An information processing method based on infrared images, medium and device

Abstract

The present application relates to the technical field of data processing, and more particularly to an information processing method based on infrared images, a medium and equipment, which corrects the temperature of an infrared thermal image through an environmental temperature parameter, eliminates the influence of environmental temperature fluctuations on infrared temperature measurement accuracy, performs time domain interframe difference processing on multiple continuous first infrared thermal images, identifies a motion area and performs mask suppression, eliminates chest wall fluctuation artifacts caused by respiratory motion, and improves the reliability of a single infrared thermal image, automatically locates a reference point by calling a model, extracts corresponding associated sub-areas of each preset body part according to automatic rules, avoids fluctuations in the range of areas caused by individual experience differences, ensures the stability and repeatability of area demarcation, and calculates composite temperature characteristics containing temperature statistical characteristics of each associated sub-area and cross-area temperature correlation characteristics through the temperature values of each pixel point in the target composite area, thereby improving the accuracy of angina infrared thermal image analysis.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to an information processing method, medium, and device based on infrared images. Background Technology

[0002] Infrared thermal imaging, as a non-invasive and visual functional imaging technology, captures infrared radiation from the human body surface and converts it into a pseudo-color temperature distribution map, which can reflect physiological changes such as local blood flow and metabolism in the body. It shows application potential in scenarios such as the auxiliary diagnosis of chest pain in traditional Chinese medicine.

[0003] However, existing methods for analyzing chest pain related to infrared thermography mostly employ a manual analysis model. Physicians manually delineate independent observation areas such as the area around the Tanzhong acupoint, anatomical regions of the chest, symmetrical areas of the back, and along the Hand Shaoyin Heart Meridian / Hand Jueyin Pericardium Meridian. Temperature data for each area is then extracted using analysis software. The range of chest pain-related areas delineated by different physicians varies significantly, and only supports independent temperature statistics for single areas, failing to match the cross-regional temperature characteristics of chest pain. Furthermore, chest wall fluctuations caused by the subject's breathing introduce motion artifacts, causing fluctuations in the temperature values ​​of lesion areas in the infrared image. Environmental temperature fluctuations, the temperature of hair and residual clothing on the skin surface, and other factors also interfere with the measurement results, leading to significant analytical bias and failing to accurately assist in traditional Chinese medicine diagnosis.

[0004] Therefore, how to achieve standardized automatic extraction of chest pain-related areas, capture cross-regional temperature linkage characteristics, and effectively suppress interfering factors such as respiratory movements in infrared thermal imaging analysis, thereby improving the accuracy of chest pain infrared image analysis, has become an urgent technical problem to be solved. Summary of the Invention

[0005] To address the aforementioned technical problems, the present invention provides an information processing method based on infrared images, which includes the following steps: S1, for any preset body part of the target object, based on the ambient temperature when the original infrared thermal image sequence corresponding to the preset body part was acquired, perform temperature reference correction on each original infrared thermal image in the original infrared thermal image sequence to obtain the first infrared thermal image sequence corresponding to the preset body part after temperature correction.

[0006] S2, perform temporal inter-frame difference processing on multiple consecutive first infrared thermal images in the first infrared thermal image sequence, identify motion regions based on the difference results, and perform mask suppression on the motion regions to obtain a single-frame infrared thermal image after motion artifact suppression corresponding to the preset body part.

[0007] S3, invoke the preset human key point detection model, locate the pixel coordinates of several preset reference points corresponding to the preset body part on the single frame infrared thermal image, and extract the associated sub-region corresponding to the preset body part from the single frame infrared thermal image according to the pixel coordinates and the preset composite region delineation rules.

[0008] S4: Traverse all preset body parts and merge the associated sub-regions corresponding to each preset body part to obtain the target composite region.

[0009] S5. Calculate the composite temperature features of the target composite region based on the temperature values ​​of each pixel within the target composite region. The composite temperature features include the temperature statistical features of each associated sub-region and the temperature correlation features between different associated sub-regions. The composite temperature features are used to characterize the infrared feature information of the target object.

[0010] The present invention also provides a non-transitory computer-readable storage medium storing at least one instruction or at least one program, wherein the at least one instruction or at least one program is loaded and executed by a processor to implement the above-described information processing method based on infrared images.

[0011] The present invention also provides an electronic device, including a processor and the aforementioned non-transitory computer-readable storage medium.

[0012] This invention has at least the following beneficial effects: By correcting the original infrared thermal image using ambient temperature parameters, the temperature values ​​of infrared thermal images acquired at different times and under different environmental conditions are unified to the same temperature reference, eliminating the adverse effects of ambient temperature fluctuations on the accuracy of infrared temperature measurement; by performing temporal inter-frame difference processing on multiple consecutive first infrared thermal images, motion regions are identified and mask suppression is applied, effectively eliminating chest wall undulation artifacts caused by respiratory movements, and outputting stable and reliable single-frame infrared thermal images while preserving the original image's spatial structure and temperature distribution characteristics; by calling preset human... The body key point detection model automatically locates the reference points and automatically extracts the associated sub-regions corresponding to each preset body part according to the preset composite region delineation rules. This avoids fluctuations in the region range caused by individual experience differences and ensures the stability and repeatability of the delineation results. By calculating the temperature value of each pixel in the target composite region, a composite temperature feature containing the temperature statistical characteristics of each associated sub-region and the cross-regional temperature correlation characteristics is calculated. This fully reflects the characteristics of cross-regional temperature linkage in chest pain, improves the accuracy of chest pain infrared thermographic analysis, and provides objective, standardized, and interpretable image analysis information for chest pain diagnosis. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a flowchart of an information processing method based on infrared images provided in Embodiment 1 of the present invention. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It is understood that, where appropriate, the terms used to distinguish similar objects can be interchanged so that the invention can also be implemented in other embodiments besides the illustrated or described embodiments. Furthermore, the terms "including," "having," and any variations are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products, or devices.

[0017] Example 1 This embodiment provides an information processing method based on infrared images, such as... Figure 1 As shown, the information processing method based on infrared images includes the following steps: S1, for any preset body part of the target object, based on the ambient temperature when the original infrared thermal image sequence corresponding to the preset body part was acquired, perform temperature reference correction on each original infrared thermal image in the original infrared thermal image sequence to obtain the first infrared thermal image sequence corresponding to the preset body part after temperature correction.

[0018] The target object refers to the individual who receives infrared thermal image acquisition and information processing.

[0019] Based on a comprehensive consideration of the pathophysiological mechanisms of chest pain, traditional Chinese medicine meridian theory, and clinical research evidence, the core pathology of chest pain lies in myocardial ischemia. The heart is located deep in the left side of the anterior chest. When myocardial ischemia occurs, changes in local metabolism and blood flow alter the surface temperature distribution in the heart's projection area. Therefore, the anterior chest is the most direct and crucial surface observation site for capturing the state of the heart.

[0020] Furthermore, during an attack of angina pectoris, the abnormal signals caused by myocardial ischemia are not limited to the anterior chest area but are also transmitted to the corresponding area on the back, i.e., the surface reflex zone of the heart, through nerve reflex pathways. Temperature abnormalities in the anterior chest usually manifest as symmetrical temperature changes in the corresponding area of ​​the back. Therefore, this embodiment uses the human body regional symmetry method to establish the symmetrical point of the Tanzhong acupoint on the back as the back observation area to capture the cross-regional temperature linkage characteristics of angina pectoris, thus forming a symmetrical temperature difference correlation characteristic between the anterior chest and the back.

[0021] Secondly, from the perspective of Traditional Chinese Medicine (TCM), both the Hand Shaoyin Heart Meridian and the Hand Jueyin Pericardium Meridian run along the inner side of the left arm, and the corresponding meridians also run along the symmetrical side of their pathways, namely the inner side of the right arm. In healthy individuals, the temperature distribution of these corresponding meridians tends to be symmetrical on both sides. However, in patients with chest pain (angina pectoris), due to abnormal cardiac function leading to disordered Qi and blood circulation in the Hand Shaoyin Heart Meridian and Hand Jueyin Pericardium Meridian, asymmetrical temperature distribution can occur on both sides. Clinical studies have confirmed that temperature changes on the left Hand Shaoyin Heart Meridian and Hand Jueyin Pericardium Meridian are more sensitive in patients with chest pain. Therefore, the temperature gradient changes along the meridian pathways of the inner left and right arms, which are the surface projection areas of the Hand Shaoyin Heart Meridian and Hand Jueyin Pericardium Meridian, are important reference indicators for diagnosing chest pain. By comparing the temperature distribution differences between the left and right arm meridian regions, a symmetrical temperature difference correlation characteristic between the left and right arm meridians can be formed, further enriching the information dimensions of cross-regional temperature linkage analysis.

[0022] Therefore, this embodiment uses the front chest, back, inner left arm, and inner right arm as preset body parts to comprehensively capture multi-dimensional temperature signals of chest pain and improve the accuracy of infrared thermography analysis.

[0023] Furthermore, the raw temperature values ​​output by infrared thermal imagers are significantly affected by ambient temperature. When the ambient temperature changes, the response characteristics of the infrared detector will drift, resulting in an overall shift in the temperature measurement results. For example, the temperature values ​​of the same target object measured by the same thermal imager at an ambient temperature of 25°C and 20°C may differ by more than 0.5°C.

[0024] Therefore, in this embodiment, an infrared thermal imager is used to acquire multiple consecutive raw infrared thermal images corresponding to each preset body part, forming a raw infrared thermal image sequence. The acquisition frequency and the number of frames in each raw infrared thermal image sequence can be set by the implementer according to the actual situation. For example, in this embodiment, the acquisition frequency is 25 frames / second and the number of frames is 5.

[0025] While acquiring the original infrared thermal image, the ambient temperature at the time of image acquisition is obtained through the temperature sensor (such as a thermocouple) built into the thermal imager. The pixel temperature value of the corresponding original infrared thermal image is compensated and corrected based on the ambient temperature. The temperature values ​​of each frame of infrared thermal image are uniformly normalized to the same temperature reference, which can eliminate the measurement error introduced by the ambient temperature fluctuation and make the infrared thermal images acquired at different times and under different environmental conditions comparable.

[0026] In one specific embodiment, S1 includes the following steps: S11, obtain the pre-stored ambient temperature-temperature offset mapping relationship, and determine the temperature offset corresponding to the preset body part based on the ambient temperature and the mapping relationship.

[0027] S12, perform offset correction on the pixel temperature values ​​of each original infrared thermal image in the original infrared thermal image sequence according to the temperature offset, and obtain the first infrared thermal image corresponding to each original infrared thermal image.

[0028] There is a defined mapping relationship between the temperature measurement error of the infrared thermal imager and the ambient temperature. By calibrating the thermal imager using a standard blackbody under multiple known ambient temperatures, a mapping relationship between ambient temperature and temperature offset can be established. In this embodiment, the thermal imager is calibrated using a standard blackbody under different combinations of ambient temperature and core temperature, and three sets of data are recorded: ambient temperature, core temperature, and the temperature measurement value output by the thermal imager. Using the ambient temperature and core temperature as inputs and the temperature measurement error as output, a support vector machine is used for fitting to obtain an error correction model.

[0029] In actual measurement, the corresponding temperature offset can be obtained by querying the mapping relationship based on the current ambient temperature.

[0030] This temperature offset is applied to every pixel in each of the original infrared thermal images in the original infrared thermal image sequence. The temperature offset is then uniformly added to or subtracted from the temperature values ​​of all pixels in the entire original infrared thermal image, achieving overall translation correction for the entire original infrared thermal image. Correspondingly, if the ambient temperature is higher than the calibration temperature, the detector response is enhanced, leading to a higher measured value, and the temperature offset needs to be subtracted; if the ambient temperature is lower than the calibration temperature, the measured value is lower, and the temperature offset needs to be added.

[0031] As described above, the original infrared thermal image is calibrated based on the ambient temperature at the time of acquisition, so that the temperature values ​​of infrared thermal images acquired at different times and under different environmental conditions are unified to the same temperature reference. This eliminates the adverse effects of ambient temperature fluctuations on the accuracy of infrared temperature measurement and provides an accurate and comparable temperature data basis for subsequent calculation of composite temperature features across frames and locations.

[0032] S2, perform temporal inter-frame difference processing on multiple consecutive first infrared thermal images in the first infrared thermal image sequence, identify motion regions based on the difference results, and perform mask suppression on the motion regions to obtain a single-frame infrared thermal image after motion artifact suppression corresponding to the preset body part.

[0033] During infrared thermography acquisition, the respiratory movements of the subject cause rhythmic rises and falls of the chest, resulting in positional shifts in temperature across different areas of the body surface between multiple thermal images. For example, a point on the chest may correspond to the temperature of the heart's projection area during exhalation, but during inhalation it may correspond to the temperature of the intercostal spaces or lung tissue. This leads to spurious temperature changes for the same pixel across consecutive frames, and this motion artifact interferes with the extraction of the true temperature features of the lesion area.

[0034] Motion artifacts caused by the rise and fall of the thoracic cavity can lead to abrupt changes in the temperature value of the same pixel between adjacent frames, while the temperature of stationary areas remains almost unchanged in a short period of time due to the thermal inertia of the tissue. Therefore, this embodiment utilizes the difference between the time-dimensional variation characteristics of motion artifacts and the stability of the actual tissue temperature in a short period of time. By performing inter-frame differential processing on multiple consecutive first infrared thermal images, the temperature change of each pixel between adjacent frames can be identified. This differential result can distinguish between moving and stationary areas.

[0035] Furthermore, the identified motion areas are located using a mask, and the temperature values ​​of the motion areas are replaced by multi-frame averaging. This eliminates artifacts caused by breathing movements, resulting in a stable and reliable single-frame infrared thermal image.

[0036] In one specific embodiment, S2 includes the following steps: S21, acquire N consecutive first infrared thermal images from the first infrared thermal image sequence, and use a specified frame from the N consecutive first infrared thermal images as a reference frame, where N > 2.

[0037] S22, perform inter-frame differential processing on N consecutive frames of the first infrared thermal image to obtain the temperature change of each pixel between adjacent frames.

[0038] S23, pixels whose temperature changes exceed a preset differential threshold are identified as moving points.

[0039] S24, generate a mask for the motion region composed of all moving points, and determine the pixel region to be replaced with temperature value based on the mask.

[0040] S25, calculate the average temperature of each pixel point corresponding to the pixel area in the first infrared thermal image of N consecutive frames.

[0041] S26, replace the temperature value of each pixel in the pixel area of ​​the reference frame with the corresponding average temperature, and keep the temperature value of the pixel outside the pixel area of ​​the reference frame unchanged, and obtain a single-frame infrared thermal image of the preset body part after motion artifact suppression.

[0042] In this case, temporal inter-frame differential processing requires at least two frames to form a comparison of changes. However, only two frames can detect changes between these two moments, which cannot cover motion information throughout the entire respiratory cycle. Therefore, this embodiment extracts N consecutive frames from the first infrared thermal image sequence, and sets N > 2, such as N = 5, to provide multi-frame information in the temporal dimension for subsequent differential processing.

[0043] Simultaneously, a designated frame is selected from N frames as a reference frame, serving as the structural and temperature basis for the final output single frame. In this embodiment, the first frame of the first infrared thermal image of N consecutive frames can be used as the designated frame.

[0044] Infrared thermal imagers exhibit minimal temperature drift over short periods, and the temperature values ​​of stationary areas should remain largely consistent across different frames. However, the chest wall fluctuations caused by respiratory movements can lead to significant temperature changes for the same pixel between inspiratory and expiratory frames. Therefore, by calculating the temperature difference pixel-by-pixel between adjacent frames, the temperature changes caused by motion can be separated from the static temperature distribution.

[0045] In the inter-frame difference results, pixels in moving regions experience significant temperature changes due to chest wall fluctuations caused by respiration, typically exceeding 0.2℃. Conversely, pixels in stationary regions exhibit almost no temperature change over a short period due to tissue thermal inertia; their temperature change originates solely from detector noise and is usually far less than 0.1℃. By setting a reasonable preset difference threshold, moving and stationary points can be distinguished. The specific value of the preset difference threshold can be set by the implementer based on actual conditions; for example, in this embodiment, it is set to 0.2℃.

[0046] Multiple moving points may be scattered across different locations in the image. The set of all moving points is integrated into a moving region, generating a corresponding binary mask Mask(x,y) to identify the pixel regions to be replaced with temperature values ​​and the regions that remain unchanged and do not require replacement. This allows subsequent processing to apply only to the target region. Correspondingly, if a pixel (x,y) belongs to the moving region, the corresponding Mask(x,y) = 1; if a pixel (x,y) does not belong to the moving region, the corresponding Mask(x,y) = 0. The regions in the mask with a value of 1 are the pixel regions to be replaced with temperature values.

[0047] For the motion region marked by the mask, breathing artifacts cause inconsistent temperature values ​​at the same pixel location across frames, and directly using the value from a single frame will introduce errors. However, in an N-frame continuous heatmap, the specific manifestation of breathing artifacts differs across frames; for example, the motion position differs between inhalation and exhalation frames, but the background temperature level of the region remains consistent across all frames. Therefore, for each pixel location within the motion region, the average temperature across N frames is calculated, and this average temperature represents the true temperature level of the region without breathing artifacts.

[0048] Furthermore, the average temperature is assigned to each pixel within the pixel region of the reference frame, while all pixel values ​​in the non-motion region of the reference frame are kept unchanged. This eliminates motion artifacts while preserving the spatial structure and resolution of the original image, ultimately outputting a stable, reliable, and spatially complete single-frame infrared thermal image after motion artifact suppression.

[0049] In one specific implementation, after determining the motion region, morphological operations such as dilation and closure can be used to post-process the motion region to fill the voids inside the motion region and smooth the region boundaries, making the mask region more complete and continuous.

[0050] As described above, by performing temporal inter-frame difference processing on multiple consecutive first infrared thermal images in the first infrared thermal image sequence, the moving region is identified, and mask suppression is applied to the moving region. The average temperature of multiple frames is used to replace the original temperature value of the moving region, which effectively eliminates the chest wall undulation artifact caused by respiratory movement. At the same time, the original spatial structure and temperature information of the non-moving region are completely preserved. Finally, a stable and reliable single-frame infrared thermal image is output, providing a high-quality image input for subsequent key point detection and associated sub-region extraction.

[0051] S3, invoke the preset human key point detection model, locate the pixel coordinates of several preset reference points corresponding to the preset body part on the single frame infrared thermal image, and extract the associated sub-region corresponding to the preset body part from the single frame infrared thermal image according to the pixel coordinates and the preset composite region delineation rules.

[0052] The preset human keypoint detection model is a pre-trained deep learning model used to automatically locate key anatomical landmarks on images. It automatically identifies and locates the pixel coordinates of preset reference points on infrared thermograms, replacing manual labeling and achieving standardized and automated region delineation. For example, in this embodiment, the preset human keypoint detection model is the HRNet model, which maintains high-resolution representation throughout the network and achieves accurate spatial localization while ensuring rich semantic information through parallel multi-resolution branches and repeated multi-scale feature fusion. Specifically, the HRNet model consists of a backbone network and four cascaded stages. The backbone network performs preliminary feature extraction on the input single-frame infrared thermogram, while the subsequent four cascaded stages are responsible for constructing and fusing multi-resolution features. Each cascaded stage consists of several high-resolution modules (HRModules). Each HRModule contains multiple parallel branches, each processing feature maps at different resolutions and exchanging information at the end of the module through a feature fusion layer. Each branch consists of several stacked residual units, which can be either BasicBlock or BottleneckBlock.

[0053] In practical applications, a preprocessed single-frame infrared thermal image is input into the HRNet model. The HRNet model performs inference through its parallel multi-resolution network structure and outputs a high-resolution thermal image. Based on the peak response location of the thermal image, the pixel coordinates of each preset reference point on the infrared thermal image can be accurately located.

[0054] Among them, the preset reference points are human anatomical landmarks that serve as coordinate references for delineating related sub-regions. Specifically, the preset reference points corresponding to the anterior chest include the point corresponding to the Tanzhong acupoint and the point corresponding to the boundary of the left thoracic anatomical region; the preset reference points corresponding to the posterior back include the point corresponding to the spinous process of the vertebral column; the preset reference points corresponding to the inner side of the left arm include the point corresponding to the Neiguan acupoint and the point corresponding to the axilla of the left arm; and the preset reference points corresponding to the inner side of the right arm include the point corresponding to the Neiguan acupoint and the point corresponding to the axilla of the right arm.

[0055] Specifically, the corresponding point for the Tanzhong acupoint is the pixel location on an infrared thermogram of the anatomical landmark located at the midpoint of the line connecting the two nipples on the midline of the anterior chest. The corresponding point for the boundary of the left thoracic anatomical region is the pixel location on an infrared thermogram of the anatomical landmark at the intersection of the left midclavicular line and the 4th intercostal space, corresponding to the boundary of the projection area of ​​the heart on the body surface. The corresponding point for the spinous processes of the spine on the back is the pixel location on an infrared thermogram of the spinous processes of the 5th to 6th thoracic vertebrae, which are the horizontally symmetrical points of the Tanzhong acupoint on the back. The corresponding point for the Neiguan acupoint on the left arm is the pixel location on an infrared thermogram of the anatomical landmark located approximately 2 inches above the wrist crease on the inner side of the left arm; the corresponding point for the Neiguan acupoint on the right arm is the pixel location on an infrared thermogram of the anatomical landmark located approximately 2 inches above the wrist crease on the inner side of the right arm. The corresponding point under the left armpit is the pixel position of the anatomical landmark located at the lower edge of the armpit on the inner side of the left armpit on the infrared thermograph; the corresponding point under the right armpit is the pixel position of the anatomical landmark located at the lower edge of the armpit on the inner side of the right armpit on the infrared thermograph.

[0056] After temperature baseline correction and motion artifact suppression, a single-frame infrared thermal image possesses a unified temperature baseline and stable image quality. However, each pixel in an infrared thermal image only contains temperature values ​​and does not include anatomical structural information. Therefore, it is necessary to automatically identify specific anatomical landmarks using a key point detection model and use the pixel coordinates corresponding to these landmarks as anchor points for subsequent region delineation.

[0057] Therefore, this embodiment calls a preset human key point detection model to automatically locate the pixel coordinates of preset reference points corresponding to each preset body part on the preprocessed single-frame infrared thermal image. Based on preset composite region delineation rules, including parameters such as the delineation method, expansion direction, and number of expansion pixels for each sub-region, it automatically calculates the pixel region range of each associated sub-region, generates corresponding pixel masks, and finally extracts the associated sub-regions corresponding to each preset body part from the single-frame infrared thermal image.

[0058] In one specific embodiment, S3 includes the following steps: S31, invoke the preset human body key point detection model, perform key point detection on a single frame infrared thermal image, and locate the pixel coordinates of several preset reference points corresponding to the preset body part.

[0059] S32, based on the pixel coordinates of each preset reference point and the preset composite region delineation rules, determine the pixel region range of the associated sub-region corresponding to the preset body part. Among them, the associated sub-region corresponding to the front chest is the front core sub-region, the associated sub-region corresponding to the back is the back reflection sub-region, and the associated sub-regions corresponding to the inner side of the left arm and the inner side of the right arm are the meridian associated sub-regions.

[0060] S33, generate a corresponding pixel mask based on the pixel area range, and extract the associated sub-region corresponding to the preset body part from the single-frame infrared thermal image based on the pixel mask.

[0061] In this process, based on the preset composite region delineation rules, the discrete reference point coordinates are converted into specific pixel region ranges, so that the associated sub-regions with actual physical areas can be extracted from a single frame infrared thermal image.

[0062] Pixel masking technology can accurately identify the boundaries of a region. Pixels with a value of 1 in the mask represent the associated sub-regions to be extracted, while those with a value of 0 are masked out. By performing a pixel-by-pixel AND operation between a single-frame infrared thermal image and the mask, the temperature values ​​of pixels outside the associated sub-regions can be set to zero, thus accurately extracting the temperature data within the associated sub-regions.

[0063] The above-mentioned method automatically locates preset reference points by calling the preset human key point detection model, and automatically extracts the associated sub-regions corresponding to each preset body part according to the preset composite region delineation rules. This transforms the delineation of associated sub-regions from relying on human experience judgment to a standardized process executed automatically by the computer, solving the problems of large differences in the range of manually delineated regions and insufficient consistency of results. It provides a standardized and repeatable region input for subsequent composite temperature feature calculation.

[0064] In one specific embodiment, the core sub-region of the chest is the area formed by extending to the left from the corresponding point of the Tanzhong acupoint to the corresponding point of the left chest anatomical division boundary.

[0065] The back reflection sub-region is the area formed by symmetrically mapping the chest core sub-region onto the back region along the horizontal direction on a single frame infrared thermogram, using the corresponding point of the spinous process of the back spine as the mapping reference point.

[0066] The meridian-related sub-region is a strip-shaped area formed by extending a preset pixel width to both sides of the line connecting the corresponding point under the left armpit to the corresponding point of the Neiguan acupoint on the left arm as the center line, and extending a preset pixel width to both sides of the line connecting the corresponding point under the right armpit to the corresponding point of the Neiguan acupoint on the right arm as the center line.

[0067] Based on the pathophysiological mechanism of chest pain, the core pathology of chest pain lies in the obstruction of the heart vessels, and the heart is located deep in the left side of the anterior chest. The Tanzhong acupoint is a landmark on the midline of the anterior chest, but the heart is actually located to the left of the Tanzhong acupoint. The square area centered on the Tanzhong acupoint cannot completely cover the surface projection area of ​​the heart.

[0068] Therefore, this embodiment uses the point corresponding to the Tanzhong acupoint as the center point, extending leftward to the corresponding point of the left thoracic anatomical region boundary, forming an irregular polygonal area that completely covers the anatomical region corresponding to the left thoracic heart projection. This core sub-region of the chest can comprehensively capture the temperature distribution of the heart region, providing a complete spatial range for subsequent calculation of temperature statistical features. Furthermore, the number of expanded pixels can be dynamically calculated based on the height and pixel ratio of the target object, or a fixed number of pixels can be used, such as expanding upwards, downwards, leftwards, and rightwards by 15 pixels each.

[0069] Secondly, during an attack of angina pectoris, the abnormal signals caused by myocardial ischemia are not limited to the anterior chest area but are also transmitted to the corresponding area of ​​the back through nerve reflex pathways. This implementation method uses the human body regional symmetry method, taking the corresponding point of the spinous process of the back spine as the mapping reference point, and symmetrically mapping the pixel range of the core sub-region of the anterior chest to the back area along the horizontal / left-right direction.

[0070] Specifically, the origin of the coordinate system is set at the left edge of the single-frame infrared thermal image, x spine This represents the x-coordinate of the point corresponding to the spinous process of the spine in the back. For each pixel in the core sub-region of the chest (x... front ,y front The x-coordinate of its mirror pixel in the back region is x back =2×x spine -x front The y-coordinate remains unchanged. back =y front The resulting back region is exactly the same size and shape as the core sub-region of the chest, ensuring a comparable anatomical temperature correspondence between the two regions.

[0071] In addition, the Hand-Shaoyin Heart Meridian and the Hand-Jueyin Pericardium Meridian both run along the inner sides of the left and right arms. In this embodiment, a line connecting the corresponding point under the left armpit to the corresponding point of the Neiguan acupoint on the left arm is used as the center line. A preset pixel width is extended to both sides along the normal direction of this center line to generate strip-shaped regions covering the pathways of the Hand-Shaoyin Heart Meridian and the Hand-Jueyin Pericardium Meridian on the left arm. Simultaneously, a preset pixel width is extended to both sides along the normal direction of the line connecting the corresponding point under the right armpit to the corresponding point of the Neiguan acupoint on the right arm, generating strip-shaped regions covering the pathways of the Hand-Shaoyin Heart Meridian and the Hand-Jueyin Pericardium Meridian on the right arm. These strip-shaped regions on both sides together constitute a meridian-related sub-region. This meridian-related sub-region can capture the temperature gradient changes along the meridian pathways on both sides and can further calculate the symmetrical temperature difference between the left and right meridian-related sub-regions, providing accurate pixel data for subsequent calculations of meridian temperature gradients and symmetrical temperature differences.

[0072] The specific value of the preset pixel width can be set by the implementer according to the actual situation. For example, it can be dynamically calculated based on the height and pixel ratio of the target object, or a fixed value of 5 pixels can be used.

[0073] The specific implementation method of dynamically calculating pixels based on the target object's height and pixel ratio is as follows: First, determine the spatial resolution of the infrared thermal imager, that is, the actual physical size of a single pixel at the target distance; then, estimate the actual physical distance between the target object's surface landmarks based on the target object's height. For example, the horizontal distance from the left clavicle midline to the anterior midline is about 1 / 16 of the height, and the longitudinal distance from the armpit to the Neiguan acupoint is about 1 / 10 of the height; finally, divide the actual physical distance by the physical size corresponding to a single pixel to convert it into the corresponding number of pixels.

[0074] The above-mentioned information provides a temperature baseline for the core area of ​​the lesion on the anterior chest, a symmetrical comparison of the reflex zones on the back, and temperature distribution information of the meridian conduction pathways on the inner sides of the left and right arms. The three information comprehensively reflect the multi-dimensional temperature characteristics of chest pain, realizing the leap from single-region analysis to multi-region composite linkage analysis, which facilitates the complete capture of cross-regional temperature linkage signals of chest pain.

[0075] S4: Traverse all preset body parts and merge the associated sub-regions corresponding to each preset body part to obtain the target composite region.

[0076] Since the three related sub-regions come from infrared thermal images from different perspectives—the front chest thermal image is from a frontal view, the back thermal image is from a rear view, and the left and right inner arm thermal images are from local perspectives—their pixel coordinate systems are different, and direct merging cannot form a meaningful spatial correspondence.

[0077] Therefore, the first step is to map each associated sub-region to a unified coordinate system. Specifically, an expanded body surface coordinate system is constructed, with the spatial location of the point corresponding to the Tanzhong acupoint as the origin, the midline of the body as the Y-axis, and the horizontal direction as the X-axis. The associated sub-regions from each viewpoint are then mapped to this coordinate system according to their anatomical positions. Correspondingly, the anterior chest core sub-region is mapped to the frontal region, the back reflex sub-region is mapped to the back region, and the meridian-related sub-regions on the inner sides of the left and right arms are mapped to the inner sides of the left and right arms, respectively.

[0078] Specifically, the thermal images of the chest and back have an approximate correspondence in human anatomical position. Based on image registration, the pixel coordinates of the chest and back thermal images can be aligned to a unified frontal coordinate system. For example, the back thermal image can be mirrored and flipped along the vertical axis before translation and registration. The thermal images of the left and right inner arms are determined in the body surface coordinate system by the coordinates of the corresponding axillary points and Neiguan acupoint points.

[0079] Furthermore, the pixel masks of each associated sub-region are bitwise ORed in a unified coordinate system to generate the overall mask of the target composite region. Specifically, each associated sub-region corresponds to a binary mask matrix, which is the same size as the infrared thermal image of the corresponding viewpoint. After mapping it to a unified body surface coordinate system, the binary mask of each associated sub-region in the body surface coordinate system is obtained. For any pixel position in the body surface coordinate system, if at least one associated sub-region mask has a value of 1 at that position, then the overall mask has a value of 1 at that position; if all associated sub-region masks have a value of 0 at that position, then the overall mask has a value of 0 at that position. Thus, the pixel range of the target composite region is determined based on the overall mask in the body surface coordinate system.

[0080] As described above, by traversing all preset body parts, the associated sub-regions corresponding to each preset body part are merged to obtain the target composite region. This unifies multiple associated sub-regions that were originally located in infrared thermograms from different perspectives into the same space, forming a comprehensive region covering the core lesion area of ​​the anterior chest, the back reflection area, and the meridian conduction areas of the inner left and right arms, providing a unified spatial framework for subsequent composite temperature feature calculations.

[0081] S5, calculate the composite temperature characteristics of the target composite region based on the temperature values ​​of each pixel within the target composite region.

[0082] Among them, the composite temperature characteristics include the temperature statistical characteristics of each associated sub-region and the temperature correlation characteristics between different associated sub-regions.

[0083] Temperature statistical features are quantitative indicators obtained by statistically analyzing the temperature values ​​of all pixels within a single associated sub-region. These include average temperature, maximum temperature, minimum temperature, and temperature gradient along a preset direction, reflecting the temperature distribution within each associated sub-region and providing basic temperature data for cross-region comparisons.

[0084] Temperature correlation characteristics are indicators obtained by quantitatively analyzing the temperature relationship between different correlated sub-regions and the temperature distribution of meridian-related sub-regions. They include temperature difference and dispersion indicators, reflecting the characteristics of cross-regional temperature linkage in chest pain.

[0085] A composite temperature feature vector is formed by arranging temperature statistical features and temperature correlation features in a preset order. This vector is used to characterize the infrared features of the target object and to establish a spatial correlation model between the correlated sub-regions. The spatial correlation model is a mathematical model of the spatial temperature relationship between the correlated sub-regions, established in the feature space using quantitative indicators such as temperature difference and dispersion. Specifically, this spatial correlation model includes at least the following indicators: symmetrical temperature difference between the anterior chest and posterior back, reflecting sagittal temperature symmetry; symmetrical temperature difference between the left and right arm meridians, reflecting coronal temperature symmetry; and dispersion of each sub-region, reflecting the degree of deviation of each region relative to the overall level. These indicators together constitute the quantitative expression of the spatial correlation model. The resulting composite temperature feature vector serves as a structured output for subsequent processing.

[0086] In one specific embodiment, S5 includes the following steps: S51, calculate the temperature statistical characteristics of each associated sub-region, wherein the temperature statistical characteristics include the average temperature, maximum temperature, minimum temperature and temperature gradient along a preset direction of each associated sub-region.

[0087] S52, calculate the temperature correlation characteristics between different related sub-regions, wherein the temperature correlation characteristics include the temperature difference between different related sub-regions, the temperature difference between the symmetrical sides of the meridian related sub-regions, and the ratio of the average temperature of each related sub-region to the overall average value of the average temperature of all related sub-regions.

[0088] S53, based on temperature statistical characteristics and temperature correlation characteristics, generates a composite temperature feature vector of the target composite region.

[0089] Among them, the average temperature reflects the overall temperature level of the associated sub-region, the highest and lowest temperatures reflect the temperature extremes of the associated sub-region, and the temperature gradient reflects the rate of temperature change along the meridian pathway, which is an important indicator for judging the state of meridian conduction function.

[0090] Specifically, the temperature gradient is the temperature change per unit length along a predetermined direction. For the left / right arm strip-shaped regions of the meridian-related sub-regions, the temperature gradient is calculated along the centerline direction: G rad-meridian =(T avg -T avg ) / L, where T avg T represents the average temperature of the meridian-related sub-region near the corresponding point on the left / right armpit. avg Let L be the average temperature of the meridian-related sub-region near the corresponding point of the left / right Neiguan acupoint, and L be the spatial distance between the left / right proximal axillary end and the proximal wrist end, in cm. Further, the mean value of the temperature gradient corresponding to the strip-shaped region of the left / right arm is calculated as the temperature gradient corresponding to the meridian-related sub-region.

[0091] Myocardial ischemia not only causes abnormal temperature in the anterior chest region, but also synchronously causes temperature changes in the back and inner left arm through nerve reflexes and meridian conduction. Therefore, by calculating the temperature difference and dispersion index between different associated sub-regions, the cross-regional temperature correlation can be quantified into calculable values, providing a basis for subsequent classification. In patients with angina pectoris, the abnormal cardiac function leads to disordered blood and qi circulation in the Hand Shaoyin Heart Meridian and Hand Jueyin Pericardium Meridian, which can result in asymmetrical temperature distribution in the left and right meridians. Therefore, for the meridian-related sub-regions, the temperature difference between the mean temperature of the left strip region and the mean temperature of the right strip region is calculated to enrich the information dimensions of temperature linkage analysis.

[0092] Specifically, the temperature difference is the difference between different associated sub-regions on the same temperature index, including three categories: average temperature difference, minimum temperature difference, and maximum temperature difference.

[0093] The dispersion index reflects the degree of dispersion of the average temperature of each associated sub-region relative to the overall average value. Specifically, it is the ratio of the average temperature of each associated sub-region to the overall average value of the average temperatures of all associated sub-regions. When the ratio is close to 1, it indicates that the temperature of the associated sub-region is close to the overall average level; when the ratio is significantly greater than 1 or less than 1, it indicates that the temperature of the associated sub-region deviates significantly from the overall average level.

[0094] The temperature statistical features include three average temperatures, three maximum temperatures, three minimum temperatures, and three temperature gradients. The temperature correlation features include four × three temperature differences and three dispersion indices. Each of these values ​​independently describes a certain dimension of the temperature distribution in the target composite region. These discrete values ​​are arranged in a preset order to form a composite temperature feature vector, so that temperature features of different types and dimensions are integrated into a unified vector form.

[0095] As described above, based on the temperature values ​​of each pixel within the target composite region, a composite temperature feature vector containing temperature statistical features and temperature correlation features is calculated, providing a comprehensive and standardized temperature basis for subsequent infrared feature information processing.

[0096] In one specific embodiment, the infrared image-based information processing method further includes the following steps: S6. The composite temperature features are compared with the preset judgment rule set to obtain the image analysis results corresponding to the target object. The preset judgment rule set includes several preset judgment rules. Each preset judgment rule corresponds to a risk level and a prompt content. Each preset judgment rule includes judgment conditions for several feature dimensions.

[0097] In one specific embodiment, S6 includes the following steps: S61, compare the composite temperature feature with each preset judgment rule, and count the number of dimensions of the composite temperature feature that satisfy the judgment conditions in each preset judgment rule.

[0098] S62, the ratio of the number of dimensions to the total number of dimensions of the judgment conditions in each preset judgment rule is determined as the matching degree between the composite temperature feature and each preset judgment rule.

[0099] S63, input the composite temperature features into the trained classification model to obtain the model analysis results; S64. According to the preset fusion strategy, the risk level and prompt content corresponding to the preset judgment rule with the highest matching degree, as well as the model analysis results, are fused to obtain the image analysis results corresponding to the target object.

[0100] Each preset judgment rule consists of multiple judgment conditions for feature dimensions. The judgment conditions are related by "AND". When a composite temperature feature satisfies all the judgment conditions of a preset judgment rule, it is considered that the composite temperature feature fully conforms to the preset judgment rule. When only some judgment conditions are satisfied, the corresponding matching degree is calculated according to the number of dimensions that satisfy the judgment conditions, and the rule analysis result is generated accordingly.

[0101] The values ​​of each dimension in the composite temperature feature vector are compared one by one with the corresponding judgment conditions in each preset judgment rule.

[0102] For example, this embodiment includes four preset judgment rules. Preset judgment rule 1 includes the following judgment conditions: the temperature difference between the core sub-region of the chest and the reflex sub-region of the back is ≥0.3℃; the temperature difference between the symmetrical sides of the meridian-related sub-regions is ≥0.3℃; and the average temperature of the core sub-region of the chest, the reflex sub-region of the back, and the meridian-related sub-region is all less than 33℃. The corresponding risk level is high risk, indicating a severe degree of stagnation in the heart meridian and obstruction of Yang Qi.

[0103] Preset judgment rule 2 includes the following judgment conditions: the highest temperature in the core sub-region of the chest ≥ 35.5℃; and the meridian temperature gradient ≥ 0.2℃ / cm. The corresponding risk level is medium to high risk, indicating stagnation of Qi and poor blood circulation. The meridian temperature gradient is the temperature change per unit length along the central line of the meridian-related sub-region.

[0104] The preset judgment rule 3 includes the following judgment conditions: the average temperature of the chest core sub-region is more than 0.2℃ lower than the average temperature of the back reflex sub-region; the dispersion index of the chest core sub-region, the back reflex sub-region, and the meridian-related sub-region are all lower than the preset dispersion threshold. The corresponding risk level is medium risk, indicating a deficiency of Yang Qi. The dispersion index is the ratio of the standard deviation of the average temperature of each related sub-region to the overall average value of the average temperature of all related sub-regions. The preset dispersion threshold is determined based on the statistical distribution of the dispersion index of multiple reference subjects; for example, it can be set as the mean of the dispersion index of multiple reference subjects minus 1.5 times the standard deviation.

[0105] The preset judgment rule 4 includes the following judgment conditions: the values ​​of each dimension in the composite temperature feature vector are all within the corresponding normal threshold range; the corresponding risk level is low risk, indicating a normal status. The normal threshold range is determined by statistical analysis of the composite temperature characteristics of multiple reference objects, with the statistical mean of each feature dimension as the center and the standard deviation of a preset multiple as the radius.

[0106] The classification model is a supervised learning model trained using labeled historical samples. The input is a composite temperature feature vector, and the output is the risk level or category label corresponding to the target object. Those skilled in the art will recognize that any classification model in the prior art falls within the protection scope of this invention, such as any of the following: support vector machine, random forest, gradient boosting tree, neural network, deep neural network, or Transformer model; further details will not be provided here.

[0107] The preset fusion strategy is used to fuse the results of rule analysis and model analysis to determine the final image analysis result.

[0108] In one implementation, when the risk level corresponding to the rule analysis result matches the risk level corresponding to the model analysis result, the risk level and corresponding prompt content are determined as the image analysis result. When the risk level corresponding to the rule analysis result does not match the risk level corresponding to the model analysis result, the final image analysis result is determined based on the matching degree corresponding to the rule analysis result and the confidence degree corresponding to the model analysis result.

[0109] In one specific implementation, determining the image analysis result based on the matching degree corresponding to the rule analysis result and the confidence degree corresponding to the model analysis result includes: obtaining the matching degree Pr corresponding to the rule analysis result and the confidence degree Pm corresponding to the model analysis result; calculating the fusion score based on the matching degree Pr and the confidence degree Pm: Score = α × Pr + (1 - α) × Pm, where α is a preset weight, 0 ≤ α ≤ 1. The final risk level is determined based on the fusion score, and corresponding prompt content is generated.

[0110] As described above, by inputting the composite temperature features into the preset decision rule set and the trained classification model respectively, rule analysis results and model analysis results are obtained. The rule analysis results have good interpretability, and the model analysis results can fully explore the nonlinear correlation between composite temperature features. Based on the rule analysis results and model analysis results, the image analysis results corresponding to the target object are determined. The combination of the two improves the accuracy, stability and robustness of the image analysis results, providing an objective, reliable and interpretable reference for the state assessment of the target object.

[0111] This embodiment corrects the temperature baseline of the original infrared thermogram by using ambient temperature parameters, unifying the temperature values ​​of infrared thermograms acquired at different times and under different environmental conditions to the same temperature baseline, thus eliminating the adverse effects of ambient temperature fluctuations on the accuracy of infrared thermometry. By performing temporal inter-frame difference processing on multiple consecutive first infrared thermograms, motion regions are identified and mask suppression is applied, effectively eliminating chest wall undulation artifacts caused by respiratory movements. While preserving the original image spatial structure and temperature distribution characteristics, a stable and reliable single-frame infrared thermogram is output. By calling a preset human key point detection model to automatically locate the reference points, and automatically extracting the associated sub-regions corresponding to each preset body part according to preset composite region delineation rules, the fluctuation of the region range caused by individual experience differences is avoided, ensuring the stability and repeatability of the delineation results. By using the temperature values ​​of each pixel in the target composite region, a composite temperature feature containing the temperature statistical characteristics of each associated sub-region and the cross-regional temperature correlation characteristics is calculated, which fully reflects the characteristics of cross-regional temperature linkage in chest pain, improving the accuracy of chest pain infrared thermogram analysis and providing objective, standardized, and interpretable image analysis information for chest pain diagnosis.

[0112] Example 2 Embodiment 2 of the present invention provides a non-transitory computer-readable storage medium, which can be disposed in an electronic device to store at least one instruction or at least one program related to implementing a method in the method embodiment. The at least one instruction or at least one program is loaded and executed by the processor to implement the information processing method based on infrared images provided in the above embodiment.

[0113] Example 3 Embodiment 3 of the present invention provides an electronic device, which includes a processor and the non-transitory computer-readable storage medium of Embodiment 2 of the present invention.

[0114] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. An information processing method based on infrared images, characterized in that, The method includes the following steps: S1, for any preset body part of the target object, according to the ambient temperature when the original infrared thermal image sequence corresponding to the preset body part was collected, the temperature reference correction is performed on each original infrared thermal image in the original infrared thermal image sequence to obtain the first infrared thermal image sequence corresponding to the preset body part after temperature correction. S2, perform temporal inter-frame difference processing on multiple consecutive first infrared thermal images in the first infrared thermal image sequence, identify motion regions based on the difference results, and perform mask suppression on the motion regions to obtain a single-frame infrared thermal image after motion artifact suppression corresponding to the preset body part. S3, call the preset human key point detection model, locate the pixel coordinates of several preset reference points corresponding to the preset body part on the single frame infrared thermal image, and extract the associated sub-region corresponding to the preset body part from the single frame infrared thermal image according to the pixel coordinates and the preset composite region delineation rules. S4, traverse all preset body parts and merge the associated sub-regions corresponding to each preset body part to obtain the target composite region; S5, calculate the composite temperature feature of the target composite region based on the temperature value of each pixel in the target composite region, wherein the composite temperature feature includes the temperature statistical features of each associated sub-region and the temperature correlation features between different associated sub-regions, and the composite temperature feature is used to characterize the infrared feature information of the target object and to establish a spatial correlation model between each associated sub-region.

2. The information processing method based on infrared images according to claim 1, characterized in that, S1 includes the following steps: S11, Obtain the pre-stored ambient temperature-temperature offset mapping relationship, and determine the temperature offset corresponding to the preset body part based on the ambient temperature and the mapping relationship. S12, the pixel temperature values ​​of each original infrared thermal image in the original infrared thermal image sequence are offset and corrected according to the temperature offset to obtain the first infrared thermal image corresponding to each original infrared thermal image.

3. The information processing method based on infrared images according to claim 1, characterized in that, S2 includes the following steps: S21, acquire N consecutive first infrared thermal images from the first infrared thermal image sequence, and use a designated frame in the N consecutive first infrared thermal images as a reference frame, where N > 2; S22, perform inter-frame differential processing on the N consecutive first infrared thermal images to obtain the temperature change of each pixel between adjacent frames. S23, the pixels whose temperature change exceeds a preset difference threshold are determined to be moving points; S24, Generate a mask for the motion region composed of all moving points, and determine the pixel region to be replaced with temperature value based on the mask; S25, calculate the average temperature of each pixel point corresponding to the pixel region in the N consecutive first infrared thermal images; S26, replace the temperature value of each pixel in the pixel area of ​​the reference frame with the corresponding average temperature, and keep the temperature value of the pixel outside the pixel area of ​​the reference frame unchanged, and obtain a single-frame infrared thermal image of the preset body part after motion artifact suppression.

4. The information processing method based on infrared images according to claim 1, characterized in that, The preset body parts include the chest, back, inner side of the left arm, and inner side of the right arm. S3 includes the following steps: S31, invoke the preset human body key point detection model, perform key point detection on the single frame infrared thermal image, and locate the pixel coordinates of several preset reference points corresponding to the preset body part. Among them, the preset base points corresponding to the front chest include the point corresponding to the Tanzhong acupoint and the point corresponding to the boundary of the left chest anatomical division; the preset base points corresponding to the back include the point corresponding to the spinous process of the back spine; the preset base points corresponding to the inner side of the left arm include the point corresponding to the Neiguan acupoint of the left arm and the point corresponding to the axilla of the left arm; the preset base points corresponding to the inner side of the right arm include the point corresponding to the Neiguan acupoint of the right arm and the point corresponding to the axilla of the right arm. S32, based on the pixel coordinates of each preset reference point and the preset composite region delineation rules, determine the pixel region range of the associated sub-region corresponding to the preset body part, wherein the associated sub-region corresponding to the front chest is the front core sub-region, the associated sub-region corresponding to the back is the back reflection sub-region, and the associated sub-regions corresponding to the inner side of the left arm and the inner side of the right arm are the meridian associated sub-regions. S33, generate a corresponding pixel mask based on the pixel region range, and extract the associated sub-region corresponding to the preset body part from the single-frame infrared thermal image based on the pixel mask.

5. The information processing method based on infrared images according to claim 4, characterized in that, The core sub-region of the chest is the area formed by extending to the left from the corresponding point of the Tanzhong acupoint to the corresponding point of the boundary of the left chest anatomical division. The back reflection sub-region is the region formed by symmetrically mapping the chest core sub-region onto the back region along the horizontal direction on the single-frame infrared thermogram, with the corresponding point of the spinous process of the back spine as the mapping reference point. The meridian-related sub-region is a strip-shaped region formed by extending a preset pixel width to both sides of the line connecting the corresponding point under the left armpit to the corresponding point of the Neiguan acupoint on the left arm as the center line, and extending a preset pixel width to both sides of the line connecting the corresponding point under the right armpit to the corresponding point of the Neiguan acupoint on the right arm as the center line.

6. The information processing method based on infrared images according to claim 1, characterized in that, S5 includes the following steps: S51, calculate the temperature statistical characteristics of each associated sub-region, wherein the temperature statistical characteristics include the average temperature, maximum temperature, minimum temperature and temperature gradient along a preset direction of each associated sub-region; S52, calculate the temperature correlation characteristics between different related sub-regions, wherein the temperature correlation characteristics include the temperature difference between different related sub-regions, the temperature difference between the symmetrical sides of the meridian related sub-regions, and the ratio of the average temperature of each related sub-region to the overall average value of the average temperature of all related sub-regions. S53, Generate a composite temperature feature vector of the target composite region based on the temperature statistical features and the temperature correlation features.

7. The information processing method based on infrared images according to claim 1, characterized in that, The infrared image-based information processing method further includes the following steps: S6, compare the composite temperature feature with the preset judgment rule set to obtain the image analysis result corresponding to the target object. The preset judgment rule set includes several preset judgment rules, each preset judgment rule corresponds to a risk level and a prompt content, and each preset judgment rule includes judgment conditions for several feature dimensions.

8. The information processing method based on infrared images according to claim 7, characterized in that, S6 includes the following steps: S61, compare the composite temperature feature with each preset judgment rule, and count the number of dimensions in which the composite temperature feature satisfies the judgment conditions in each preset judgment rule. S62, the ratio of the number of dimensions to the total number of dimensions of the judgment conditions in each preset judgment rule is determined as the matching degree between the composite temperature feature and each preset judgment rule, and the rule analysis result is obtained. S63, input the composite temperature features into the trained classification model to obtain the model analysis results; S64, according to the preset fusion strategy, the risk level and prompt content corresponding to the preset judgment rule with the highest matching degree, as well as the model analysis result, are fused to obtain the image analysis result corresponding to the target object.

9. A non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores at least one instruction or at least one program segment, characterized in that, The at least one instruction or the at least one program segment is loaded and executed by the processor to implement the information processing method based on infrared images as described in any one of claims 1-8.

10. An electronic device, characterized in that, Includes a processor and the non-transitory computer-readable storage medium as described in claim 9.

Images