WO2021246130A1 - Gas leak location identification device, gas leak location identification system, gas leak location identification method, gas leak location estimation model generation device, gas leak location estimation model generation method, and program - Google Patents

Abstract

This gas leak location identification device comprises: an image input unit for receiving input of a gas distribution image including an image of a region of gas that has leaked into a space; and a gas leak location identification unit which, using an inference model that has performed machine learning using, as training data, a combination of a training gas distribution image including an image of a region of gas that has leaked into the air and a gas leak location indicating the location of the gas leak source in the gas region shown on the training gas distribution image, identifies a gas leak location indicating the location of the leak source in the gas region on the gas distribution image received by the image input unit.

Description

Gas leak position identification device, gas leak position identification system, gas leak position identification method, gas leak position estimation model generation device, gas leakage position estimation model generation method, and program

This disclosure relates to a device, a system, a method, and a program for identifying the position of a leak source in a method of detecting a gas leaked into a space using an image.

Gas plants, petrochemical plants, thermal power plants, steel-related facilities, etc. handle a large amount of gas during operation. In such facilities, the danger of gas leaks is recognized due to aging deterioration of the facilities and operational mistakes, and gas detection devices are installed to minimize gas leaks before a major accident occurs. In addition to gas detection devices that utilize the fact that the electrical characteristics of the probe change when gas molecules come into contact with the detection probe, in recent years, gas visualization imaging devices that utilize infrared absorption by gas have been introduced.

In a gas visualization imager, electromagnetic waves mainly in the infrared region called blackbody radiation emitted from a background object with an absolute temperature of 0 K or higher are absorbed by the gas, or blackbody radiation is generated from the gas itself. The presence of gas is detected by capturing changes in the amount of electromagnetic waves. By taking a picture of the monitored space with a gas visualization imager, gas leaks can be captured as images, so gas leaks can be detected earlier than with a detection probe type that can only monitor grid-like locations. However, the location of the gas can be accurately grasped.

For example, in the image processing apparatus for gas detection disclosed in Patent Document 1, the leak position is specified by dividing the image into blocks and determining whether or not each block is in the gas region over a plurality of frames. Further, for example, in the image processing apparatus for gas detection disclosed in Patent Document 2, from the first frequency component data showing the temperature change due to the leaked gas with respect to the infrared image obtained by capturing the monitored object at a plurality of times. However, the frequency is low, and the second frequency component data indicating the temperature change of the background to be monitored is removed from the image data indicating the infrared image.

International Publication No. 2016/143754 International Publication No. 2017/0743430

However, as in Patent Document 1, when the leak position is specified by a determined rule (algorithm), it becomes difficult to accurately specify the leak position if the preconditions on which the rule is based are broken. In general, equipment such as piping is installed in a place to be monitored, and the shape and arrangement are often complicated. In addition, in a monitoring device that constantly monitors equipment, there are many restrictions on the arrangement of the image pickup device, so it is often difficult to keep all possible gas leak sources within the field of view without obstruction. Therefore, a part of the leak source or the gas cloud may be shielded and a part of the image of the gas cloud in the detection image may be lost, making it difficult to identify the position of the leak source. Also, when the background temperature near the leak source and the gas temperature are close, the change in the infrared signal due to the gas is hardly detected near the leak source, which is substantially the same as when a part of the gas cloud is shielded. , It becomes difficult to identify the location of the leak source. In addition, even when the wind direction changes, it is often the case that the leak position is erroneously specified according to the determined rules.

In view of the above problems, an aspect of the present disclosure is to provide an identification device and an identification method for identifying a gas leak position indicating the position of a gas leak source regardless of the state of the gas leak.

In the gas leak position identification device according to one aspect of the present disclosure, the existence range of the gas leaked in the space is drawn as a gas region, and a gas distribution image in which a part of the gas region is shielded is used as an input. The teacher is a combination of the receiving image input unit, the teacher gas distribution image in which a part of the gas region is shielded, and the gas leak position indicating the position of the gas leak source in the gas region shown in the teacher gas distribution image. Using a machine-learned estimation model as data, it includes a gas leak position identification unit that identifies a gas leak position indicating the position of a leak source in the gas region of the gas distribution image received by the image input unit.

According to the above aspect, the gas leak position is identified based on a guessing model that can identify the gas leak position based on the characteristics of the gas distribution image. Therefore, by properly designing the teacher data, the gas leak position can be identified regardless of the gas leak state.

It is a functional block diagram of the gas detection system 100 which concerns on Embodiment 1. FIG. It is a schematic diagram which shows the relationship between a monitoring object 300 and an image generation part 10. It is a schematic diagram which shows the outline of the logical structure of a machine learning model. It is a schematic diagram which shows an example of the monitoring object 300, the captured image, and the gas distribution image which concerns on Embodiment 1. FIG. It is a flowchart which shows the operation of the gas detection apparatus 20 in a learning phase. This is an example of a gas distribution image as teacher data. This is an example of generating a gas distribution image as teacher data. This is an example of generating a gas distribution image as teacher data. It is a flowchart which shows the operation of the gas detection apparatus 20 in the operation phase. It is a schematic diagram which shows the positional relationship between a monitoring target 300 and a gas leak position. It is a functional block diagram of the gas detection system 101 which concerns on modification 1. FIG. It is an example of the leakage position image which concerns on the modification 1. It is an example of the teacher data which concerns on the modification 1. It is a schematic diagram which shows the positional relationship between a monitoring target 300 and a gas leak position. It is a functional block diagram of the gas detection system 110 which concerns on modification 2. FIG. It is a functional block diagram of the gas detection system 200 which concerns on Embodiment 2. FIG. It is a schematic diagram which shows the positional relationship between a monitoring target 300 and a gas leak position. It is a schematic diagram which shows the positional relationship between a monitoring target 300 and a gas leak position.

<< Embodiment 1 >>
Hereinafter, the gas detection system 100 according to the first embodiment will be described with reference to the drawings.

FIG. 1 is a functional block diagram of the gas detection system 100 according to the first embodiment. As shown in FIG. 1, the gas detection system 100 includes an image generation unit 10 for capturing an image of a monitored object, a gas detection device 20 for detecting gas based on an image acquired by the image generation unit 10, and a display unit 24. Have. The image generation unit 10 and the display unit 24 are configured to be connectable to the gas detection device 20, respectively.

<Image generation unit 10>
The image generation unit 10 is a device or system that captures an image of a monitored object and provides an image to the gas detection device 20. In the first embodiment, the image generation unit 10 is a so-called infrared camera that detects and images infrared light having a wavelength of 3.2 to 3.4 μm, and is a hydrocarbon such as methane, ethane, ethylene, and propylene. It is possible to detect system gas. The image generation unit 10 is not limited to this, and may be an image pickup device capable of detecting the gas to be monitored. For example, if the monitoring target is a gas that can be detected by visible light such as white smoked water vapor, it is general. It may be a visible light camera. In the present specification, the gas refers to a gas leaked from a closed space such as a pipe or a tank, and is not intentionally diffused into the atmosphere.

As shown in the schematic diagram of FIG. 2, the image generation unit 10 is installed so that the monitoring target 300 is included in the field of view range 310 of the image generation unit 10. The image generation unit 10 outputs the captured image as a video signal to the gas detection device 20. The video signal is, for example, a signal for transmitting an image of 30 frames per second.

<Structure of gas detection device 20>
The gas detection device 20 is a device that acquires an image of a monitored object from the image generation unit 10, detects a gas region based on the image, and notifies the user of the gas detection through the display unit 24. The gas detection device 20 is realized as, for example, a computer including a general CPU (Central Processing Unit), a RAM, and a program executed by these. As will be described later, the gas detection device 20 may further include a GPU (Graphics Processing Unit) and a RAM as arithmetic units. As shown in FIG. 1, the gas detection device 20 includes an image acquisition unit 201, a gas region extraction unit 211, a gas region image acquisition unit 212, a leak position information acquisition unit 213, a machine learning unit 2141, and a learning model holding unit 2142. The result output unit 215 is provided. The gas region extraction unit 211 has the function of the image input unit of the present disclosure. Further, the machine learning unit 2141 and the learning model holding unit 2142 constitute a gas leak position identification unit 214. The gas region extraction unit 211 and the gas leak position identification unit 214 constitute the gas leak position identification device 21.

The image acquisition unit 201 is a circuit that acquires an image of the monitored object from the image generation unit 10. In the embodiment, the image acquisition unit 201 acquires a video signal from the image generation unit 10, restores the video signal to an image, and outputs the video signal to the gas region extraction unit 211 as a moving image composed of a plurality of frames. The image is an infrared photograph of the monitored object, and has the intensity of infrared rays as a pixel value.

The gas region extraction unit 211 is a circuit that performs gas detection processing on the moving image output by the image acquisition unit 201 to generate a gas distribution image including the gas region. A known method can be used for the gas detection process. Specifically, for example, the method described in International Publication No. 2017/0743430 (Patent Document 2) can be used. Then, a gas distribution image as a moving image obtained by cutting out a region including a gas region from each frame of the moving image is generated. Specifically, as shown in the schematic bird's-eye view of FIG. 4A, when the gas 320 leaking from the equipment 300 is imaged by the image generation unit 10, the gas is as shown in the frame example of FIG. 4B. The gas region 420 corresponding to is imaged. Here, a part of the gas region 420 is obscured by the image 400 of the equipment and is not shown in the image. Therefore, as the gas distribution image, an image including a part 520 of the gas region is generated as shown in the schematic diagram of FIG. 4 (c). The gas region extraction unit 211 may perform processing such as gain adjustment after cutting out a region including the gas region from each frame of the moving image, and the pixel value of the moving image output by the image acquisition unit 201 may be used. Instead, the difference in pixel values may be mapped to obtain a gas distribution image.

If the size of the gas distribution image or the number of frames as a moving image is excessive, the amount of calculation for machine learning and judgment based on machine learning becomes large. In the first embodiment, the number of pixels of the gas distribution image is 224 × 224 pixels, and the number of frames is 16.

The gas region image acquisition unit 212 is a circuit for acquiring a gas distribution image having the same format as the gas distribution image generated by the gas region extraction unit 211 and having a known gas leakage position. Here, the gas distribution image is an image of an image of a gas cloud leaking from one leakage source, and there is only one gas leakage position, which is the position of the leakage source, with respect to the gas distribution image. If the acquired image is not in the same format as the gas distribution image generated by the gas region extraction unit 211, the gas region image acquisition unit 212 cuts out, enlarges, or reduces the image so as to have the same format. May be done. Further, for example, when the acquired image is three-dimensional voxel data, it may be converted into a two-dimensional image of a viewpoint from one point.

The leak position information acquisition unit 213 is a circuit that acquires the gas leakage position corresponding to the gas distribution image acquired by the gas region image acquisition unit 212. The gas leak position is designated as the coordinates in the gas distribution image acquired by the gas region image acquisition unit 212. If the acquired gas leak position is a coordinate in space, it is converted to the coordinate in the gas distribution image.

The machine learning unit 2141 executes machine learning based on the combination of the gas distribution image received by the gas region image acquisition unit 212 and the gas leakage position corresponding to the gas distribution image received by the leak position information acquisition unit 213. It is a circuit that generates a machine learning model. The machine learning model predicts the gas leak position indicating the position of the gas leak source based on the combination of the feature amount of the gas distribution image, for example, the outer peripheral shape of the gas region, the gas shading distribution, and the time change of these. Formed to do. As machine learning, for example, a convolutional neural network (CNN) can be used, and known software such as PyTorch can be used. FIG. 3 is a schematic diagram showing an outline of the logical configuration of the machine learning model. The machine learning model includes an input layer 51, an intermediate layer 52-1, an intermediate layer 52-2, ..., An intermediate layer 52-n, and an output layer 53, and the interlayer filter is optimized by learning. For example, when the number of pixels of the gas distribution image is 224 × 224 pixels and the number of frames is 16, the input layer 51 accepts a 224 × 224 × 16 three-dimensional tensor in which the pixel value of the gas distribution image is input. The intermediate layer 52-1 is, for example, a convolution layer, and receives a 224 × 224 × 16 three-dimensional tensor generated by a convolution operation from the data of the input layer 51. The intermediate layer 52-2 is, for example, a pooling layer, and accepts a three-dimensional tensor obtained by resizing the data of the intermediate layer 52-1. The intermediate layer 52-n is, for example, a fully connected layer, and the data of the intermediate layer 52- (n-1) is converted into a two-dimensional vector showing coordinate values. The configuration of the intermediate layer is an example, and the number n of the intermediate layers is about 3 to 5, but the number n is not limited to this. Further, although the number of neurons in each layer is drawn as the same in FIG. 3, each layer may have an arbitrary number of neurons. The machine learning unit 2141 receives a moving image as a gas distribution image as an input, performs learning with the gas leakage position as the correct answer, generates a machine learning model, and outputs it to the learning model holding unit 2142. The machine learning unit 2141 may be realized by the GPU and software when the gas detection device 20 includes a GPU and a RAM as arithmetic units.

The learning model holding unit 2142 holds the machine learning model generated by the machine learning unit 2141 and outputs the gas leakage position corresponding to the gas distribution image generated by the gas region extraction unit 211 using the machine learning model. Is. The gas leak position is specified and output as a coordinate value in the input gas distribution image.

The determination result output unit 215 is a circuit for generating an image for displaying on the display unit 24 by superimposing the gas leak position output by the learning model holding unit 2142 on the moving image acquired by the image acquisition unit 201. Further, the determination result output unit 215 may have a function of outputting information indicating a gas leak position. The information indicating the gas leak position is, for example, numerical information indicating a position in an image or a position in space.

<Other configurations>
The display unit 24 is a display device such as a liquid crystal display or an organic EL display.

<Operation>
Hereinafter, the operation of the gas detection device 20 in the present embodiment will be described with reference to the drawings.

<Learning phase>
FIG. 5 is a flowchart showing the operation of the gas detection device 20 in the learning phase.

First, as teacher data, a combination of the gas distribution image and the gas leak position is created (step S110). As the gas distribution image, an image in which the gas leak position is known can be used. FIG. 6A is a schematic diagram of a gas distribution image in which the gas leak position is known, and FIG. 6B is a combination of the gas distribution image and the coordinates of the gas leak position. As the training data, for example, a part of the gas region in the gas distribution image or an image in which the gas leak position is hidden (shielded) behind the equipment may be used, and similarly, a part of the gas region or The gas leak position can be detected with high accuracy for the gas distribution image in which the gas leak position is hidden behind the equipment. The teacher data is not limited to this, and a gas distribution image in which the gas region and the gas leak position are not shielded may be used, or the wind direction around the leak source is not constant (for example, a moving image after the gas leak occurs). It may be a gas distribution image (the wind direction has changed until the last frame acquisition of the image). With this configuration, it is possible to detect the gas leak position when the same phenomenon occurs in the moving image captured in the operation phase described later. FIG. 6C is a schematic diagram of a gas distribution image in which a part of the gas region is shielded, and FIG. 6D is a combination of the gas distribution image and the coordinates of the gas leak position. Further, FIG. 6 (e) is a schematic diagram of a gas distribution image in which the gas leak position and the surrounding gas region are shielded, and FIG. 6 (f) shows the gas distribution image and the coordinates of the gas leak position. It is a combination of.

The partially shielded image as teacher data is not limited to the case of acquiring from the actual equipment, and may be formed by processing or simulating the gas distribution image. FIG. 7 shows an example of generating a partially shielded image. 7 (a) is a combination of the base gas distribution image and the gas leak position, and FIGS. 7 (b) to 7 (f) are one of the gas regions in the gas distribution image of FIG. 7 (a). It is a combination of the gas distribution image that shields the part and the gas leak position. By using gas distribution images with different shapes and positions of the shield portion as teacher data, even if there is a shield such as a pillar or pipe between the gas leak position and the image generation unit 10, the gas can be accurately measured. It is possible to generate a machine learning model in which the leakage position can be detected. Further, the gas distribution image may be based on a simulation. For example, a gas leakage simulation is performed using a three-dimensional structural model of the equipment, and a viewpoint from a predetermined one point is obtained from the three-dimensional voxel data which is the simulation result. You may generate an image of. More specifically, for example, modeling is performed to lay out a structure in a three-dimensional space based on the three-dimensional voxel data of the equipment. Next, conditions such as the type of gas, flow rate, wind speed, wind direction, shape of gas leak source, diameter, and position are determined, and a three-dimensional fluid simulation is performed to calculate the gas distribution state. A viewpoint is set in space for the three-dimensional model formed in this way, and a two-dimensional image from the viewpoint is generated as a gas distribution image. Specifically, for one straight line passing through the viewpoint in space corresponding to one pixel of the gas distribution image, the gas concentration thickness product is calculated for the gas existing from the viewpoint to the surface of the equipment. The light intensity for each pixel is calculated with, and the pixel value is calculated. FIG. 8 (a) is a schematic diagram showing the positional relationship between the voxel data 301 showing equipment and the voxel data 321 showing gas, which are formed by fluid simulation, respectively, and are shown in FIGS. 8 (b) and 8 (c), respectively. Shows a gas distribution image and a gas leak position obtained from the viewpoint 11 with the direction d1 as the central axis, and a gas distribution image and a gas leak position obtained from the viewpoint 12 with the direction d2 as the central axis.

Next, the combination of the gas distribution image and the gas leak position is input to the gas detection device 20 (step S120). The gas distribution image is input to the gas region image acquisition unit 212, and the corresponding gas leak position is input to the leak position information acquisition unit 213.

Next, input data to the convolutional neural network and execute machine learning (step S130). As a result, the parameters are optimized by trial and error by deep learning, and a machine-learned model is formed. The formed machine-learned model is held in the learning model holding unit 2142.

By the above operation, a machine-learned model that outputs the gas leak position based on the feature amount of the gas distribution image is formed.

<Operation phase>
FIG. 9 is a flowchart showing the operation of the gas detection device 20 in the learning phase.

First, the gas region is detected from each frame of the captured image, and the gas distribution image including the gas region and its surroundings is cut out (step S210). The detection of the gas region is performed by a known method based on the time-series change of the brightness in the captured image, its frequency, and the like. Then, a part thereof is cut out from each frame of the captured image so as to include all the pixels in which gas is detected, and a gas distribution image is generated as a frame of the gas distribution image.

Next, the gas leak position is identified from the gas distribution image using the trained model (step S220). By using the machine-learned model formed by step S130, the coordinates of the gas leak position with respect to the gas distribution image are identified. Since the learning model holding unit 2142 outputs the gas leak position as a coordinate value with respect to the gas distribution image, the gas leak position identification unit 214 has a correspondence relationship between the coordinates of the initial captured image and the gas distribution image from the gas region extraction unit 211. Based on, the gas leak position is output as a coordinate value with respect to the initially captured image. As shown in the schematic diagram of FIG. 10A, the gas leak position 330 of the gas 320 is identified as the gas leak position 430 corresponding to the gas region 420 in the captured image, and is a linear (or conical) leak suspected region. Identified as being present within 330.

Next, the identified gas leak position is superimposed on the initial captured image, and the gas leak position is displayed (step S230). By this treatment, the leak position of the gas can be specified.

When identifying the gas leak position, if there is 3D voxel data indicating the configuration of the equipment, the range may be further limited by using this. As mentioned above, the gas leak location can be identified as being within the spatially linear (or conical) region 330. Here, when the three-dimensional voxel data of the equipment 300 exists, as shown in the schematic diagram of FIG. 10B, a linear (or conical) leakage suspected region 330 is formed on the three-dimensional voxel data. By superimposing, the positional relationship between the equipment and the gas leakage position becomes clearer. Further, when the equipment is present in or near the suspected leakage area 330, if gas is detected at the position corresponding to the equipment in the gas distribution image, it is highly possible that the equipment position is the gas leakage position. On the other hand, if gas is not detected at the position corresponding to the equipment in the gas distribution image, it is judged that there is a high possibility that the gas leakage position exists behind the position hidden by the equipment position. be able to.

<Summary>
With the above configuration, in a gas detection device that detects gas using an image, if the gas region is imaged, the gas leak position can be identified. In particular, when the teacher data is used sufficiently and appropriately, an image of a part of the gas region and the gas leak position being shielded by a structure such as a pillar or a pipe, or a situation where the wind direction is not constant. Even when it is not easy to identify the position of the leak source only from the shape of the gas region, such as the image captured by the above, the leak position of the gas can be identified. Therefore, gas leaks do not depend on the imaging conditions, including cases where the camera, which is the image generator, cannot be installed so that the suspected leak location can be seen without a shield, or where it is easily affected by wind. The location can be identified.

<< Modification 1 >>
In the first embodiment, the gas leak position corresponding to the gas distribution image is a coordinate value. However, the gas leak position may be a probability distribution image showing the probability of being a gas leak position.

FIG. 11 is a functional block diagram of the gas detection system 101 including the gas detection device 26 according to the first modification. As shown in FIG. 11, the gas detection device 26 replaces the leak position information acquisition unit 213, the machine learning unit 2141, and the learning model holding unit 2142 with the leak position image acquisition unit 223, the machine learning unit 2241, and the learning model holding unit. 2242 is provided. The machine learning unit 2241 and the learning model holding unit 2242 constitute a gas leak position identification unit 224, and the gas region extraction unit 211 and the gas leak position identification unit 224 constitute a gas leak position identification device 22.

The leak position image acquisition unit 223 is a circuit that acquires a gas leak position image corresponding to the gas distribution image acquired by the gas region image acquisition unit 212. As shown in FIGS. 12A and 12B, the gas leak position image is an image in which the probability of the gas leak position is mapped to each coordinate in the gas distribution image acquired by the gas region image acquisition unit 212. Is. For example, if the gas leak position can be specified in one pixel or a narrow range, the image is as shown in FIG. 12 (a), and if the gas leak position is limited to a certain range, the image is as shown in FIG. 12 (b). It becomes an image. More specifically, the probability p of the gas leak position is as follows with respect to the coordinates where the distance from the center coordinate of the region where the gas leak position can exist (hereinafter referred to as "gas leak region") is r. Calculate with the formula and map p.

Here, α and σ are values determined according to the accuracy of the gas leak position.

The calculation of p is not limited to the above formula, and the following formula may be used.

Therefore, for example, when there is a gas distribution image with a known gas leak position shown in FIG. 13 (a), the gas distribution image shown in FIG. 13 (b) and the gas leak position image shown in FIG. 13 (c) are used. Pairs are entered as teacher data.

The machine learning unit 2241 executes machine learning based on the combination of the gas distribution image received from the gas region image acquisition unit 212 and the gas leak position image received from the leak position image acquisition unit 223, and generates a machine learning model. It is a circuit to do. As machine learning, for example, a convolutional neural network (CNN) can be used. The trained model generated by the machine learning unit 2241 is a model in which the output of the output layer is a gas leak position image. That is, for example, when the gas distribution image is a moving image of 224 × 224 pixels and 16 frames, the input layer accepts a 224 × 224 × 16 three-dimensional tensor, and the output layer receives the gas leak position for each pixel of the gas distribution image. Outputs a 224 × 224 matrix storing the probabilities of. The gas leak position image is not limited to this case. For example, the leak position image acquisition unit 223 accepts a pair of the center coordinate of the gas leak region, the value of α, and the value of σ and outputs the gas leak position image. May be good. The machine learning unit 2241 receives a gas distribution image as an input, performs learning with the gas leak image as the correct answer, generates a machine learning model, and outputs it to the learning model holding unit 2242.

The learning model holding unit 2242 holds the machine learning model generated by the machine learning unit 2241, and outputs a gas leakage position image corresponding to the gas distribution image generated by the gas region extraction unit 211 using the machine learning model. It is a circuit.

In the first modification, in the learning mode, the gas distribution image is input and the learning is performed with the gas leak position image as the correct answer to generate a machine learning model. Then, in the operation mode, the generated machine learning model is used to output a gas leak position image based on the gas distribution image.

With the above configuration, when identifying the gas leak position from the gas distribution image using the trained model, an image that maps the probability of the gas leak position for each coordinate of the gas distribution image can be obtained. Therefore, as shown in the schematic diagram of FIG. 14 (a), the gas leak position is specified as the gas leak region 450 in the captured image, and exists in the cylindrical (or conical) leak suspected region 350 in the space. Identified as. Here, it is shown that when the gas leak region 450 is obtained as the gas leak regions 431, 432, and 433 having different probabilities of including the gas leak region, each corresponds to the leak suspected regions 351, 352, and 353 in the space. ing.

When identifying the gas leak position, if there is 3D voxel data indicating the configuration of the equipment, the range may be further limited by using this. When the three-dimensional voxel data of the equipment 300 exists, as shown in the schematic diagram of FIG. 14 (b), the cylindrical (or conical) leak suspected region 350 is superimposed on the three-dimensional voxel data. As a result, the positional relationship between the equipment and the gas leakage position becomes clearer. Further, when the equipment is present in or near the suspected leakage area 350, if gas is detected at the position corresponding to the equipment in the gas distribution image, it is highly possible that the equipment position 355 is the gas leakage position. On the other hand, if gas is not detected at the position corresponding to the equipment in the gas distribution image, there is a high possibility that the gas leakage position exists behind the equipment hidden by the equipment position 355. You can judge.

<Summary>
With the above configuration, in the gas detection device that detects gas using an image, if the gas region is imaged, the range of the gas leak position and its probability can be identified. In addition, since the gas leakage position is indicated by its existence range and existence probability, it is possible to prevent a decrease in the identification accuracy of the leakage position due to lack of teacher data or overfitting of biased teacher data, and easily improve the identification accuracy. Is possible. Specifically, with respect to the first embodiment, the standard deviation of the difference between the pixel on the gas distribution image corresponding to the gas leak position and the identified gas leak position is as follows.

On the other hand, with respect to the modified example 1, the standard deviation of the difference between the pixel on the gas distribution image corresponding to the gas leak position and the pixel having the highest probability value in the identified gas leak position image is as follows. Became.

<< Modification 2 >>
In the first embodiment and the first modification, one gas detection device is used to identify the gas leakage position in the operation mode using the machine learning model generated in the learning mode. However, machine learning and gas leak location identification need not be performed on the same hardware and may be performed using different hardware.

FIG. 15 is a functional block diagram of the gas detection system 110 according to the second modification. As shown in FIG. 15, the gas detection system 110 includes an image generation unit 10 for capturing an image of a monitored object, a gas detection device 27 for detecting gas based on an image acquired by the image generation unit 10, and a learning data creation device. It has 30 and a display unit 24. The image generation unit 10, the display unit 24, and the learning data creation device 30 are configured to be connectable to the gas detection device 27, respectively.

The gas detection device 27 is a device that acquires an image of the monitored object captured from the image generation unit 10, detects the gas region based on the image, and notifies the user of the gas detection through the display unit 24. The gas detection device 27 is realized as, for example, a computer including a general CPU and RAM, and a program executed by these. The gas detection device 27 includes an image acquisition unit 201, a gas region extraction unit 211, a learning model holding unit 2142, and a determination result output unit 215. The learning model holding unit 2142 constitutes a gas leak position identification unit 234. Further, the gas region extraction unit 211 and the gas leak position identification unit 234 constitute the gas leak position identification device 23. The learning data creation device 30 is realized as, for example, a computer including a general CPU, a GPU, a RAM, and a program executed by these. The learning data creating device 30 includes a gas region image acquisition unit 212, a leak position information acquisition unit 213, and a machine learning unit 2141.

The gas detection device 27 carries out only the operation mode operation of the gas detection device 20 according to the first embodiment. Further, the learning data creating device 30 carries out only the learning mode operation of the gas detecting device 20 according to the first embodiment. The gas detection device 27 and the learning data creation device 30 are connected by, for example, a LAN, and the learned model formed by the learning data creation device 30 is stored in the learning model holding unit 2142 of the gas detection device 272. The storage of the trained model in the training model holding unit 2142 is not limited to duplication by the network, and may be performed using, for example, a removable medium, an optical disk, a ROM, or the like.

<Summary>
With the above configuration, in a gas detection device that detects gas using an image, if the gas region is imaged, the gas leak position can be identified. Also, since machine learning and gas leak location identification are performed on different hardware, the gas detector does not need to have resources for machine learning. Therefore, the gas detection device only needs to have the resources for operating the trained model, and can be realized by using a simple device such as a notebook computer, a smartphone, or a tablet. Further, since it is possible to operate the trained model constructed by machine learning using one learning data creation device 30 with a plurality of gas detection devices 27, it is possible to easily manufacture the gas detection device 27. It will be possible.

<< Embodiment 2 >>
In the first embodiment and each modification, a case where the gas leak position is two-dimensionally identified by using one image generation unit has been described. However, the gas leak position may be identified three-dimensionally by using two image generation units having different viewpoints.

The gas detection system 200 according to the second embodiment is characterized in that the gas leak position is three-dimensionally specified by using two gas leak position identification devices.

FIG. 16 is a functional block diagram showing the configuration of the gas detection system 200 according to the second embodiment. As shown in FIG. 16, the gas detection system 200 includes a first image generation unit 10-1 for imaging a monitored object, a second image generation unit 10-2 for imaging a monitored object, and a gas. It has a leak position identification system 25 and a display unit 24. The image generation units 10-1, 10-2, and the display unit 24 are configured to be connectable to the gas leak position identification system 25, respectively.

The gas leak position identification system 25 acquires an image of the monitored object captured from the image generation units 10-1 and 10-2, detects the gas region based on the image, and notifies the user of the gas detection through the display unit 24. It is a device to do. The gas leak position identification system 25 includes image acquisition units 211-1 and 201-2, gas region extraction units 211-1 and 211-2, learning model holding units 2142-1 and 2142-2, and visual field information holding units 226. The result output unit 227 is provided. The gas leak position identification unit 234-1 including the gas region extraction unit 211-1 and the learning model holding unit 2142-1 constitutes the gas leak position identification device 23-1. Similarly, the gas leak position identification unit 234-2 including the gas region extraction unit 211-2 and the learning model holding unit 2142-2 constitutes the gas leak position identification device 23-2. Although not shown in FIG. 12, in the second embodiment, the gas detection system 200 includes the learning data creation device 30 as in the modified example 2, and the learning model holding units 2142-1 and 2142-2 are each provided. , Holds the trained model generated by the training data creation device 30. The gas leak position identification system 25 is realized as, for example, a computer including a general CPU and RAM, and a program executed by these. Of the gas leak position identification system 25, each of the gas leak position identification device 23-1 and the gas leak position identification device 23-2 may be realized as a single computer independently of other configurations. At this time, the image acquisition unit 211-1 and the gas leak position identification device 23-1 are realized by a single computer, and the image acquisition unit 201-2 and the gas leak position identification device 23-2 are realized by another computer. Then, the remaining field information holding unit 226 and the determination result output unit 227 may be realized by another computer.

The image acquisition unit 211-1, the gas region extraction unit 211-1, and the learning model holding unit 2142-1 are based on the image captured by the image generation unit 10-1 as in the first embodiment and the first modification. The leak position is identified as the coordinate value of the image captured by the image generation unit 10-1. Similarly, the image acquisition unit 201-2, the gas region extraction unit 211-2, and the learning model holding unit 2142-2 determine the gas leak position based on the image captured by the image generation unit 10-2 in the image generation unit 10-. It is identified as the coordinate value of the image captured by 2.

The field of view information holding unit 226 holds the three-dimensional voxel data of the equipment to be monitored, and the installation position coordinates and the field of view orientation information of each of the image generation units 10-1 and 10-2. Further, the visual field information holding unit 226 may have three-dimensional voxel data of the equipment to be monitored.

The determination result output unit 227 is a learning model holding unit 2142-1 and a learning model based on the installation position coordinates and the viewing direction orientation information of each of the image generation units 10-1 and 10-2 held by the field information holding unit 226. The gas leakage position identified by each of the holding units 2142-2 is calculated as a relative position from the installation position coordinates of the image generation units 10-1 and 10-2. Specifically, it will be described with reference to the schematic diagram of FIG. 17 (a). For example, if the learning model holding unit 2142-1 identifies the gas leak position 436 for the image acquired from the image generation unit 10-1, it can be estimated that the gas leak position exists in the suspected leak region 346. Similarly, if the learning model holding unit 2142-2 identifies the gas leak position 437 for the image acquired by the image generation unit 10-2, it can be estimated that the gas leak position exists in the suspected leak region 347. Therefore, it can be estimated that the gas leak position exists in the region 330 included in both the region 346 and the region 347. As shown in the schematic diagram of FIG. 17B, when the region included in both the region 346 and the region 347 does not exist, it is presumed that the region 346 and the region 347 exist in the region 311 near the point where the region 346 and the region 347 are in close contact with each other. You may.

When identifying the gas leak position, if there is 3D voxel data indicating the configuration of the equipment, the range may be further limited by using this. When the 3D voxel data of the equipment 300 exists, by superimposing the two suspected leak areas on the 3D voxel data, the positional relationship between the equipment and the two suspected leak areas becomes clearer, and the estimated gas leakage occurs. The positional relationship between the position and the equipment becomes clearer. Further, as described in the first embodiment, which of the equipment and the leak position is closer to the image generation unit 10 can be determined by whether or not gas is detected at the gas leak position, so that the estimated gas leak position is used. If the gas leak position and the position of the equipment and the image generation unit are inconsistent, it can be determined that the accuracy of gas leak position identification by at least one gas leak position identification device is low.

<Summary>
With the above configuration, in a gas detection system that detects gas using an image, if the gas region is imaged, the gas leak position can be identified. Further, when one gas leak can be imaged at the same time by two image acquisition units, it is possible to spatially specify the gas leak position.

<< Modification 3 >>
In the second embodiment, as in the first embodiment, each gas leak position identification device outputs the gas leak position as a coordinate value, but as in the first modification, each gas leak position identification device is at the gas leak position. A certain probability may be output.

Since the operation of the gas leak position identification device is the same as that of the modified example 1, the final judgment result output will be described below. FIG. 18A is a schematic diagram of the determination result output process. For example, if the learning model holding unit 2142-1 identifies the gas leak position 438 with respect to the image acquired from the image generation unit 10-1, it can be estimated that the gas leak position exists in the suspected leak region 348. Similarly, if the learning model holding unit 2142-2 identifies the gas leakage position 439 for the image acquired by the image generation unit 10-2, it can be estimated that the gas leakage position exists in the suspected leakage area 349. Therefore, it can be estimated that the gas leak position exists in the region 332 included in both the region 348 and the region 349.

When identifying the gas leak position, if there is 3D voxel data indicating the configuration of the equipment, the range may be further limited by using this. When the 3D voxel data of the equipment 300 exists, as shown in the schematic diagram of FIG. 18B, the equipment and the two suspected leaks are suspected by superimposing the two suspected leak areas on the 3D voxel data. The positional relationship with the area becomes clearer, and the positional relationship between the estimated gas leakage position and the equipment becomes clearer. Further, as described in the first embodiment, which of the equipment and the leak position is closer to the image generation unit 10 can be determined by whether or not gas is detected at the gas leak position, so that the estimated gas leak position is used. If the gas leak position and the position of the equipment and the image generation unit are inconsistent, it can be determined that the accuracy of gas leak position identification by at least one gas leak position identification device is low.

<Summary>
With the above configuration, in a gas detection system that detects gas using an image, if the gas region is imaged, the gas leak position can be identified. Further, when one gas leak can be imaged at the same time by two image acquisition units, it is possible to spatially specify the gas leak position. Furthermore, since the gas leakage position is indicated by its existence range and existence probability, it is possible to prevent a decrease in the identification accuracy of the leakage position due to lack of teacher data or overfitting of biased teacher data, and easily improve the identification accuracy. Is possible.

<< Other variants of the embodiment >>
(1) In the modified example 1 and the modified example 3, the gas leak position image in which the probability of the gas leak position is mapped to each coordinate in the gas distribution image is used as the information indicating the gas leak position. However, in the first and third modifications, by using the image including the data only for the coordinates corresponding to the gas leak position in the gas distribution image as the information indicating the gas leak position, the gas leak position coordinates are substantially used. May be used.

(2) In the modified example 1 and the modified example 3, the gas leak position image as the teacher data is used, in which the probability of the gas leak position decreases as the distance from the reference coordinate 1 increases. However, the gas leak position image is not limited to this, and any distribution may be used. Further, for a gas distribution image showing gas leaked from two or more leak sources at the same time, an image having the maximum probability at each coordinate corresponding to the leak source may be used. By doing so, when gas leaks from a plurality of adjacent leak sources at the same time to form one gas cloud, there is a possibility that the number of leak sources is 2 or more based on the probability distribution which is the output. Can be shown to exist.

(3) In the second embodiment and the third modification, the gas leak position identification device uses the trained model formed by the machine learning model generation device as in the second modification, but the first embodiment and the first modification 1 are used. Similarly, machine learning and operation may be performed using the same device. On the contrary, in the first modification, the machine learning and the operation may be performed by using individual devices as in the second modification.

(4) In each embodiment and modification, the captured image is an infrared image having a wavelength of 3.2 to 3.4 μm, but the image is not limited to this, as long as the presence of gas to be detected can be confirmed. Any image such as an infrared image, a visible image, and an ultraviolet image in another wavelength range may be used. Further, the method for detecting the gas region is not limited to the above-mentioned one, and may be any processing capable of detecting the gas region.

(5) Although the present invention has been described based on the above embodiment, the present invention is not limited to the above embodiment, and the following cases are also included in the present invention.

For example, in the present invention, the gas leakage position identification device may be a device using an FPGA (Field Programmable Gate Array) or an ASIC (Application Specific Integrated Circuit) as a processor.

Further, some or all of the components constituting each of the above devices may be composed of one system LSI (Large Scale Integration). A system LSI is a super-multifunctional LSI manufactured by integrating a plurality of components on one chip, and specifically, is a computer system including a microprocessor, ROM, RAM, and the like. .. These may be individually integrated into one chip, or may be integrated into one chip so as to include a part or all of them. The LSI may be referred to as an IC, a system LSI, a super LSI, or an ultra LSI depending on the degree of integration. A computer program that achieves the same operation as each of the above devices is stored in the RAM. When the microprocessor operates according to the computer program, the system LSI achieves its function. For example, the present invention also includes a case where the accessibility method of the present invention is stored as an LSI program, and the LSI is inserted into a computer to execute a predetermined program.

The method of making an integrated circuit is not limited to LSI, and may be realized by a dedicated circuit or a general-purpose processor. After manufacturing the LSI, an FPGA that can be programmed or a reconfigurable processor that can reconfigure the connection and settings of the circuit cells inside the LSI may be used.

Furthermore, if an integrated circuit technology that replaces an LSI appears due to advances in semiconductor technology or another technology derived from it, it is naturally possible to integrate functional blocks using that technology.

In addition, the division of functional blocks in the block diagram is an example, and multiple functional blocks can be realized as one functional block, one functional block can be divided into multiple, and some functions can be transferred to other functional blocks. You may. Further, the functions of a plurality of functional blocks having similar functions may be processed by a single hardware or software in parallel or in a time division manner.

Further, the order in which the above steps are executed is for exemplifying in order to specifically explain the present invention, and may be an order other than the above. Further, a part of the above steps may be executed simultaneously with other steps (parallel).

Further, at least a part of the functions of the gas leak position identification device according to each embodiment and the modification thereof may be combined. Further, the numbers used above are all exemplified for the purpose of specifically explaining the present invention, and the present invention is not limited to the exemplified numbers.

Further, the present invention also includes various modifications in which modifications within the range that can be conceived by those skilled in the art are made to the present embodiment.

≪Summary≫
(1) The gas leak position identification device according to one aspect of the present disclosure has an image input unit that accepts a gas distribution image in which an image of a gas region leaked in space is included in the image as an input, and a gas leaked into the air. Machine-learned estimation using the combination of the teacher gas distribution image in which the image of is included in the image and the gas leak position indicating the position of the gas leak source in the gas region shown in the teacher gas distribution image as teacher data. Using a model, the gas leak position identification unit for identifying the gas leak position indicating the position of the leak source in the gas region of the gas distribution image received by the image input unit is provided.

Further, in the gas leak identification method according to one aspect of the present disclosure, an image of a gas region leaked into space is received as an input, and an image of a gas region leaked into air is included in the image. Using a machine-learned estimation model using the combination of the teacher gas distribution image included in the teacher gas distribution image and the gas leak position indicating the position of the gas leak source in the gas region shown in the teacher gas distribution image as teacher data, The gas leak position indicating the position of the leak source in the gas region of the received gas distribution image is identified.

Further, the program according to one aspect of the present disclosure is a program for causing a computer to perform a gas leak position identification process, and the gas leak position identification process includes an image of a gas region leaked in space in an image. A gas that accepts a gas distribution image as an input and shows the position of the gas leakage source in the gas region shown in the teacher gas distribution image in which the image of the gas leaked into the air is included in the image and the gas distribution image. Using a machine-learned estimation model using the combination with the leak position as teacher data, the gas leak position indicating the position of the leak source in the gas region of the received gas distribution image is identified.

According to the gas leak position identification device, the gas leak position identification method, and the program according to one aspect of the present disclosure, the gas leak position is identified based on an estimation model that can identify the gas leak position based on the characteristics of the gas distribution image. Therefore, by properly designing the teacher data, the gas leak position can be identified regardless of the gas leak state.

(2) In the gas leak device identification device according to one aspect of the present disclosure, the gas distribution image input to the image input unit includes a gas distribution image in which a part of the gas region is shielded, and the teacher gas. The distribution image may include a gas distribution image in which a part of the gas region is shielded.

With the above configuration, it is possible to identify the gas leak position even for a gas distribution image in which a part of the gas region is shielded by an obstacle.

(3) In the gas leak device identification device according to one aspect of the present disclosure, the gas distribution image received by the image input unit and the teacher gas distribution image included in the teacher data are both moving images including a plurality of frames. It is an image, and the estimation model is formed with a change in the gas region between frames of the gas distribution image as one of the feature quantities, and the gas leak position identification unit changes the gas region between the frames of the gas distribution image. The gas leak position may be identified by using it as one of the feature quantities.

With the above configuration, the time-series change of the feature amount in the gas region of the gas distribution image can be further used as the feature amount, so that the output of the guess model becomes more accurate.

(4) In the gas leak device identification device according to one aspect of the present disclosure, the estimation model uses the coordinate values in the teacher gas distribution image as the gas leak position in the teacher data, and the gas leak position identification unit uses gas. As the leak position, the coordinate value in the gas distribution image received by the image input unit may be output.

With the above configuration, since the output of the guess model is a coordinate value, the amount of calculation can be reduced and the output process of the gas leak position becomes simple.

(5) In the gas leak device identification device according to one aspect of the present disclosure, the estimation model uses a probability distribution in which each pixel of the teacher gas distribution image indicates a gas leak source as a gas leak position in the teacher data. The gas leak position identification unit may output a probability distribution indicating the probability that each pixel of the gas distribution image received by the image input unit is a gas leak source as the gas leak position.

With the above configuration, the deviation between the spatial position indicated by the gas leak position and the actual leak source can be reduced, and the influence of overfitting can be suppressed even when the teacher data is biased or insufficient.

(6) The gas leak device identification system according to one aspect of the present disclosure includes a first image acquisition unit that images a gas facility from a first viewpoint and generates a gas distribution image from the obtained image, and the first image acquisition unit. A second image acquisition unit that images the gas equipment from a second viewpoint different from the viewpoint of the above and generates a gas distribution image from the obtained image, and a gas leak position identification device according to one aspect of the present disclosure. A first gas leak position identification device and a second gas leak position identification device are provided, and the first image acquisition unit and the second image acquisition unit include a common spatial region in their respective imaging ranges. It may be arranged.

With the above configuration, it is possible to monitor a common spatial region from two different viewpoints, and when a gas leak occurs in the common spatial region, the position of the leak source can be identified from each of the different viewpoints. , The location of the leak source can be limited to a narrow spatial range.

(7) The gas leak device identification system according to one aspect of the present disclosure further includes a three-dimensional leak position identification unit, and the first gas leak position identification device acquires a gas distribution image from the first image acquisition unit. The corresponding gas leakage position is identified, and the second gas leakage position identification device acquires a gas distribution image from the second image acquisition unit to identify the corresponding gas leakage position, and the three-dimensional leakage position. The identification unit identifies the first spatial region indicated by the gas leakage position identified by the first gas leakage position identification device based on the imaging range of the first image acquisition unit, and the identification unit of the second image acquisition unit. The second spatial region indicated by the gas leak position identified by the second gas leak position identification device is specified based on the imaging range, and is included in both the first spatial region and the second spatial region. The spatial region may be identified as the leak source.

With the above configuration, it is possible to monitor a common space area from two different viewpoints, and when a gas leak occurs in the common space area, the position of the leak source can be identified three-dimensionally.

(8) The gas leak device identification system according to one aspect of the present disclosure includes a first image acquisition unit that captures a space in which gas leaks from a first viewpoint and generates a gas distribution image from the obtained image. A second image acquisition unit that captures the space from a second viewpoint different from the first viewpoint and generates a gas distribution image from the obtained image, and the gas distribution image from the first image acquisition unit. A first gas leakage position identification unit that acquires and identifies the corresponding gas leakage position, and a second gas leakage position identification that acquires the gas distribution image from the second image acquisition unit and identifies the corresponding gas leakage position. Based on the imaging range of the unit and the first image acquisition unit, the first spatial region indicated by the gas leakage position identified by the first gas leakage position identification device is specified, and the second image acquisition unit is imaged. The second space region indicated by the gas leak position identified by the second gas leak position identification device is specified based on the range, and the space included in both the first space region and the second space region. It is provided with a three-dimensional leakage position identification unit that identifies the region as the leakage source of the gas.

According to the gas leak device identification system according to one aspect of the present disclosure, it is possible to monitor a common space area from two different viewpoints, and when a gas leak occurs in the common space area, each of the different viewpoints. Since the position of the leak source can be identified from, the position of the leak source can be limited to a narrow spatial range.

(9) The gas leak device identification system according to one aspect of the present disclosure maintains the relationship between the three-dimensional structural information of the equipment in the space, the three-dimensional structural information, the first viewpoint, and the second viewpoint. The equipment information holding unit may be further provided, and the three-dimensional leakage position identification unit may further identify the relative positional relationship between the gas leakage position and the equipment.

With the above configuration, it is possible to identify which equipment is the source of the gas leak, so it is possible to promptly deal with the equipment.

(10) The gas leak position estimation model generation device according to one aspect of the present disclosure has an image input unit that accepts a gas distribution image in which an image of a gas region leaked in space is included in the image as an input, and the gas distribution image. Teacher data on the combination of the gas leak position input unit that accepts the gas leak position indicating the position of the gas leak source in the gas region shown in 1 as an input, the gas distribution image, and the gas leak position corresponding to the gas distribution image. It is provided with a machine learning unit that performs machine learning as an input and generates an estimation model that outputs a gas leak position indicating a leak source in a gas region by inputting a gas distribution image.

Further, the gas leak position estimation model generation method according to one aspect of the present disclosure accepts a gas distribution image in which an image of a gas region leaked in space is included in the image as an input, and the gas region shown in the gas distribution image. Accepts the gas leak position indicating the position of the gas leak source as input, machine-learns the combination of the gas distribution image and the gas leak position corresponding to the gas distribution image as teacher data, and inputs the gas distribution image. Generate an inference model that outputs the gas leak location indicating the leak source in the gas region.

Further, the program according to one aspect of the present disclosure is a program for causing a computer to perform a gas leak position estimation model generation process, and in the gas leak position estimation model generation process, an image of a gas region leaked into the space is an image. Accepts the gas distribution image contained in the gas as an input, accepts the gas leak position indicating the position of the gas leak source in the gas region shown in the gas distribution image as an input, and corresponds to the gas distribution image and the gas distribution image. Machine learning is performed using the combination with the gas leak position as teacher data, and a guess model that outputs the gas leak position indicating the leak source in the gas region is generated by inputting the gas distribution image.

According to the gas leak position estimation model generation device, the gas leakage position estimation model generation method, and the program according to one aspect of the present disclosure, an estimation model capable of identifying the gas leakage position based on the characteristics of the gas distribution image is formed. Therefore, by properly designing the teacher data, it is possible to generate a guess model that can identify the gas leak position regardless of the gas leak state.

(11) The gas leak position identification device according to one aspect of the present disclosure includes an image input unit that accepts a gas distribution image in which an image of a gas region leaked in space is included in the image as an input, and a gas leaked in space. Machine learning is performed using the combination of the teacher gas distribution image in which the image of the region is included in the image and the gas leak position indicating the position of the gas leakage source in the gas region shown in the teacher gas distribution image as teacher data. , A machine learning unit that generates an estimation model that outputs a gas leakage position indicating the leakage source of the gas region by inputting a gas distribution image, and a gas of the gas distribution image received by the image input unit using the estimation model. It is provided with a gas leak position identification unit for identifying a gas leak position indicating the position of the leak source in the region.

Further, in the gas leak position identification method according to one aspect of the present disclosure, an image of a gas region leaked in space is received as an input, and an image of a gas region leaked in space is an image. Machine learning is performed using the combination of the teacher gas distribution image contained therein and the gas leak position indicating the position of the gas leak source in the gas region shown in the teacher gas distribution image as teacher data, and the gas distribution image is obtained. A guess model that outputs the gas leak position indicating the leak source in the gas region is generated as an input, and the guess model is used to identify the gas leak position that indicates the position of the leak source in the gas region of the received gas distribution image. ..

Further, the program according to one aspect of the present disclosure is a program for causing a computer to perform a gas leak position identification process, and the gas leak position identification process includes an image of a gas region leaked in space in an image. The gas distribution image is accepted as an input, and the image of the gas region leaked in the space is included in the image, and the position of the gas leakage source in the gas region shown in the teacher gas distribution image is shown. Machine learning is performed using the combination with the gas leakage position as teacher data, and an estimation model that outputs the gas leakage position indicating the leakage source in the gas region is generated by inputting the gas distribution image, and the estimation model is used to receive the above. Identify the gas leak location that indicates the location of the leak source in the gas region of the gas distribution image.

According to the gas leak position estimation model generation device, the gas leakage position estimation model generation method, and the program according to one aspect of the present disclosure, a guess model capable of identifying the gas leakage position based on the characteristics of the gas distribution image is formed. Identify the location of the gas leak. Therefore, by appropriately designing the teacher data, it is possible to generate a guess model that can identify the gas leak position regardless of the gas leak state and identify the gas leak position.

The gas leak position identification device, the gas leak position identification method, and the program according to the present disclosure can identify the gas leak position regardless of the state of the gas leak, and can safely investigate the site. It is useful.

100, 101, 110 Gas detection system 10 Image generation unit 20, 26, 27 Gas detection device 201 Image acquisition unit 21, 22, 23 Gas leak identification device 211 Gas region extraction unit 212 Gas region image acquisition unit 213 Leakage position information acquisition unit 214, 224, 234 Gas leak position identification unit 2141, 2241 Machine learning unit 2142, 2242 Learning model holding unit 215, 227 Judgment result output unit 24 Display unit 25 Gas leak position identification system 226 Field information holding unit

Claims (17)

1. An image input unit that accepts a gas distribution image in which an image of a gas region leaked into space is included in the image as an input.
   A combination of a teacher gas distribution image in which an image of a gas region leaked in space is included in the image and a gas leak position indicating the position of a gas leak source in the gas region shown in the teacher gas distribution image is taught. A gas leak position including a gas leak position identification unit that identifies a gas leak position indicating the position of a leak source in the gas region of the gas distribution image received by the image input unit using a machine-learned estimation model as data. Identification device.
2. The gas distribution image input to the image input unit includes a gas distribution image in which a part of the gas region is shielded.
   The teacher gas distribution image includes a gas distribution image in which a part of the gas region is shielded.
   The gas leak position identification device according to claim 1.
3. The gas distribution image received by the image input unit and the teacher gas distribution image included in the teacher data are both moving images including a plurality of frames.
   The guess model is formed by using the change in the gas region between frames of the gas distribution image as one of the features.
   The gas leak position identification device according to claim 1 or 2, wherein the gas leak position identification unit identifies a gas leak position by using a change in a gas region between frames of a gas distribution image as one of the feature quantities.
4. The guess model uses the coordinate values in the teacher gas distribution image as the gas leak position in the teacher data.
   The gas leak position identification device according to any one of claims 1 to 3, wherein the gas leak position identification unit outputs a coordinate value in a gas distribution image received by the image input unit as a gas leak position.
5. The estimation model uses a probability distribution in which each pixel of the teacher gas distribution image indicates a gas leak source as a gas leak position in the teacher data.
   The gas leak position identification unit outputs a probability distribution indicating the probability that each pixel of the gas distribution image received by the image input unit is a gas leak source as a gas leak position. Any one of claims 1 to 3. The gas leak location identification device described in the section.
6. A first image acquisition unit that captures an image of gas equipment from a first viewpoint and generates a gas distribution image from the obtained image.
   A second image acquisition unit that images the gas equipment from a second viewpoint different from the first viewpoint and generates a gas distribution image from the obtained image.
   The first image acquisition unit and the said The second image acquisition unit is a gas leak position identification system arranged so as to include a common spatial area in each imaging range.
7. Further equipped with a 3D leak position identification unit,
   The first gas leak position identification device acquires a gas distribution image from the first image acquisition unit and identifies the corresponding gas leak position.
   The second gas leak position identification device acquires a gas distribution image from the second image acquisition unit and identifies the corresponding gas leak position.
   The three-dimensional leakage position identification unit identifies the first spatial region indicated by the gas leakage position identified by the first gas leakage position identification device based on the imaging range of the first image acquisition unit, and the second. The second spatial region indicated by the gas leak position identified by the second gas leak position identification device is specified based on the imaging range of the image acquisition unit, and the first spatial region and the second spatial region are defined. The gas leak position identification system according to claim 6, wherein a spatial region included in any of the above is identified as a leak source.
8. A first image acquisition unit that images the space where gas leaks from the first viewpoint and generates a gas distribution image from the obtained image.
   A second image acquisition unit that images the space from a second viewpoint different from the first viewpoint and generates a gas distribution image from the obtained image.
   A first gas leak position identification unit that acquires the gas distribution image from the first image acquisition unit and identifies a corresponding gas leak position, and a first gas leak position identification unit.
   A second gas leak position identification unit that acquires the gas distribution image from the second image acquisition unit and identifies the corresponding gas leak position, and a second gas leak position identification unit.
   The first spatial region indicated by the gas leak position identified by the first gas leak position identification device is specified based on the image pickup range of the first image acquisition unit, and is based on the image pickup range of the second image acquisition unit. The second space region indicated by the gas leak position identified by the second gas leak position identification device is specified, and the space region included in both the first space region and the second space region is defined as described above. A three-dimensional leak location identification unit that identifies as a gas leak source,
   A gas leak location identification system.
9. Further provided with equipment information holding unit that holds the three-dimensional structure information of the equipment in the space and the relationship between the three-dimensional structure information and the first viewpoint and the second viewpoint.
   The gas leak position identification system according to claim 7 or 8, wherein the three-dimensional leak position identification unit further identifies the relative positional relationship between the gas leak position and the equipment.
10. An image input unit that accepts a gas distribution image in which an image of a gas region leaked into space is included in the image as an input.
    A gas leak position input unit that accepts a gas leak position indicating the position of a gas leak source in the gas region shown in the gas distribution image as an input, and a gas leak position input unit.
    Machine learning is performed using the combination of the gas distribution image and the gas leak position corresponding to the gas distribution image as teacher data, and a guess model is generated that outputs the gas leak position indicating the leak source in the gas region by inputting the gas distribution image. A gas leak position estimation model generator equipped with a machine learning unit.
11. An image input unit that accepts a gas distribution image in which an image of a gas region leaked into space is included in the image as an input.
    A combination of a teacher gas distribution image in which an image of a gas region leaked in space is included in the image and a gas leak position indicating the position of a gas leak source in the gas region shown in the teacher gas distribution image is taught. A machine learning unit that performs machine learning as data and generates an estimation model that outputs a gas leak position indicating the leak source in the gas region by inputting a gas distribution image.
    A gas leak position identification device including a gas leak position identification unit that identifies a gas leak position indicating the position of a leak source in a gas region of the gas distribution image received by the image input unit using the estimation model.
12. The image of the gas region leaked into the space is included in the image. The gas distribution image is accepted as input.
    A combination of a teacher gas distribution image in which an image of a gas region leaked in space is included in the image and a gas leak position indicating the position of a gas leak source in the gas region shown in the teacher gas distribution image is taught. A gas leak position identification method for identifying a gas leak position indicating the position of a leak source in the gas region of the received gas distribution image using a machine-learned estimation model as data.
13. A program that causes a computer to perform gas leak location identification processing.
    The gas leak position identification process is performed.
    The image of the gas region leaked into the space is included in the image. The gas distribution image is accepted as input.
    A combination of a teacher gas distribution image in which an image of a gas region leaked in space is included in the image and a gas leak position indicating the position of a gas leak source in the gas region shown in the teacher gas distribution image is taught. A program that identifies the location of a gas leak that indicates the location of the leak source in the gas region of the received gas distribution image using a machine-learned estimation model as data.
14. The image of the gas region leaked into the space is included in the image. The gas distribution image is accepted as input.
    The gas leak position indicating the position of the gas leak source in the gas region shown in the gas distribution image is accepted as an input.
    Machine learning is performed using the combination of the gas distribution image and the gas leak position corresponding to the gas distribution image as teacher data, and a guess model is generated that outputs the gas leak position indicating the leak source in the gas region by inputting the gas distribution image. How to generate a gas leak position estimation model.
15. A program that causes a computer to generate a gas leak position estimation model.
    The gas leak position estimation model generation process is performed.
    The image of the gas region leaked into the space is included in the image. The gas distribution image is accepted as input.
    The gas leak position indicating the position of the gas leak source in the gas region shown in the gas distribution image is accepted as an input.
    Machine learning is performed using the combination of the gas distribution image and the gas leak position corresponding to the gas distribution image as teacher data, and a guess model is generated that outputs the gas leak position indicating the leak source in the gas region by inputting the gas distribution image. Program to do.
16. The image of the gas region leaked into the space is included in the image. The gas distribution image is accepted as input.
    A combination of a teacher gas distribution image in which an image of a gas region leaked in space is included in the image and a gas leak position indicating the position of a gas leak source in the gas region shown in the teacher gas distribution image is taught. Machine learning is performed as data, and a guess model that outputs the gas leak position indicating the leak source in the gas region is generated by inputting the gas distribution image.
    A gas leak position identification method for identifying a gas leak position indicating the position of a leak source in a gas region of the received gas distribution image using the guess model.
17. A program that causes a computer to perform gas leak location identification processing.
    The gas leak position identification process is performed.
    The image of the gas region leaked into the space is included in the image. The gas distribution image is accepted as input.
    A combination of a teacher gas distribution image in which an image of a gas region leaked into the air is included in the image and a gas leak position indicating the position of a gas leak source in the gas region shown in the teacher gas distribution image is taught. Machine learning is performed as data, and a guess model that outputs the gas leak position indicating the leak source in the gas region is generated by inputting the gas distribution image.
    A program for identifying a gas leak position indicating the position of a leak source in the gas region of the received gas distribution image using the guess model.

Images