WO2022186306A1

WO2022186306A1 - Fire detecting device, and fire detecting method

Info

Publication number: WO2022186306A1
Application number: PCT/JP2022/008985
Authority: WO
Inventors: 悠子川添; 正雄長谷川
Original assignee: 株式会社Ihi
Priority date: 2021-03-02
Filing date: 2022-03-02
Publication date: 2022-09-09
Also published as: JPWO2022186306A1

Abstract

A fire detecting device (3, 3B) for detecting a fire on the Earth using observation data from a radiometer (100) mounted on an artificial satellite (1) is provided with: an image generating unit (30) for generating a pseudo-color composite image on the basis of the observation data; an extracting unit (33) for extracting an image of a zone corresponding to a fire detection area from within the pseudo-color composite image; a processing unit (34) for expanding pixel data of each pixel in the image extracted by the extracting unit; and a determining unit (35, 35B) for determining the presence or absence of a fire on the basis of the pixel data expanded by the processing unit.

Description

Fire detection device and fire detection method

The present disclosure relates to a fire detection device and fire detection method.
This application claims priority based on Japanese Patent Application No. 2021-032432 filed in Japan on March 2, 2021, the content of which is incorporated herein.

A known method is to detect forest fire smoke using satellite observation data. For example, Non-Patent Document 1 discloses a method of generating an RGB pseudo-color composite image from observation data and detecting forest fire smoke from the data of this pseudo-image.

However, due to atmospheric influences such as the amount of moisture in the atmosphere, it becomes difficult to distinguish between data that indicate the occurrence of fires and data that does not indicate the occurrence of fires in the pseudo-color composite image, which reduces the accuracy of smoke-based fire detection. Sometimes.

The present disclosure has been made in view of such circumstances, and aims to provide a fire detection device and a fire that can improve the accuracy of fire detection when performing fire detection from pseudo-color composite image data. It is to provide a detection method.

(1) A first aspect of the present disclosure is a fire detection device that detects a fire on the earth using observation data of a radiometer mounted on an artificial satellite, wherein pseudo-color synthesis is performed based on the observation data An image generating unit for generating an image, an extracting unit for extracting an image of a range corresponding to a fire detection area from the pseudo-color composite image, and decompressing pixel data for each pixel in the image extracted by the extracting unit. and a determination unit that determines whether or not there is a fire based on the pixel data decompressed by the processing unit.

(2) In the fire detection device of (1) above, the processing unit normalizes the decompressed pixel data, and the determination unit determines whether or not there is a fire based on the normalized pixel data. may

(3) The fire detection device of (1) above, wherein the pixel data includes a smoke aerosol reflectance index, which is an index related to reflectance to smoke, and a moisture index related to the amount of moisture in the atmosphere and on the ground, The processing unit may decompress either or both of the smoke aerosol reflectance index and the moisture index among the pixel data for each pixel in the image extracted by the extraction unit.

(4) In the fire detection device of (3) above, the processing unit may normalize the expanded index out of the smoke aerosol reflectance index and the moisture index.

(5) The fire detection device according to any one of (1) to (4) above, wherein the position is suspected to be a fire based on information of an observation band indicating brightness temperature among the observation data. A fire candidate position specifying unit that specifies a fire candidate position and a fire detection area creating unit that creates a predetermined area including the fire candidate position as a fire detection area may be provided.

(6) In the fire detection device according to any one of (1) to (5) above, the determination unit receives pixel data or a value based on the pixel data as input data, and determines whether the input data indicates a fire. The presence or absence of a fire may be determined by inputting the pixel data decompressed by the processing unit or a value based on the pixel data into the learning model.

(7) A second aspect of the present disclosure is a fire detection method for a fire detection device that detects a fire on the earth using observation data of a radiometer mounted on a satellite, comprising: extracting an image of a range corresponding to a fire detection area from the pseudo-color composite image; and decompressing pixel data of each pixel in the extracted image. and determining whether or not there is a fire based on the decompressed pixel data.

As described above, according to the present disclosure, it is possible to improve the accuracy of fire detection when performing fire detection from pseudo-color composite image data.

BRIEF DESCRIPTION OF THE DRAWINGS It is a figure which shows an example of schematic structure of the fire detection system which concerns on 1st Embodiment. 1 is a schematic configuration diagram of a fire detection device according to a first embodiment; FIG. 4 is a flow chart of the fire detection device according to the first embodiment; It is a figure which shows an example of schematic structure of the fire detection system which concerns on 2nd Embodiment. It is a schematic block diagram of the fire detection apparatus which concerns on 2nd Embodiment. FIG. 10 is a diagram showing determination thresholds (boundary lines) according to the second embodiment; It is a flow chart of a fire detection device concerning a 2nd embodiment.

The fire detection device according to this embodiment will be described below with reference to the drawings.

<First Embodiment>
FIG. 1 is a diagram showing an example of a schematic configuration of a fire detection system A including a fire detection device according to the first embodiment. A fire detection system A includes a satellite 1 , a ground receiving station 2 and a fire detection device 3 .

Artificial satellite 1 is a geostationary satellite placed in a geostationary orbit at an altitude of about 36,000 km above the equator, and revolves at the same period as the rotation of the earth. For example, the artificial satellite 1 is the geostationary meteorological satellite "Himawari". This artificial satellite 1 is equipped with a visible and infrared radiometer 100, and can measure radiance (observation data) in a predetermined wavelength band (observation band) at regular intervals.

Therefore, the artificial satellite 1 can acquire observation data with a higher time resolution than a low-orbit satellite (satellite that orbits the earth at an altitude of several thousand kilometers from the ground). When acquiring observation data, the artificial satellite 1 transmits the acquired observation data to a ground receiving station 2 on the earth. As an example of the present embodiment, in the following description, a case where the artificial satellite 1 is Himawari 8 will be explained, but the present invention is not limited to this, and other artificial satellites may be used.

The ground receiving station 2 is a communication facility that is installed at a predetermined location on the ground and receives observation data from the artificial satellite 1. The ground receiving station 2 includes at least a radio antenna and a radio communication device for wireless communication with the satellite 1, receives observation data intermittently transmitted from the satellite 1, and outputs the observation data to the fire detection device 3. .

The fire detection device 3 uses the observation data of the radiometer 100 mounted on the satellite 1 to detect fires in the fire detection area on the earth. The fire detection device 3 may be, for example, a data center or other information processing device. Although the fire detection device 3 acquires the observation data from the ground receiving station 2, it is not limited to this, and may acquire the observation data from a server of the Meteorological Agency, for example.

Here, as an example, the visible/infrared radiometer 100 of the artificial satellite 1 has a total of 16 bands, including 3 visible bands and 13 near-infrared/infrared bands. Among them, when the satellite 1 is Himawari 8, the fire detection device 3 uses band B1 with band number 1, band B3 with band number 3, band B6 with band number 6, band B7 with band number 7, band Observation data of band B14 of number 14 and band B15 of band number 15 are used.

The configuration of the fire detection device 3 according to the first embodiment will be described below. FIG. 2 is a schematic configuration diagram of the fire detection device 3 according to the first embodiment.

The fire detection device 3 includes a fire candidate position specifying unit 10 and a fire detection unit 20. These components are realized by executing a program (software) by a hardware processor such as a CPU (Central Processing Unit). In addition, some or all of these components are hardware such as LSI (Large Scale Integrated Circuit), ASIC (Application Specific Integrated Circuit), FPGA (Field-Programmable Gate Array), GPU (Graphics Processing Unit), etc. (including circuitry), or by cooperation of software and hardware. The program may be stored in advance in a storage device such as a HDD (Hard Disk Drive) or flash memory (a storage device with a non-transitory storage medium), or may be stored in a removable storage such as a DVD or CD-ROM. It may be stored in a medium (non-transitory storage medium) and installed in the storage device by loading the storage medium into the drive device. The storage device is configured by, for example, an HDD, flash memory, EEPROM (Electrically Erasable Programmable Read Only Memory), ROM (Read Only Memory), or RAM (Random Access Memory).

The fire candidate position identifying unit 10 identifies a position where a fire is suspected to occur (hereinafter referred to as a "fire candidate position") using the observation data of band B7 in the observation band. For example, the fire candidate position specifying unit 10 uses the observation data of the band B7 as brightness temperature information, and specifies the fire candidate position based on the brightness temperature. Then, the fire candidate position identification unit 10 transmits position information indicating the identified fire candidate position to the fire detection unit 20 .

The fire detection unit 20 includes an image generation unit 30, a mask unit 31, a fire detection area creation unit 32, an extraction unit 33, a processing unit 34, and a determination unit 35.

The image generation unit 30 generates a pseudo-color composite image of RGB colors based on the observation data. For example, the image generator 30 may use the technique disclosed in Non-Patent Document 1 to generate a pseudo-color composite image.

An example of a method for generating a pseudo-color composite image will be described below. The image generator 30 generates a pseudo-color composite image based on observation data of band B1, band B3, band B6, band B14, and band B15. Specifically, first, the image generation unit 30 utilizes the difference in reflectance for smoke aerosol between the band B1 and the band B3, and obtains the difference between the band B1 and the band B3 to calculate the smoke aerosol reflectance AE (Aerosol Enhancement : AE). This smoke aerosol reflectance AE is the reflectance with the ground cover reflectance suppressed. For example, in order to obtain the smoke aerosol reflectance AE more effectively, the image generation unit 30 calculates the smoke aerosol reflectance AE using the difference between twice the reflectance of the band B3 and the reflectance of the band B1. You may

Next, the image generator 30 obtains a smoke aerosol reflectance index (hereinafter referred to as "SARI": Smoke Aerosol Reflectance Index) based on the smoke aerosol reflectance AE. SARI is a value for image-enhancing the smoke aerosol reflectance AE with an exponential function so that differences in density can be identified, and is an index obtained by exponentially enhancing the smoke aerosol reflectance AE. This SARI indicates a correlation with smoke density. In this way, SARI is an index that emphasizes smoke aerosol reflectance AE, which is a value in which the reflectance of main coverings such as soil, vegetation, and cities is suppressed by taking the difference between band B3 and band B1. be.

The image generation unit 30 generates a pseudo-color composite image using the difference in brightness temperature between band B14 and band B15. The image generator 30 generates a water index (hereinafter referred to as “WI”) for distinguishing between fire smoke, clouds, and snow and ice. This WI is information about the amount of water in the atmosphere and on the ground. WI is calculated using the method described in Non-Patent Document 1, for example.

The image generator 30 generates a pseudo-color composite image based on SARI and WI. For example, the image generation unit 30 generates a pseudo-color composite image expressing smoke in red by assigning SARI to "R", reflectance of band B6 to "G", and WI to "B". Here, in the false-color composite image, band B6 (mid-infrared reflectance) shows no reflection in smoke and reflection in thick clouds. By assigning band B6 to "G" in this way, smoke can be represented in red and thick clouds can be represented in white. Thus, in the pseudo-color composite image, smoke is red, soil and forests are green, thick clouds are pink to white, and thin clouds, fog, snow and ice, and water surfaces are blue. This allows the identification of forest fire smoke with a pseudo-color composite image.

The fire detection area creating unit 32 creates a fire detection area ROI (Region Of Interest) based on the fire candidate positions from the fire candidate position specifying unit 10 . For example, the fire detection area ROI is a rectangular area having fire candidate positions. For example, the fire detection area ROI is a square area (for example, 25 [km]×25 [km]) centered on the fire candidate position.

The mask section 31 includes a first mask section 310 and a second mask section 320 .

The first mask unit 310 determines the presence or absence of clouds in each pixel data of the pseudo-color composite image based on WI and band B6. The mask unit 31 performs mask processing to exclude pixel data determined to have clouds in the pseudo-color composite image so as not to be detected. This is because the ground surface data is not reflected in the pseudo-color composite image if there are clouds.

The second mask unit 320 excludes pixel data corresponding to the sea area from the pseudo-color composite image so as not to be detected. Considering that the sea shows a high brightness temperature depending on the incident angle from the sun, and that the possibility of fires breaking out on the sea is low, it is excluded by masking so as not to be detected.

The extracting unit 33 extracts an image of the range of the fire detection area ROI created by the fire detection area creating unit 32 (hereinafter referred to as "extracted image") from the pseudo-color composite image.

The processing unit 34 decompresses pixel data for each pixel in the extraction image extracted by the extraction unit 33 . Here, the pixel data to be decompressed is either or both of SARI and WI data. Expansion is a process of expanding the distribution of pixel data for each pixel in the extracted image so that it is distributed over a wider range. In the following, the pixel data before being decompressed may be referred to as "first data".

For example, the processing unit 34 determines the range (a, b) to be stretched. In this embodiment, as an example, the processing unit 34 determines a as the "minimum value in the data" and b as the "maximum value in the data". Then, the processing unit 34 decompresses a to "0 level" and b to "255 level". For example, this stretching may be histogram stretching.

Next, the processing unit 34 normalizes the pixel data after decompressing the first data (hereinafter referred to as "second data"). Here, the second data is decompressed SARI and WI data, decompressed SARI and non-decompressed WI data, and non-decompressed SARI and decompressed WI data. data of For example, the processing unit 34 normalizes the second data by setting the average value of the second data for each pixel to "1". Pixel data after normalizing the second data is referred to as third data. However, it is not limited to this, and the processing unit 34 may omit the normalization processing.
When the processing unit 34 decompresses both SARI and WI data, only one of the decompressed SARI and WI may be normalized.

The determination unit 35 determines whether or not there is a fire within the fire detection area ROI based on the pixel data decompressed by the processing unit 34 . For example, the determination unit 35 may determine whether or not there is a fire in the fire detection area ROI based on the pixel data (i.e., the third data) decompressed and normalized by the processing unit 34, Whether or not there is a fire in the fire detection area ROI may be determined based on the decompressed pixel data (that is, the second data).

For example, the determination unit 35 sets a determination value for each pixel, and compares this determination value with a threshold (hereinafter referred to as "determination threshold") to determine the presence or absence of fire for each pixel. The determination threshold may be a single value or a value representing a predetermined range. Here, the determination value may be either or both of at least decompressed SARI and WI among the second data and the third data. That is, the determination unit 35 may use SARI and WI of the second data or the third data as determination values. In this case, the determination unit 35 compares the SARI with a first determination threshold (e.g., first range), and compares WI with a second determination threshold (e.g., second range). By performing processing for each pixel, the presence or absence of fire is determined for each pixel. However, the determination value is not limited to this, and may be a value calculated using SARI and WI of the second data or the third data. In this case, for example, since one determination value is set for each pixel, the determination unit 35 performs a process of comparing the determination value and the determination threshold value for each pixel, thereby determining whether or not there is a fire for each pixel. You may For example, when the determination value is within a predetermined range indicated by the determination threshold value, it may be determined as a fire, and when the determination value is outside the predetermined range, it may be determined as a fire. That is, the method of comparing the determination value and the determination threshold may vary depending on the value of the determination value and the setting method of the determination threshold, and can be arbitrarily adjusted by the user.

In addition, the determination unit 35 creates a composite image using the second data or the third data that has undergone decompression and normalization processing, and applies a determination threshold to this composite image to determine whether the fire has occurred. The presence or absence may be determined.

The operation flow of the fire detection device 3 according to the first embodiment will be described below with reference to FIG. FIG. 3 is a flow chart of the fire detection device 3 according to the first embodiment.

The fire detection device 3 acquires observation data from the satellite 1 (step S101). The fire detection device 3 calculates SARI and WI based on the observation data (step S102). Also, the fire detection device 3 creates a fire detection area ROI based on the observation data (step S103). The fire detection device 3 generates a pseudo-color composite image based on the observation data, SARI, and WI (step S104). The fire detection device 3 masks the pixel data corresponding to the clouds and the sea in the pseudo-color composite image (step S105). The fire detection device 3 extracts an image within the range of the fire detection area ROI from the masked pseudo-color composite image as an extraction image (step S106).

The fire detection device 3 decompresses either or both of the SARI and WI data for each pixel in the extracted extraction image (step S107). The fire detection device 3 then normalizes the decompressed data (step S108). The fire detection device 3 sets a judgment value from the expanded and normalized values (step S109), and judges whether or not there is a fire using the judgment value and a preset judgment threshold (step S110).

As described above, the fire detection device 3 according to the first embodiment includes the image generation unit 30, the extraction unit 33, the processing unit 34, and the determination unit 35. The image generator 30 generates a pseudo-color composite image based on observation data from the artificial satellite 1 . The processing unit 34 extracts an image of a range corresponding to the fire detection area from the pseudo-color composite image. The processing unit 34 decompresses one or both of the SARI and WI data for each pixel in the image extracted by the extraction unit 33 . The determination unit 35 determines whether or not there is a fire based on the data decompressed by the processing unit 34 .

With such a configuration, it is possible to distinguish between data indicating the occurrence of fire in the pseudo-color composite image and data not indicating the occurrence of fire even if there is an influence on the atmosphere, and fire detection is performed from the data of the pseudo-color composite image. In some cases, the accuracy of the fire detection can be improved.

<Second embodiment>
FIG. 4 is a diagram showing an example of a schematic configuration of a fire detection system B including a fire detection device 3B according to the second embodiment. A fire detection device 3B according to the second embodiment differs from the first embodiment in that it uses a learning model to determine the presence or absence of a fire. In the following description, portions having functions similar to those described in the first embodiment are given the same names and symbols, and specific descriptions of their configurations and functions may be omitted. .

A fire detection system B includes a satellite 1, a ground receiving station 2, and a fire detection device 3B.

FIG. 5 is a diagram showing an example of a schematic configuration of a fire detection device 3B according to the second embodiment. As shown in FIG. 5, the fire detection device 3B includes a fire candidate position identification unit 10, a learning model creation unit 40, and a fire detection unit 20B. These components are implemented by, for example, a hardware processor such as a CPU executing a program (software). Also, some or all of these components may be realized by hardware (circuitry) such as LSI, ASIC, FPGA, GPU, etc., or by cooperation of software and hardware may be The program may be stored in advance in a storage device such as a HDD (Hard Disk Drive) or flash memory (a storage device with a non-transitory storage medium), or may be stored in a removable storage such as a DVD or CD-ROM. It may be stored in a medium (non-transitory storage medium) and installed in the storage device by loading the storage medium into the drive device. The storage device is configured by, for example, an HDD, flash memory, EEPROM, ROM, or RAM.

The learning model creation unit 40 sets an appropriate judgment threshold by using machine learning. As an example, the learning model creation unit 40 includes a preprocessing unit 41 and a learning unit 42 .

The preprocessing unit 41 creates learning data for machine learning by the learning unit 42 . This learning data is calculated from the judgment value (hereinafter referred to as the "first past judgment value") calculated from the observation data of the location where the fire actually broke out and the observation data of the location where the fire did not actually break out. and the determined value (hereinafter referred to as "second past determination value"). For example, the preprocessing unit 41 acquires observation data from past fires from the Meteorological Agency server or the like. In addition, the observation data when there was a fire in the past is called "past observation data."

The preprocessing unit 41 generates a pseudo-color composite image using past observation data. Note that the method of generating the pseudo-color composite image by the preprocessing unit 41 is the same as the method of generating the pseudo-color composite image by the image generation unit 30, so the description is omitted. In the following description, the pseudo-color composite image generated by the preprocessing unit 41 is referred to as a “learning pseudo-color composite image” to distinguish it from the pseudo-color composite image generated by the image generation unit 30 .

The preprocessing unit 41 extracts an image of a predetermined area (hereinafter referred to as "learning extracted image") from the learning pseudo-color composite image. Here, the size, shape, etc. of the predetermined area are preferably the same as those of the fire detection area ROI. Also, the predetermined area is a fire area, which is an area including a place where a fire actually broke out.

The preprocessing unit 41 applies the same processing as the processing performed by the processing unit 34 to the pixel data of each pixel in the learning extraction image. The preprocessing unit 41 decompresses one or both of SARI and WI, which are pixel data for each pixel in the learning extraction image. Further, the preprocessing unit 41 normalizes the decompressed data out of the SARI and WI when the processing unit 34 executes normalization processing. Then, the preprocessing unit 41 obtains a determination value for each pixel from the pixel data after performing the same processing as the processing performed by the processing unit 34 . Here, the location of the fire is known. Therefore, based on the location where the fire occurred, the preprocessing unit 41 determines the determination value (first past determination value) of the pixel corresponding to the location where the fire occurred, among the determination values for each pixel, and the determination value of the pixel corresponding to the location where the fire occurred. A determination value (second past determination value) of a pixel corresponding to the location is determined. Then, the preprocessing unit 41 transmits the first past determination value labeled indicating that there is a fire and the second past determination value labeled indicating that there is no fire to the learning unit 42 as learning data. do.

The learning unit 42 uses the learning data generated by the preprocessing unit 41 to machine-learn a learning model in which the input is the judgment value and the output is the presence or absence of fire (for example, a support vector machine (VM), etc. supervised learning). For example, in this learning model, machine learning is performed using learning data to draw a boundary line H as shown in FIG. 6 that divides the first past judgment value and the second past judgment value. Then, in the learned model after learning, when an input judgment value (a judgment value that is not learning data) is input as input data, the input data indicates that there is a fire based on the boundary line H. Output whether it is a value or not. The learning unit 42 transmits the constructed learned model to the fire detection unit 20B. Note that this boundary line H corresponds to the determination threshold. For example, in the example shown in FIG. 6, x indicates that there is a fire, and black squares indicate that there is no fire.

The fire detection unit 20B includes an image generation unit 30, a mask unit 31, a fire detection area creation unit 32, an extraction unit 33, a processing unit 34, and a determination unit 35B.

The determination unit 35B has a learned model constructed by the learning unit 42 . The determination unit 35B determines whether or not there is a fire for each pixel by inputting the determination value set by the processing unit 34 into the learned model as input data. Further, the determination unit 35B may specify the location of the fire based on the pixels of the input data determined to have the fire.

The operation flow of the fire detection device 3B according to the second embodiment will be described below with reference to FIG. FIG. 7 is a flow chart of the fire detection device 3B according to the second embodiment of the fire detection device 3B.

The fire detection device 3B acquires observation data from the satellite 1 (step S201). The fire detection device 3B calculates SARI and WI based on the observation data (step S202). Also, the fire detection device 3B creates a fire detection area ROI based on the observation data (step S203). The fire detection device 3B generates a pseudo-color composite image based on the observation data, SARI, and WI (step S204). The fire detection device 3B masks the pixel data corresponding to the clouds and the sea in the pseudo-color composite image (step S205). The fire detection device 3B extracts an image of the range of the fire detection area ROI from the masked pseudo-color composite image as an extraction image (step S206).

The fire detection device 3B decompresses either or both of the SARI and WI data for each pixel in the extracted extraction image (step S207). The fire detection device 3B then normalizes the decompressed data (step S208). The fire detection device 3B sets a determination value from the decompressed and normalized values (step S209), and inputs the determination value to the learned model to determine whether or not there is a fire for each pixel (step S210). .

Note that in the second embodiment, the learning model creation unit 40 is included in the fire detection device 3B, but is not limited to this, and may be a device separate from the fire detection device 3B. That is, the fire detection device 3 should have at least a trained model, and a device different from the fire detection device 3 may create the trained model.

According to at least one embodiment described above, by decompressing the pixel data of each pixel in the pseudo-color composite image generated based on the observation data, the data indicating the occurrence of the fire in the pseudo-color composite image and the fire The distinction from the data that does not show the occurrence of fire becomes clear, and it is possible to suppress the deterioration of the accuracy of fire detection due to smoke. As a result, the accuracy of fire detection can be improved.

In addition, the trained model uses past judgment values generated based on decompressed pixel data as learning data to detect fires without depending on the atmospheric environment or land in the fire detection area. This eliminates the need to collect and learn learning data for each fire detection area. In other words, the boundary line H is a threshold that can be used all over the world.

Further, the

fire detection devices

3 and 3B of at least one embodiment described above are, for example, computers, and are configured to be executed by one or more processors and the one or more processors. and a memory storing one or more programs. One or more programs cause the

fire detection devices

3 and 3B to generate a pseudo-color composite image based on the observation data, and out of the pseudo-color composite image, an image (extracted image) of a range corresponding to the fire detection area. is extracted, pixel data for each pixel in the extracted image is decompressed, and the presence or absence of fire is determined based on the decompressed pixel data.
Among the components provided in the

fire detection device

3 or 3B, one or more components may be configured by one computer, that is, the

fire detection device

3 or 3B may be configured by a plurality of computers, or the

fire detection device

3 or 3B may be configured in one computer.

The

fire detection device

3 or 3B of the above embodiment includes the fire candidate position specifying unit 10, the mask unit 31, and the fire detection area creation unit 32, and these configurations are essential elements for the fire detection device 3 or 3B. is not. Instead of the fire candidate position specifying unit 10 and the fire detection area creating unit 32, the

fire detection device

3 or 3B may acquire the fire candidate positions and the fire detection area ROI from an external device. In the configuration of the present disclosure, the mask section 31 may be excluded from the

fire detection device

3 or 3B if mask processing is unnecessary.

Throughout the specification, when a part is referred to as "including", "having" or "comprising" an element, this does not mean excluding the other element, unless specifically stated to the contrary. It means that it can further include a component of

In addition, the term "... unit" described in the specification means a unit that processes at least one function or operation, which may be embodied as hardware or software, or a combination of hardware and software. It may be embodied in combination.

Although the embodiment of the present invention has been described in detail with reference to the drawings, the specific configuration is not limited to this embodiment, and includes design within the scope of the gist of the present invention.

The execution order of each process such as actions, procedures, steps, and stages in the devices, systems, programs, and methods shown in the claims, the specification, and the drawings is etc., and it should be noted that they can be implemented in any order unless the output of the previous process is used in the subsequent process. Regarding the operation flow in the claims, the specification, and the drawings, even if the description is made using "first," "next," etc. for convenience, it means that it is essential to carry out in this order. not a thing

A, B Fire detection system 1 Artificial satellite 2

Ground receiving station

3, 3B Fire detection device 10 Fire candidate position specifying unit 20 Fire detection unit 30 Image generation unit 31 Mask unit 32 Fire detection area creation unit 33 Extraction unit 34

Processing unit

35, 35B Determination unit 40 Learning model creation unit

Claims

A fire detection device that detects fires on the earth using observation data from a radiometer mounted on a satellite,
an image generation unit that generates a pseudo-color composite image based on the observation data;
an extraction unit for extracting an image of a range corresponding to a fire detection area from the pseudo-color composite image;
a processing unit for decompressing pixel data for each pixel in the image extracted by the extracting unit;
a determination unit that determines whether there is a fire based on the pixel data decompressed by the processing unit;
A fire detection device comprising:
The processing unit normalizes the decompressed pixel data,
The determination unit determines whether or not there is a fire based on the normalized pixel data.
The fire detection device according to claim 1.
The pixel data includes a smoke aerosol reflectance index, which is an index for reflectance to smoke, and a moisture index, which is an index for atmospheric and surface water content;
The processing unit decompresses one or both of the smoke aerosol reflectance index and the moisture index out of the pixel data for each pixel in the image extracted by the extraction unit.
The fire detection device according to claim 1.
The processing unit normalizes the expanded exponent of the smoke aerosol reflectance exponent and the moisture exponent.
The fire detection device according to claim 3.
a fire candidate position specifying unit that specifies a fire candidate position, which is a position where a fire is suspected to occur, based on information of an observation band indicating brightness temperature among the observation data;
a fire detection area creation unit that creates a predetermined area including the fire candidate position as a fire detection area;
The fire detection device according to any one of claims 1 to 4, comprising:
The determination unit has a learning model that receives pixel data or a value based on the pixel data as input data and outputs a determination result as to whether or not the input data indicates a fire, and the pixel data decompressed by the processing unit. or determining whether there is a fire by inputting a value based on the pixel data into the learning model;
The fire detection device according to any one of claims 1 to 5.
A fire detection method for a fire detection device that detects a fire on the earth using observation data from a radiometer mounted on a satellite, comprising:
generating a pseudo-color composite image based on the observed data;
a step of extracting an image of a range corresponding to a fire detection area from the pseudo-color composite image;
decompressing pixel data for each pixel in the extracted image;
determining whether there is a fire based on the decompressed pixel data;
fire detection methods including;