide warning device may build a spectral library based on the method previously described in FIG. 5 and store the spectral library in the memory 130. The processor 150 may check models clustered by class in the spectral library pre-stored in the memory 130 and the class to which the cluster similar to the spectrum of the current spectral image belongs. Here, when the spectral library is stored and operated in the server device 200, the processor 150 transmits the cluster extracted based on the spectrum of the spectral image at the current time to the server device 200, and Comparison results can be received from (200).

The processor 150 may check whether an abnormal spectrum occurs in step 605. For example, the processor 150 may check the nearest centroid in the process of determining which cluster the spectrum of the spectral image at the current time corresponds to. The processor 150 can check whether the spectral image spectrum at the current time is outside a certain predefined range based on the standard deviation of the nearest cluster while a specific predefined climate condition is satisfied.

If the spectral image spectrum at the current time is outside a certain range, the processor 150 may determine that an abnormal spectrum has occurred, and in step 607, may regard it as a landslide risk situation and issue a risk warning accordingly. For example, a spectrum collected under certain climatic conditions (e.g., rainy season) is closest to a previously stored cluster in the spectroscopic library (e.g., a cluster corresponding to a rainy day), while at the same time the center line of that cluster is μ(average) + 1.5σ. If it is outside the (standard deviation) range, the processor 150 may warn of a landslide risk. Here, the coefficient 1.5 in front of the standard deviation is a device that adjusts the sensitivity of the warning and can be changed by the operator or manager of the landslide warning device 100. In relation to landslide risk warning, the processor 150 sends a landslide risk warning message to a predefined manager terminal device or an electronic device of a manager managing a village adjacent to the area where a landslide is to occur, or a management office managing an adjacent village. can be transmitted. Alternatively, the processor 150 may transmit a landslide risk warning message to a designated server device.

Thereafter, in step 609, the processor 150 may check whether the landslide risk warning function is terminated. For example, the processor 150 may automatically terminate the landslide risk warning function according to settings after a landslide risk warning. Alternatively, the processor 150 may end the landslide risk warning function when a response to the landslide risk warning is received from the devices that transmitted the warning message. If no termination event for a separate landslide risk warning occurs, the processor 150 may branch to step 601 and re-perform the following operations.

Meanwhile, in step 605, if an abnormal spectrum does not occur, the processor 150 may perform the function specified in step 611. For example, the processor 150 may transmit an estimated value estimating a low probability of a landslide occurring for spectral image collection at the current time to a designated terminal, electronic device, or server device 200. Afterwards, the processor 150 may branch to step 601 and re-perform the following operations.

The landslide occurrence prediction method and the landslide warning device supporting the method according to the embodiment of the present invention described above are based on the spectrum of a spectral image taken of a mountain slope (e.g., a valley or valley water) corresponding to a landslide occurrence risk area. Landslide warnings can be made through machine learning.

As described above, although this specification contains details of numerous specific implementations, these should not be construed as limitations on the scope of any invention or what may be claimed, but rather may be specific to particular embodiments of a particular invention. It should be understood as a description of features.

Additionally, although operations are depicted in the drawings in a specific order, this should not be construed as requiring that those operations be performed in the specific order or sequential order shown or that all of the depicted operations must be performed to obtain desirable results. In certain cases, multitasking and parallel processing may be advantageous. Additionally, the separation of various system components in the above-described embodiments should not be construed as requiring such separation in all embodiments, and the described program components and systems may generally be integrated together into a single software product or packaged into multiple software products. You must understand that it is possible.

The present description sets forth the best mode of the invention and provides examples to illustrate the invention and to enable any person skilled in the art to make and use the invention. The specification prepared in this way does not limit the present invention to the specific terms presented. Accordingly, although the present invention has been described in detail with reference to the above-described examples, those skilled in the art may make modifications, changes, and variations to the examples without departing from the scope of the present invention.

Therefore, the scope of the present invention should not be determined by the described embodiments, but by the scope of the patent claims.

10: Landslide system environment
50: Mountain slope
100: Landslide warning device
110: communication circuit,
120: spectral camera
130: memory
140: sensor unit
150: processor
200: Server device
210: Server communication circuit
230: Server memory
250: server processor

Claims (10)

A spectroscopic camera that acquires spectral images of at least a portion of a mountain slope;
A processor functionally connected to the spectral camera,
The processor,
Control the spectral camera to collect a plurality of spectral images of the mountain slope,
Perform clustering model learning on the spectra of the plurality of spectral images to generate a landslide prediction reference model,
Store the landslide prediction reference model,
An unsupervised learning-based landslide warning device, characterized in that it is set to estimate whether a landslide has occurred by performing analysis on a spectral image acquired at a specific time based on the landslide prediction standard model. According to paragraph 1,
The processor,
Check weather information,
An unsupervised learning-based landslide warning device, characterized in that it is configured to collect a plurality of spectral images for each weather information and store the plurality of spectral images together with the weather information. According to paragraph 2,
The processor is
An unsupervised learning-based landslide warning device, characterized in that it is set to generate a class for each weather information by classifying the spectral images based on weather information at the time of collecting the plurality of spectral images. According to paragraph 3,
The processor is
Unsupervised, characterized in that it is set to generate the landslide prediction reference model including a plurality of clusters classified by weather information by performing artificial neural network clustering model learning based on the spectrum of spectral images divided into classes for each weather information. Learning-based landslide warning device. According to paragraph 4,
The processor is
Detecting the cluster with the highest similarity by comparing the spectrum of the spectral image acquired at the specific time when landslide prediction is required with the clusters included in the landslide prediction standard model,
An unsupervised learning-based landslide warning device, characterized in that it is set to estimate whether a landslide has occurred depending on whether the spectrum of the spectral image acquired at the specific time is outside a predefined standard range for the detected cluster. According to clause 5,
The processor is
An unsupervised learning-based landslide warning device that is set to transmit a risk warning about the possibility of a landslide to a designated manager device or an electronic device of a designated management office when the landslide is predicted to occur. Collecting, by a processor of a landslide warning device, a plurality of spectral images of the mountain slope by controlling a spectral camera arranged to acquire spectral images of at least a portion of the mountain slope;
Generating a landslide prediction reference model by performing clustering model learning on the spectra of the plurality of spectral images;
Saving the landslide prediction reference model;
An unsupervised learning-based landslide occurrence prediction method comprising: estimating whether a landslide has occurred by performing analysis on a spectral image acquired at a specific time based on the landslide prediction reference model. In clause 7,
The processor further includes collecting weather information,
The step of collecting the plurality of spectral images is
collecting a plurality of spectral images for each weather information;
An unsupervised learning-based landslide occurrence prediction method comprising: storing the plurality of spectral images together with the weather information. According to clause 8,
The step of generating the landslide prediction reference model is,
A step of performing artificial neural network clustering model learning based on the spectrum of spectral images divided into classes for each weather information, and generating the landslide prediction reference model including a plurality of clusters for each weather information,
The step of estimating whether a landslide has occurred is,
detecting a cluster with the highest similarity by comparing the spectrum of the spectral image acquired at the specific point in time when landslide prediction is required with clusters included in the landslide prediction standard model;
Estimating whether a landslide has occurred based on whether the spectrum of the spectral image acquired at the specific time deviates from a predefined reference range for the detected cluster; unsupervised learning-based landslide comprising a. How to predict occurrence. According to clause 9,
An unsupervised learning-based landslide occurrence prediction method further comprising transmitting a risk warning about the possibility of a landslide occurring to a designated manager device or an electronic device of a designated management office.

Images