Temperature measurement method and system for correcting emissivity difference of organism by adopting AI modeling

Abstract

The invention discloses a temperature measurement method and system for correcting organism emissivity difference through AI modeling, and relates to the technical field of organism emissivity difference temperature measurement, and the method comprises the steps: collecting the multiband infrared radiation intensity data of the surface of an organism through a multispectral imaging device; the multi-band infrared radiation intensity data are input into a space-time convolutional neural network, a wavelength-time dependent dynamic emissivity function is output, the space-time convolutional neural network comprises a physical constraint layer, and the physical constraint layer takes the Planck radiation law as priori knowledge to be embedded into a network structure; a differentiable physical engine is adopted to optimize the dynamic emissivity function, neural network parameters are adjusted by minimizing a loss function containing a physical constraint term, and the physical constraint term represents the gradient difference between the actually measured radiation intensity and the theoretical radiation intensity; and generating pathological tissue emissivity disturbance data by using a generative adversarial network, and adding the disturbance data into a training data set to enhance the generalization ability of the model.

Description

Technical Field

[0001] This invention relates to the field of bioluminescence difference temperature measurement technology, and in particular to a method and system for measuring bioluminescence difference using AI modeling and correction. Background Technology

[0002] Biological emissivity differential thermometry is a method that uses the infrared radiation characteristics of biological surfaces to measure their surface temperature. Since the emissivity of biological tissues is not constant but varies with wavelength, time, and tissue state, accurate measurement of biological surface temperature requires consideration of these changing factors. Therefore, improving the intelligence and safety of biological emissivity differential thermometry using advanced technologies has become one of the urgent problems to be solved.

[0003] In the field of bioluminescence differential thermometry, traditional infrared thermometry methods assume that the surface emissivity of organisms is constant or remains unchanged over a wide wavelength range, ignoring the actual changes in emissivity with wavelength, time, and tissue state. This leads to inaccurate measurement results, especially in pathological conditions. Furthermore, existing thermometry techniques perform poorly when dealing with pathological tissues because the emissivity characteristics of pathological tissues differ significantly from those of normal tissues, and existing training datasets are insufficient to cover all possible situations. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method for temperature measurement that uses AI modeling to correct differences in biological emissivity. This addresses the problem that traditional infrared thermometry methods assume that the emissivity of the biological surface is constant or remains unchanged over a wide wavelength range, ignoring the actual changes in emissivity with wavelength, time, and tissue state. This leads to inaccurate measurement results, especially in pathological conditions. Furthermore, existing thermometry techniques perform poorly when dealing with pathological tissues because the emissivity characteristics of pathological tissues differ significantly from those of normal tissues, and existing training datasets are insufficient to cover all possible scenarios.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a method for thermometry that uses AI modeling to correct for differences in emissivity in biological organisms, comprising:

[0008] Multi-band infrared radiation intensity data of the surface of organisms are collected using a multispectral imaging device.

[0009] The multi-band infrared radiation intensity data is input into a spatiotemporal convolutional neural network, which outputs a wavelength-time dependent dynamic emissivity function. The spatiotemporal convolutional neural network includes a physical constraint layer, which embeds Planck's radiation law as prior knowledge into the network structure.

[0010] The dynamic emissivity function is optimized using a differentiable physics engine. The neural network parameters are adjusted by minimizing a loss function that includes physical constraints, which characterize the gradient difference between the measured radiation intensity and the theoretical radiation intensity.

[0011] Generative adversarial networks are used to generate perturbation data of pathological tissue emissivity, and the perturbation data is added to the training dataset to enhance the generalization ability of the model.

[0012] The optimized dynamic emissivity function is substituted into the temperature inversion equation for iterative calculation to obtain the corrected surface temperature distribution of the organism.

[0013] As a preferred embodiment of the temperature measurement method for correcting biological emissivity differences using AI modeling as described in this invention, the multispectral imaging device specifically includes the following steps for acquiring multi-band infrared radiation intensity data of the biological surface:

[0014] A multispectral imaging device was used to collect infrared radiation intensity data from the surface of a living organism within multiple preset wavelength ranges. Each wavelength range corresponds to one spectral channel, and the collected infrared radiation intensity data is denoted as follows: Where, λ i Let represent the i-th wavelength, x and y represent spatial coordinates, and t represent time;

[0015] For the collected Data preprocessing;

[0016] The preprocessing includes removing ambient background radiation, dark current noise, and lens response inhomogeneities to obtain corrected radiation intensity data.

[0017] As a preferred embodiment of the thermometry method for correcting biological emissivity differences using AI modeling as described in this invention, the spatiotemporal convolutional neural network outputs a wavelength-time dependent dynamic emissivity function, specifically including the following steps:

[0018] Using a spatiotemporal convolutional neural network for the input The data undergoes feature extraction, and the network structure includes multiple spatiotemporal convolutional layers and physical constraint layers;

[0019] In the physical constraint layer, Planck's radiation law is embedded as prior physical knowledge;

[0020] Theoretical radiation intensity is calculated using a physical constraint layer. and compared with the measured radiation intensity Compare and generate error terms.

[0021] The spatiotemporal convolutional neural network adjusts the network parameters through the backpropagation algorithm to minimize the error term and finally outputs the dynamic emissivity function ε(λ,t);

[0022] The dynamic emissivity function is used to describe the relationship between emissivity and wavelength λ and time t.

[0023] As a preferred embodiment of the temperature measurement method using AI modeling to correct differences in bioluminescence described in this invention, the optimization of the dynamic emissivity function by the differentiable physics engine specifically includes the following steps:

[0024] A differentiable physics engine is used to optimize the dynamic emissivity function ε(λ,t), and a loss function L containing physical constraints is constructed to guide parameter updates.

[0025] The loss function L includes the data fitting term L data and physical constraint term L phys Two parts;

[0026] The data fitting term L data Defined as the mean square error between the measured radiation intensity and the model-predicted radiation intensity;

[0027] The physical constraint term L phys Defined as the difference between the measured radiation intensity gradient and the theoretical radiation intensity gradient;

[0028] By minimizing the loss function L using gradient descent, the parameters of the spatiotemporal convolutional neural network are updated, making the dynamic emissivity function ε(λ,t) more consistent with physical laws.

[0029] As a preferred embodiment of the temperature measurement method for correcting biological emissivity differences using AI modeling as described in this invention, the generation of pathological tissue emissivity perturbation data via generative adversarial network specifically includes the following steps:

[0030] A generative adversarial network (GAN) is constructed, in which the generator is used to generate emissivity perturbation data simulating pathological tissues, and the discriminator is used to distinguish between real data and generated data.

[0031] The generator takes a random noise vector z as input and outputs emissivity perturbation data δε(λ,t).

[0032] The discriminator receives the true emissivity data ε(λ,t) or the generated perturbation data δε(λ,t) and outputs the probability that it is the true data.

[0033] Through adversarial training, the perturbation data generated by the generator can approximate the emissivity variation characteristics of real pathological tissues.

[0034] The generated perturbation data δε(λ,t) is added to the original emissivity data to obtain the enhanced training data;

[0035] The enhanced data is added to the training set to improve the model's ability to adapt to pathological tissues.

[0036] As a preferred embodiment of the temperature measurement method using AI modeling to correct biological emissivity differences as described in this invention, the iterative calculation of substituting the dynamic emissivity function into the temperature inversion equation specifically includes the following steps:

[0037] The optimized dynamic emissivity function ε(λ,t) is substituted into the temperature inversion equation to calculate the temperature distribution on the surface of the organism.

[0038] The temperature inversion equation is based on Planck's law and emissivity correction;

[0039] An iterative algorithm is used to optimize the temperature. The initial temperature value T0 is obtained by inversion when the emissivity is constant ε0 under the gray body assumption.

[0040] Subsequently, T0 is substituted into the dynamic emissivity function ε(λ,t) to calculate the new corrected radiation intensity;

[0041] By minimizing I corr Compared with the measured I′ λ The error between (x,y,t) is used to iteratively update the temperature value until the convergence condition is met, thus obtaining the corrected surface temperature distribution T(x,y,t) of the organism.

[0042] As a preferred embodiment of the temperature measurement method using AI modeling to correct differences in bioluminescence described in this invention, the spatial filtering and smoothing of the temperature distribution data specifically includes the following steps:

[0043] The obtained temperature distribution data T(x,y,t) is spatially filtered, and a two-dimensional Gaussian filter is used to remove local noise.

[0044] A Gaussian filter is convolved with the temperature distribution data to obtain the smoothed temperature distribution T. smooth (x,y,t);

[0045] Edge enhancement processing is performed on the smoothed temperature data, and the Sobel operator is used to detect temperature gradient change regions and enhance abnormal regions in the temperature distribution.

[0046] Output the corrected image of the organism's surface temperature.

[0047] Secondly, the present invention provides a temperature measurement system that uses AI modeling to correct differences in emissivity in biological organisms, comprising:

[0048] The system includes a data acquisition module, a model training module, an emissivity optimization module, a temperature inversion module, and an image processing module.

[0049] The data acquisition module is used to acquire multi-band infrared radiation intensity data of the surface of organisms through a multispectral imaging device, and to preprocess the data to remove noise and background radiation.

[0050] The model training module is used to input the preprocessed data into the spatiotemporal convolutional neural network, combine the output wavelength-time dependent dynamic emissivity function of the physical constraint layer, and enhance the training set using a generative adversarial network.

[0051] The emissivity optimization module is used to optimize the dynamic emissivity function using a differentiable physics engine, and adjusts the parameters by minimizing the loss function to make it conform to physical laws.

[0052] The temperature inversion module is used to substitute the optimized dynamic emissivity function into the temperature inversion equation for iterative calculation to obtain the corrected surface temperature distribution of the organism.

[0053] The image processing module is used to perform spatial filtering and smoothing on the temperature distribution data, enhance the temperature gradient change area, and output a clear temperature image.

[0054] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the method for measuring temperature using AI modeling to correct differences in bioluminescence as described in the first aspect of the present invention.

[0055] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the method for measuring temperature using AI modeling to correct differences in bioluminescence as described in the first aspect of the present invention.

[0056] The beneficial effects of this invention are as follows: By inputting multi-band infrared radiation intensity data into a spatiotemporal convolutional neural network and outputting a wavelength-time dependent dynamic emissivity function, physical laws are combined with the AI ​​model, avoiding the physical inconsistency problem that occurs in purely data-driven models. The dynamic emissivity function can accurately reflect the radiation characteristics of biological tissues at different wavelengths and times, breaking through the limitations of the traditional gray body assumption. A differentiable physics engine is used to optimize the dynamic emissivity function, and the neural network parameters are adjusted by minimizing the loss function containing physical constraint terms, which significantly improves the physical rationality of the model and avoids unreasonable outputs produced by black box models. Generative adversarial networks are used to generate pathological tissue emissivity perturbation data, enhancing the model's generalization ability and effectively improving the model's applicability under complex biological conditions such as pathological tissues, thus expanding the application boundaries of the temperature measurement system. Attached Figure Description

[0057] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0058] Figure 1 This is a flowchart of the temperature measurement method using AI modeling to correct differences in the emissivity of organisms in Example 1.

[0059] Figure 2 This is a schematic diagram of the temperature measurement system that uses AI modeling to correct differences in the emissivity of organisms in Example 1. Detailed Implementation

[0060] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0061] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0062] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0063] Example, refer to Figure 1 and Figure 2This embodiment of the invention provides a method for thermometry that uses AI modeling to correct for differences in emissivity in organisms, comprising the following steps:

[0064] S1. Collect multi-band infrared radiation intensity data on the surface of organisms using a multispectral imaging device;

[0065] Furthermore, a multispectral imaging device is used to collect infrared radiation intensity data from the surface of a living organism within multiple preset wavelength ranges. Each wavelength range corresponds to a spectral channel, and the collected infrared radiation intensity data is denoted as... Where, λ i Let represent the i-th wavelength, x and y represent spatial coordinates, and t represent time;

[0066] For the collected Data preprocessing;

[0067] Preprocessing includes removing ambient background radiation, dark current noise, and lens response inhomogeneities, resulting in corrected radiance data.

[0068] It should be noted that the use of multispectral imaging technology to collect multi-band infrared radiation from the surface of organisms breaks through the limitation of the assumption of constant emissivity in traditional single-band or broadband infrared thermometry methods. It can more comprehensively reflect the wavelength dependence of the radiation characteristics of biological tissue surfaces. By acquiring radiation intensity in multiple preset spectral channels and combining spatial and temporal dimensions, a high-dimensional infrared radiation dataset is constructed, providing rich and physically meaningful input features for subsequent AI models. In addition, preprocessing of the raw data to remove environmental background radiation, dark current noise, and lens response inhomogeneity effectively improves data quality and ensures the stability and accuracy of the subsequent modeling process.

[0069] S2. Input multi-band infrared radiation intensity data into a spatiotemporal convolutional neural network and output a wavelength-time dependent dynamic emissivity function. The spatiotemporal convolutional neural network contains a physical constraint layer, which embeds Planck's radiation law as prior knowledge into the network structure.

[0070] Furthermore, a spatiotemporal convolutional neural network is used for the input... The data undergoes feature extraction, and the network structure includes multiple spatiotemporal convolutional layers and physical constraint layers;

[0071] Within the physical constraint layer, Planck's radiation law is embedded as prior physical knowledge, and its expression is:

[0072]

[0073] Among them, B λ(T) represents the radiation intensity of a blackbody at temperature T and wavelength λ, h is Planck's constant, c is the speed of light, and k is Boltzmann's constant;

[0074] Theoretical radiation intensity is calculated using a physical constraint layer. and compared with the measured radiation intensity Compare and generate error terms. The expression is:

[0075]

[0076] The spatiotemporal convolutional neural network adjusts its parameters through backpropagation to minimize the error term, ultimately outputting a dynamic emissivity function ε(λ,t), which has the following form:

[0077]

[0078] The dynamic emissivity function is used to describe the relationship between emissivity and wavelength λ and time t;

[0079] It should be noted that by introducing a spatiotemporal convolutional neural network with a physical constraint layer, the modeling transformation of the emissivity function from a static constant to a dynamic function was achieved. The Planck radiation law was embedded in the physical constraint layer, so that the neural network not only relies on data when learning the emissivity variation law, but is also guided by physical laws, thereby avoiding modeling results that violate physical common sense. The design improves the interpretability and physical consistency of the model, and at the same time, it enables the emissivity function to more accurately reflect the radiation characteristics of biological tissues at different wavelengths and times, providing a key correction basis for subsequent temperature inversion.

[0080] S3. A differentiable physics engine is used to optimize the dynamic emissivity function. The neural network parameters are adjusted by minimizing the loss function containing physical constraints. The physical constraints characterize the gradient difference between the measured radiation intensity and the theoretical radiation intensity.

[0081] Furthermore, the differentiable physics engine optimizes the dynamic emissivity function by including the following steps:

[0082] A differentiable physics engine is used to optimize the dynamic emissivity function ε(λ,t), and a loss function L containing physical constraints is constructed to guide parameter updates.

[0083] The loss function L includes the data fitting term L data and physical constraint term L phys The two parts are expressed as follows:

[0084] L=αL data +βL phys ;

[0085] Where α and β are weighting coefficients;

[0086] Data fitting term L data Defined as the mean square error between the measured radiation intensity and the model-predicted radiation intensity, its expression is:

[0087]

[0088] Physical constraint term L phys Defined as the difference between the measured radiation intensity gradient and the theoretical radiation intensity gradient, expressed as:

[0089]

[0090] The loss function L is minimized by gradient descent, thereby updating the parameters of the spatiotemporal convolutional neural network and making the dynamic emissivity function ε(λ,t) more in line with physical laws.

[0091] It should be noted that by introducing a differentiable physics engine to further optimize the emissivity function, the coupling relationship between the model and physical laws is strengthened. By constructing a composite loss function that includes data fitting terms and physical constraint terms, the model can approximate the measured data while also meeting the gradient consistency requirements in the radiative transfer process. The dual constraint mechanism not only improves the model's generalization ability but also enhances its robustness to abnormal data and noise. In addition, since the loss function is differentiable, the optimization process is more efficient and stable, which helps to obtain a more physically reasonable and mathematically convergent dynamic emissivity function.

[0092] S4. Generate pathological tissue emissivity perturbation data using generative adversarial networks, and add the perturbation data to the training dataset to enhance the model's generalization ability.

[0093] Furthermore, a generative adversarial network (GAN) is constructed, in which a generator is used to generate emissivity perturbation data simulating pathological tissues, and a discriminator is used to distinguish between real data and generated data.

[0094] The generator takes a random noise vector z as input and outputs emissivity perturbation data δε(λ,t), expressed as:

[0095] δε(λ,T)=G(z);

[0096] Where G represents the generator function;

[0097] The discriminator receives the true emissivity data ε(λ,t) or the generated perturbation data δε(λ,t) and outputs the probability that it is the true data.

[0098] Through adversarial training, the perturbation data generated by the generator can approximate the emissivity variation characteristics of real pathological tissues.

[0099] The generated perturbation data δε(λ,t) is added to the original emissivity data to obtain the enhanced training data:

[0100] ε aug (λ,t)=ε(λ,t)+δε(λ,t);

[0101] The enhanced data is added to the training set to improve the model's adaptability to pathological tissues;

[0102] It should be noted that by constructing a Generative Adversarial Network (GAN) to generate emissivity perturbation data of pathological tissues, the problem of scarce pathological samples and uneven data distribution in actual training is solved. The generator can learn and simulate the changing characteristics of emissivity of real pathological tissues through adversarial training, thereby generating representative perturbation data. Adding this perturbation data to the training set not only expands the data diversity but also improves the model's adaptability to special tissue states. The method significantly enhances the applicability and robustness of the system in the clinical environment, enabling the temperature measurement system to maintain high accuracy and stability when facing pathological states.

[0103] S5. Substitute the optimized dynamic emissivity function into the temperature inversion equation and perform iterative calculations to obtain the corrected surface temperature distribution of the organism.

[0104] Furthermore, the optimized dynamic emissivity function ε(λ,t) is substituted into the temperature inversion equation to calculate the temperature distribution on the surface of the organism.

[0105] The temperature inversion equation, based on Planck's law and emissivity correction, is expressed as follows:

[0106]

[0107] Wherein, the nonlinear mapping function f represents the nonlinear mapping relationship from the corrected radiation intensity to obtain the temperature, T(x,y,t) represents the temperature of the organism's surface at spatial coordinates (x,y) and time t, and I′ λ (x,y,t) represents the corrected infrared radiation intensity data at wavelength λ, spatial coordinates (x,y), and time t.

[0108] An iterative algorithm is used to optimize the temperature. The initial temperature value T0 is obtained by inversion under the gray body assumption when the emissivity is constant ε0, and the expression is:

[0109]

[0110] Subsequently, T0 is substituted into the dynamic emissivity function ε(λ,t) to calculate the new corrected radiation intensity, expressed as:

[0111] I corr=ε(λ,T0,t)B λ (T0);

[0112] By minimizing I corr Compared with the measured I′ λ The error between (x,y,t) is used to iteratively update the temperature value until the convergence condition is met, and the corrected surface temperature distribution of the organism T(x,y,t) is obtained.

[0113] Spatial filtering and smoothing of temperature distribution data specifically includes the following steps:

[0114] The obtained temperature distribution data T(x,y,t) is spatially filtered using a two-dimensional Gaussian filter to remove local noise. The kernel function expression is as follows:

[0115]

[0116] Where σ is the standard deviation of the Gaussian kernel, G(x,y) represents the value of the two-dimensional Gaussian filter at the spatial coordinates (x,y), and x and v represent the two dimensions in the spatial coordinate system, respectively. The normalization coefficients of the Gaussian function are... This represents the exponential decay characteristic of the filter weights with respect to the square of the distance;

[0117] A Gaussian filter is convolved with the temperature distribution data to obtain the smoothed temperature distribution T. smooth (x,y,t);

[0118] Edge enhancement processing is performed on the smoothed temperature data, and the Sobel operator is used to detect temperature gradient change regions and enhance abnormal regions in the temperature distribution.

[0119] Output the corrected image of the organism's surface temperature;

[0120] It should be noted that by substituting the optimized dynamic emissivity function into the temperature inversion equation based on Planck's law, and combining iterative optimization strategies, a closed-loop calculation process from emissivity correction to temperature inversion is realized. This method breaks through the limitation of the assumption of constant emissivity in traditional temperature measurement, thus improving the accuracy of temperature inversion. At the same time, the introduction of spatial filtering and smoothing techniques, such as Gaussian filtering and Sobel edge enhancement, further enhances the visual effect and diagnostic value of the temperature image. The output corrected temperature image not only has higher spatial resolution but also retains detailed information in the temperature anomaly region, providing reliable data support for medical diagnosis and thermal monitoring.

[0121] This embodiment also provides a temperature measurement system that uses AI modeling to correct for differences in emissivity in biological organisms, including:

[0122] The system includes a data acquisition module, a model training module, an emissivity optimization module, a temperature inversion module, and an image processing module.

[0123] The data acquisition module is used to acquire multi-band infrared radiation intensity data of the surface of organisms through a multispectral imaging device, and to preprocess the data to remove noise and background radiation.

[0124] The model training module is used to input preprocessed data into the spatiotemporal convolutional neural network, combine the output wavelength-time dependent dynamic emissivity function of the physical constraint layer, and use generative adversarial networks to enhance the training set.

[0125] The emissivity optimization module is used to optimize the dynamic emissivity function using a differentiable physics engine. It adjusts the parameters by minimizing the loss function to make it conform to physical laws.

[0126] The temperature inversion module is used to substitute the optimized dynamic emissivity function into the temperature inversion equation for iterative calculation to obtain the corrected surface temperature distribution of the organism.

[0127] The image processing module is used to perform spatial filtering and smoothing on temperature distribution data, enhance the temperature gradient change area, and output a clear temperature image.

[0128] This embodiment also provides a computer device suitable for the use of AI modeling to correct biological emissivity differences in temperature measurement, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the use of AI modeling to correct biological emissivity differences in temperature measurement as proposed in the above embodiment.

[0129] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0130] This embodiment also provides a storage medium storing a computer program. When executed by a processor, the program implements the temperature measurement method for correcting biological emissivity differences using AI modeling, as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0131] In summary, this invention combines physical laws with an AI model by inputting multi-band infrared radiation intensity data into a spatiotemporal convolutional neural network and outputting a wavelength-time dependent dynamic emissivity function. This avoids the physical inconsistencies that arise in purely data-driven models. The dynamic emissivity function accurately reflects the radiation characteristics of biological tissues at different wavelengths and times, overcoming the limitations of the traditional gray-body assumption. A differentiable physics engine is used to optimize the dynamic emissivity function, and the neural network parameters are adjusted by minimizing the loss function containing physical constraints, significantly improving the model's physical rationality and avoiding unreasonable outputs from black-box models. Furthermore, generative adversarial networks are used to generate pathological tissue emissivity perturbation data, enhancing the model's generalization ability and effectively improving its applicability under complex biological conditions such as pathological tissues, thus expanding the application boundaries of temperature measurement systems.

[0132] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A thermometry method using AI modeling to correct for differences in emissivity in organisms, characterized in that: include: Multi-band infrared radiation intensity data of the surface of organisms are collected using a multispectral imaging device. The multi-band infrared radiation intensity data is input into a spatiotemporal convolutional neural network, which outputs a wavelength-time dependent dynamic emissivity function. The spatiotemporal convolutional neural network includes a physical constraint layer, which embeds Planck's radiation law as prior knowledge into the network structure. The dynamic emissivity function is optimized using a differentiable physics engine. The neural network parameters are adjusted by minimizing a loss function that includes a physical constraint term, which represents the gradient difference between the measured radiation intensity and the theoretical radiation intensity. Generative adversarial networks are used to generate perturbation data of pathological tissue emissivity, and the perturbation data is added to the training dataset to enhance the generalization ability of the model. The optimized dynamic emissivity function is substituted into the temperature inversion equation for iterative calculation to obtain the corrected surface temperature distribution of the organism.

2. The method for measuring temperature using AI modeling to correct differences in bioemissivity as described in claim 1, characterized in that: The multispectral imaging device acquires multi-band infrared radiation intensity data of the surface of organisms, specifically including the following steps: A multispectral imaging device was used to collect infrared radiation intensity data from the surface of a living organism within multiple preset wavelength ranges. Each wavelength range corresponds to one spectral channel, and the collected infrared radiation intensity data is denoted as follows: Where, λ i Let represent the i-th wavelength, x and y represent spatial coordinates, and t represent time; For the collected Data preprocessing; The preprocessing includes removing ambient background radiation, dark current noise, and lens response inhomogeneities to obtain corrected radiation intensity data.

3. The thermometry method using AI modeling to correct differences in bioemissivity as described in claim 2, characterized in that: The output wavelength-time dependent dynamic emissivity function of the spatiotemporal convolutional neural network specifically includes the following steps: Using a spatiotemporal convolutional neural network for the input The data undergoes feature extraction, and the network structure includes multiple spatiotemporal convolutional layers and physical constraint layers; In the physical constraint layer, Planck's radiation law is embedded as prior physical knowledge; Theoretical radiation intensity is calculated using a physical constraint layer. and compared with the measured radiation intensity Compare and generate error terms. The spatiotemporal convolutional neural network adjusts the network parameters through the backpropagation algorithm to minimize the error term and finally outputs the dynamic emissivity function ε(λ,t); The dynamic emissivity function is used to describe the relationship between emissivity and wavelength λ and time t.

4. The thermometry method using AI modeling to correct differences in bioemissivity as described in claim 3, characterized in that: The differentiable physics engine optimizes the dynamic emissivity function by including the following steps: A differentiable physics engine is used to optimize the dynamic emissivity function ε(λ,t), and a loss function L containing physical constraints is constructed to guide parameter updates. The loss function L includes the data fitting term L data and physical constraint term L phys Two parts; The data fitting term L data Defined as the mean square error between the measured radiation intensity and the model-predicted radiation intensity; The physical constraint term L phys Defined as the difference between the measured radiation intensity gradient and the theoretical radiation intensity gradient. By minimizing the loss function L using gradient descent, the parameters of the spatiotemporal convolutional neural network are updated, making the dynamic emissivity function ε(λ,t) more consistent with physical laws.

5. The thermometry method using AI modeling to correct differences in bioemissivity as described in claim 4, characterized in that: The generative adversarial network generates pathological tissue emissivity perturbation data specifically through the following steps: A generative adversarial network (GAN) is constructed, in which the generator is used to generate emissivity perturbation data simulating pathological tissues, and the discriminator is used to distinguish between real data and generated data. The generator takes a random noise vector z as input and outputs emissivity perturbation data δε(λ,t). The discriminator receives the true emissivity data ε(λ,t) or the generated perturbation data δε(λ,t) and outputs the probability that it is the true data. Through adversarial training, the perturbation data generated by the generator approximates the emissivity variation characteristics of real pathological tissues. The generated perturbation data δε(λ,t) is added to the original emissivity data to obtain the enhanced training data; The enhanced data is added to the training set to improve the model's ability to adapt to pathological tissues.

6. The thermometry method using AI modeling to correct differences in bioemissivity as described in claim 5, characterized in that: The iterative calculation of substituting the dynamic emissivity function into the temperature inversion equation specifically includes the following steps: The optimized dynamic emissivity function ε(λ,t) is substituted into the temperature inversion equation to calculate the temperature distribution on the surface of the organism. The temperature inversion equation is based on Planck's law and emissivity correction; An iterative algorithm is used to optimize the temperature. The initial temperature value T0 is obtained by inversion when the emissivity is constant ε0 under the gray body assumption. Subsequently, T0 is substituted into the dynamic emissivity function ε(λ,t) to calculate the new corrected radiation intensity; By minimizing I corr Compared with actual measurement The error between the values ​​is used to iteratively update the temperature value until the convergence condition is met, thus obtaining the corrected surface temperature distribution T(x,y,t) of the organism.

7. The thermometry method using AI modeling to correct differences in bioemissivity as described in claim 6, characterized in that: The spatial filtering and smoothing of the temperature distribution data specifically includes the following steps: The obtained temperature distribution data T(x,y,t) is spatially filtered, and a two-dimensional Gaussian filter is used to remove local noise. A Gaussian filter is convolved with the temperature distribution data to obtain the smoothed temperature distribution T. smooth (x,y,t); Edge enhancement processing is performed on the smoothed temperature data, and the Sobel operator is used to detect temperature gradient change regions and enhance abnormal regions in the temperature distribution. Output the corrected image of the organism's surface temperature.

8. A thermometry system for correcting biological emissivity differences using AI modeling, based on the thermometry method for correcting biological emissivity differences using AI modeling as described in any one of claims 1 to 7, characterized in that: include: The system includes a data acquisition module, a model training module, an emissivity optimization module, a temperature inversion module, and an image processing module. The data acquisition module is used to acquire multi-band infrared radiation intensity data of the surface of organisms through a multispectral imaging device, and to preprocess the data to remove noise and background radiation. The model training module is used to input the preprocessed data into the spatiotemporal convolutional neural network, combine the output wavelength-time dependent dynamic emissivity function of the physical constraint layer, and enhance the training set using a generative adversarial network. The emissivity optimization module is used to optimize the dynamic emissivity function using a differentiable physics engine, and adjusts the parameters by minimizing the loss function to make it conform to physical laws. The temperature inversion module is used to substitute the optimized dynamic emissivity function into the temperature inversion equation for iterative calculation to obtain the corrected surface temperature distribution of the organism. The image processing module is used to perform spatial filtering and smoothing on the temperature distribution data, enhance the temperature gradient change area, and output a clear temperature image.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the temperature measurement method using AI modeling to correct differences in bioluminescence as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the temperature measurement method using AI modeling to correct differences in bioluminescence as described in any one of claims 1 to 7.

Images