Infrared thermal imaging-based data feature extraction and analysis method and system
By combining infrared thermal imaging technology with machine learning, the difference between body surface temperature and core body temperature is quantified. Taking into account environmental and individual characteristics, a core body temperature prediction model is established, which solves the problems of accuracy and applicability of infrared thermal imaging technology in body temperature detection and achieves efficient body temperature prediction under different environments and individuals.
Patent Information
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- ZHEJIANG LAB
- Filing Date
- 2025-06-04
- Publication Date
- 2026-07-09
AI Technical Summary
Existing infrared thermal imaging technology cannot accurately quantify the difference between body surface temperature and core body temperature in body temperature detection, and it fails to effectively consider the influence of ambient temperature and individual characteristics, which limits its application environment and applicability.
By establishing a data feature extraction and analysis method based on infrared thermal imaging, user data is acquired and a core body temperature prediction model is built using machine learning. Considering facial temperature distribution, personal characteristics, and environmental characteristics, the model is fitted with fitting coefficients using coupled feature expressions and adaptively adjusted to improve prediction accuracy.
The accuracy and individual adaptability of core body temperature prediction have been improved under different ambient temperatures, meeting the needs of rapid monitoring of multiple people and long-term monitoring of a single person, and realizing rapid and real-time body temperature prediction.
Smart Images
Figure CN2025098932_09072026_PF_FP_ABST
Abstract
Description
A Data Feature Extraction and Analysis Method and System Based on Infrared Thermal Imaging Technical Field
[0001] This invention relates to the field of body temperature monitoring, and specifically to a data feature extraction and analysis method and system based on infrared thermal imaging. Background Technology
[0002] Non-contact core body temperature detection methods, such as infrared thermal imaging, have played a crucial role in the rapid screening and control of some infectious diseases, particularly in densely populated public places like hospitals, schools, and airports. However, non-contact detection of core body temperature often suffers from low accuracy and susceptibility to environmental influences. Since infrared thermal imaging infers core body temperature solely from surface temperature, the unquantifiable difference between surface and core temperatures is a key reason why it struggles to accurately predict core body temperature. This difference is influenced by complex factors, closely related to environmental factors, individual characteristics, and personal condition. Current standards recommend a limited application temperature range of only 20-24°C for infrared thermal imaging, significantly restricting its application scenarios.
[0003] Existing technologies have been researched and improved to address the above problems; however, they generally suffer from the following shortcomings:
[0004] (1) In the process of detecting core body temperature by infrared thermal imaging, the difference between body surface temperature and core body temperature was not accurately quantified, and the role of ambient temperature in this process was ignored (coupled effect between ambient temperature and body surface temperature), which limited the application range of ambient temperature for infrared thermal imaging body temperature monitoring.
[0005] (2) Existing technologies ignore the influence of individual differences on the surface-core body temperature pattern and do not take into account the errors caused by different individual characteristics in the process of calculating core body temperature, which will affect the population adaptability of infrared thermal imaging body temperature monitoring.
[0006] Therefore, there is an urgent need to propose effective solutions to the above problems. Summary of the Invention
[0007] The purpose of this invention is to address the shortcomings of existing technologies by proposing a data feature extraction and analysis method and system based on infrared thermal imaging.
[0008] To achieve the above objectives, the present invention provides the following technical solution: In a first aspect, the present invention provides a data feature extraction and analysis method based on infrared thermal imaging, the method comprising the following steps:
[0009] Acquire user data, including infrared thermal imaging facial temperature distribution characteristics, personal characteristics, environmental characteristics, and actual core body temperature, and establish a database; determine whether it is a multi-person rapid monitoring scenario or a single-person long-term monitoring scenario;
[0010] For scenarios involving rapid monitoring of multiple users, after establishing a database, all user data in the database is used as the dataset. A fitting coefficient K is obtained by fitting the data using a first coupling feature expression. The fitting coefficient K is then substituted into a second coupling feature expression to calculate the coupling features of all users in the database. Based on all user data in the database as the training dataset, a core body temperature prediction model is established using coupling features, facial temperature distribution features, personal features, and environmental features as inputs, and real core body temperature as the label. The model is trained using machine learning methods. The infrared thermal imaging facial temperature distribution features, personal features, and environmental features of the target user are obtained. Based on the second coupling feature expression after substituting the fitting coefficient K, the target user's coupling features are obtained and input into the core body temperature prediction model to predict the target user's core body temperature. This process is repeated multiple times to obtain the core body temperatures of multiple target users.
[0011] For long-term monitoring of a single individual, after establishing a database, the infrared thermal imaging facial temperature distribution characteristics, personal characteristics, and environmental characteristics of the target user are obtained. User data with similar personal characteristics to the target user are extracted from the database and used as a dataset. A fitting coefficient K is obtained by fitting the target user's data using a first coupling feature expression. The fitting coefficient K is then substituted into a second coupling feature expression to calculate the coupling features between all users in the database and the target user. Based on the similarity to the target user's personal characteristics, weights are assigned to all users in the database, and adaptive adjustments are made based on these weights. This dataset serves as a training dataset, using coupling features, facial temperature distribution characteristics, personal characteristics, and environmental characteristics as inputs, and real core body temperature as the label. Machine learning methods are used to train and establish a personalized core body temperature prediction model. The infrared thermal imaging facial temperature distribution characteristics, personal characteristics, environmental characteristics, and coupling features of the target user are then input into the personalized core body temperature prediction model to predict the target user's core body temperature.
[0012] The first coupling characteristic expression is: T c =(K+1)T max -KT a The second coupling characteristic expression is: T op =(K+1)T max -KT a , among which, T max The maximum temperature of the facial ROI, T a For ambient temperature, T c For true core body temperature, T op K represents the coupling characteristic, and K is the fitting coefficient.
[0013] Furthermore, the infrared thermal imaging facial temperature distribution characteristics include the maximum temperature T of the facial ROI. max Maximum temperature T of forehead ROI FEmax Maximum temperature T at the inner canthus of the eye (ROI) CEmax Maximum temperature T at the mouth ROI Mmax ;
[0014] The personal characteristics include gender, age, race, BMI, and body fat percentage; where gender and race are assigned values {0,1,…,n} by category, and age, BMI, and body fat percentage are assigned values {0,1,…,n} by numerical range.
[0015] The environmental characteristics include ambient temperature T. a .
[0016] Furthermore, the infrared thermal imaging facial temperature distribution features of users and target users in the database are obtained by extracting faces, dividing facial regions, and calculating temperature values from the users' infrared thermal imaging images.
[0017] The personal characteristics of users in the database are obtained through manual collection, while the personal characteristics of target users are obtained through manual collection or predicted using machine learning methods based on the user's infrared thermal images.
[0018] The environmental characteristics of users and target users in the database are collected through auxiliary environmental parameter acquisition devices;
[0019] The database contains users' actual core body temperature (T). c It was measured using a medical thermometer.
[0020] In some specific embodiments, the method further includes: performing feature selection on features in the training dataset to obtain a key feature set, training the set using machine learning methods, and using the key feature set of the target user as input features of the core body temperature prediction model or the personalized core body temperature prediction model to predict the core body temperature of the target user; wherein the key feature set includes at least coupling features.
[0021] Preferably, the feature selection method includes: Pearson correlation coefficient and recursive feature elimination.
[0022] Preferably, the machine learning methods include: random forest, neural network, and support vector machine.
[0023] Furthermore, in a single-person long-term monitoring scenario, the step of extracting user data with similar personal characteristics from the database through the target user's personal characteristics as a dataset includes: calculating the similarity between the personal characteristics of all users in the database and the target user, setting a threshold, and including user data with similarity greater than the threshold into the dataset;
[0024] The step of assigning weights to all users in the database based on their similarity to the target user's personal characteristics, and adaptively adjusting the weights to form a training dataset, includes: normalizing personal characteristics, calculating the similarity between the personal characteristics of all users in the database and the target user, assigning weights to all users in the database based on the similarity so that users with relatively high similarity have relatively high weights, and resampling according to the weights to obtain the training dataset.
[0025] The similarity calculation methods include: weighted Euclidean distance, Manhattan distance, Mahalanobis distance, Langevin distance, and Minkowski distance.
[0026] In some specific embodiments, the method further includes: after predicting the target user's core body temperature, measuring the target user's actual core body temperature using a medical thermometer, and incorporating the target user's infrared thermal imaging facial temperature distribution characteristics, personal characteristics, environmental characteristics, and actual core body temperature into a database for updating.
[0027] In a second aspect, the present invention provides a core body temperature prediction system based on infrared thermal imaging, the system comprising:
[0028] The data acquisition unit includes a personal feature acquisition module, a facial temperature data acquisition module, an environmental data acquisition module, and a core body temperature acquisition module. These modules are used to acquire personal features, infrared thermal imaging facial temperature distribution features, environmental features, and real core body temperature data measured by a medical thermometer, respectively, and to establish a database. In a single-person long-term monitoring scenario, the data in the database is input into the first similarity measurement module, and in a multi-person rapid monitoring scenario, it is input into the coupling feature calculation module.
[0029] The feature extraction unit includes a first similarity measurement module and a coupled feature calculation module. The first similarity measurement module is used to calculate the similarity between the target user and the personal features of users in the database in a single-person long-term monitoring scenario, set a threshold, and select user data with similarity greater than the threshold to input into the coupled feature calculation module to calculate coupled features. The coupled feature calculation module is used to calculate coupled features based on facial temperature distribution features and environmental features. The data processed by the feature extraction unit is used as a training dataset to input into the model building unit in a multi-person rapid monitoring scenario, and as an input data adaptive adjustment unit in a single-person long-term monitoring scenario.
[0030] The data adaptive adjustment unit is used to receive the data processed by the feature extraction unit in a single-person long-term monitoring scenario, perform data adaptive adjustment, and obtain a training dataset. The data adaptive adjustment unit includes a second similarity measurement module and a data resampling module. The second similarity measurement module evaluates the similarity between the target user and the personal features of users in the database. The data resampling module resamples the data according to the similarity to obtain the training dataset. Users with relatively high similarity have relatively high weights.
[0031] The model building unit is used to train a core body temperature prediction model based on a training dataset using machine learning methods. The training dataset consists of personal characteristics, facial temperature distribution characteristics, environmental characteristics, coupling characteristics, and real core body temperature.
[0032] The model application unit includes a core body temperature calculation module, which calculates the predicted core body temperature based on the target user's personal characteristics, facial temperature distribution characteristics, environmental characteristics, and coupling characteristics according to the core body temperature prediction model.
[0033] As a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is used to implement the above-described data feature extraction and analysis method based on infrared thermal imaging.
[0034] Compared with the prior art, the present invention has at least the following beneficial effects:
[0035] (1) The facial temperature-ambient temperature coupling feature proposed in this invention is an important feature obtained after analyzing the human body-environment heat transfer mechanism. It quantifies the difference between the body surface temperature and the core body temperature affected by the ambient temperature, thereby improving the accuracy of calculating the core body temperature using infrared thermal imaging facial temperature under different ambient temperatures and improving the environmental adaptability of infrared thermal imaging body temperature monitoring technology.
[0036] (2) The personalized core body temperature prediction model proposed in this invention can adaptively adjust the model structure to obtain the prediction model that best matches the target user. It takes into account the differences between different individuals and can establish the optimal prediction model for different individuals, thereby improving the individual adaptability of infrared thermal imaging body temperature monitoring technology.
[0037] (3) This invention can meet the needs of different monitoring scenarios. For long-term monitoring of a single person, it provides the most suitable coupling features and personalized models for the target user to ensure the most accurate body temperature prediction results. For rapid monitoring of multiple people, it can achieve rapid and real-time body temperature prediction while ensuring accuracy. Attached Figure Description
[0038] Figure 1 is a flowchart of the data feature extraction and analysis method based on infrared thermal imaging proposed in this invention;
[0039] Figure 2 is a structural block diagram of the core body temperature prediction system based on infrared thermal imaging in a multi-person rapid monitoring scenario proposed in this invention.
[0040] Figure 3 is a structural block diagram of the core body temperature prediction system based on infrared thermal imaging in the single-person long-term monitoring scenario proposed in this invention. Detailed Implementation
[0041] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings.
[0042] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The singular forms “a,” “the,” and “the” used in this invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.
[0043] This invention proposes a data feature extraction and analysis method based on infrared thermal imaging, including both rapid multi-person monitoring and long-term single-person monitoring scenarios. In rapid multi-person monitoring, the time spent retraining the model is unnecessary; the key is to quickly obtain prediction results. Therefore, a universal prediction model can be used to obtain prediction results quickly and in real-time while ensuring a certain level of accuracy. For long-term single-person monitoring, a personalized core body temperature prediction model is established based on individual characteristics. After initially obtaining the individual's characteristic data, a certain amount of time is needed to complete the adaptive adjustment of the model training dataset and the model training process. Once the model is trained, it becomes the optimal prediction model for that target user, improving the accuracy of the entire monitoring process.
[0044] Figure 1 shows a flowchart of the data feature extraction and analysis method based on infrared thermal imaging proposed in this invention, which specifically includes the following steps:
[0045] S1. Establish a user database, which includes user data such as infrared thermal imaging facial temperature distribution characteristics, personal characteristics, environmental characteristics, and real core body temperature data measured by medical thermometers; obtain the facial temperature-ambient temperature coupling feature expression of the coefficients to be fitted based on the heat transfer mechanism.
[0046] S2. Obtain the monitoring scenario requirements. If it is a multi-person rapid monitoring scenario, skip to step S3 and end after completing step S7; if it is a single-person long-term monitoring scenario, skip to step S8.
[0047] S3. Using the first coupling feature expression, the fitting coefficient K of the coupling feature expression is obtained by fitting all user data in the database; based on the second coupling feature expression, the coupling features of users in the database are calculated according to facial temperature distribution features and environmental features.
[0048] S4. Based on machine learning algorithms, use all user data in the database as the training dataset to establish a core body temperature prediction model; the input to the model training is all the features in steps S1 and S3 of the dataset, and the output of the model training is the actual core body temperature in the dataset.
[0049] In some embodiments, during the model building process, feature selection can be performed on the input features to remove features that have little impact on core body temperature prediction, and retain important features to form a key feature set as the final input features for model training.
[0050] S5. Obtain the infrared thermal imaging facial temperature distribution characteristics, personal characteristics and environmental characteristics of the target user. Based on the second coupling feature expression, substitute the fitting coefficient K obtained in step S3, and calculate the coupling characteristics of the target user according to the facial temperature distribution characteristics and environmental characteristics.
[0051] S6. Obtain the infrared thermal imaging facial temperature distribution features, personal features, environmental features, and coupling features of the target user in step S4 as model input features, and predict the core body temperature of the target user based on the established core body temperature prediction model.
[0052] In some embodiments, the model input features are the key feature set obtained through feature selection in step S4.
[0053] S7. Repeat steps S5 and S6 to calculate the core body temperature of different target users;
[0054] S8. Obtain the infrared thermal imaging facial temperature distribution characteristics, personal characteristics, and environmental characteristics of the target user;
[0055] S9. Extract user data with similar personal characteristics from the database based on the target user's personal characteristics. Based on the first coupling feature expression, fit the fitting coefficient K of the target user's coupling feature expression. Based on the second coupling feature expression, calculate the target user's coupling features and the coupling features of users in the database based on facial temperature distribution features and environmental features.
[0056] S10. Based on the similarity of personal characteristics between the target user and the users in the database, different weights are assigned to the samples in the database, and the data distribution in the database is adaptively adjusted to form a training dataset. A personalized core body temperature prediction model is established based on the machine learning algorithm and the adjusted training dataset. The input to the model training is all the features in steps S1 and S9 in the dataset, and the output of the model training is the real core body temperature in the dataset.
[0057] In some embodiments, during the model building process, feature selection can be performed on the input features to remove features that have little impact on core body temperature prediction, and retain important features to form a key feature set as the final input features for model training.
[0058] S11. Obtain the infrared thermal imaging facial temperature distribution characteristics, personal characteristics, environmental characteristics, and coupling characteristics of the target user as model input features, and predict the core body temperature of the target user based on the established personalized core body temperature prediction model.
[0059] In some embodiments, the model input features are the key feature set obtained through feature selection in step S10.
[0060] In some embodiments, S12 is also included, which can selectively use a medical thermometer to measure the target user's true core body temperature and save all data of the target user to a database.
[0061] Repeat step S11, continuously inputting the updated characteristics of the target user, and calculate the target user's core body temperature in real time.
[0062] In step S1, the infrared thermal imaging facial temperature distribution features include the maximum temperature (T) of the facial ROI (Region of Interest). max ), maximum temperature of forehead ROI (T) FEmax ), Maximum temperature of the inner canthus region of the eye (T) CEmax ), Maximum ROI temperature at the mouth (T) Mmax The facial temperature distribution features are obtained by extracting faces from infrared thermal imaging images, dividing facial regions, and calculating temperature values before inputting them into the database. The personal features include gender, age, ethnicity, BMI (Body Mass Index), and body fat percentage, and are converted to a format readable by the model. Gender and ethnicity are assigned values {0,1,…,n} by category, and age, BMI, and body fat percentage are assigned values within ranges {0,1,…,n}. These personal features are collected manually and then input into the database. The environmental features include ambient temperature (T). a The actual core body temperature data (T) is collected by auxiliary environmental parameter acquisition equipment and then input into the database. cThe temperature is measured by a medical thermometer and then entered into the database.
[0063] In step S1, the user data in the database includes facial temperature distribution characteristics, personal characteristics, ambient temperature, and actual core body temperature. During the database establishment phase, a large amount of actual core body temperature data of users measured by medical thermometers is recorded and stored for model training and coefficient fitting of coupled features.
[0064] In steps S1, S3, and S9, the second coupling feature (T) op The expression is T op =(K+1)T max -KT a , among which, T max The maximum temperature of the facial ROI, T a The ambient temperature is used. The coefficients K to be fitted for the coupling feature are obtained by fitting the following equation (the expression for the first coupling feature) using the least squares method or other algorithms: T c =(K+1)T max -KT a , among which, T max The maximum temperature of the facial ROI, T a For ambient temperature, T c This represents the actual core body temperature. In scenarios involving rapid monitoring of multiple users, to improve efficiency and save model training time, the K value is set to the value of T for all users in the database. c T max T a The data fitting results allow for the calculation of coupling features for all users in the database before acquiring target user data, enabling the training of the core body temperature prediction model. In long-term single-person monitoring scenarios, the fitting coefficient K of the coupling feature is based on T, a set of users with similar personal characteristics in the database. c T max T a The data is fitted to ensure that the resulting coupling features are most accurately adapted to the target individual.
[0065] In steps S4 and S10, the key feature set includes at least coupling features. The most relevant and discriminative key feature set can be selected using feature selection methods such as Pearson correlation coefficient and recursive feature elimination. The machine learning algorithm can be one capable of fitting various complex nonlinear mappings, including random forests, neural networks, and support vector machines.
[0066] In steps S6 and S11, the infrared thermal imaging facial temperature distribution characteristics, personal characteristics, environmental characteristics, and coupling characteristics of the target user are obtained through real-time processing of the infrared thermal image and auxiliary environmental parameter acquisition equipment. In multi-person rapid monitoring scenarios, personal characteristics may be impossible to collect manually; therefore, the target user's personal characteristics (such as gender and age) can be predicted using infrared thermal images and machine learning algorithms. In single-person monitoring scenarios, the target user's personal characteristic data can be collected manually in advance and input into the model; other data are obtained through real-time processing of the infrared thermal image and auxiliary environmental parameter acquisition equipment.
[0067] In step S9, user data with similar personal characteristics are extracted from the database based on the target user's personal characteristics, the similarity between all users in the database and the target user's personal characteristics is calculated, a threshold is set, and user data with similarity greater than the threshold is included in the dataset.
[0068] In step S10, the calculation of personal feature similarity is obtained by normalizing features such as gender, age, race, BMI, and body fat percentage, and then using weighted Euclidean distance, Manhattan distance, Mahalanobis distance, Langevsky distance, Minkowski distance, or other similarity calculation methods. Based on the similarity calculation results, different weights are assigned to the data of different users in the database, so that users with similar features have greater weights. The data is then resampled according to the weights to form the training dataset of the model.
[0069] This invention also provides a core body temperature prediction system based on infrared thermal imaging, comprising the following modules: a data acquisition unit, including a personal feature acquisition module, a facial temperature data acquisition module, an environmental data acquisition module, and a core body temperature acquisition module, used to acquire personal features, infrared thermal imaging facial temperature distribution features, environmental features, and real core body temperature data measured by a medical thermometer, and to establish a database; in a single-person long-term monitoring scenario, the data in the database is input into a first similarity measurement module, and in a multi-person rapid monitoring scenario, it is input into a coupling feature calculation module;
[0070] The feature extraction unit includes a first similarity measurement module and a coupled feature calculation module. The first similarity measurement module is used to calculate the similarity between the target user and the personal features of users in the database in a single-person long-term monitoring scenario, set a threshold, and select user data with similarity greater than the threshold to input into the coupled feature calculation module to calculate coupled features. The coupled feature calculation module is used to calculate coupled features based on facial temperature distribution features and environmental features. The data processed by the feature extraction unit is used as a training dataset to input into the model building unit in a multi-person rapid monitoring scenario, and as an input data adaptive adjustment unit in a single-person long-term monitoring scenario.
[0071] The data adaptive adjustment unit is used to receive the data processed by the feature extraction unit in a single-person long-term monitoring scenario, perform data adaptive adjustment, and obtain a training dataset. The data adaptive adjustment unit includes a second similarity measurement module and a data resampling module. The second similarity measurement module evaluates the similarity between the target user and the personal features of users in the database. The data resampling module resamples the data according to the similarity to obtain the training dataset, so that users with relatively high similarity have relatively high weights.
[0072] The model building unit is used to train a core body temperature prediction model based on a training dataset using machine learning methods. The training dataset consists of personal characteristics, facial temperature distribution characteristics, environmental characteristics, coupling characteristics, and real core body temperature.
[0073] The model application unit includes a core body temperature calculation module, which calculates the predicted core body temperature based on the target user's personal characteristics, facial temperature distribution characteristics, environmental characteristics, and coupling characteristics according to the core body temperature prediction model.
[0074] In some embodiments, modules can be added or removed according to different monitoring scenario requirements to provide a corresponding system structure.
[0075] (1) In the scenario of rapid monitoring of multiple people, as shown in Figure 2, the core body temperature prediction system based on infrared thermal imaging includes a data acquisition unit, a feature extraction unit, a model building unit, and a model application unit.
[0076] The data acquisition unit includes a personal feature acquisition module, a facial temperature data acquisition module, an environmental data acquisition module, and a core body temperature acquisition module, which are used to acquire personal features, infrared thermal imaging facial temperature distribution features, environmental features, and real core body temperature data measured by a medical thermometer, respectively.
[0077] The feature extraction unit includes a coupled feature calculation module, which calculates facial temperature-ambient temperature coupled features based on facial temperature distribution features and environmental features.
[0078] The model building unit includes a training dataset, a machine learning algorithm, and a core body temperature prediction model. The training dataset consists of personal features, facial temperature distribution features, environmental features, coupling features, and core body temperature. The core body temperature prediction model is built based on the machine learning algorithm and the training dataset.
[0079] The model application unit includes a core body temperature calculation module, which calculates the predicted core body temperature based on the core body temperature prediction model according to the personal characteristics, facial temperature distribution characteristics, environmental characteristics, and coupling characteristics of multiple users.
[0080] (2) In the scenario of long-term monitoring of a single person, as shown in Figure 3, the core body temperature prediction system based on infrared thermal imaging includes a data acquisition unit, a feature extraction unit, a data adaptive adjustment unit, a model building unit, and a model application unit.
[0081] The data acquisition unit includes a personal feature acquisition module, a facial temperature data acquisition module, an environmental data acquisition module, and a core body temperature acquisition module, which are used to acquire personal features, infrared thermal imaging facial temperature distribution features, environmental features, and real core body temperature data measured by a medical thermometer, respectively.
[0082] The feature extraction unit includes a similarity measurement algorithm and a coupling feature calculation module. The similarity measurement algorithm is used to evaluate the similarity between the target user and the personal features of users in the database, and selects user data with high similarity to fit the coefficients of the coupling feature expression of the target user. The coupling feature calculation module calculates the facial temperature-ambient temperature coupling feature based on facial temperature distribution features and environmental features.
[0083] The data adaptive adjustment unit includes a similarity measurement algorithm and a data resampling algorithm. The similarity measurement algorithm is used to evaluate the similarity between the target user and the personal characteristics of users in the database. The data resampling algorithm resamples the data based on the similarity to obtain the training dataset, and user data with high similarity has higher weight.
[0084] The model building unit includes a training dataset, a machine learning algorithm, and a personalized core body temperature prediction model. The training dataset is obtained after being processed by the data adaptive adjustment unit and consists of personal characteristics, facial temperature distribution characteristics, environmental characteristics, coupling characteristics, and core body temperature. A personalized core body temperature prediction model is built based on the machine learning algorithm and the training dataset.
[0085] The model application unit includes a core body temperature calculation module, which calculates the predicted core body temperature based on a personalized core body temperature prediction model according to the individual characteristics of a single user, facial temperature distribution characteristics, environmental characteristics, and coupling characteristics.
[0086] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is used to implement the data feature extraction and analysis method based on infrared thermal imaging.
[0087] The computer-readable storage medium may be a hard disk or memory, or any other device with data processing capabilities, such as a plug-in hard disk, smart media card (SMC), SD card, flash card, etc.
[0088] Furthermore, the computer-readable storage medium may include both internal storage units of any device with data processing capabilities and external storage devices. The computer-readable storage medium is used to store the computer program and other programs and data required by the device with data processing capabilities, and may also be used to temporarily store data that has been output or will be output.
[0089] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only.
[0090] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.
Claims
1. A data feature extraction and analysis method based on infrared thermal imaging, characterized in that, The method includes the following steps: Acquire user data, including infrared thermal imaging facial temperature distribution characteristics, personal characteristics, environmental characteristics, and actual core body temperature, and establish a database; determine whether it is a multi-person rapid monitoring scenario or a single-person long-term monitoring scenario; For scenarios involving rapid monitoring of multiple users, after establishing a database, all user data in the database is used as the dataset. A fitting coefficient K is obtained by fitting the data using a first coupling feature expression. The fitting coefficient K is then substituted into a second coupling feature expression to calculate the coupling features of all users in the database. Based on all user data in the database as the training dataset, a core body temperature prediction model is established using coupling features, facial temperature distribution features, personal features, and environmental features as inputs, and real core body temperature as the label. The model is trained using machine learning methods. The infrared thermal imaging facial temperature distribution features, personal features, and environmental features of the target user are obtained. Based on the second coupling feature expression after substituting the fitting coefficient K, the target user's coupling features are obtained and input into the core body temperature prediction model to predict the target user's core body temperature. This process is repeated multiple times to obtain the core body temperatures of multiple target users. For long-term monitoring of a single individual, after establishing a database, the infrared thermal imaging facial temperature distribution characteristics, personal characteristics, and environmental characteristics of the target user are obtained. User data with similar personal characteristics to the target user are extracted from the database and used as a dataset. A fitting coefficient K is obtained by fitting the target user's data using a first coupling feature expression. The fitting coefficient K is then substituted into a second coupling feature expression to calculate the coupling features between all users in the database and the target user. Based on the similarity to the target user's personal characteristics, weights are assigned to all users in the database, and adaptive adjustments are made based on these weights. This dataset serves as a training dataset, using coupling features, facial temperature distribution characteristics, personal characteristics, and environmental characteristics as inputs, and real core body temperature as the label. Machine learning methods are used to train and establish a personalized core body temperature prediction model. The infrared thermal imaging facial temperature distribution characteristics, personal characteristics, environmental characteristics, and coupling features of the target user are then input into the personalized core body temperature prediction model to predict the target user's core body temperature. The first coupling characteristic expression is: T c =(K+1)T max -KT a The second coupling characteristic expression is: T op =(K+1)T max -KT a , among which, T max The maximum temperature of the facial ROI, T a For ambient temperature, T c For true core body temperature, T op This is a coupling feature.
2. The data feature extraction and analysis method based on infrared thermal imaging according to claim 1, characterized in that, The infrared thermal imaging facial temperature distribution characteristics include the maximum temperature T of the facial ROI. max Maximum temperature T of forehead ROI FEmax Maximum temperature T at the inner canthus of the eye (ROI) CEmax Maximum temperature T at the mouth ROI Mmax ; The personal characteristics include gender, age, race, BMI, and body fat percentage; where gender and race are assigned values {0,1,…,n} by category, and age, BMI, and body fat percentage are assigned values {0,1,…,n} by numerical range. The environmental characteristics include ambient temperature T. a .
3. The data feature extraction and analysis method based on infrared thermal imaging according to claim 1, characterized in that, The infrared thermal imaging facial temperature distribution features of users and target users in the database are obtained by extracting faces, dividing facial regions, and calculating temperature values from the infrared thermal imaging images of users. The personal characteristics of users in the database are obtained through manual collection, while the personal characteristics of target users are obtained through manual collection or predicted using machine learning methods based on the user's infrared thermal images. The environmental characteristics of users and target users in the database are collected through auxiliary environmental parameter acquisition devices; The database contains users' actual core body temperature (T). c It was measured using a medical thermometer.
4. The data feature extraction and analysis method based on infrared thermal imaging according to claim 1, characterized in that, The method further includes: performing feature selection on features in the training dataset to obtain a key feature set, training the model using machine learning methods, and using the key feature set of the target user as input features of the core body temperature prediction model or personalized core body temperature prediction model to predict the core body temperature of the target user; wherein the key feature set includes at least coupling features.
5. The data feature extraction and analysis method based on infrared thermal imaging according to claim 4, characterized in that, The feature selection methods include: Pearson correlation coefficient and recursive feature elimination.
6. The data feature extraction and analysis method based on infrared thermal imaging according to claim 1, characterized in that, The machine learning methods include: random forest, neural network, and support vector machine.
7. The data feature extraction and analysis method based on infrared thermal imaging according to claim 1, characterized in that, In a single-person long-term monitoring scenario, the step of extracting user data with similar personal characteristics from the database based on the target user's personal characteristics as a dataset includes: calculating the similarity between the personal characteristics of all users in the database and the target user, setting a threshold, and including user data with similarity greater than the threshold into the dataset; The step of assigning weights to all users in the database based on their similarity to the target user's personal characteristics, and adaptively adjusting the weights to form a training dataset, includes: normalizing personal characteristics, calculating the similarity between the personal characteristics of all users in the database and the target user, assigning weights to all users in the database based on the similarity so that users with relatively high similarity have relatively high weights, and resampling according to the weights to obtain the training dataset. The similarity calculation methods include: weighted Euclidean distance, Manhattan distance, Mahalanobis distance, Langevin distance, and Minkowski distance.
8. The data feature extraction and analysis method based on infrared thermal imaging according to claim 1, characterized in that, The method further includes: after predicting the target user's core body temperature, measuring the target user's actual core body temperature using a medical thermometer, and incorporating the target user's infrared thermal imaging facial temperature distribution characteristics, personal characteristics, environmental characteristics, and actual core body temperature into the database for updating.
9. A core body temperature prediction system based on infrared thermal imaging, characterized in that, It is used to implement the method as described in any one of claims 1-8, wherein the system comprises: The data acquisition unit includes a personal feature acquisition module, a facial temperature data acquisition module, an environmental data acquisition module, and a core body temperature acquisition module. These modules are used to acquire personal features, infrared thermal imaging facial temperature distribution features, environmental features, and real core body temperature data measured by a medical thermometer, respectively, and to establish a database. In a single-person long-term monitoring scenario, the data in the database is input into the first similarity measurement module, and in a multi-person rapid monitoring scenario, it is input into the coupling feature calculation module. The feature extraction unit includes a first similarity measurement module and a coupled feature calculation module. The first similarity measurement module is used to calculate the similarity between the target user and the personal features of users in the database in a single-person long-term monitoring scenario, set a threshold, and select user data with similarity greater than the threshold to input into the coupled feature calculation module to calculate coupled features. The coupled feature calculation module is used to calculate coupled features based on facial temperature distribution features and environmental features. The data processed by the feature extraction unit is used as a training dataset to input into the model building unit in a multi-person rapid monitoring scenario, and as an input data adaptive adjustment unit in a single-person long-term monitoring scenario. The data adaptive adjustment unit is used to receive the data processed by the feature extraction unit in a single-person long-term monitoring scenario, perform data adaptive adjustment, and obtain a training dataset. The data adaptive adjustment unit includes a second similarity measurement module and a data resampling module. The second similarity measurement module evaluates the similarity between the target user and the personal features of users in the database. The data resampling module resamples the data according to the similarity to obtain the training dataset. Users with relatively high similarity have relatively high weights. The model building unit is used to train a core body temperature prediction model based on a training dataset using machine learning methods. The training dataset consists of personal characteristics, facial temperature distribution characteristics, environmental characteristics, coupling characteristics, and real core body temperature. The model application unit includes a core body temperature calculation module, which calculates the predicted core body temperature based on the target user's personal characteristics, facial temperature distribution characteristics, environmental characteristics, and coupling characteristics according to the core body temperature prediction model.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it is used to implement the method as described in any one of claims 1-8.