A thermal sensation assessment method and system based on infrared thermal imaging technology

Abstract

The present invention relates to the field of thermal sensation assessment technology, and in particular to a thermal sensation assessment method and system based on infrared thermal imaging technology, which obtain a facial infrared thermal image sequence of a target object captured by an infrared thermal imager; divide at least one pair of symmetrical physiological region pairs based on the left-right symmetrical physiological regions of the human face; calculate the dynamic symmetry index of each current symmetrical physiological region pair in the infrared thermal image sequence acquired in real time, compare the dynamic symmetry index with a normal symmetry reference range in a clean environment, and generate a comprehensive symmetry breaking index (SBI); quantify dust interference by calculating the dynamic symmetry index of the facial symmetrical physiological region pairs and comparing them with the reference to generate a comprehensive symmetry breaking index (SBI), adjust the confidence of the thermal sensation assessment result or switch the assessment strategy according to the SBI level, and have the functions of identifying and quantifying the interference of dust on infrared radiation transmission, adjusting the thermal sensation assessment result according to the degree of interference, and improving the assessment accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of thermal sensation evaluation, and in particular to a thermal sensation evaluation method and system based on infrared thermal imaging technology. Background Art

[0002] In industrial production environments, such as wood processing, textile manufacturing, or metal polishing workshops, worker thermal comfort is crucial to work efficiency and health and safety. Existing technologies typically use infrared thermal imaging technology to non-contactly monitor workers' skin temperature. This technology, combined with thermal perception assessment models, guides real-time adjustments in environmental control systems (such as ventilation and air conditioning). Specifically, an infrared thermal imager periodically captures infrared thermal radiation images of exposed areas, such as the worker's face or neck, extracts temperature characteristics (such as average temperature and standard deviation of temperature distribution), and calculates a thermal perception index based on a correlation model between calibrated skin temperature and subjective thermal perception under clean air conditions. This index is then used to control environmental equipment. This technology effectively maintains a thermally comfortable environment in environments with high air cleanliness.

[0003] However, the above technologies face significant challenges in actual industrial applications. Production activities generate a large amount of suspended particulate matter (such as wood chips, fibers, and metal dust), which form a dynamically changing aerosol system in the air. When the infrared thermal imager is measuring, dust particles will interact with the infrared radiation emitted by the skin: part of the radiation is absorbed or scattered, resulting in attenuation of the effective signal; at the same time, the dust itself emits interfering radiation due to temperature or friction heating, which in turn causes a dynamic deviation between the apparent temperature measured by the thermal imager and the actual skin temperature of the person. The existing thermal perception assessment model based on calibration under clean air cannot directly process this interfered temperature data, thereby avoiding assessment errors caused by dust interference and ensuring that the environmental control system can be effectively adjusted based on accurate thermal perception information.

[0004] In view of the above problems, the existing technology is in urgent need of improvement. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the existing technology and to propose a thermal perception evaluation method and system based on infrared thermal imaging technology.

[0006] In a first aspect, the present invention provides a thermal sensation assessment method based on infrared thermal imaging technology, the method comprising the following steps:

[0007] S1: Acquire a facial infrared thermal image sequence of the target object captured by an infrared thermal imager;

[0008] S2: dividing at least one pair of symmetrical physiological regions based on the left-right symmetrical physiological regions of the human face;

[0009] S3: Calculating the dynamic symmetry index of each current pair of symmetrical physiological regions in the real-time acquired infrared thermal image sequence, comparing the dynamic symmetry index with a normal symmetry reference range in a clean environment, and generating a comprehensive symmetry breaking index (SBI). The SBI is used to quantify the degree of interference of dust on infrared radiation transmission.

[0010] S4: The comprehensive symmetry breaking index (SBI) is divided into several levels, and the confidence level of the thermal sensation assessment results is dynamically adjusted or the thermal sensation assessment strategy is switched according to the SBI level.

[0011] Specifically, this method quantifies the degree of dust interference with infrared radiation transmission by analyzing the differences in infrared thermal image features between left and right symmetrical regions of the human face, thereby adjusting the thermal perception assessment results based on infrared thermal imaging. In industrial production environments, suspended dust absorbs and scatters infrared radiation, causing the apparent temperature measured by the infrared thermal imager to deviate from the true skin temperature. This interference can be spatially non-uniform, disrupting the inherent left-right thermal symmetry of the human face. The method first acquires a sequence of infrared thermal images of the target face (S1), which serves as the basic data for assessment. Next, based on the physiological symmetry of the human face, pairs of left-right symmetrical physiological regions are identified (S2). Within each image frame, symmetry metrics (such as temperature difference and correlation) are calculated between these pairs of symmetrical regions to generate dynamic symmetry indices (S3). These real-time indices are compared with a normal symmetry baseline established in a clean environment, and the degree of deviation is calculated to generate a comprehensive symmetry breaking index (SBI). The SBI value reflects the intensity of dust interference with infrared measurements in the current environment; higher SBI values ​​indicate greater interference. Finally, the use of thermal perception assessment results is dynamically adjusted based on the SBI level (S4). When the SBI is low, the interference is considered minimal, and the reliability of the apparent temperature-based assessment results is high. When the SBI is high, the interference is considered significant, and the confidence in the assessment results is reduced, or a switch is made to another assessment strategy that is not affected by dust or is less affected, such as an assessment model based on environmental sensor data. This method can thus identify and quantify dust interference, avoiding the use of distorted temperature data for assessment when interference is severe, thereby improving the accuracy and robustness of thermal sensation assessment in dusty environments.

[0012] Furthermore, the present application also proposes that step S1 includes:

[0013] S11: periodically capturing original images containing faces using a fixedly mounted infrared thermal imager;

[0014] S12: Noise is filtered out of the original image and an image registration technology based on feature point matching is used to align a multi-frame infrared thermal image sequence to the same spatial coordinate system to obtain an infrared thermal image sequence of the target object's face.

[0015] Furthermore, the present application also proposes that step S2 includes:

[0016] S21: Using facial recognition algorithm to locate the face area in infrared thermal image sequence;

[0017] S22: Based on a preset anatomical landmark or template, the facial midline is used as the symmetry axis to automatically divide the face into a number of left-right symmetrical physiological region pairs, including left and right forehead regions, left and right cheek regions, and left and right eye canthus regions.

[0018] Furthermore, the present application also proposes that step S3 includes:

[0019] S31: Repeat step S2 for each frame of the infrared thermal image sequence to divide the human face into symmetrical physiological area pairs;

[0020] S32: calculating the symmetry index of each pair of symmetrical physiological regions in each frame to obtain a real-time dynamic symmetry index;

[0021] S33: Compare the dynamic symmetry index with the normal symmetry reference range under a clean environment, and calculate the weighted average or maximum value of the selected index's deviation from the reference range, which is the comprehensive symmetry breaking index SBI.

[0022] Furthermore, the present application also proposes that the symmetry index includes temperature difference, temperature distribution correlation coefficient and similarity index.

[0023] Furthermore, the present application also proposes that the establishment of the normal symmetry reference range in the clean environment includes:

[0024] When the dust sensor is used to assist in determining whether the air environment is clean, a multi-frame reference infrared thermal image sequence is collected;

[0025] Calculate the average temperature difference, temperature distribution correlation coefficient, and structural similarity index of each symmetry index in each pair of symmetrical physiological regions;

[0026] Statistically calculate the distribution range of each symmetry index and establish a normal symmetry benchmark.

[0027] Furthermore, the present application also proposes that step S4 includes:

[0028] S41: divided into no interference / slight interference, moderate interference, and severe interference levels based on the SBI threshold;

[0029] S42: Use SBI as an influencing factor for the confidence of the thermal sensation assessment result. If SBI is at the no interference / slight interference level, the confidence of the thermal sensation assessment result calculated based on the apparent temperature is determined to be high. As SBI increases, the confidence of the assessment result is reduced.

[0030] Furthermore, the present application also proposes that the switching thermal sensation assessment strategy includes:

[0031] Setting a first threshold and a second threshold, wherein the first threshold is smaller than the second threshold;

[0032] When the SBI is less than or equal to the first threshold, the confidence level of the thermal assessment result is high, and the original apparent temperature is used for thermal assessment. When the SBI is greater than the first threshold and less than or equal to the second threshold, the environmental control adjustment based on the measurement is temporarily suspended, and the previous high-confidence control adjustment mode is maintained or the preset conservative control mode is adopted. When the SBI is greater than the second threshold, the infrared thermal imaging assessment result is disabled, and the system switches to a robust thermal assessment model based on environmental sensor data.

[0033] Outputs an alarm message to indicate possible measurement interference and recommends checking the environment or performing manual confirmation.

[0034] Furthermore, the present application also proposes that the environmental sensor data information includes:

[0035] Obtain data from dust concentration sensors and ambient temperature and humidity sensors;

[0036] The data are input into a robust thermal sensation evaluation model under dust interference conditions that has been calibrated offline to generate a revised thermal sensation index.

[0037] In a second aspect, a thermal sensation assessment system based on infrared thermal imaging technology is provided, the system comprising

[0038] A data acquisition module, which acquires a facial infrared thermal image sequence of a target object captured by an infrared thermal imager;

[0039] A symmetry analysis module, which divides at least one pair of symmetrical physiological regions based on the left and right symmetrical physiological regions of the human face;

[0040] An interference judgment module calculates the dynamic symmetry index of each current pair of symmetrical physiological regions in the real-time infrared thermal image sequence, compares the dynamic symmetry index with the normal symmetry reference range in a clean environment, and generates a comprehensive symmetry breaking index (SBI). The SBI is used to quantify the degree of dust interference with infrared radiation transmission;

[0041] The evaluation and adjustment module divides the comprehensive symmetry breaking index (SBI) into several levels and dynamically adjusts the confidence level of the thermal sensation evaluation results or switches the thermal sensation evaluation strategy according to the SBI level.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] This technical solution utilizes symmetry breaking in facial infrared thermal image sequences as an indicator of dust interference, avoiding the difficulty of directly physically modeling and correcting complex and dynamically changing dust interference. It can effectively identify the degree of interference with infrared thermal imaging measurements in industrial environments with dynamically changing and unevenly distributed production dust. This method can improve the reliability and accuracy of thermal sensation assessment results in the presence of dust interference, enabling environmental control systems to more accurately respond to workers' true thermal comfort needs. It avoids misjudgments caused by measurement distortion (such as misjudging a low temperature as cold or a high temperature as hot) and erroneous control actions (such as unnecessary cooling or heating), thereby improving the stability and effectiveness of environmental control systems, contributing to improved workplace comfort and potentially reducing energy consumption.

[0044] From the above, it can be seen that the present application provides a thermal sensation assessment method and system based on infrared thermal imaging technology, which quantifies dust interference by calculating the dynamic symmetry index of symmetrical physiological regions of the face and comparing it with the benchmark to generate a comprehensive symmetry breaking index SBI. The confidence of the thermal sensation assessment result is adjusted or the assessment strategy is switched according to the SBI level. It has the functions of identifying and quantifying the interference of dust on infrared radiation transmission, adjusting the thermal sensation assessment result according to the degree of interference, and improving the assessment accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a flow chart of a thermal sensation assessment method based on infrared thermal imaging technology proposed in the present invention.

[0046] Figure 2 This is a structural diagram of a thermal sensation assessment system based on infrared thermal imaging technology proposed by the present invention.

[0047] In the figure: 201, data acquisition module; 202, symmetry analysis module; 203, interference judgment module; 204, evaluation and adjustment module. DETAILED DESCRIPTION

[0048] The embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and are not to be construed as limiting the present invention.

[0049] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the specified features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0050] Application Scenario: In industrial production environments, such as wood processing, textile manufacturing, and metal polishing workshops, production activities generate large amounts of suspended particulate matter (such as wood chips, fibers, and metal dust), which forms a dynamically changing aerosol system in the air. During infrared thermal imaging camera measurements, dust particles interact with infrared radiation emitted by the skin: some of this radiation is absorbed or scattered, resulting in effective signal attenuation. Simultaneously, the dust itself emits interfering radiation due to temperature rise or frictional heating, leading to a dynamic deviation between the apparent temperature measured by the thermal imager and the actual skin temperature of the person. Existing thermal assessment models calibrated under clean air conditions cannot directly handle this interfering temperature data, potentially leading to assessment errors and compromising the accurate adjustment of environmental control systems. Therefore, effectively identifying and quantifying the interference of dust on infrared radiation transmission and adjusting thermal assessment strategies accordingly to avoid assessment errors caused by dust interference and ensure that environmental control systems can effectively adjust based on accurate thermal information are urgent technical challenges.

[0051] like Figure 1 A thermal sensation assessment method based on infrared thermal imaging technology is shown, and the method includes the following steps:

[0052] S1: Acquire a facial infrared thermal image sequence of the target object captured by an infrared thermal imager;

[0053] S2: dividing at least one pair of symmetrical physiological regions based on the left-right symmetrical physiological regions of the human face;

[0054] S3: Calculating the dynamic symmetry index of each current pair of symmetrical physiological regions in the real-time acquired infrared thermal image sequence, comparing the dynamic symmetry index with a normal symmetry reference range in a clean environment, and generating a comprehensive symmetry breaking index (SBI). The SBI is used to quantify the degree of interference of dust on infrared radiation transmission.

[0055] S4: The comprehensive symmetry breaking index (SBI) is divided into several levels, and the confidence level of the thermal sensation assessment results is dynamically adjusted or the thermal sensation assessment strategy is switched according to the SBI level.

[0056] Step S1 acquires a sequence of infrared thermal images of the target object's face. Specifically, a fixed-mounted infrared thermal imager periodically captures original images containing the face, filters the original images for noise, and uses image registration technology based on feature point matching to align multiple frames of the infrared thermal image sequence to the same spatial coordinate system, thereby obtaining a sequence of infrared thermal images of the target object's face.

[0057] Step S2 divides the facial region into pairs. Specifically, a facial recognition algorithm is used to locate the facial region in the infrared thermal image sequence. Based on preset anatomical landmarks or templates, the facial midline is used as the axis of symmetry to automatically divide the facial region into pairs of symmetrical left and right physiological regions, including the left and right forehead regions, the left and right cheek regions, and the left and right eye corner regions.

[0058] Step S3 calculates the comprehensive symmetry breaking index SBI. Specifically, the operation of step S2 is repeated for each frame of the infrared thermal image sequence acquired to divide the human face into symmetrical physiological area pairs, and the symmetry measurement index of each symmetrical physiological area pair in each frame is calculated to obtain a real-time dynamic symmetry index. The symmetry index includes temperature difference, temperature distribution correlation coefficient and similarity index. The dynamic symmetry index is compared with the normal symmetry reference range in a clean environment, and the weighted average value or maximum value of the selected index deviating from the reference range is calculated, which is the comprehensive symmetry breaking index SBI. The establishment of the normal symmetry reference range in a clean environment includes: collecting a multi-frame reference infrared thermal image sequence when the air environment is clean with the assistance of a dust sensor; calculating the average temperature difference, temperature distribution correlation coefficient and structural similarity index of each symmetry index in each symmetrical physiological area pair; statistically analyzing the distribution range of each symmetry index to establish a normal symmetry benchmark.

[0059] Step S4 adjusts the thermal sensation assessment result. Specifically, the SBI value is divided into no interference / slight interference, moderate interference, and severe interference levels, and the SBI is used as an influencing factor of the confidence of the thermal sensation assessment result. If the SBI is at the no interference / slight interference level, the confidence of the thermal sensation assessment result calculated based on the apparent temperature is determined to be high. As the SBI increases, the confidence of the assessment result is reduced. The switching thermal sensation assessment strategy includes: setting a first threshold and a second threshold, and the first threshold is less than the second threshold; when the SBI is less than or equal to the first threshold, the confidence of the thermal sensation assessment result is high, and the original apparent temperature is used for thermal sensation assessment; when the SBI is greater than the first threshold and less than or equal to the second threshold, the environmental control adjustment based on the measurement is temporarily suspended, and the previous high-confidence control adjustment mode is maintained or the preset conservative control mode is adopted; when the SBI is greater than the second threshold, the infrared thermal imaging assessment result is disabled, and the robust thermal sensation assessment model based on the environmental sensor data is switched; an alarm message is output to indicate that there may be measurement interference, and it is recommended to check the environment or perform manual confirmation. The environmental sensor data information includes: obtaining data from a dust concentration sensor or an ambient temperature and humidity sensor; inputting the data into a robust thermal sensation evaluation model under dust interference conditions that has been calibrated offline to generate a corrected thermal sensation index.

[0060] Specifically, this method quantifies the degree of dust interference with infrared radiation transmission by analyzing the differences in infrared thermal image features between left and right symmetrical regions of the human face, thereby adjusting the thermal perception assessment results based on infrared thermal imaging. In industrial production environments, suspended dust absorbs and scatters infrared radiation, causing the apparent temperature measured by the infrared thermal imager to deviate from the true skin temperature. This interference can be spatially non-uniform, disrupting the inherent left-right thermal symmetry of the human face. The method first acquires a sequence of infrared thermal images of the target face (S1), which serves as the basic data for assessment. Next, based on the physiological symmetry of the human face, pairs of left-right symmetrical physiological regions are identified (S2). Within each image frame, symmetry metrics (such as temperature difference and correlation) are calculated between these pairs of symmetrical regions to generate dynamic symmetry indices (S3). These real-time indices are compared with a normal symmetry baseline established in a clean environment, and the degree of deviation is calculated to generate a comprehensive symmetry breaking index (SBI). The SBI value reflects the intensity of dust interference with infrared measurements in the current environment; higher SBI values ​​indicate greater interference. Finally, the use of thermal perception assessment results is dynamically adjusted based on the SBI level (S4). When the SBI is low, the interference is considered minimal, and the reliability of the apparent temperature-based assessment results is high. When the SBI is high, the interference is considered significant, and the confidence in the assessment results is reduced, or a switch is made to another assessment strategy that is not affected by dust or is less affected, such as an assessment model based on environmental sensor data. This method can thus identify and quantify dust interference, avoiding the use of distorted temperature data for assessment when interference is severe, thereby improving the accuracy and robustness of thermal sensation assessment in dusty environments.

[0061] In some embodiments, the system is equipped with a fixed-mounted infrared thermal imager that periodically (e.g., once per second) captures infrared images of a worker's face. The images are subjected to median filtering to remove noise, and consecutive frames are aligned using a SIFT-based image registration algorithm. A facial recognition algorithm locates the facial region and automatically divides the face into three symmetrical pairs: the forehead, cheeks, and eye corners, based on a preset template and the facial midline. For each image frame, the average temperature difference and the Pearson correlation coefficient of the pixel temperature distribution are calculated for each symmetrical pair. These values ​​are compared with a baseline range established using data collected in a cleanroom environment. For example, the baseline range might be set as a forehead temperature difference of less than 0.3°C and a correlation coefficient greater than 0.95. The degree to which the current temperature difference and correlation coefficient deviate from the baseline range is calculated and weighted summed to obtain the SBI. For example, an SBI of 0.5 or less is considered mild interference, between 0.5 and 1.5 (inclusive) is considered moderate interference, and greater than 1.5 is considered severe interference. When the SBI is less than or equal to 0.5, the apparent temperature measured by the infrared thermal imager is input into a standard PMV model for thermal perception assessment. When the SBI is between 0.5 and 1.5 (inclusive), environmental control adjustments based on this assessment are suspended, and the control state when the previous SBI was less than 0.5 is maintained. When the SBI is greater than 1.5, the infrared thermal imaging assessment results are disabled, and the data from the temperature, humidity, and dust concentration sensors in the workshop are used instead. A pre-calibrated thermal perception model that takes into account the influence of dust is input for assessment, and an alarm message is output to prompt the operator.

[0062] This application further proposes a method for obtaining a sequence of infrared thermal images of a target object's face, the method comprising the following steps:

[0063] S11: periodically capturing original images containing faces using a fixedly mounted infrared thermal imager;

[0064] S12: Noise is filtered out of the original image and an image registration technology based on feature point matching is used to align a multi-frame infrared thermal image sequence to the same spatial coordinate system to obtain an infrared thermal image sequence of the target object's face.

[0065] Step S11, by fixing the infrared thermal imager in a specific location, such as above or to the side of the work area, reduces the impact of camera motion on image stability. Periodic capture involves capturing a number of frames at a preset interval, such as per second, to obtain a continuous stream of image data. Capturing the original image containing the face identifies the target area for subsequent processing.

[0066] Furthermore, step S12 processes the captured original image. Noise filtering can use filtering algorithms in the spatial domain or frequency domain, such as median filtering or Gaussian filtering, to remove random noise in the image caused by environmental interference or the sensor itself, thereby improving the signal-to-noise ratio of the image. Image registration technology based on feature point matching is used, such as using algorithms such as SIFT, SURF, or ORB to extract stable feature points in the image. By matching feature points between different frames, the geometric transformation parameters (such as translation, rotation, and scaling) between frames are calculated. Using the calculated transformation parameters, the image pixels of subsequent frames are mapped to the spatial coordinate system of the first frame or a reference frame, thereby achieving alignment of multiple frames. This ensures that the same facial physiological area in the image sequence corresponds to the same image position at different time points.

[0067] Specifically, in industrial production environments, due to the presence of suspended particulate matter such as dust, the original images captured by the infrared thermal imager may contain noise, and workers may move their heads, resulting in changes in facial position and posture between consecutive frames. These problems will affect the subsequent accurate division of symmetrical physiological areas of the face and the calculation of symmetry indicators. In this solution, in step S11, the infrared thermal imager is fixedly installed and periodically captured to obtain a sequence of original infrared thermal images containing the worker's face. Then, in step S12, these original images are subjected to noise filtering to remove random interference caused by dust and other particles, thereby improving image quality. Subsequently, an image registration technology based on feature point matching is used to correct the inter-frame displacement and rotation caused by the worker's head movement, and the image sequence is aligned to a unified spatial reference. In this way, a clear and inter-frame aligned infrared thermal image sequence of the target object's face is obtained, which provides reliable input data for the accurate division of symmetrical physiological areas and calculation of dynamic symmetry indicators in the subsequent steps, and solves the technical problems of high noise in the original image and misalignment between frames.

[0068] In some specific embodiments, a fixed-mounted uncooled infrared thermal imager with a frame rate of 25 frames per second and a spatial resolution of 320x240 pixels can be used. The thermal imager periodically captures infrared images containing the worker's face. The captured original image is first filtered through a 3x3 median filter to remove noise. Subsequently, an image registration algorithm based on ORB feature point extraction and matching is applied to the filtered image sequence. The first frame of the sequence is selected as the reference frame, and the ORB feature points between each subsequent frame and the reference frame are extracted. The affine transformation matrix is ​​estimated by the RANSAC algorithm, and the matrix is ​​used to perform geometric transformations on the subsequent frames to align them to the spatial coordinate system of the reference frame. In this way, a sequence of infrared thermal images of the target object's face with inter-frame alignment and reduced noise is obtained for subsequent symmetry analysis.

[0069] This application further proposes that step S2 includes:

[0070] S21: Using facial recognition algorithm to locate the face area in infrared thermal image sequence;

[0071] S22: Based on a preset anatomical landmark or template, the facial midline is used as the symmetry axis to automatically divide the face into a number of left-right symmetrical physiological region pairs, including left and right forehead regions, left and right cheek regions, and left and right eye canthus regions.

[0072] Among them, step S21 uses a facial recognition algorithm, whose function is to accurately locate the facial area in the infrared thermal image sequence. As a result, subsequent processing can be focused on the target area. Step S22 automatically divides a number of bilaterally symmetrical physiological area pairs based on preset anatomical landmarks or templates, with the facial midline as the axis of symmetry. The preset anatomical landmarks or templates provide a reference for the facial structure. The facial midline serves as the axis of symmetry to guide the division process, ensuring that the divided area pairs have physiological symmetry. The divided area pairs include the left and right forehead areas, the left and right cheek areas, and the left and right eye corner areas. These areas are parts of the face with bilateral symmetry.

[0073] Specifically, it is necessary to accurately identify the facial area and divide it into symmetrical physiological area pairs to support the subsequent calculation of symmetry indicators. A facial recognition algorithm is used to locate the facial area in the infrared thermal image sequence. This step finds the facial boundary in the infrared thermal image sequence and limits the processing range to the facial area. As a result, subsequent processing is only performed on the face. Furthermore, based on the preset anatomical landmarks or templates, the facial midline is determined as the symmetry axis. Based on the symmetry axis and the landmarks or templates, several left-right symmetrical physiological area pairs are automatically divided. These area pairs include left and right forehead areas, left and right cheek areas, and left and right eye corner areas. Through these steps, the facial area is accurately identified and divided into specific left-right symmetrical area pairs. This provides basic data for the subsequent calculation of dynamic symmetry indicators.

[0074] In some embodiments, a deep learning-based facial detection model is used to identify a facial bounding box in an infrared thermal image. Within this bounding box, a facial landmark detection algorithm is used to locate facial landmarks, such as the inner and outer canthi of the eyes, and the tip of the nose. The facial midline is estimated as a vertical line connecting the midpoint of the inner canthi of the eyes and the tip of the nose. The left and right forehead regions are defined as the areas above the left and right eyebrows and lateral to the facial midline. The left and right cheek regions are defined as the areas below the left and right eyes and lateral to the facial midline. The left and right eye corner regions are defined as the small areas around the left and right eye corners. These regions are automatically segmented based on the detected landmarks and the calculated facial midline.

[0075] This application further proposes that step S3 includes:

[0076] S31: Repeat step S2 for each frame of the infrared thermal image sequence to divide the human face into symmetrical physiological area pairs;

[0077] S32: calculating the symmetry metrics of each pair of symmetrical physiological regions in each frame to obtain a real-time dynamic symmetry index;

[0078] S33: Compare the dynamic symmetry index with the normal symmetry reference range under a clean environment, and calculate the weighted average or maximum value of the selected index's deviation from the reference range, which is the comprehensive symmetry breaking index SBI.

[0079] For each frame in the acquired infrared thermal image sequence, a facial recognition algorithm is executed to locate the facial region. Based on preset anatomical landmarks or templates, with the facial midline as the axis of symmetry, several pairs of bilaterally symmetrical physiological regions are automatically segmented, such as the forehead, cheeks, and canthus. For each segmented pair of symmetrical physiological regions in each frame, a symmetry metric is calculated. These metrics may include the average temperature difference between the left and right regions, the correlation coefficient of the temperature distribution, and the structural similarity index. The symmetry indices calculated for each pair of symmetrical physiological regions in each frame are combined to form a dynamic symmetry index reflecting the current state of facial symmetry. These dynamic symmetry indices are compared with a normal symmetry baseline range established in a clean environment. The degree to which the current index value deviates from this baseline range is calculated. The deviation degrees of multiple symmetrical physiological region pairs and the deviation information from multiple frames can be integrated into a single value, the comprehensive symmetry violation index (SBI), using a weighted average or maximum value. The weighted average can be weighted based on the sensitivity of different regions or indicators to interference. The maximum value directly reflects the most severe symmetry violation.

[0080] Specifically, this technical solution, by processing a sequence of continuously acquired infrared thermal images, addresses the problem of accurately calculating an index reflecting the degree of interference in a dynamically changing environment. First, each frame in the sequence undergoes independent facial region segmentation and symmetry region pairing, ensuring that subsequent calculations are based on accurate region definitions. Next, symmetry metrics, such as temperature difference, correlation coefficient, or similarity index, are calculated for each symmetry region pair frame by frame, generating dynamic data reflecting the symmetry state at each moment. This frame-by-frame data captures the temporal evolution of interference. Finally, these dynamic symmetry metrics are compared to a normal baseline range established in an interference-free environment to quantify their deviations. A composite symmetry breaking index (SBI) is generated by calculating the weighted average or maximum of these deviations. The SBI integrates symmetry information from multiple frames and regions into a single value that directly quantifies the degree of interference caused by environmental factors such as dust on infrared radiation transmission. Through these steps, the solution enables real-time and accurate assessment of interference levels, providing a reliable basis for subsequent adjustments to the confidence level of thermal assessment results or switching assessment strategies, thereby avoiding assessment errors caused by interference.

[0081] In some specific embodiments, a sequence of 100 infrared thermal images is acquired. For each frame in the sequence, a face detection algorithm is first executed to locate the face frame. Then, within the face frame, the left and right cheek regions and the left and right forehead regions are automatically segmented based on a preset template and the facial midline. For each frame, the average temperature difference and temperature distribution correlation coefficient of the left and right cheek regions, as well as the average temperature difference and temperature distribution correlation coefficient of the left and right forehead regions, are calculated. These calculated values ​​are compared with a baseline range established in a clean environment. For example, in a clean environment, the cheek temperature difference is less than 0.5°C and the correlation coefficient is greater than 0.95. The deviation value of each indicator that exceeds the baseline range is calculated for each frame. For example, if the cheek temperature difference in a frame is 1.0°C, the deviation value is 1.0 - 0.5 = 0.5°C. The weighted average of the deviation values ​​for all regions and all indicators in the 100 frames is calculated, or the maximum of all deviation values ​​is directly taken as the final comprehensive symmetry breaking index (SBI). For example, if the maximum deviation value is 0.8, the SBI is 0.8. This SBI value is then used to determine the interference level.

[0082] The present application further proposes that symmetry indicators include temperature difference, temperature distribution correlation coefficient and similarity index.

[0083] Specifically, to address the issue of lack of a specific measurement method, which prevents the comprehensive capture of dust's impact on infrared image symmetry, this approach uses temperature difference, temperature distribution correlation coefficient, and similarity index as dynamic symmetry metrics. After acquiring a series of infrared thermal images of the target subject's face, each frame is processed. First, pairs of bilaterally symmetrical physiological regions of the human face are segmented. Then, for each pair of symmetrical physiological regions in each frame, the temperature difference, temperature distribution correlation coefficient, and similarity index are calculated. The temperature difference directly reflects the overall temperature difference within the symmetrical region, which can be caused by uneven absorption and scattering of infrared radiation by dust. The temperature distribution correlation coefficient assesses the consistency of the spatial temperature distribution pattern within the symmetrical region, which can be altered by localized interference caused by dust. The similarity index comprehensively considers the brightness, contrast, and structural information of the symmetrical region, reflecting the impact of dust on image detail and texture. The calculated temperature difference, temperature distribution correlation coefficient, and similarity index are used as the dynamic symmetry metrics for that pair of symmetrical physiological regions in that frame. By calculating these indicators for each pair of symmetrical physiological regions across multiple image frames and comparing them with a baseline range of normal symmetry in a clean environment, a comprehensive symmetry breaking index (SBI) can be calculated. The SBI quantifies the degree to which dust interferes with infrared radiation transmission. Using these specific, multi-dimensional symmetry indices can more accurately reflect the extent to which dust disrupts the symmetry of infrared thermal images, thereby improving the accuracy of the SBI and providing a reliable basis for subsequent dynamic adjustments to the confidence level of thermal assessment results or switching assessment strategies.

[0084] In some specific embodiments, the calculation of the symmetry index is illustrated using the left and right cheek regions as an example. First, pixel temperature data for the left and right cheek regions is extracted from the current frame of infrared thermal imagery. The average temperature of all pixels in the left cheek region, T_left, and the average temperature of all pixels in the right cheek region, T_right, are calculated, and the temperature difference is calculated as |T_left - T_right|. The pixel temperature data for the left cheek region is arranged into a vector V_left, and the pixel temperature data for the right cheek region is arranged into a vector V_right. The Pearson correlation coefficient R between V_left and V_right is calculated. The image blocks of the left and right cheek regions are used as input, and their structural similarity index (SSIM) is calculated. The calculated temperature difference, correlation coefficient R, and similarity index SSIM are used as dynamic symmetry indicators for the left and right cheek regions in the image frame. This calculation can be repeated for multiple pairs of symmetrical physiological regions (such as the left and right foreheads, left and right canthi), and the indicators of different region pairs are weighted averaged or maximized to obtain the overall dynamic symmetry index for the image frame. These indicators are then used to calculate the comprehensive symmetry breaking index (SBI). For example, a function can be set to perform a weighted summation of the temperature difference, (1-R) ​​and (1-SSIM) to obtain a comprehensive value. The larger the value, the more serious the symmetry breaking.

[0085] This application further proposes that the establishment of a normal symmetry reference range in a clean environment includes:

[0086] When the dust sensor is used to assist in determining that the air environment is clean, a multi-frame reference infrared thermal image sequence is collected;

[0087] Calculate the average temperature difference, temperature distribution correlation coefficient, and structural similarity index of each symmetry index in each pair of symmetrical physiological regions;

[0088] Calculate the distribution range of each symmetry index and establish a normal symmetry benchmark.

[0089] Among them, a specific method is provided for establishing a normal symmetry benchmark range in a clean environment. This method first uses a dust sensor to determine whether the current air environment meets clean conditions, ensuring that the collected data is not interfered with by dust. This is the prerequisite for establishing a reliable benchmark. After confirming the cleanliness of the environment, a multi-frame reference infrared thermal image sequence is collected. These multi-frame data can reflect the normal fluctuation range of temperature and distribution in the symmetrical areas of the human face in a clean environment. Next, various symmetry indicators are calculated for the collected reference image sequence, including the average temperature difference, the temperature distribution correlation coefficient, and the structural similarity index. These indicators quantify the degree of similarity between the left and right symmetrical areas in a clean environment. Finally, these calculated symmetry indicators are statistically analyzed to determine their distribution range in a clean environment, thereby establishing a normal symmetry benchmark.

[0090] Specifically, in industrial production environments, workers' thermal comfort is particularly important for work efficiency and health and safety. This technical solution uses dust sensors to assist in determining when the air environment is clean, and collects a multi-frame reference infrared thermal image sequence. The average temperature difference, temperature distribution correlation coefficient, and structural similarity index of each symmetry index in each symmetrical physiological region are calculated. The distribution range of each symmetry index is statistically analyzed to establish a normal symmetry benchmark. This benchmark range provides a reliable reference for comparing dynamic symmetry indicators with the benchmark, so that the generated comprehensive symmetry breaking index SBI can accurately reflect the degree of interference of dust on infrared radiation transmission, thereby supporting the adjustment of the confidence level of subsequent thermal sensation assessment results or the switching of assessment strategies.

[0091] In some specific embodiments, establishing a normal symmetry baseline range in a clean environment can be achieved by setting a dust concentration threshold, such as 100 μg / m³. When the dust concentration in the air detected by the dust sensor remains below the threshold for a continuous period of time (e.g., 5 minutes), the environment is determined to be clean. At this point, the infrared thermal imager is activated to capture a sequence of infrared thermal images of the target subject's face at a rate of one frame per second, for example, for 60 frames. For each captured image frame, the facial region is identified and divided into pairs of symmetrical physiological regions, such as the left and right forehead regions, left and right cheek regions, and left and right eye corner regions. The average temperature difference, temperature distribution correlation coefficient, and structural similarity index of each symmetrical region pair in each frame are calculated. For example, for the left and right forehead regions, the difference between the average temperature of the left forehead and the average temperature of the right forehead are calculated; the correlation coefficient between the pixel temperature distributions of the left and right forehead regions is calculated; and the structural similarity index of the image blocks of the left and right forehead regions is calculated. The average temperature differences of all left and right forehead regions calculated from these 60 frames were statistically analyzed to determine their distribution range. For example, the mean and standard deviation were calculated, and the mean plus or minus two standard deviations was used as the normal baseline range. Similar statistics were performed on the average temperature differences, temperature distribution correlation coefficients, and structural similarity indices of the left and right cheeks and left and right canthus regions to establish their respective normal baseline ranges. This established a normal symmetry baseline range for a clean environment, providing a reference for subsequent comparisons of symmetry indicators in real-time monitoring.

[0092] The present application further proposes dynamically adjusting the confidence level of the thermal sensation evaluation result or switching the thermal sensation evaluation strategy according to the level of the comprehensive symmetry breaking index SBI, wherein step S4 includes:

[0093] S41: divided into no interference / slight interference, moderate interference, and severe interference levels based on the SBI threshold;

[0094] S42: Use SBI as an influencing factor for the confidence of the thermal sensation assessment result. If SBI is at the no interference / slight interference level, the confidence of the thermal sensation assessment result calculated based on the apparent temperature is determined to be high. As SBI increases, the confidence of the assessment result is reduced.

[0095] Specifically, the technical solution refines the specific method of adjusting the confidence of the thermal sensation evaluation results based on the comprehensive symmetry breaking index SBI. Step S41 divides the SBI value range into different levels by setting the SBI threshold. These levels represent different degrees of interference of dust on infrared radiation transmission, such as no interference / slight interference, moderate interference, and severe interference. This division provides a framework for quantifying the degree of interference. Step S42 uses the interference level divided by S41 to dynamically adjust the confidence of the thermal sensation evaluation results. When the SBI is at the no interference / slight interference level, it indicates that the dust has little interference with infrared radiation. At this time, the thermal sensation evaluation results calculated based on the apparent temperature are considered to have high confidence. As the SBI value increases, that is, the interference level develops from moderate to severe, the confidence of the evaluation results decreases. By associating SBI with specific interference levels and directly mapping these levels to the confidence of the evaluation results, the solution provides a clear mechanism to judge and quantify the reliability of thermal sensation evaluation results under different dust interference conditions. This solves the problem of difficulty in accurately determining the reliability of assessment results under varying levels of interference, making subsequent environmental control decisions based on the assessment results more robust. S41 provides a basis for classifying interference levels, and S42 provides rules for adjusting confidence levels based on these classifications.

[0096] This application further proposes to divide the comprehensive symmetry breaking index (SBI) into several levels, and dynamically adjust the confidence level of the thermal sensation evaluation result or switch the thermal sensation evaluation strategy according to the SBI level. The switching thermal sensation evaluation strategy includes:

[0097] Setting a first threshold and a second threshold, wherein the first threshold is smaller than the second threshold;

[0098] When the SBI is less than or equal to the first threshold, the confidence level of the thermal assessment result is high, and the original apparent temperature is used for thermal assessment. When the SBI is greater than the first threshold and less than or equal to the second threshold, the environmental control adjustment based on the measurement is temporarily suspended, and the previous high-confidence control adjustment mode is maintained or the preset conservative control mode is adopted. When the SBI is greater than the second threshold, the infrared thermal imaging assessment result is disabled, and the system switches to a robust thermal assessment model based on environmental sensor data.

[0099] Outputs an alarm message to indicate possible measurement interference and recommends checking the environment or performing manual confirmation.

[0100] Specifically, to address the reduced accuracy of thermal assessment in dust-interference environments, this solution introduces a comprehensive symmetry breaking index (SBI) as a measure of the degree of dust interference on infrared radiation transmission. Based on this index, the thermal assessment strategy is dynamically adjusted or switched. First, two SBI thresholds are set to categorize the degree of interference into low, medium, and high levels. When the SBI value is at a low interference level (less than or equal to the first threshold), the infrared thermal imaging data is minimally affected by dust, resulting in a highly reliable assessment result. The system then directly uses the thermal assessment results based on the raw apparent temperature to guide environmental control. When the SBI value is at a moderate interference level (greater than the first threshold and less than or equal to the second threshold), the infrared thermal imaging data is subject to some influence, posing a risk for direct control. The system then suspends environmental control adjustments based on the current measurement results, maintaining the previous control state or adopting a pre-set conservative control mode, thereby avoiding erroneous adjustments caused by inaccurate data. When the SBI value reaches a severe interference level (greater than the second threshold), indicating that the infrared thermal imaging data is unreliable, the system completely abandons the infrared thermal imaging assessment results and switches to a robust thermal assessment model that uses environmental sensor data (such as dust concentration, temperature and humidity). This ensures that usable assessment results can still be obtained under severe interference conditions. Furthermore, when the SBI value reaches a moderate or severe interference level, the system outputs an alarm message, prompting the operator to check the environment or perform manual confirmation, thereby improving the system's safety and practicality.

[0101] In some specific embodiments, the first threshold is set to 0.1, and the second threshold is set to 0.3. When the calculated comprehensive symmetry breaking index (SBI) is less than or equal to 0.1, the system uses facial apparent temperature data acquired by the infrared thermal imager, combined with a thermal perception model calibrated in a clean environment, to calculate the thermal perception index and adjust ventilation or air conditioning equipment based on this index. When the SBI is greater than 0.1 and less than or equal to 0.3, the system does not adjust environmental control equipment based on the currently calculated thermal perception index, but instead maintains the control mode determined when the SBI was less than or equal to 0.1. Simultaneously, the system displays a warning message on the user interface: "Moderate dust interference, control adjustment suspended." When the SBI is greater than 0.3, the system ignores the infrared thermal imager data and instead reads data from the dust concentration sensor and the ambient temperature and humidity sensor. This data is input into a robust thermal perception assessment model pre-calibrated at different dust concentrations to obtain a revised thermal perception index, which is then used to perform environmental control. Simultaneously, the system displays a warning message on the user interface: "Severe dust interference, switched to backup assessment mode, please check the environment." As a result, the system can dynamically adjust the evaluation and control strategies according to the actual degree of dust interference, improving the accuracy of thermal sensation assessment and the robustness of environmental control in industrial environments.

[0102] This application further proposes that environmental sensor data information includes:

[0103] Obtain data from dust concentration sensors and ambient temperature and humidity sensors;

[0104] The data are input into the robust thermal sensation evaluation model under offline calibrated dust interference conditions to generate a revised thermal sensation index.

[0105] Data from dust concentration sensors and ambient temperature and humidity sensors provide alternative environmental inputs. Dust concentration data directly reflects the intensity of environmental factors that may interfere with infrared measurements. Ambient temperature and humidity data are fundamental parameters affecting human thermal perception and are generally unaffected by the optical interference of airborne dust. These data are essential inputs for thermal assessment. Inputting this data into an offline calibrated robust thermal assessment model under dust interference conditions is the core of robust assessment. Using a model that has been pre-calibrated or trained in dust-interfering environments enables more accurate processing of data from environmental sensors and accounts for the potential impact of dust on human thermal perception and sensor data interpretation. The robustness of the model ensures that the system can still provide a reliable thermal assessment even if infrared data fails. Offline calibration ensures the accuracy of the model before practical application. The output of the assessment process is a corrected thermal sensation index. The thermal sensation index calculated based on environmental sensor data and the robustness model is an effective estimate in the presence of dust interference. The term "corrected" indicates that it may differ from the standard thermal sensation index calculated in a clean environment, or may be more accurate after accounting for interference factors. This index can be used to guide the adjustment of environmental control systems.

[0106] Specifically, when infrared thermal imaging evaluation is unavailable due to interference, the system switches to a thermal sensation evaluation scheme based on environmental sensor data. The system obtains data from dust concentration sensors and ambient temperature and humidity sensors. These data are input into a robust thermal sensation evaluation model that has been calibrated or trained in advance under environmental conditions with dust interference. The model processes the sensor data and takes into account the possible impact of dust on human thermal perception or sensor data interpretation, thereby generating a corrected thermal sensation index. This corrected thermal sensation index is an effective estimate in the current environment with dust interference and is used to guide the adjustment of the environmental control system. In this way, the problem of being unable to perform effective thermal sensation evaluation when infrared thermal imaging evaluation is interfered with is solved, and a backup and robust evaluation method based on environmental sensor data is provided to ensure the system's ability to continue operating in complex environments.

[0107] In some embodiments, the system deploys an optical dust concentration sensor and an integrated temperature and humidity sensor. The dust concentration sensor outputs the dust mass concentration per unit volume (e.g., mg / m³) by measuring light scattering or attenuation by air. The temperature and humidity sensor outputs ambient temperature (e.g., °C) and relative humidity (e.g., %RH). These sensor data are periodically transmitted to a processing unit (e.g., an industrial PC or PLC) via an industrial communication bus (e.g., Modbus RTU). The processing unit runs a pre-trained machine learning model (e.g., a support vector regression model) that uses dust concentration, ambient temperature, and relative humidity as input features. This model is trained and calibrated offline by collecting sensor data and simultaneously recording personnel's subjective thermal sensation ratings in controlled or real-world industrial environments with varying dust concentration, temperature, and humidity combinations, or using other dust-unaffected reference methods (e.g., calculations based on a standard PMV model based on metabolic rate, clothing, and air velocity). When infrared thermal imaging assessment is deemed unavailable, the system activates the model, feeding it real-time sensor data. The model then outputs a modified predicted mean vote (PMV) index. This index reflects a person's expected thermal sensation under current environmental conditions, accounting for the potential impact of dust. This provides a reliable indicator of thermal sensation, even if infrared measurements are disrupted. This information can be used to adjust ventilation or air conditioning system operating parameters to maintain thermal comfort.

[0108] refer to Figure 2 This application further proposes a thermal sensation evaluation system based on infrared thermal imaging technology, which is applied to the steps of any of the above methods. The system includes

[0109] The data acquisition module 201 acquires a sequence of infrared thermal images of the target object's face captured by an infrared thermal imager;

[0110] Symmetry analysis module 202, which divides at least one pair of symmetrical physiological regions based on the left-right symmetrical physiological regions of the human face;

[0111] Interference judgment module 203 calculates the dynamic symmetry index of each current symmetrical physiological region pair in the real-time infrared thermal image sequence, compares the dynamic symmetry index with the normal symmetry reference range under a clean environment, and generates a comprehensive symmetry breaking index (SBI). The SBI is used to quantify the degree of dust interference with infrared radiation transmission;

[0112] The evaluation adjustment module 204 divides the comprehensive symmetry breaking index SBI into several levels, and dynamically adjusts the confidence level of the thermal sensation evaluation result or switches the thermal sensation evaluation strategy according to the SBI level.

[0113] Among them, the data acquisition module 201 is configured to obtain a sequence of infrared thermal images of the target object's face captured by an infrared thermal imager. The symmetry analysis module 202 is configured to divide at least one pair of symmetrical physiological regions based on the left-right symmetrical physiological regions of the human face. The interference judgment module 203 is configured to calculate the dynamic symmetry index of each current symmetrical physiological region pair in the infrared thermal image sequence acquired in real time, compare the dynamic symmetry index with the normal symmetry reference range under a clean environment, and generate a comprehensive symmetry breaking index SBI, which is used to quantify the degree of interference of dust on infrared radiation transmission. The evaluation and adjustment module 204 is configured to divide the comprehensive symmetry breaking index SBI into several levels, and dynamically adjust the confidence of the thermal sensation evaluation result or switch the thermal sensation evaluation strategy according to the level of SBI.

[0114] In some specific embodiments, the data acquisition module 201 receives a sequence of 640x480 pixel infrared thermal image frames from a fixed-mounted infrared thermal imager at a frequency of 1 Hz. The symmetry analysis module 202 uses a pretrained convolutional neural network model to detect the facial region in the image and further identifies pixel sets in the left and right forehead regions and the left and right cheek regions. For each image frame, the interference determination module 203 calculates the Pearson correlation coefficient between the average temperature difference and temperature distribution of the left and right forehead regions, as well as the Pearson correlation coefficient between the average temperature difference and temperature distribution of the left and right cheek regions. The normal symmetry baseline range in a clean environment is set as: an average forehead temperature difference of less than 0.8°C with a correlation coefficient greater than 0.9; an average cheek temperature difference of less than 0.6°C with a correlation coefficient greater than 0.92. The SBI is calculated as the weighted average of the deviations of each indicator from its upper limit of the normal baseline range. The assessment and adjustment module 204 sets the first threshold of the SBI to 0.15 and the second threshold to 0.4. When the SBI is less than or equal to 0.15, the confidence level of the thermal assessment result is set to high, and the standard thermal model based on infrared apparent temperature is used. When the SBI is greater than 0.15 and less than or equal to 0.4, the confidence level of the assessment result decreases, and the system maintains the previous high-confidence control mode. When the SBI is greater than 0.4, the infrared thermal imaging assessment result is disabled, and the thermal assessment model based on dust concentration and ambient temperature and humidity sensor data is switched to.

[0115] The basic principles, main features, and advantages of the present invention are shown and described above. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. Various changes and improvements are possible without departing from the spirit and scope of the present invention, and such changes and improvements fall within the scope of the invention as claimed.

Claims

1. A thermal sensation assessment method based on infrared thermal imaging technology, characterized in that: The method comprises the following steps: S1: Acquire a facial infrared thermal image sequence of the target object captured by an infrared thermal imager; S2: dividing at least one pair of symmetrical physiological regions based on the left-right symmetrical physiological regions of the human face; S3: Calculating the dynamic symmetry index of each current pair of symmetrical physiological regions in the real-time acquired infrared thermal image sequence, comparing the dynamic symmetry index with a normal symmetry reference range in a clean environment, and generating a comprehensive symmetry breaking index (SBI). The SBI is used to quantify the degree of interference of dust on infrared radiation transmission. S4: The comprehensive symmetry breaking index (SBI) is divided into several levels, and the confidence level of the thermal sensation assessment results is dynamically adjusted or the thermal sensation assessment strategy is switched according to the SBI level.

2. The thermal sensation evaluation method based on infrared thermal imaging technology according to claim 1, characterized in that: Step S1 includes: S11: periodically capturing original images containing faces using a fixedly mounted infrared thermal imager; S12: Noise is filtered out of the original image and an image registration technology based on feature point matching is used to align a multi-frame infrared thermal image sequence to the same spatial coordinate system to obtain an infrared thermal image sequence of the target object's face.

3. The thermal sensation evaluation method based on infrared thermal imaging technology according to claim 1, characterized in that: Step S2 includes: S21: Using facial recognition algorithm to locate the face area in infrared thermal image sequence; S22: Based on a preset anatomical landmark or template, the facial midline is used as the symmetry axis to automatically divide the face into a number of left-right symmetrical physiological region pairs, including left and right forehead regions, left and right cheek regions, and left and right eye canthus regions.

4. The thermal sensation evaluation method based on infrared thermal imaging technology according to claim 1, characterized in that: Step S3 includes: S31: Repeat step S2 for each frame of the infrared thermal image sequence to divide the human face into symmetrical physiological area pairs; S32: calculating the symmetry metrics of each pair of symmetrical physiological regions in each frame to obtain a real-time dynamic symmetry index; S33: Compare the dynamic symmetry index with the normal symmetry reference range under a clean environment, and calculate the weighted average or maximum value of the selected index's deviation from the reference range, which is the comprehensive symmetry breaking index SBI.

5. The thermal sensation evaluation method based on infrared thermal imaging technology according to claim 4, characterized in that: The symmetry index includes temperature difference, temperature distribution correlation coefficient and similarity index.

6. The thermal sensation evaluation method based on infrared thermal imaging technology according to claim 5, characterized in that: The establishment of the normal symmetry reference range in the clean environment includes: When the dust sensor is used to assist in determining that the air environment is clean, a multi-frame reference infrared thermal image sequence is collected; Calculate the average temperature difference, temperature distribution correlation coefficient, and structural similarity index of each symmetry index in each pair of symmetrical physiological regions; Calculate the distribution range of each symmetry index and establish a normal symmetry benchmark.

7. The thermal sensation evaluation method based on infrared thermal imaging technology according to claim 1, characterized in that: Step S4 includes: S41: divided into no interference / slight interference, moderate interference, and severe interference levels based on the SBI threshold; S42: Use SBI as an influencing factor for the confidence of the thermal sensation assessment result. If SBI is at the no interference / slight interference level, the confidence of the thermal sensation assessment result calculated based on the apparent temperature is determined to be high. As SBI increases, the confidence of the assessment result is reduced.

8. The thermal sensation evaluation method based on infrared thermal imaging technology according to claim 7, characterized in that: The switching thermal sensation assessment strategy includes: Setting a first threshold and a second threshold, wherein the first threshold is smaller than the second threshold; When the SBI is less than or equal to the first threshold, the confidence level of the thermal assessment result is high, and the original apparent temperature is used for thermal assessment. When the SBI is greater than the first threshold and less than or equal to the second threshold, the environmental control adjustment based on the measurement is temporarily suspended, and the previous high-confidence control adjustment mode is maintained or the preset conservative control mode is adopted. When the SBI is greater than the second threshold, the infrared thermal imaging assessment result is disabled, and the system switches to a robust thermal assessment model based on environmental sensor data. Outputs an alarm message to indicate possible measurement interference and recommends checking the environment or performing manual confirmation.

9. The thermal sensation evaluation method based on infrared thermal imaging technology according to claim 8, characterized in that: The environmental sensor data information includes: Obtain data from dust concentration sensors and ambient temperature and humidity sensors; The data are input into a robust thermal sensation evaluation model under dust interference conditions that has been calibrated offline to generate a revised thermal sensation index.

10. A thermal sensation assessment system based on infrared thermal imaging technology, characterized in that: The system is applied to the steps of any one of the methods described in claims 1 to 9 above, and the system includes: A data acquisition module, which acquires a facial infrared thermal image sequence of a target object captured by an infrared thermal imager; A symmetry analysis module, which divides at least one pair of symmetrical physiological regions based on the left and right symmetrical physiological regions of the human face; An interference judgment module calculates the dynamic symmetry index of each current pair of symmetrical physiological regions in the real-time infrared thermal image sequence, compares the dynamic symmetry index with the normal symmetry reference range in a clean environment, and generates a comprehensive symmetry breaking index (SBI). The SBI is used to quantify the degree of dust interference with infrared radiation transmission; The evaluation and adjustment module divides the comprehensive symmetry breaking index (SBI) into several levels and dynamically adjusts the confidence level of the thermal sensation evaluation results or switches the thermal sensation evaluation strategy according to the SBI level.

Images