KR102532738B1 - Apparatus and method for detecting a possibility of wildfire - Google Patents

Abstract

The present invention relates to an apparatus for detecting the possibility of a wildfire and a method for detecting the possibility of a wildfire. The apparatus for detecting the possibility of a wildfire according to some embodiments comprises: a data collection module for collecting satellite image data; a feature extracting module for extracting a first image feature related to smoke caused by a wildfire, the temperature feature of an infrared band and the wavelength feature of a visible band from the satellite image data at first coordinates; and a determination module for generating the wildfire possibility index of the first coordinates based on the extracted first image feature, the temperature feature and the wavelength feature, wherein the temperature feature is calculated based on the result of comparing the temperature measured at the first coordinates and the temperature measured at coordinates around the first coordinates, and the wavelength feature is can be calculated based on first wavelength data regarding a first wavelength and second wavelength data regarding a second wavelength shorter than the first wavelength.

Description

Forest fire possibility detection device and method {APPARATUS AND METHOD FOR DETECTING A POSSIBILITY OF WILDFIRE}

The present invention relates to an apparatus and method for detecting a potential forest fire.

Currently, hundreds of artificial satellites around the world are performing their own tasks in the earth's air for the purposes of meteorology, communication, broadcasting, agriculture, space development, and military.

Recently, forest fires have occurred frequently due to climate change and increased human activity. Disasters caused by forest fires destroy forests that have been afforestation for a long time, bring economic losses, and may also cause human casualties. In addition, it is difficult to obtain accurate information about fires in areas where direct observation is difficult.

Therefore, detection of the possibility of forest fires through artificial satellites is in the limelight, but with current technology, it is difficult to accurately detect the possibility of forest fires due to variables such as clouds and water vapor.

The problem to be solved by the present invention is to provide a forest fire possibility detection device.

Another problem to be solved by the present invention is to provide a method for detecting a possibility of forest fire.

The objects of the present invention are not limited to the above-mentioned objects, and other objects and advantages of the present invention not mentioned above can be understood by the following description and will be more clearly understood by the examples of the present invention. It will also be readily apparent that the objects and advantages of the present invention may be realized by means of the instrumentalities and combinations indicated in the claims.

An apparatus for detecting a possibility of forest fire according to some embodiments of the present invention includes a data collection module for collecting satellite image data, a first image feature related to smoke due to a forest fire from satellite image data, and an infrared band at a first coordinate. A feature extraction module for extracting a temperature feature and a wavelength feature of a visible band and a determination module for generating a forest fire likelihood index of a first coordinate based on the extracted first image feature, temperature feature, and wavelength feature, wherein the temperature feature, It is calculated based on a comparison result of the temperature measured at the first coordinate and the temperature measured at a coordinate near the first coordinate, and the wavelength characteristics include first wavelength data for a first wavelength among visible band data, It may be calculated based on second wavelength data for a second wavelength having a relatively shorter wavelength than the wavelength.

Meanwhile, the feature extraction module may further extract second image features related to clouds distributed in the air at the second coordinates.

On the other hand, the feature extraction module calculates the shape, movement direction, and movement speed of the cloud at the second coordinate to generate a predicted shape of the cloud at the time when a predetermined time has elapsed, and the first coordinate at the time when the predetermined time has elapsed Based on a comparison result between the first image feature extracted from and the predicted shape of the generated cloud, it may be verified whether the extracted first image feature corresponds to an image feature related to smoke.

Meanwhile, the feature extraction module extracts a first carbon dioxide index from data of a carbon dioxide-sensitive band at a first coordinate at which a first image feature is extracted, and at a second coordinate at which a second image feature is extracted, a carbon dioxide-sensitive band A second carbon dioxide index is extracted from the data of , and a first image feature is determined based on whether a threshold value of the extracted first carbon dioxide index and the second carbon dioxide index exceeds a threshold value and a comparison result between the first carbon dioxide index and the second carbon dioxide index. It can be verified whether it corresponds to the relevant image feature.

On the other hand, it further includes a learning module for learning a deep learning model using training data including at least one of a first image feature, a temperature feature, and a wavelength feature, and the determination module includes the first image feature, temperature feature, and wavelength feature. can be input into a deep learning model to generate a forest fire probability index.

Meanwhile, the learning module may determine bands of the first wavelength and the second wavelength within the visible band by using the training data.

A forest fire possibility detection method according to some embodiments of the present invention includes the steps of collecting satellite image data, a first image feature related to smoke from a forest fire, and a temperature feature of an infrared band from the satellite image data at a first coordinate. , Extracting a wavelength feature of the visible band and generating a forest fire likelihood index of a first coordinate based on the extracted first image feature, temperature feature, and wavelength feature, wherein the temperature feature is measured at the first coordinate It is calculated based on a comparison result of the measured temperature and the temperature measured at a coordinate near the first coordinate, and the wavelength characteristic is the first wavelength data for the first wavelength among visible band data and a relatively shorter wavelength than the first wavelength. It can be calculated based on the second wavelength data for the second wavelength.

Meanwhile, the method may further include extracting second image features related to clouds distributed in the air from the second coordinates.

Meanwhile, in the second coordinates, calculating the shape, movement direction, and movement speed of the cloud to generate the predicted shape of the cloud at the time when a predetermined time has elapsed, and the first coordinate extracted from the first coordinate at the time when the predetermined time has elapsed. The method may further include verifying whether or not the extracted first image feature corresponds to an image feature related to smoke based on a comparison result between one image feature and the predicted shape of the generated cloud.

Meanwhile, extracting a first carbon dioxide index from data of a band sensitive to carbon dioxide at a first coordinate from which a first image feature is extracted, and extracting a first carbon dioxide index from data of a band sensitive to carbon dioxide at a second coordinate from which a second image feature is extracted. 2 The first image feature is an image feature related to smoke based on the step of extracting the carbon dioxide index, whether the extracted first carbon dioxide index and the second carbon dioxide index exceed the threshold value, and the comparison result between the first carbon dioxide index and the second carbon dioxide index. A step of verifying whether or not corresponds to may be further included.

By extracting image features, temperature features, and wavelength features and using them together, the possibility of forest fires can be detected more accurately.

In addition to the above description, specific effects of the present invention will be described together while explaining specific details for carrying out the present invention.

1 illustrates a wildfire potential detection system in accordance with some embodiments.
2 is a block diagram of a device for detecting a potential forest fire in accordance with some embodiments.
3 is a block diagram of a feature extraction module in accordance with some embodiments.
4A and 4B are conceptual diagrams for explaining a process of verifying whether a first image feature corresponds to an image feature related to smoke, according to some embodiments.
5 is a conceptual diagram illustrating a process of extracting temperature features according to some embodiments.
6 is a flow diagram of a method for detecting a potential forest fire in accordance with some embodiments.
7 is a flowchart of a method for detecting a possibility of a forest fire according to some other embodiments.
8 is a flowchart of a method for detecting a possibility of a forest fire according to some other embodiments.

Terms or words used in this specification and claims should not be construed as being limited to a general or dictionary meaning. According to the principle that an inventor may define a term or a concept of a word in order to best describe his/her invention, it should be interpreted as meaning and concept consistent with the technical spirit of the present invention. In addition, the embodiments described in this specification and the configurations shown in the drawings are only one embodiment in which the present invention is realized, and do not represent all of the technical spirit of the present invention, so they can be replaced at the time of the present application. It should be understood that there may be many equivalents and variations and applicable examples.

Terms such as first, second, A, and B used in this specification and claims may be used to describe various components, but the components should not be limited by the terms. Terms are only used to distinguish one component from another. For example, a first element may be termed a second element, and similarly, a second element may be termed a first element, without departing from the scope of the present invention. The term 'and/or' includes a combination of a plurality of related recited items or any one of a plurality of related recited items.

Terms used in this specification and claims 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. It should be understood that terms such as "include" or "having" in this application do not exclude in advance the possibility of existence or addition of features, numbers, steps, operations, components, parts, or combinations thereof described in the specification. .

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which the present invention belongs.

Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning in the context of the related art, and unless explicitly defined in the present application, they should not be interpreted in an ideal or excessively formal meaning. don't In addition, each configuration, process, process or method included in each embodiment of the present invention may be shared within a range that does not contradict each other technically.

Also, the term "module" used in the specification means a software or hardware component, and a "module" performs certain roles. However, "module" is not meant to be limited to software or hardware. A "module" may be configured to reside on an addressable storage medium and may be configured to reproduce one or more processors. Thus, as an example, a "module" may include components such as software components, object-oriented software components, class components, and task components, processes, functions, properties, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuitry, data, databases, data structures, tables, arrays and variables. Functionality provided within components and “modules” may be combined into fewer components and “modules” or further separated into additional components and “modules”.

Hereinafter, with reference to FIGS. 1 to 8 , a forest fire possibility detection device and method according to some embodiments of the present invention will be described.

1 illustrates a forest fire potential detection system 1 according to some embodiments.

The forest fire possibility detection system 1 may include a satellite image capture device 110 , an external data server 115 and a forest fire possibility detection device 120 .

The satellite image capture device 110 is a device for capturing satellite image data, and may include, for example, an artificial satellite located on the earth's surface. The orbit and movement method of the satellite image capture device 110 may vary, and for some examples, the satellite image capture device 110 may include a low-orbit satellite, a polar-orbit satellite, and a geostationary orbit satellite.

The satellite image capture device 110 transmits a signal directly from a passive sensor and/or satellite that observes energy re-radiated after reflecting or absorbing solar energy by an object on the earth and a signal reflected back by an object on the earth. It may include an active sensor or the like that measures .

For some examples, the satellite image capture device 110 may include an optical sensor, a multi-spectral sensor, a near-infrared sensor, an infrared sensor, a radar sensor, and the like. Satellite image data captured by the satellite image capture device 110 may include an optical signal, a multi-spectral signal, a near-infrared signal, an infrared signal, a radar image, and the like. In this case, as some examples, the radar image may include a synthetic aperture radar image, that is, a synthetic aperture radar (SAR) image, but is not limited thereto.

Satellite image data photographed by the satellite image capture device 110 may include data of various wavelength bands (bands). For example, satellite image data includes ultraviolet (UV) band data, visible band data, near infrared band data, infrared band data, specific components (eg, carbon dioxide, carbon monoxide, etc.) ) may include data of a band sensitive to .

The satellite image capture device 110 may capture satellite image data containing the image of the earth at a predetermined location on the earth's surface, and the captured satellite image data may be transmitted to the forest fire possibility detection device 120 . At this time, the forest fire possibility detection system 1 may include a separate network and/or communication network for satellite communication between the satellite image capture device 110 and the forest fire possibility detection device 120.

The external data server 115 may be one or more servers that store other data.

Other data may include, for example, weather information (eg, climate information, weather information, precipitation information), wind speed information, cloud status information, typhoon information, tidal wave information, fire information, forest fire history information, and the like. Forest fire history information may include, for example, forest fire history information according to seasons, forest fire history information by year, information on areas damaged by past forest fires, and the like.

The forest fire possibility detection device 120 may receive data from the satellite image capture device 110 and the external data server 115, and store, manage, and/or analyze the received data.

For some examples, the forest fire possibility detection device 120 may generate forest fire-related analysis data based on received satellite image data and other data.

The detailed configuration and operating method of the forest fire possibility detection device 120 will be described in detail with reference to FIG. 2 .

2 is a block diagram of a wildfire potential detection device 120 in accordance with some embodiments.

Referring to FIG. 2 , a forest fire possibility detection device 120 may include a data collection module 210 , a feature extraction module 220 , a determination module 230 and a learning module 240 .

The data collection module 210 may collect data for generating analysis data related to forest fires.

For some examples, the data collection module 210 may collect satellite image data and other data. At this time, as described above, the satellite image data may be captured by the satellite image capture device 110, and other data may be data stored and managed by the external data server 115.

As described above, satellite image data may include data of various wavelength bands (bands). For example, satellite image data includes ultraviolet (UV) band data, visible band data, near infrared band data, infrared band data, specific components (eg, carbon dioxide, carbon monoxide, etc.) ) may include data of a band sensitive to .

The feature extraction module 220 may extract features related to forest fires at the first coordinates from satellite image data and/or other data. In this case, the first coordinate may mean a coordinate of interest for generating analysis data related to a forest fire.

In some examples, wildfire-related features may include, but are not limited to, image features, temperature features, wavelength features, and the like. A detailed configuration of the feature extraction module will be described later with reference to FIGS. 3 to 5 .

The determination module 230 may generate analysis data related to forest fires based on the features extracted by the feature extraction module 220 .

In some examples, the decision module 230 may generate analytical data related to a wildfire using a predetermined mathematical formula for a relationship between the extracted features and/or other collected data and a wildfire probability index.

For example, the determination module 230 may generate analysis data related to a forest fire by inputting a first image feature, a temperature feature, and a wavelength feature related to smoke from a forest fire into a deep learning model received from the learning module 240. . The deep learning model will be described later in the learning module 240.

The analysis data related to forest fires may include, but are not limited to, a forest fire likelihood index, which means the probability that a forest fire is occurring at a predetermined coordinate, a forest fire spread rate, a forest fire evolution rate, a forest fire final damage area, and the like. . For convenience of explanation, the analysis data related to forest fires is described as a forest fire probability index.

For some examples, when at least one of the extracted first image feature, temperature feature, and wavelength feature falls short of a predetermined criterion, the determination module 230 controls the data collection module 210 to control the data collection module 210 without determining a forest fire likelihood index, and the satellite satellite The image may be collected again and/or the feature extraction module 220 may be controlled to extract features from the satellite image again.

In some other examples, when at least one of the extracted first image feature, temperature feature, and wavelength feature does not meet a predetermined criterion, the determination module 230 sets a weight of an item corresponding to a feature that does not meet the criterion in the deep learning model. can be given low. However, it is not limited to the above example.

The learning module 240 may perform learning using training data for generating a forest fire probability index.

In some examples, the training module 240 may train a deep learning model to generate a wildfire probability index. For example, the learning module 240 may train a deep learning model using training data including collected satellite image data, other data, and/or extracted forest fire-related features.

At this time, the learned deep learning model uses satellite image data, other data, and / or features related to forest fires as input data based on machine learning, and outputs a forest fire probability index as output data. there is.

Deep learning technology, a type of machine learning, learns by going down to a multi-level deep level based on data. That is, deep learning represents a set of machine learning algorithms that extract core data from a plurality of data while stepping up.

The deep learning model may use various known deep learning structures. For example, the deep learning model may use a structure such as a convolutional neural network (CNN), a recurrent neural network (RNN), a deep belief network (DBN), or a graph neural network (GNN).

Specifically, CNN (Convolutional Neural Network) extracts the basic features of an object when a person recognizes an object, and then performs complex calculations in the brain to recognize the object based on the result. It is a simulated model.

RNN (Recurrent Neural Network) is widely used in natural language processing, etc., and is an effective structure for processing time-series data that changes over time.

DBN (Deep Belief Network) is a deep learning structure composed of multiple layers of RBM (Restricted Boltzman Machine), a deep learning technique. When a certain number of layers is obtained by repeating RBM (Restricted Boltzman Machine) learning, a DBN (Deep Belief Network) having a corresponding number of layers may be configured.

GNN (Graphic Neural Network, hereinafter, GNN) represents an artificial neural network structure implemented in a way to derive similarities and feature points between modeling data using modeling data modeled on the basis of data mapped between specific parameters. .

On the other hand, learning of the artificial neural network of the deep learning model can be performed by adjusting the weight of the connection line between nodes (and adjusting the bias value if necessary) so that a desired output is produced for a given input. In addition, the artificial neural network may continuously update a weight value by learning. In addition, a method such as back propagation may be used to learn the artificial neural network.

In this case, unsupervised learning, semi-supervised learning, supervised learning, and the like may be used as the machine learning method of the artificial neural network. In addition, the deep learning model can be controlled to automatically update the artificial neural network structure for outputting analysis data after learning according to settings.

The deep learning model learned in the learning module 240 may be delivered to the decision module 230 and stored separately or an existing model may be updated.

As another example, the learning module 240 may determine the band of the first wavelength and the second wavelength, which are the basis of the wavelength feature, using training data such as collected satellite image data and other data.

Before a wildfire occurs, plants that live in the area can reflect some of the visible band (wavelength range) and absorb some. In some instances, plants may reflect a first wavelength and absorb a second wavelength. At this time, when a forest fire occurs, plants existing in an area where a forest fire occurs may be burned due to the fire. When the vegetation is burned, it can be predicted that reflectance and/or absorption of the first wavelength and the second wavelength will be different in the area compared to before a forest fire occurs.

Using this principle, the learning module 240 may compare data before and after a forest fire to determine a first wavelength band and a second wavelength band that generate a difference in reflectivity and/or absorbance at predetermined coordinates.

In this way, by using the wavelength features extracted using the data of the visible band, it has a remarkable effect that the possibility of a forest fire can be detected even using only the data of the visible band without using the data of the near-infrared band or the infrared band.

As another few examples, the learning module 240 may determine a band sensitive to carbon dioxide by using training data such as collected satellite image data and other data. For example, the learning module 240 may determine a band sensitive to carbon dioxide among a plurality of bands of satellite image data by using existing forest fire history information, satellite image data of an area where a large amount of carbon dioxide is generated, and the like as training data.

3 is a block diagram of a feature extraction module 220 in accordance with some embodiments.

Referring to FIG. 3 , the feature extraction module 220 may include an image feature extractor 310 , a temperature feature extractor 320 , and a wavelength feature extractor 330 .

The image feature extraction unit 310 may include a first image feature extraction unit 311 , a second image feature extraction unit 312 and a verification unit 313 . In this case, the first image feature may be an image feature related to smoke from a forest fire, and the second image feature may mean an image feature related to clouds distributed in the air.

The first image feature extractor 311 may extract a first image feature related to smoke due to a forest fire from satellite image data collected at a first coordinate, that is, a coordinate of interest for generating a forest fire probability index.

In some examples, the first image feature associated with smoke from a forest fire may include size or extent of smoke, direction, outline, and the like. For example, the first image feature extractor 311 may extract, as the first image feature, a shape that gradually increases in size over time and/or a shape that spreads in all directions without a flow in a certain direction. It is not limited thereto.

For some examples, the first image feature extractor 311 may extract the first image feature from satellite image data using a deep learning model that recognizes an object from an image learned by the learning module 240 .

As another example, the first image feature extractor 311 may extract a first image feature from satellite image data using a predetermined object recognition model including an algorithm for recognizing an object from an image. The object recognition model may include, for example, FCN, Mask R-CNN, U-Net, image contour analysis model, etc., but is not limited thereto.

The second image feature extractor 312 may extract second image features related to clouds distributed in the air from satellite image data collected at the second coordinates.

For some examples, the second image feature related to clouds may include a size, a curvature, a change in curvature, an outline, and the like of the object. For example, the second image feature extractor 312 may extract, as the second image feature, a shape that has a constant shape but moves in a predetermined direction over time, but is not limited thereto.

For some examples, the second image feature extractor 312 may extract second image features from satellite image data using a deep learning model that recognizes an object from an image learned by the learning module 240 .

As another example, the second image feature extractor 312 may extract second image features from satellite image data using a predetermined object recognition model including an algorithm for recognizing an object from an image. The object recognition model may include, for example, FCN, Mask R-CNN, U-Net, image contour analysis model, etc., but is not limited thereto.

The verifier 313 may verify whether the first image feature extracted by the first image feature extractor 311 corresponds to an image feature related to acting.

In satellite imagery, smoke generated by forest fires and clouds distributed in the atmosphere are expressed in similar shapes and colors, so they may be confused and not clearly distinguished. Therefore, the extracted first image feature is an image related to smoke. A process of verifying whether or not it corresponds to the characteristic may be necessary. The present invention has a remarkable effect of detecting the possibility of a forest fire more accurately by verifying whether or not the first image feature extracted through the method according to the following exemplary embodiments corresponds to an image feature related to smoke.

For some examples, the verification unit 313 may verify whether the first image feature extracted based on the predicted shape of the cloud corresponds to an image feature related to smoke. It will be described in detail with reference to FIGS. 4a and 4b.

4A and 4B are conceptual diagrams for explaining a process of verifying whether a first image feature corresponds to an image feature related to smoke, according to some embodiments.

FIG. 4A shows image features S1 and C1 at a specific point in time, and FIG. 4B shows the image feature S2 and predicted shape C2 of clouds at a point in time after a certain time has elapsed from FIG. 4A. .

4A shows a first image feature S1 related to smoke extracted from a first coordinate and a second image feature C1 related to clouds distributed in the air extracted from a second coordinate.

Referring to FIGS. 3 and 4A , the verification unit 313 calculates the shape (eg, contour), movement direction (eg, northwest), and movement speed of the cloud in the image feature C1 related to the extracted cloud. can

For some examples, when the cloud-related image feature C1 is extracted, the verification unit 313 continuously extracts image features corresponding to the cloud image feature C1 at predetermined time intervals, and implements it as, for example, a time-lapse video. there is.

As another example, the verification unit 313 may generate a cloud motion vector.

The verification unit 313 may calculate a rate of change in the shape of the cloud, a moving direction, a moving speed, and the like of the cloud-related image feature C1 from the implemented time-lapse image and/or the cloud movement vector.

The verification unit 313 may generate a predicted shape of a cloud at a point in time when a predetermined time elapses based on the calculated shape change rate, movement direction, and movement speed of the cloud.

FIG. 4B shows a predicted form C2 of a generated cloud and a first image feature S2 related to smoke extracted from a first coordinate at a time point corresponding to the predicted form C2 of a cloud.

The verifier 313 may compare the predicted shape C2 of the generated cloud with the extracted first image feature S2.

As some examples, as shown in FIG. 4B , the predicted shape C2 of the cloud was predicted to move in a northwesterly direction than the image feature C1 of the cloud due to atmospheric convection, but the first image feature S2 If the region only spreads and does not move in a specific direction, the verifying unit 313 may confirm that the extracted first image feature S2 is an image feature related to smoke.

In some other examples, as shown in FIG. 4B , while the predicted shape C2 of the cloud moves to the northwest, it is predicted to maintain the shape of the existing image feature C1 of the cloud, but the first image feature S2 is the original image feature S2. If the shape of the first image feature S1 extracted in S1 is extracted in an irregularly diffused form without maintaining the shape, the verification unit 313 determines that the extracted first image feature S2 is an image feature related to smoke. can

As another example, when the wind speed is rather strong, the area located on the windward side of the cloud image feature (C1) has a large degree of change in the shape of the cloud, and the area located on the opposite side to the windward side has a large cloud shape. Although the degree of change of is predicted to be small, the first image feature (S2) does not have a difference in the degree of change between the area located on the windward side and the area located on the opposite side to the windward side, and the wind simply blows. In the case of only spreading in a thin direction, the verifying unit 313 may confirm that the extracted first image feature S2 is an image feature related to smoke.

The verification unit 313 may verify whether the first image feature corresponds to an image feature related to smoke by mutually combining the method of comparing the predicted cloud shape C2 and the extracted first image feature S2. there is.

Referring back to FIG. 3 , as another example, the verification unit 313 may verify whether the first image feature extracted based on the carbon dioxide index corresponds to an image feature related to smoke.

Smoke from wildfires can have a higher proportion of carbon dioxide in its composition than atmospheric clouds. Among the satellite image data, by using the data of the carbon dioxide sensitive band (hereinafter referred to as "A" band) to estimate and use the carbon dioxide level, it is determined whether the extracted first image feature corresponds to the image feature related to smoke. This has the significant effect of being able to verify and thereby more accurately detect the possibility of forest fires.

At this time, the specific wavelength region value of the A band may be input by the manager and/or manager of the forest fire possibility detection device (FIG. 2, 120), or may be determined through a predetermined learning process by the learning module (FIG. 2, 240). .

For some examples, the verifier 313 may extract the carbon dioxide index from each of the first coordinates from which the first image feature is extracted and the second coordinates from which the second image feature is extracted. That is, the verification unit 313 extracts the first carbon dioxide index from the A-band data at the first coordinates where the first image feature is extracted, and extracts the first carbon dioxide index from the A-band data at the second coordinates where the second image feature is extracted. A second carbon dioxide index may be extracted.

The verifier 313 determines the extracted first carbon dioxide index when the extracted first carbon dioxide index exceeds the threshold value, the extracted second carbon dioxide index is below the threshold value, and the difference between the first carbon dioxide index and the second carbon dioxide index is equal to or greater than the reference value. 1 It can be confirmed that the image feature is an image feature related to acting.

The temperature feature extraction unit 320 may extract temperature features from infrared band data among collected satellite image data.

The temperature characteristic may be calculated based on a comparison result of the temperature measured at the first coordinate and the temperature measured at a coordinate near the first coordinate. Referring to FIG. 5, a process of extracting temperature features will be described in detail.

5 is a conceptual diagram illustrating a process of extracting temperature features according to some embodiments.

Referring to FIG. 5 , the first coordinate, which is a target for generating a forest fire possibility index, is represented by P1, the coordinates near the first coordinate by P2, and other coordinates by P3.

In the infrared band data, the higher the temperature, the brighter the temperature. As the coordinates of the ignition point, P1, are darker than the nearby coordinates, P2, and P2 is darker than the other coordinates, P3.

When a forest fire occurs, it can be predicted that the forest fire will spread from the ignition point P1 to the surroundings, and eventually, coordinates that are shown darkly around the ignition point P1 in the infrared band data will expand.

As shown in FIG. 5 , the temperature feature extractor 320 may extract a temperature feature based on a temperature difference between coordinates represented by a contrast difference between coordinates in infrared band data.

For some examples, the temperature feature extractor 320 calculates a contrast difference D11 between P1 and P2 at a first time point, calculates a contrast difference D12 between P2 and P3 at a first time point, and calculates a contrast difference D12 between P2 and P3 at a second time point P1 A contrast difference (D21) between P2 and P2 may be calculated, and a contrast difference (D22) between P2 and P3 may be calculated at the second time point. Thereafter, the temperature feature extractor 320 may extract the temperature feature by substituting each calculated contrast difference into Equation 1 below, but is not limited thereto.

<Equation 1>

Here, TF means the temperature characteristic.

That is, assuming that the first time point is the initial stage of ignition, the temperature of P1 will be high and the temperatures of P2 and P3 will be low at the first time point. Assuming that the second time point is an ignition diffusion time point, the temperature of P1 and P2 will be high and the temperature of P3 will be relatively low at the second time point. At this time, D11 and D22 may be relatively large, and D12 and D21 may be relatively low. As a result, it may be determined that a forest fire is more likely to occur at the coordinates of interest as the temperature feature has a higher value.

In this way, by using the temperature feature, it is possible to estimate the speed and area of a forest fire along with the possibility of a forest fire, which has a remarkable effect that coping with a forest fire can be made easier.

Referring back to FIG. 3 , the wavelength feature extraction unit 330 may extract wavelength features related to forest fires from visible band data among collected satellite image data.

For some examples, the wavelength characteristics may be calculated based on first wavelength data for a first wavelength among visible band data and second wavelength data for a second wavelength having a relatively shorter wavelength than the first wavelength.

For example, the wavelength feature extractor 330 may calculate the wavelength feature using Equation 2 below.

<Equation 2>

Here, WF means a wavelength characteristic related to a forest fire, and Q1 and Q2 are data for a first wavelength and data for a second wavelength, respectively, among visible band data, that is, measured reflectivity of the first and second wavelengths can mean

For some examples, bands (wavelength ranges) of the first wavelength and the second wavelength may be previously determined among visible light bands. For some examples, the bands of the first wavelength and the second wavelength may be set in advance by a learning process of the learning module (FIG. 2, 240).

Before a wildfire occurs, plants in the area can reflect some of the visible band and absorb some. In some instances, plants may reflect a first wavelength and absorb a second wavelength. At this time, when a forest fire occurs, plants existing in an area where a forest fire occurs may be burned due to the fire. When the vegetation is burned, it can be predicted that reflectance and/or absorption of the first wavelength and the second wavelength will be different in the area compared to before a forest fire occurs.

Using this principle, the learning module (FIG. 2, 240) compares data before and after a forest fire to determine a band of a first wavelength and a second wavelength that generate a difference in reflectance and/or absorbance of predetermined coordinates. there is.

In this way, by using the wavelength features extracted using the data of the visible band, it has a remarkable effect that the possibility of a forest fire can be detected even using only the data of the visible band without using the data of the near-infrared band or the infrared band.

6 is a flow diagram of a method for detecting a potential forest fire in accordance with some embodiments. Each step of FIG. 6 may be performed by the forest fire possibility detection device 120 of FIGS. 1 and 2 . Hereinafter, redundant contents are omitted and briefly described.

First, satellite image data may be collected (S110).

In this case, the satellite image data may include data of various wavelength bands (bands) as described above. For example, satellite image data includes ultraviolet (UV) band data, visible band data, near infrared band data, infrared band data, specific components (eg, carbon dioxide, carbon monoxide, etc.) ) may include data of a band sensitive to .

Subsequently, a first image feature related to smoke caused by a forest fire is extracted from the satellite image data collected at the first coordinates (S120), and from among the satellite image data collected at the first coordinates, infrared band data related to the forest fire is extracted. A temperature feature may be extracted (S130), and a wavelength feature related to a forest fire may be extracted from visible band data among satellite image data collected at the first coordinate (S140).

For some examples, it is possible to extract features of smoke from a forest fire from satellite image data using an object recognition model including an algorithm for recognizing an object from an image. A detailed explanation is omitted.

For some examples, the temperature characteristic may be calculated based on a comparison result of a temperature measured at the first coordinate and a temperature measured at a coordinate near the first coordinate. For some examples, the wavelength characteristics may be calculated based on first wavelength data for a first wavelength among visible band data and second wavelength data for a second wavelength having a relatively shorter wavelength than the first wavelength.

Subsequently, a forest fire possibility index of the first coordinates may be generated based on the extracted first image feature, temperature feature, and wavelength feature (S150).

In some examples, a forest fire probability index may be generated by inputting a first image feature, a temperature feature, and a wavelength feature to a deep learning model to be trained. A detailed explanation is omitted.

7 is a flowchart of a method for detecting a possibility of a forest fire according to some other embodiments. Each step of FIG. 7 may be performed by the forest fire possibility detection device 120 of FIGS. 1 and 2 .

First, satellite image data may be collected (S110). Step S110 has been described above with reference to FIG. 6, and is omitted here.

Subsequently, a first image feature related to smoke caused by a forest fire is extracted from the satellite image data collected at the first coordinates (S120), and from among the satellite image data collected at the first coordinates, infrared band data related to the forest fire is extracted. A temperature feature may be extracted (S130), and a wavelength feature related to a forest fire may be extracted from visible band data among satellite image data collected at the first coordinate (S140).

Steps S120 to S140 have been described above with reference to FIG. 6 and will be omitted here.

Subsequently, a second image feature related to clouds distributed in the air may be extracted from the second coordinate (S121).

For some examples, features related to the shape of clouds may be extracted from satellite image data using an object recognition model including an algorithm for recognizing an object from an image. A detailed explanation is omitted.

Subsequently, the shape, movement direction, and movement speed of the cloud may be calculated from the second coordinates to generate the predicted shape of the cloud when a predetermined time elapses ( S122 ).

For some examples, when image features related to clouds are extracted, image features corresponding to cloud image features may be continuously extracted at predetermined time intervals and implemented as, for example, a time-lapse video. As another example, cloud movement vectors may be generated.

The shape change rate, movement direction, and movement speed of clouds can be calculated from the implemented time-lapse image and/or cloud movement vector. A detailed explanation is omitted.

Then, based on the comparison result of the first image feature extracted from the first coordinates and the predicted shape of the generated cloud at the time point when a predetermined time has elapsed, whether the extracted first image feature corresponds to the image feature related to smoke can be verified (S123).

In some examples, the predicted shape of the cloud was predicted to move in a northwesterly direction than the image feature of the cloud due to atmospheric convection, but the first image feature only spreads in the area and does not move in a specific direction. It can be confirmed that the first image feature is an image feature related to smoke.

In some other examples, the predicted shape of the cloud was predicted to maintain the shape of the image feature of the existing cloud while moving to the northwest, but the first image feature does not maintain the shape of the previously extracted first image feature and diffuses irregularly. When extracted in the form of smoke, it can be confirmed that the extracted first image feature is an image feature related to smoke.

As another example, when the wind speed is rather strong, the area located on the windward side of the cloud image features has a large degree of change in the shape of the cloud, and the area located on the opposite side of the windward side has a high degree of change in the shape of the cloud Although predicted to be small, the first image feature does not exist in the difference between the degree of change in the area located on the windward side and the area located on the opposite side of the windward side, and simply spreads in the windward direction. In this case, it can be confirmed that the extracted first image feature is an image feature related to smoke. A detailed explanation is omitted.

Subsequently, a forest fire possibility index of the first coordinates may be generated based on the extracted first image feature, temperature feature, and wavelength feature (S150). Step S150 has been described above with reference to FIG. 6, and is omitted here.

8 is a flowchart of a method for detecting a possibility of a forest fire according to some other embodiments. Each step of FIG. 8 may be performed by the forest fire possibility detection device 120 of FIGS. 1 and 2 .

First, satellite image data may be collected (S110). Step S110 has been described above with reference to FIG. 6, and is omitted here.

Subsequently, a first image feature related to smoke caused by a forest fire is extracted from the satellite image data collected at the first coordinates (S120), and from among the satellite image data collected at the first coordinates, infrared band data related to the forest fire is extracted. A temperature feature may be extracted (S130), and a wavelength feature related to a forest fire may be extracted from visible band data among satellite image data collected at the first coordinate (S140).

Steps S120 to S140 have been described above with reference to FIG. 6 and will be omitted here.

Subsequently, a second image feature related to clouds distributed in the air may be extracted from the second coordinate (S121). Step S121 has been described above with reference to FIG. 7 and is omitted here.

Subsequently, a first carbon dioxide index may be extracted from the data of the carbon dioxide-sensitive band at the first coordinates from which the first image feature is extracted (S124), and at the second coordinates from which the second image feature is extracted, the carbon dioxide-sensitive band A second carbon dioxide index may be extracted from the data of (S125).

Smoke from wildfires can have a higher proportion of carbon dioxide in its composition than atmospheric clouds. Among the satellite image data, by using the data of the band sensitive to carbon dioxide (hereinafter referred to as "A" band) to estimate and use the carbon dioxide level, it is determined whether the extracted first image feature corresponds to the image feature related to smoke. This has the significant effect of being able to verify and thereby more accurately detect the possibility of forest fires.

For some examples, the carbon dioxide index may be extracted from each of the first coordinates from which the first image feature is extracted and the second coordinates from which the second image feature is extracted. That is, at the first coordinates where the first image feature is extracted, the first carbon dioxide index is extracted from the data of the A band, and at the second coordinate where the second image feature is extracted, the second carbon dioxide index is extracted from the data of the A band. can do. A detailed explanation is omitted.

Next, it is verified whether the first image feature corresponds to the image feature related to smoke based on whether the extracted first carbon dioxide index and the second carbon dioxide index exceed the threshold value and the comparison result between the first carbon dioxide index and the second carbon dioxide index. It can (S126).

In some examples, when the extracted first carbon dioxide index exceeds the threshold value, the extracted second carbon dioxide index is less than the threshold value, and the difference between the first carbon dioxide index and the second carbon dioxide index is greater than or equal to a reference value, the extracted first image feature It can be confirmed that this is an image feature related to acting.

Subsequently, a forest fire possibility index of the first coordinates may be generated based on the extracted first image feature, temperature feature, and wavelength feature (S150). Step S150 has been described above with reference to FIG. 6, and is omitted here.

The above description is merely an example of the technical idea of the present embodiment, and various modifications and variations can be made to those skilled in the art without departing from the essential characteristics of the present embodiment. Therefore, the present embodiments are not intended to limit the technical idea of the present embodiment, but to explain, and the scope of the technical idea of the present embodiment is not limited by these embodiments. The scope of protection of this embodiment should be construed according to the claims below, and all technical ideas within the scope equivalent thereto should be construed as being included in the scope of rights of this embodiment.

Claims (10)

A data collection module for collecting satellite image data;
a feature extraction module for extracting a first image feature related to smoke from a forest fire, a temperature feature of an infrared band, and a wavelength feature of a visible band from the satellite image data at a first coordinate at a first point in time; and
A determination module for generating a forest fire likelihood index of the first coordinate based on the extracted first image feature, temperature feature, and wavelength feature,
The temperature characteristic is calculated based on a comparison result of a temperature measured at the first coordinate and a temperature measured at a coordinate near the first coordinate,
The wavelength characteristics are calculated based on first wavelength data for a first wavelength among data of the visible band and second wavelength data for a second wavelength having a relatively shorter wavelength than the first wavelength,
The feature extraction module,
Further extracting second image features related to clouds distributed in the air from the second coordinates of the first viewpoint,
From the second coordinates at the first time point, a shape change rate, a moving direction, and a moving speed of the cloud are calculated to generate a predicted shape of the cloud at a second time point when a predetermined time has elapsed compared to the first time point; ,
Based on a comparison result between the first image feature extracted from the first coordinates of the second viewpoint and the generated prediction shape of the cloud at the second viewpoint, the first image feature extracted at the first viewpoint A wildfire probable detection device that verifies whether or not this plume corresponds to image features related to smoke.
delete delete delete According to claim 1,
Further comprising a learning module for learning a deep learning model using training data including at least one of the first image feature, the temperature feature, and the wavelength feature,
The decision module,
A forest fire possibility detection device for generating the forest fire possibility index by inputting the first image feature, temperature feature, and wavelength feature to the deep learning model.
According to claim 5,
The learning module,
Forest fire possibility detection device for determining bands of the first wavelength and the second wavelength within the visible band using the training data.
Collecting satellite image data;
extracting a first image feature related to smoke from a forest fire, a temperature feature of an infrared band, and a wavelength feature of a visible band from the satellite image data at a first coordinate of a first time point;
extracting a second image feature related to clouds distributed in the air from the second coordinate of the first viewpoint;
From the second coordinates at the first time point, calculating the shape change rate, movement direction, and movement speed of the cloud, and generating a predicted shape of the cloud at a second time point when a predetermined time has elapsed compared to the first time point. step;
Based on a comparison result between the first image feature extracted from the first coordinates of the second viewpoint and the generated prediction shape of the cloud at the second viewpoint, the first image feature extracted at the first viewpoint verifying whether or not this corresponds to an image characteristic related to acting; and
Generating a forest fire likelihood index of the first coordinate based on the extracted first image feature, temperature feature, and wavelength feature,
The temperature characteristic is calculated based on a comparison result of a temperature measured at the first coordinate and a temperature measured at a coordinate near the first coordinate,
The wavelength feature is calculated based on first wavelength data for a first wavelength among data of the visible band and second wavelength data for a second wavelength having a relatively shorter wavelength than the first wavelength. delete delete delete

Images