Core body temperature estimation system, core body temperature estimation method, and core body temperature estimation program

Abstract

This invention provides a core body temperature estimation system, a core body temperature estimation method, and a core body temperature estimation program that can estimate core body temperature at a low cost without interfering with movement during exercise. [Solution] The present invention provides a core body temperature estimation system 1 for estimating the core body temperature of a subject U during exercise, comprising: an image acquisition unit 201 that acquires a first visible image G10 and a thermal image G20 including the face of the subject U; a region extraction unit 202 that extracts a second visible image G11 of all or a specific part of the face region from the first visible image G10; a surface temperature acquisition unit 205 that acquires surface temperature information corresponding to the second visible image G11 based on the thermal image G20; a core body temperature acquisition unit 206 that inputs the surface temperature information to a core body temperature estimation model M and acquires the estimated core body temperature of the subject U estimated by the core body temperature estimation model M; and an output unit 207 that outputs the estimated core body temperature of the subject.

Description

[Technical Field]

[0001] The present invention relates to a core body temperature estimation system, a core body temperature estimation method, and a core body temperature estimation program for estimating core body temperature. [Background Art]

[0002] In recent years, due to the influence of temperature rise caused by climate change, core body temperature easily rises during exercise, creating an environment where heat stroke is likely to occur. For this reason, it has become important to monitor the core body temperature of a subject to prevent heat stroke during exercise, regardless of whether the exercise is indoors or outdoors. Normally, measuring core body temperature requires invasive measurement on the body such as in the rectum or ear. Invasive measurement not only interferes with movement during exercise, but is also difficult to implement from the viewpoint of safety, making measurement during exercise difficult.

[0003] For this reason, methods have been proposed in which a wearable sensor that can be worn during exercise measures the body surface temperature of a subject, and estimates the subject's core body temperature from the measured body surface temperature. The present applicant has also proposed a measurement device using a wearable sensor (Patent Document 1). According to this measurement device, it is possible to estimate the subject's core body temperature without invading the body, and accurately prevent heat stroke during exercise. [Prior Art Documents] [Patent Documents]

[0004] [Patent Document 1] Japanese Patent Application No. 2017-217224 Publication [Summary of the Invention] [Problem to be Solved by the Invention]

[0005] However, depending on the sport, wearable sensors may interfere with movement during exercise, as they may involve physical contact between competitors or require delicate body movements. Furthermore, in school sports events such as club activities or athletic festivals, it is difficult to provide wearable sensors to all participants due to cost considerations.

[0006] Therefore, in view of the above circumstances, the present invention aims to provide a core body temperature estimation system, a core body temperature estimation method, and a core body temperature estimation program that can estimate core body temperature without interfering with movement during exercise and while keeping costs down. [Means for solving the problem]

[0007] In light of the current situation, the inventors, after diligent study, conceived the idea of ​​using visible and thermal images of a subject during exercise.

[0008] In other words, we discovered that by extracting the facial region where the skin is always exposed in a visible image containing a subject during exercise, obtaining the surface temperature corresponding to the facial region in a thermal image containing the subject during exercise, and inputting this surface temperature into a machine learning-based core body temperature estimation model to estimate the subject's core body temperature, we were able to estimate the subject's core body temperature, thus completing the present invention.

[0009] This invention encompasses the following inventions. (1) A core body temperature estimation system for estimating the core body temperature of a subject during exercise, comprising: an image acquisition unit that acquires a visible image and a thermal image including the subject's face; a region extraction unit that extracts a visible image of all or a specific part of the face region from the visible image; a surface temperature acquisition unit that acquires surface temperature information corresponding to the visible image of all or a specific part of the face region extracted by the region extraction unit based on the thermal image; a core body temperature acquisition unit that inputs the surface temperature information acquired by the surface temperature acquisition unit into a core body temperature estimation model that has been trained using the surface temperature of all or a specific part of a person's face as an explanatory variable and the person's core body temperature as an objective variable, and acquires the estimated core body temperature of the subject estimated by the core body temperature estimation model; and an output unit that outputs the estimated core body temperature of the subject acquired by the core body temperature acquisition unit.

[0010] (2) The core body temperature estimation system according to (1), wherein the visible image and the thermal image acquired by the image acquisition unit are images taken with the same field of view and the same scale.

[0011] (3) The core body temperature estimation system according to (1) or (2), further comprising a region extraction unit for extracting a portion of the subject's face included in a visible image of all or a specific portion of the facial region, wherein the surface temperature acquisition unit acquires surface temperature information corresponding to the portion of the subject's face extracted by the region extraction unit based on the thermal image.

[0012] (4) The core body temperature estimation system according to (1), further comprising a personal identification unit that identifies the subject based on a visible image of all or a specific part of the facial region extracted by the region extraction unit, wherein the core body temperature acquisition unit acquires an estimated core body temperature for each subject identified by the personal identification unit.

[0013] (5) A method for estimating the core body temperature of a subject during exercise, wherein an information processing device acquires a visible image and a thermal image including the subject's face, extracts a visible image of all or a specific part of the face region from the visible image, acquires surface temperature information corresponding to all or a specific part of the visible image of the face region based on the thermal image, inputs the surface temperature information acquired by the surface temperature acquisition unit into a core body temperature estimation model that has been trained using the surface temperature of all or a specific part of the person's face as an explanatory variable and the person's core body temperature as an objective variable, obtains the subject's estimated core body temperature, and outputs the subject's estimated core body temperature obtained from the core body temperature estimation model.

[0014] (6) A core body temperature estimation program for estimating the core body temperature of a subject during exercise, wherein the computer functions as an image acquisition unit that acquires a visible image and a thermal image including the subject's face; a region extraction unit that extracts a visible image of all or a specific part of the face region from the visible image; a surface temperature acquisition unit that acquires surface temperature information corresponding to the visible image of all or a specific part of the face region extracted by the region extraction unit based on the thermal image; a core body temperature estimation model that has been trained using machine learning with the surface temperature of all or a specific part of a person's face as an explanatory variable and the person's core body temperature as the objective variable, and inputs the surface temperature information acquired by the surface temperature acquisition unit to acquire the estimated core body temperature of the subject estimated by the core body temperature estimation model; and an output unit that outputs the estimated core body temperature of the subject acquired by the core body temperature acquisition unit. [Effects of the Invention]

[0015] According to the core body temperature estimation system, core body temperature estimation method, and core body temperature estimation program of the present invention described above, core body temperature can be estimated at a low cost without interfering with the movements of a subject during exercise.

[0016] Further, when the visible image and the thermal image are both captured with the same field of view and at the same scale, the surface temperature acquisition unit can acquire surface temperature information corresponding to the face region without adjusting the position corresponding to the face region.

[0017] Further, according to the configuration further including a region extraction unit, facial regions are extracted from a visible image in which all or a specific part of the face region has been extracted, and surface temperature information corresponding to each of the facial regions is acquired, whereby the core body temperature of a subject during exercise can be estimated with high accuracy.

[0018] Further, according to the configuration further including an individual identification unit, even in an image captured with a large number of subjects, each individual subject can be identified and the core body temperature of each subject can be estimated respectively. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] [Figure 1] A block diagram showing a core body temperature estimation system according to an embodiment of the present invention. [Figure 2] An explanatory diagram showing processing operations in the core body temperature estimation system. [Figure 3] An explanatory diagram showing facial regions extracted by the region extraction unit. [Figure 4] A flow chart showing the processing procedure up to output of an estimated core body temperature. [Figure 5] A box-and-whisker plot showing the mean absolute error distribution of subject U in multiple regression analysis. [Figure 6] A box-and-whisker plot showing the mean absolute error distribution of subject U in a neural network. [Figure 7] A table showing a comparison of correlation coefficients between surface temperature and core body temperature for each facial region PA. [Figure 8] A box-and-whisker plot showing a comparison of mean absolute error of subject U between multiple regression analysis and a neural network. [Figure 9] A graph showing transitions of estimated core body temperature and measured core body temperature of subject U in multiple regression analysis. [Figure 10]Graph showing transitions of the estimated core body temperature and measured core body temperature of subject U in a neural network. Mode for Carrying Out the Invention

[0020] Hereinafter, representative embodiments of the present invention will be described in detail with reference to the accompanying drawings.

[0021] The core body temperature estimation system 1 according to the present embodiment acquires a visible image and a thermal image that include a subject during exercise as an image, and estimates core body temperature using the visible image and the thermal image. A face region is extracted from the visible image, and the surface temperature of the face region is acquired from the thermal image corresponding to the face region. Then, the surface temperature is input to a core body temperature estimation model obtained by machine-learning a training data set necessary for estimating core body temperature, and the core body temperature of the subject is estimated. The estimated core body temperature is output to an external device or the like that notifies the subject during exercise of the risk of heat stroke.

[0022] As shown in FIG. 1, the core body temperature estimation system 1 is implemented by an information processing device 10. The core body temperature estimation system 1 may be implemented by a computing device such as a server, a computer, a smartphone, or a tablet, or may be implemented by combining a plurality of these computing devices. The information processing device 10 includes an information processing section 20, a storage section 30, and a communication section 40.

[0023] The information processing unit 20 can be composed of an arithmetic processing unit including a circuit consisting of a CPU (Central Processing Unit), GPU (Graphics Processing Unit), ASIC (Application Specific Integrated Circuit), PLD (Programmable Logic Device), or FPGA (Field Programmable Gate Array). In other words, the functions of the information processing unit 20 are realized by the arithmetic processing unit. The storage unit 30 can be composed of a storage device consisting of an HDD or SSD, and RAM, etc. The communication unit 40 can be composed of a communication processing unit including various communication circuits that perform wired and / or wireless communication processing with external devices, the Internet, LAN (Local Area Network), cellular network, and other communication networks N.

[0024] The memory unit 30 stores a core body temperature estimation program for the information processing device 10 to perform the core body temperature estimation method, as well as files and data used to execute the core body temperature estimation program. It also stores a core body temperature estimation model M that has been trained using machine learning to estimate the core body temperature of a subject. The algorithm of the core body temperature estimation model M may use multiple regression analysis (Linear) or a neural network (NN), and any machine learning algorithm that can estimate core body temperature using the surface temperature of the subject U's face during exercise as input is acceptable.

[0025] As shown in Figure 1, the information processing unit 20 includes an image acquisition unit 201, a region extraction unit 202, a body part extraction unit 203, a personal identification unit 204, a surface temperature acquisition unit 205, a core body temperature acquisition unit 206, and an output unit 207. The information processing unit 20 functions using a core body temperature estimation model M and a core body temperature estimation program stored in the storage unit 30.

[0026] As shown in Figure 2, the image acquisition unit 201 acquires a first visible image G10 and a thermal image G20, including the face of the subject U. Specifically, the image acquisition unit 201 is connected to a communication network N via a communication unit 40 and communicates with an imaging device that images the subject U during movement, thereby acquiring the first visible image G10 and the thermal image G20. Alternatively, the image acquisition unit 201 may communicate with an image storage that temporarily stores the first visible image G10 and the thermal image G20, which were captured by the imaging device of the subject U during movement, and acquire the first visible image G10 and the thermal image G20. Preferably, the first visible image G10 and the thermal image G20 are images captured with the same field of view and scale. Preferably, the imaging device is a spectral camera that captures the first visible image G10 and the thermal image G20 simultaneously. However, if the first visible image G10 and the thermal image G20 can be captured with the same field of view and scale, they may be captured with different imaging devices. Furthermore, the first visible image G10 and thermal image G20 are preferably images taken from the front of the subject U during exercise.

[0027] As shown in Figure 2, the region extraction unit 202 extracts a second visible image G11 from the first visible image G10 that represents all or a specific part of the face region FA. Specifically, the region extraction unit 202 extracts the face region FA with a large amount of exposed skin from the first visible image G10. The second visible image G11 may be a region that includes all facial components such as the eyes, nose, mouth, eyebrows, and ears, or it may be a specific part of the face. A specific part of the face refers to one or more facial components and their surroundings. After this, the range of the face region FA is not particularly limited as long as the surface temperature of subject U, for which the core body temperature estimation model M is needed to estimate subject U's core body temperature, can be obtained.

[0028] The surface temperature acquisition unit 205 acquires surface temperature information corresponding to the visible image of the face region FA extracted by the region extraction unit 202, based on the thermal image G20. Specifically, the surface temperature acquisition unit 205 acquires the surface temperature represented by the pixel corresponding to the face region FA in the second visible image G11, among the pixels constituting the thermal image G20 in which the subject U was captured during movement. By acquiring the temperature information corresponding to the face region FA in the second visible image G11, the surface temperature distribution of the face region FA is obtained. In this example, the surface temperature acquisition unit 205 calculates the instantaneous average value of the surface temperature distribution of the face region FA and acquires this as the surface temperature information of the face region FA. In other examples, the surface temperature acquisition unit 205 may calculate the instantaneous median, maximum value, minimum value, variance, or standard deviation of the surface temperature distribution of the face region FA and acquire it as the surface temperature information of the face region FA. Alternatively, the surface temperature acquisition unit 205 may calculate two or more values ​​from the instantaneous mean, median, maximum, minimum, variance, and standard deviation for the surface temperature distribution of the face region FA, and acquire these as surface temperature information for the face region FA.

[0029] The surface temperature acquisition unit 205 divides the face region FA into multiple parts and acquires surface temperature information from the corresponding pixels for each part, thereby improving the accuracy of the core body temperature of the subject U that is subsequently estimated. Therefore, it is preferable to further include a part extraction unit 203 that extracts the face parts PA of the subject U included in the second visible image G11, which may be all or a specific part of the face region FA. Specifically, as shown in Figures 2 and 3, the part extraction unit 203 divides the face region FA in the second visible image G11 into multiple face parts PA and extracts the multiple face parts PA. When dividing the face region FA in the second visible image G11, the part extraction unit 203 may use a well-known machine learning model for dividing the face region of a person in an image, such as a Face Segmentation model. The facial area PA may be divided into individual facial components such as eyes, nose, mouth, eyebrows, and ears. Alternatively, if a facial component is a pair, it may be further divided into right eye, left eye, upper lip, and lower lip. Or, a continuous area of ​​skin on the face, such as the forehead, cheeks, and chin, may be divided as a single facial area. How the facial region FA is divided into multiple facial area PAs can be adjusted according to the direction, field of view, and scale of imaging of the subject U during movement. The surface temperature acquisition unit 205 acquires surface temperature information from the pixels corresponding to each facial area PA among the pixels that make up the thermal image G20.

[0030] Furthermore, depending on the sport, there may be multiple subjects U. In this case, it is preferable to further provide a personal identification unit 204 that identifies each subject U individually. The personal identification unit 204 identifies subject U based on a second visible image G11 of all or a specific part of the face region FA from the first visible image G10 using a region extraction unit 202. The personal identification unit 204 may use well-known technologies, such as face recognition using machine learning models, identification by feature point extraction, person identification using deep learning, as well as conventional image processing technologies such as pattern matching, contour extraction, and identification using color and texture information.

[0031] As shown in Figure 2, the core body temperature acquisition unit 206 inputs surface temperature information acquired by the surface temperature acquisition unit 205 to the core body temperature estimation model M and acquires the core body temperature of the subject U estimated by the core body temperature estimation model M. In this example, a pre-trained model is used as the core body temperature estimation model M, which uses the surface temperature of all or a specific part of a person's face as an explanatory variable and the measured core body temperature as the objective variable. At this time, the surface temperature information acquired by the surface temperature acquisition unit 205 is input to the core body temperature estimation model M and becomes the data to be estimated for estimating the core body temperature of the subject U. In the training stage of the core body temperature estimation model M, the surface temperature information acquired by the surface temperature acquisition unit 205 is used as input to the training data. Furthermore, for measuring the core body temperature, which is the objective variable, well-known measurement means such as an oral capsule sensor can be used.

[0032] As shown in Figure 2, the output unit 207 outputs the estimated core body temperature of subject U acquired by the core body temperature acquisition unit 206. Specifically, the output unit 207 outputs the core body temperature information acquired by the core body temperature acquisition unit 206 to a monitor terminal (not shown) held by a monitor who monitors subject U during exercise, via the communication unit 40. Examples of such monitors include the organizing committee or referees in charge of the athletic competition, medical staff on standby at the athletic competition, or management staff of the athlete subject U. This allows the monitor to grasp the heat stress situation of subject U during exercise in real time based on the estimated core body temperature output to the monitor terminal and take appropriate action quickly.

[0033] Preferably, the monitoring terminal has a function to notify the user via its screen or speaker when the estimated core body temperature output from the output unit 207 exceeds a predetermined threshold. For example, the risk of heatstroke can be prevented by providing notifications via vibration, sound, or messages to the monitoring terminal. Furthermore, the estimated core body temperature of the subject U acquired by the core body temperature acquisition unit 206 can be stored chronologically in the memory unit 30 or an external storage device and used for condition management and performance analysis.

[0034] Furthermore, the output unit 207 can upload estimated core body temperature data to a cloud server via the communication unit 40, allowing it to be used for statistical analysis and big data analysis. This enables the analysis of core body temperature fluctuation trends under specific sports or environmental conditions, providing insights useful for managing the risk of heatstroke.

[0035] The following describes the processing steps of the core body temperature estimation system 1, in which the core body temperature estimation method functions on the information processing device 10, from acquiring a first visible image G10 and a thermal image G20 taken from the front of a subject U during exercise, to outputting the estimated core body temperature, based on Figure 4.

[0036] First, the image acquisition unit 201 acquires the first visible image G10 and the thermal image G20 (S401), and stores the first visible image G10 and the thermal image G20 in the storage unit 30 (S402). The first visible image G10 and the thermal image G20 are images captured with the same field of view and the same scale, and are stored in the storage unit 30 in association with each other.

[0037] Next, the region extraction unit 202 extracts a second visible image G11 from the first visible image G10 that includes all or a specific part of the face region FA of the subject U (S403) and stores it in the storage unit 30 (S404). The surface temperature acquisition unit 205 acquires surface temperature information corresponding to the second visible image G11 based on the thermal image stored in the storage unit 30 (S405) and inputs it into the core body temperature estimation model M (S406).

[0038] Next, the core body temperature acquisition unit 206 acquires the estimated core body temperature estimated by the core body temperature estimation model M (S407), and finally, the output unit 207 outputs the estimated core body temperature to the storage unit 30 or an external device via the communication unit 40 (S408).

[0039] According to the core body temperature estimation system 1, which utilizes this core body temperature estimation method, core body temperature can be estimated at low cost without interfering with the subject's movements during exercise. [Examples]

[0040] Next, we will describe the performance evaluation of the core body temperature estimation system 1 according to an embodiment of the present invention. Note that the following description is not limited to the present invention.

[0041] The performance evaluation in this embodiment was conducted using data collected in an actual exercise environment to demonstrate the effectiveness of the core body temperature estimation system 1, which estimates the core body temperature of a subject U during exercise. The evaluation was carried out by comparing the error between the estimated core body temperature estimated using the core body temperature estimation model M, based on surface temperature information obtained from the face region FA (face region PA), and the actually measured core body temperature. The core body temperature estimation model M utilized multiple regression analysis and a neural network.

[0042] (1) Evaluation environment The performance evaluation of Core Body Temperature Estimation System 1 was based on data from exercise in a hot environment. The experimental environment consisted of a room temperature of 40°C and 40% humidity. Subject U exercised on a treadmill, and temperature data (thermal image G20) from around the face was collected. The subject's true core body temperature was measured using an oral capsule sensor. The accuracy of core body temperature estimation was evaluated using 30 minutes of exercise data under these conditions. Core body temperature estimation performance was evaluated using the Mean Absolute Error (MAE) with leave-one-subject-out cross-validation.

[0043] (2) Feature selection The core body temperature estimation model M used surface temperatures obtained from each facial area PA (Patient Aspect) of the facial region FA as input features. Specifically, surface temperatures were obtained for each facial area PA, such as lips, neck, nose, and skin, and the core body temperature estimation model M was trained based on these features. To select facial area PA-specific features for the core body temperature estimation model M, 15 comprehensive explanatory variable sets were created using the surface temperatures of the four facial area PAs: lips, neck, nose, and skin. Feature selection was performed by comparing the mean absolute errors when each explanatory variable set was fed into multiple regression analysis and a neural network. As a result, the explanatory variable set that minimized the mean absolute error in both multiple regression analysis and the neural network was the set using the lip and skin area-specific features.

[0044] The mean absolute error distributions for subjects U in the multiple regression analysis and neural network are shown in Figures 5 and 6, respectively. To facilitate comparison of the mean absolute error distributions for subjects U, the mean absolute error distributions are shown when using four explanatory variable sets consisting of individual features for each facial region PA, namely upper lip region PA1, lower lip region PA2, neck region PA3, nose region PA4, and skin region PA5, and when using only the facial region FA (mean surface temperature) as an explanatory variable. Referring to Figures 5 and 6, it was confirmed that the explanatory variable set of skin region PA5, upper lip region PA1, and lower lip region PA2 resulted in a mean absolute error of less than 0.60°C for 75% of subjects U in both the multiple regression analysis and the neural network, indicating that the mean absolute error was generally lower compared to other explanatory variable sets. In particular, the comparison with the facial region FA showed that there were many subjects with a mean absolute error of 0.55°C or less, verifying the effectiveness of region extraction.

[0045] As can be seen from the table shown in Figure 7, among the surface temperatures of each facial region PA obtained from four regions—the lip region PA12 (composed of the upper lip region PA1 and the lower lip region PA2), the neck region PA3, the nose region PA4, and the skin region PA5—the neck region PA3 and the nose region PA4 were not selected as features for estimating core body temperature because they showed a smaller positive correlation with core body temperature compared to the other regions.

[0046] (3) Effects of adding past statistics Next, we evaluated the effect of adding past surface temperature statistics to the core body temperature estimation model M. For pairs using facial region PA features, specifically lip region PA12 and skin region PA5, which had the smallest mean absolute error, we examined the effect of inputting past statistics as training data for the core body temperature estimation model M. Figure 8 compares the performance of estimation using only facial region PA surface temperature as training data with the performance of estimation using past facial region PA surface temperature statistics, i.e., the mean, median, maximum, minimum, variance, and standard deviation over the past 5 minutes, as well as the training data. As can be seen from Figure 8, in both the multiple regression analysis and the neural network, including past statistics reduced the variability of the mean absolute error distribution. Furthermore, the change in the mean of the mean absolute error decreased by approximately 0.10°C in both the multiple regression analysis (from 0.505°C to 0.392°C) and the neural network (from 0.491°C to 0.396°C), indicating that the overall mean absolute error decreased due to the consideration of time-series changes.

[0047] (4) Impact of estimation accuracy To evaluate the impact on estimation accuracy in the core body temperature estimation model M, multiple regression analysis and neural networks were compared. As can be seen in Figures 9 and 10, when the estimation accuracy of each regression model was examined, the neural network shown in Figure 10 showed a gradual increase in the estimated core body temperature, while the multiple regression analysis shown in Figure 9 showed a tendency to increase while fluctuating up and down. This difference is thought to be partly due to the fact that the skin surface temperature obtained by thermal imaging G20 itself was fluctuating up and down.

[0048] Furthermore, both the multiple regression analysis and the neural network showed an overall increase in the estimated core body temperature, but it remained lower than the actually measured core body temperature. One possible reason for this is that there were times when the facial region FA or facial area PA of subject U was not accurately detected, and in those cases, the surface temperature of the neck or part of the clothing was calculated as the surface temperature of the facial area PA.

[0049] Therefore, to investigate the impact of images with reduced facial PA detection accuracy on the performance of core body temperature estimation, a comparison was made with the case where these images were removed. As a result, the mean absolute error was 0.35°C when images with reduced facial PA detection accuracy were not removed, and 0.39°C when images with reduced facial PA detection accuracy were removed. In other words, it was confirmed that high estimation accuracy is maintained even when images with reduced facial PA detection accuracy are included. This is because statistical values ​​for surface temperature for each area over a certain period are calculated and used as features. By utilizing these statistical values, it is thought that the decrease in estimation accuracy can be suppressed even when false detections occur or when the number of pixels for each detected facial PA is small.

[0050] The core body temperature estimation system 1 of this embodiment has been confirmed to be able to estimate the core body temperature of subject U during exercise with high accuracy, and can be said to be effective in preventing heatstroke and managing physical condition. In the future, it will be possible to utilize data from a wider variety of exercise environments and subjects to enable evaluation under a wide range of conditions. Furthermore, although this system was used for subject U during exercise, in other embodiments it can also be applied to estimate the core body temperature of workers working in high-temperature work environments.

[0051] Although embodiments of the present invention have been described above, the present invention is not limited in any way to these embodiments, and can be implemented in various forms without departing from the spirit of the invention. For example, in this embodiment, multiple regression analysis and neural networks were used, but the present invention is not limited thereto, and can also be combined with methods such as decision tree analysis, support vector machines (SVMs), and reinforcement learning algorithms. [Explanation of Symbols]

[0052] 1. Core Body Temperature Estimation System 10 Information Processing Devices 20 Information Processing Section 30 Storage section 40 Communications Department 201 Image Acquisition Unit 202 Region extraction part 203 Part extraction part 204 Personal identification section 205 Surface temperature acquisition section 206 Core body temperature acquisition department 207 Output section FA facial region G10 First visible image G11 Second visible image G20 Thermal Images M Core Body Temperature Estimation Model N Communication Network PA (Patient Protection) - Face Area PA1 Upper lip area PA2 lower lip area PA3 neck area PA4 Nose area PA5 Skin area PA12 Lip area U Target Persons

Claims

1. A core body temperature estimation system that estimates the core body temperature of a subject during exercise, An image acquisition unit that acquires a visible image and a thermal image including the face of the subject, A region extraction unit that extracts a visible image of all or a specific part of the facial region from the aforementioned visible image, A surface temperature acquisition unit acquires surface temperature information corresponding to the visible image of all or a specific part of the face region extracted by the region extraction unit based on the thermal image, A core body temperature acquisition unit inputs surface temperature information acquired by the surface temperature acquisition unit into a core body temperature estimation model that has been trained using the surface temperature of all or a specific part of a person's face as an explanatory variable and the person's core body temperature as the dependent variable, and acquires the estimated core body temperature of the subject estimated by the core body temperature estimation model. The system includes an output unit that outputs the estimated core body temperature of the subject obtained by the core body temperature acquisition unit, Core body temperature estimation system.

2. The visible image and the thermal image acquired by the image acquisition unit are images captured with the same field of view and the same scale. The core body temperature estimation system according to claim 1.

3. The system further includes a region extraction unit that extracts a portion of the subject's face included in a visible image of all or a specific portion of the facial region, The surface temperature acquisition unit acquires surface temperature information corresponding to the facial area of ​​the subject extracted by the area extraction unit, based on the thermal image. The core body temperature estimation system according to claim 1 or 2.

4. The system further includes a personal identification unit that identifies the subject based on the visible image of all or a specific part of the facial region extracted by the region extraction unit, The core body temperature acquisition unit acquires the estimated core body temperature for each subject whose individual has been identified by the personal identification unit. The core body temperature estimation system according to claim 1.

5. A method for estimating the core body temperature of a subject during exercise, Information processing device, A visible image and thermal image of the subject, including the face, are acquired. From the aforementioned visible image, extract a visible image of all or a specific part of the facial region. Based on the thermal image, surface temperature information corresponding to the visible image of all or a specific part of the facial region is obtained. The surface temperature information acquired by the surface temperature acquisition unit is input to a machine learning model for estimating core body temperature, which uses the surface temperature of all or a specific part of a person's face as an explanatory variable and the person's core body temperature as the dependent variable, in order to obtain the estimated core body temperature of the subject. Outputs the estimated core body temperature of the subject obtained from the core body temperature estimation model. Method for estimating core body temperature.

6. A core body temperature estimation program that estimates the core body temperature of a subject during exercise, Computer An image acquisition unit that acquires a visible image and a thermal image, including the face of the subject. A region extraction unit that extracts a visible image of all or a specific part of the facial region from the aforementioned visible image. A surface temperature acquisition unit acquires surface temperature information corresponding to the visible image of all or a specific part of the facial region extracted by the region extraction unit based on the thermal image. A core body temperature acquisition unit that inputs surface temperature information acquired by the surface temperature acquisition unit into a core body temperature estimation model that has been trained using the surface temperature of all or a specific part of a person's face as an explanatory variable and the person's core body temperature as the dependent variable, and acquires the estimated core body temperature of the subject estimated by the core body temperature estimation model, and This unit functions as an output unit that outputs the estimated core body temperature of the subject obtained by the core body temperature acquisition unit. Core body temperature estimation program.

Images