Infrared thermal image temperature field reconstruction system based on atlas analysis

Abstract

The application provides an infrared thermal image temperature field reconstruction system based on atlas analysis, and belongs to the technical field of infrared thermal image temperature field reconstruction. The data preprocessing module synchronously processes infrared thermal images and calibration data, divides temperature measurement sub-regions, and constructs a temperature measurement data set. The reference field solving module mines historical data and solves the theoretical temperature reference field in combination with a heat conduction equation. The initial atlas generation module sets a differentiated confidence interval and generates an initial three-dimensional temperature field atlas. The dynamic atlas correction module constructs an adaptive correction tensor and corrects the atlas pixel by pixel. The reconstruction positioning module maps real-time data, completes temperature field reconstruction and abnormal heat zone positioning. The application effectively suppresses interference, improves temperature measurement accuracy, realizes intelligent reconstruction of complex target temperature fields and automatic positioning of abnormal heat zones, and provides stable and reliable technical support for non-contact thermal state detection in the fields of industry and power.

Description

Technical Field

[0001] This invention belongs to the field of infrared thermal imaging temperature field reconstruction technology, specifically an infrared thermal imaging temperature field reconstruction system based on spectral analysis. Background Technology

[0002] Infrared thermal imaging temperature measurement technology, with its advantages of non-contact, visualization and real-time monitoring, has been widely used in industrial equipment condition monitoring, power fault diagnosis and building energy efficiency testing, and is a core means of temperature field analysis and anomaly identification.

[0003] Current infrared thermal imaging temperature field reconstruction methods largely rely on direct conversion of single-frame grayscale values, failing to fully integrate multi-source environmental and spatial calibration information. This makes it difficult to eliminate temperature measurement errors caused by atmospheric attenuation, uneven surface emissivity, and detector noise. Traditional methods often employ globally uniform temperature measurement parameters without differentiating them for the thermal characteristics and temperature measurement accuracy requirements of different target regions. This results in insufficient temperature measurement accuracy in critical areas and wasted resources in non-critical areas. Furthermore, existing technologies lack in-depth analysis of historical operating data, failing to establish a theoretical temperature benchmark that closely matches the target's heat conduction patterns, leading to significant deviations between the reconstructed temperature field and the actual heat distribution. In addition, most systems only perform numerical temperature calculations, failing to construct three-dimensional feature maps for joint analysis of spatial and temperature dimensions. Abnormal hot zone location relies on manual judgment, resulting in low automation and identification accuracy, making it difficult to meet the demands for high-precision, dynamic, and intelligent infrared thermal imaging temperature field reconstruction and fault monitoring under complex operating conditions. Summary of the Invention

[0004] This invention provides an infrared thermal imaging temperature field reconstruction system based on spectral analysis to address the deficiencies in existing technologies.

[0005] This invention provides an infrared thermal image temperature field reconstruction system based on spectral analysis, comprising: The data preprocessing module is used to simultaneously preprocess the raw infrared thermal image frame sequence and multi-source environmental calibration data, divide the target temperature measurement sub-regions, fuse calibration parameters to extract the radiation feature vectors of each sub-region, and construct a structured temperature measurement partition dataset.

[0006] The reference field solution module is used to mine historical temperature measurement datasets under the same operating conditions, extract the regional thermal conductivity coefficient, detector radiation response function, and internal heat source intensity and the topological relationship with the temperature measurement space, and solve the theoretical temperature reference field of each temperature measurement sub-region based on the thermal conduction differential equation.

[0007] The initial map generation module is used to configure differentiated temperature confidence intervals for each sub-region based on temperature measurement accuracy requirements and environmental fluctuation characteristics, and generate an initial three-dimensional temperature field feature map.

[0008] The dynamic temperature field feature map module is used to collect data on the emissivity distribution of the target surface, the atmospheric radiation attenuation coefficient, and the dark current noise of the detector. It constructs a multi-field coupled adaptive correction tensor and performs pixel-by-pixel correction on the initial three-dimensional temperature field feature map to obtain the dynamic temperature field feature map.

[0009] The reconstruction and positioning module is used to acquire real-time infrared thermal image frame data and map it to the spatial coordinate system of dynamic temperature field feature map. It calculates the temperature mean deviation and spatial structure similarity of the corresponding sub-regions to complete temperature field reconstruction and abnormal hot zone positioning.

[0010] According to the present invention, an infrared thermal image temperature field reconstruction system based on spectral analysis is provided. The original infrared thermal image frame sequence includes a grayscale thermal image stream, frame synchronization timestamps, and detector gain profiles. Multi-source environmental calibration data includes a blackbody calibration coefficient matrix, an atmospheric transmittance parameter table, a target surface emissivity map, and lidar three-dimensional point cloud data.

[0011] According to the infrared thermal imaging temperature field reconstruction system based on spectral analysis provided by the present invention, the process of dividing the target temperature measurement sub-region by the data preprocessing module includes: The U-Net network is used to perform pixel-level semantic segmentation on infrared grayscale thermal images to identify the target subject, background and interference thermal areas.

[0012] The Canny edge detection algorithm is used to extract the contour features of the target subject, and spatial projection correction is performed by combining it with 3D point cloud data from LiDAR.

[0013] Based on contour features, the target body is divided into several continuous temperature measurement sub-regions, generating an initial temperature measurement partition sequence.

[0014] The target temperature measurement sub-region is obtained by removing non-effective temperature measurement sub-regions in the initial sequence whose area is smaller than a preset threshold and whose gray-level variance is lower than the noise level.

[0015] According to the infrared thermal imaging temperature field reconstruction system based on spectral analysis provided by the present invention, the process of constructing a structured temperature measurement zoning dataset by the data preprocessing module includes: Based on the pixel coordinates of the sub-regions and the point cloud data of the lidar, the actual physical size and temperature measurement distance of each sub-region are calculated. By combining the blackbody calibration coefficient matrix, the gray values ​​of the sub-regions are converted into the original radiance values, and then into radiance values. By introducing atmospheric transmittance parameters, the standard radiative transfer model is used to correct the radiance value for atmospheric attenuation. Calculate the radiation flux density and gray-level gradient variance of each sub-region to generate a radiation feature vector. The radiation feature vector includes the average gray value of the sub-region, the gray-level gradient variance, the radiation flux density, and the coordinates of the geometric center of the sub-region. Each temperature measurement sub-region is assigned a unique spatial identifier, and its spatial coordinates, radiation feature vector, and calibration parameters are encapsulated into structured data units, which are then arranged according to spatial topological relationships to form a structured temperature measurement partition dataset.

[0016] According to the infrared thermal imaging temperature field reconstruction system based on spectral analysis provided by the present invention, the process of the reference field solving module mining historical temperature measurement datasets under the same operating conditions and extracting the regional thermal conductivity coefficient, internal heat source intensity, detector radiation response function, and the spatial topological relationship of the temperature measurement includes: Select historical temperature measurement records that match the material, structure, and environmental conditions of the current temperature measurement target to ensure consistency between historical data and current temperature measurement conditions.

[0017] The historical structured temperature measurement zoning dataset is analyzed to extract the heat conduction parameters of each sub-region. Combined with the temperature change data in the historical temperature measurement records, the internal heat source intensity data of each sub-region is calculated by inverting the heat conduction equation. The temperature distribution pattern, heat conduction characteristics and internal heat source intensity distribution range of each sub-region are statistically analyzed.

[0018] The detector's radiation response function is obtained by fitting the correspondence between the detector's output grayscale value and the actual temperature using the least squares method.

[0019] Extract the spatial adjacency relationships and heat flow transfer paths of historical temperature measurement zones to construct a temperature measurement spatial topology map.

[0020] According to the infrared thermal imaging temperature field reconstruction system based on spectral analysis provided by the present invention, the process of the reference field solving module solving the theoretical temperature reference field of each temperature measurement sub-region based on the heat conduction differential equation includes: Establish the two-dimensional steady-state heat conduction differential equation for the target region.

[0021] By combining boundary conditions and historical thermal conductivity data, the two-dimensional steady-state thermal conductivity differential equation is solved discretically using the finite difference method to obtain the theoretical temperature reference value for each sub-region.

[0022] By superimposing the ambient background radiation temperature correction term, the final theoretical temperature reference field is generated.

[0023] According to the infrared thermal imaging temperature field reconstruction system based on spectral analysis provided by the present invention, the process of the initial spectral generation module generating an initial three-dimensional temperature field feature map includes: Construct a three-dimensional coordinate system with spatial coordinates (x, y) as the horizontal axis and temperature T as the vertical axis.

[0024] The theoretical temperature reference values ​​of each temperature measurement sub-region are mapped to a three-dimensional coordinate system, and a continuous temperature reference surface is generated using bilinear interpolation.

[0025] Based on the temperature measurement accuracy requirements of each sub-region, differentiated temperature confidence intervals are configured for the temperature reference surface.

[0026] Mark the upper and lower boundaries of the confidence interval in the three-dimensional coordinate system to form an initial three-dimensional temperature field feature map.

[0027] According to the infrared thermal imaging temperature field reconstruction system based on spectral analysis provided by the present invention, the process of constructing a multi-field coupled adaptive correction tensor by dynamic spectral correction includes: Calculate the emissivity correction coefficient based on the emissivity distribution data of the target surface; Calculate the atmospheric radiation correction coefficient based on atmospheric temperature and humidity data; Calculate the radiance value corresponding to the dark current noise based on the detector's dark current noise data; The emissivity correction coefficient, atmospheric radiation correction coefficient, and dark current noise radiance value are arranged according to spatial dimensions to construct a 3×N-dimensional multi-field coupled adaptive correction tensor, where N is the number of thermometric sub-regions.

[0028] According to the infrared thermal image temperature field reconstruction system based on spectral analysis provided by the present invention, the process of dynamically correcting the initial three-dimensional temperature field feature map pixel by pixel includes: The temperature values ​​of the initial three-dimensional temperature field feature map are converted into radiance values, and then pixel-wise operations are performed with the multi-field coupled adaptive correction tensor to obtain the corrected radiance values. The corrected radiance values ​​are converted into temperature values ​​using the Planck inverse function, resulting in the corrected temperature reference surface. The width of the temperature confidence interval is dynamically adjusted based on the noise level of each sub-region; A correction coefficient thermodynamic layer is overlaid on the three-dimensional temperature field feature map to show the degree of correction in each region; The multi-field coupled adaptive correction tensor is updated once every preset time interval. When a sudden change in ambient temperature or humidity is detected, the correction tensor update is triggered.

[0029] According to the infrared thermal imaging temperature field reconstruction system based on spectral analysis provided by the present invention, the process by which the reconstruction and positioning module completes temperature field reconstruction and abnormal thermal zone positioning includes: Real-time infrared thermal image frame data is converted into grayscale temperature maps and mapped to the spatial coordinate system of dynamic temperature field feature map.

[0030] Calculate the deviation between the actual mean temperature and the theoretical temperature reference value for each temperature measurement sub-region.

[0031] The spatial structural similarity between the feature maps of the real-time temperature field and the dynamic temperature field is calculated using the structural similarity index.

[0032] When the average temperature deviation exceeds a preset threshold or the structural similarity is lower than a preset threshold, the sub-region is marked as an abnormal hot zone and an alarm message is output.

[0033] This invention provides an infrared thermal imaging temperature field reconstruction system based on spectral analysis. Through synchronous preprocessing of multi-source data and pixel-level semantic segmentation, it accurately divides temperature measurement sub-regions and extracts radiation features, constructing a structured dataset and significantly improving the effectiveness and standardization of the original data. Relying on historical operating condition data mining and solving the thermal conduction differential equation, a theoretical temperature reference field that closely matches the actual thermal characteristics of the target is generated, ensuring the theoretical accuracy of the temperature field reconstruction from the source. A three-dimensional temperature field feature map combined with differentiated confidence interval settings is used to accommodate the temperature measurement accuracy requirements of different regions, making the temperature field presentation more intuitive and adaptable. Pixel-by-pixel dynamic correction is achieved through multi-field coupling adaptive correction tensors, effectively offsetting interference factors such as emissivity, atmospheric radiation, and detector noise, significantly improving temperature measurement accuracy and environmental adaptability. Finally, through real-time data mapping and joint judgment of deviation and similarity, the system automatically completes accurate temperature field reconstruction and rapid location of abnormal hot areas without manual intervention. While improving temperature measurement accuracy and reconstruction effect, it also enables intelligent monitoring and fault early warning of complex target thermal states, providing stable and reliable technical support for non-contact thermal detection in industries such as industry and power. Attached Figure Description

[0034] The invention will now be further described with reference to the accompanying drawings.

[0035] Figure 1 This is a schematic diagram of the structure of an infrared thermal imaging temperature field reconstruction system based on spectral analysis in this invention; Figure 2 This is a flowchart illustrating the process of dividing the target temperature measurement sub-region in the data preprocessing module of this invention. Figure 3 This is a schematic diagram of the process by which the data preprocessing module constructs a structured temperature measurement zoning dataset in this invention; Figure 4 This is a schematic diagram of the process by which the reference field solving module solves the theoretical temperature reference field for each temperature measurement sub-region in this invention; Figure 5 This is a schematic diagram of the process by which the initial map generation module generates the initial three-dimensional temperature field feature map in this invention. Detailed Implementation

[0036] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0037] like Figures 1 to 5As shown in the figure, an infrared thermal imaging temperature field reconstruction system based on spectral analysis provided by this invention includes a data preprocessing module, a reference field solving module, an initial spectral generation module, and a dynamic spectral correction and reconstruction positioning module.

[0038] The data preprocessing module is used to simultaneously preprocess the raw infrared thermal image frame sequence and multi-source environmental calibration data, divide the target temperature measurement sub-regions, fuse calibration parameters to extract the radiation feature vectors of each sub-region, and construct a structured temperature measurement partition dataset.

[0039] The raw infrared thermal image frame sequence includes a 16-bit grayscale thermal image stream output by an uncooled focal plane detector, a frame synchronization timestamp, and a detector gain profile. The response band of the uncooled focal plane detector covers 8μm to 14μm, the frame synchronization timestamp accuracy is not less than 1 millisecond, and the detector gain profile includes gain coefficients, bias voltages, and integration time parameters corresponding to different temperature ranges. The acquisition frame rate of the raw frame sequence is set to 25 frames / second to 50 frames / second according to the thermal response speed of the temperature measurement target.

[0040] The multi-source environmental calibration data includes a blackbody calibration coefficient matrix, an atmospheric transmittance parameter table, a target surface emissivity map, and lidar 3D point cloud data. The blackbody calibration coefficient matrix is ​​generated using a standard blackbody furnace with an accuracy of ±0.01℃. Blackbody radiation data is collected every 5℃ within the temperature measurement range of -20℃ to 150℃. The matrix has an M×3 dimension, with three columns corresponding to the upper limit of the segmented temperature interval, the gain coefficient K, and the bias coefficient B, respectively. In use, the temperature value is initially estimated based on the grayscale value G of the sub-region, matched to its corresponding temperature interval, and the gain coefficient K and bias coefficient B corresponding to that interval are extracted and substituted into the radiation intensity conversion formula to achieve segmented linear calibration across the entire temperature measurement range.

[0041] The atmospheric transmittance parameter table is generated by combining real-time ambient temperature, relative humidity, and atmospheric pressure data collected from meteorological stations with the MODTRAN atmospheric radiative transfer model. It covers the infrared band from 0.7μm to 14μm and has a resolution of 0.1μm.

[0042] The emissivity map of the target surface was obtained by scanning the target surface by material partitions using an emissivity meter under the same environmental conditions. The scanning step size was consistent with the size of the smallest material partition on the target surface. After the measurement was completed, the emissivity values ​​of each material partition were mapped to the corresponding pixels by combining the spatial registration relationship between the lidar 3D point cloud data and the infrared thermal image, generating a material partition emissivity mapping map consistent with the pixel resolution of the infrared thermal imager.

[0043] The 3D point cloud data of the lidar is acquired using a line lidar with a point cloud density of no less than 100 points / square centimeter. After acquisition, the point cloud is spatially aligned with the infrared thermal image frame using a point cloud registration algorithm, and the registration error is controlled within 1 pixel.

[0044] The process of dividing the target temperature measurement sub-region includes: The U-Net network is invoked, and the input infrared grayscale thermal image frame is used as the network input. The encoder performs multi-scale feature extraction on the thermal image.

[0045] The decoder upsamples the extracted features and incorporates a channel attention mechanism to enhance the feature channels corresponding to the target hot areas in the thermal image, suppressing the feature responses of the background and interference areas, and improving the recognition accuracy of the target subject.

[0046] During the network training phase, an infrared thermal image dataset containing more than 10,000 annotated images was used as the training dataset. The annotations included three types of regions: target subject, background, and interference hotspots. A semantic segmentation mask was output to clearly distinguish the target subject, background, and interference hotspots, laying the foundation for subsequent contour extraction.

[0047] Based on the target subject mask image obtained from semantic segmentation, the Canny edge detection algorithm is launched, and a 5×5 Gaussian filter kernel is used to perform Gaussian filtering on the mask image to smooth the image and eliminate noise interference.

[0048] Edge detection is performed using a dual threshold, with a fixed ratio of 2:1 between the high and low thresholds. The low threshold is adaptively adjusted based on the noise level of the current infrared thermal image, and its value is between 10% and 15% of the average gray level of the image. The high threshold is adjusted synchronously with the low threshold. The initial edge contour is obtained through dual threshold filtering.

[0049] A morphological closing operation is performed on the initial contour using a 3×3 structuring element. The contour is first expanded and then eroded to fill in small breaks and eliminate burrs, resulting in a continuous and complete target body contour.

[0050] By combining the 3D point cloud data acquired by the lidar, the projection distortion of the extracted target contour is corrected. The 2D coordinates of each pixel in the infrared thermal image are converted into 3D coordinates in the camera coordinate system through the camera intrinsic parameter matrix (obtained through camera calibration), thus completing the mapping from pixel coordinates to camera coordinates.

[0051] By using the camera and lidar extrinsic parameter matrices obtained through hand-eye calibration, the three-dimensional coordinates in the camera coordinate system are converted into three-dimensional coordinates in the lidar coordinate system, thereby achieving spatial alignment between infrared thermal images and lidar point cloud data.

[0052] The converted coordinates are matched with the 3D point cloud data of the LiDAR. Based on the 3D shape of the target surface reflected by the point cloud data, the contour projection distortion caused by camera viewing angle deviation and changes in the curvature of the target surface is corrected to ensure that the corrected target contour is completely consistent with the spatial position and shape of the actual target.

[0053] Based on the corrected target body contour, a region growing algorithm is used to divide a continuous temperature measurement sub-region, determine the geometric center of the target body contour, and use the pixel corresponding to the geometric center as the seed point for region growing.

[0054] Region growth rules are defined, with gray-level similarity and spatial continuity as the core criteria. The growth threshold is set to a gray-level difference of no more than 5 gray levels between adjacent pixels. That is, when the difference between the gray-level value of a pixel adjacent to the seed point and the gray-level value of the seed point is within 5 gray levels, and the pixel is located within the outline of the target subject, it is included in the growth region. The growth operation is continuously executed. When the area of ​​the current growth region reaches the preset maximum area of ​​the sub-region, growth stops, and an independent initial temperature measurement sub-region is generated. The growth process is repeated until all pixels within the outline of the target subject are divided into the corresponding initial temperature measurement sub-regions, forming an initial temperature measurement partition sequence.

[0055] The initial temperature measurement zone sequence undergoes validity screening using two thresholds: an area threshold (5×5 pixels) and a grayscale variance threshold. Small regions smaller than the area threshold are eliminated to prevent them from affecting measurement accuracy. The grayscale variance threshold, representing the noise level, is calculated by pre-collecting 100 frames of background infrared thermal images without targets and averaging the grayscale variance of all background pixels. Regions with grayscale variances below the threshold are removed (these regions lack valid temperature information and are considered invalid temperature measurement areas). After eliminating invalid sub-regions, the remaining valid temperature measurement sub-regions are numbered sequentially by row and column, clearly defining the spatial identifier of each sub-region to obtain the target temperature measurement sub-regions that meet the temperature measurement requirements.

[0056] The process of constructing a structured temperature measurement zoning dataset includes: Based on the pixel coordinates of the sub-regions and the LiDAR point cloud data, the actual physical size and temperature measurement distance of each sub-region are calculated. The three-dimensional point cloud data of the LiDAR corresponding to all pixels in each sub-region are extracted. The minimum bounding rectangle of the sub-region in three-dimensional space is calculated to obtain the actual length and width and calculate the physical area. At the same time, the straight-line distance from the geometric center of the sub-region to the infrared thermal imager lens is calculated as the temperature measurement distance.

[0057] By combining the blackbody calibration coefficient matrix, the grayscale values ​​of the sub-regions are converted into the original radiant intensity values. For each pixel within the sub-region, based on its grayscale value G, the gain coefficient K and bias coefficient B in the blackbody calibration coefficient matrix are used to calculate the radiant intensity value using the formula... The original radiation intensity value is calculated, and then the radiation intensity is converted into radiance through the detector's optical system parameters. The conversion coefficient is provided by the detector's factory calibration document.

[0058] By introducing atmospheric transmittance parameters, atmospheric attenuation correction is applied to the original radiation intensity value. Based on the sub-regional temperature measurement distance and the current atmospheric transmittance parameter table, the atmospheric transmittance for the corresponding distance and infrared band is found. The standard radiative transfer model is used for correction: in, The radiance measured by the detector (by (conversion) Radiation along the upward path of the atmosphere. This is downward atmospheric radiation. Let be the target emissivity. If path radiation and environmental reflections (such as in close-range, low-humidity scenarios) are ignored, then it simplifies to... .

[0059] The radiant flux density and gray-level gradient variance of each sub-region are calculated. The radiant flux density is obtained by integrating the corrected radiant intensity value within the sub-region and then dividing it by the actual physical area of ​​the sub-region. The gray-level gradient variance is obtained by calculating the gray-level gradient magnitude of each pixel in the x and y directions within the sub-region and then calculating the variance of all gradient magnitudes.

[0060] A radiation feature vector is generated, which includes the average gray value of the sub-region, the gray-level gradient variance, the radiation flux density, and the geometric center coordinates of the sub-region. The average gray value of the sub-region is the arithmetic mean of the gray values ​​of all pixels in the sub-region, and the geometric center coordinates of the sub-region are the average of the coordinates of all pixels in the sub-region. The origin of the coordinate system is the upper left corner of the infrared thermal image, the positive x-axis is horizontal to the right, and the positive y-axis is vertical downward.

[0061] Each temperature measurement sub-region is assigned a unique spatial identifier, which uses 16-bit binary encoding. The first 8 bits represent the row number of the sub-region, and the last 8 bits represent the column number of the sub-region. Its spatial coordinates, radiation feature vector, and calibration parameters are encapsulated into structured data units and arranged according to spatial topology to form a structured temperature measurement partition dataset. The spatial topology is arranged in ascending order of the row and column numbers of the sub-regions.

[0062] The reference field solution module is used to mine historical temperature measurement datasets under the same operating conditions, extract the regional thermal conductivity coefficient, detector radiation response function, and internal heat source intensity and the topological relationship with the temperature measurement space, and solve the theoretical temperature reference field of each temperature measurement sub-region based on the thermal conduction differential equation.

[0063] The process of mining historical temperature measurement datasets under the same operating conditions and extracting the spatial topological relationships between regional thermal conductivity, internal heat source intensity, detector radiation response function, and temperature measurement topology includes: Select historical temperature measurement records that are consistent with the current temperature measurement target in terms of material, structure, and environmental conditions to ensure consistency between historical data and current temperature measurement conditions. The selection criteria include that the deviation of parameters such as the material grade, structural dimensions, operating power, ambient temperature range, ambient humidity range, and atmospheric pressure range of the temperature measurement target does not exceed 5%. For continuously operating industrial equipment, it is also necessary to ensure that the time interval between the historical temperature measurement record collection time and the current time does not exceed 7 days.

[0064] The historical structured temperature measurement zoning dataset is analyzed to extract the thermal conductivity parameters of each sub-region. The thermal conductivity parameters include the thermal conductivity coefficients in the x and y directions. For isotropic materials, the thermal conductivity coefficients in the two directions are equal, while for anisotropic materials, the coefficients in the two directions are extracted separately.

[0065] By combining temperature change data from historical temperature measurement records, the internal heat source intensity data corresponding to each sub-region is calculated using the principle of heat conduction correlation. The specific implementation process is as follows: Initiate the temperature measurement data acquisition process, ensuring that the target is under the preset standard operating conditions (consistent with the selected historical temperature measurement conditions). Use the same acquisition parameters as the data preprocessing module to continuously acquire temperature change data of the target for 10 minutes. The acquisition frequency is consistent with the acquisition frame rate of the original infrared thermal image frame sequence (25 frames / second to 50 frames / second) to ensure that each frame of data corresponds to the real-time temperature value of each temperature measurement sub-region.

[0066] Steady-state determination processing is performed on all temperature data within 10 minutes. The temperature fluctuation value of each temperature measurement sub-region during the entire acquisition cycle is calculated. The temperature fluctuation value is defined as the difference between the maximum and minimum values ​​of all temperature data within 10 minutes in that sub-region. A continuous time period in which the temperature fluctuation of all temperature measurement sub-regions is less than ±0.1℃ is selected as the steady-state data segment. If the temperature fluctuation of a single sub-region exceeds the standard, the data segment corresponding to that sub-region is removed, and only the continuous data in which all sub-regions meet the steady-state conditions are retained.

[0067] After the steady-state data are determined, the finite element method is used to invert and solve for the intensity of the internal heat source. The specific steps are as follows: The target area is divided into finite element meshes according to temperature measurement sub-regions. Each temperature measurement sub-region corresponds to one finite element element. The meshing accuracy matches the size of the temperature measurement sub-region, ensuring that the geometric boundary of each element completely coincides with the boundary of the temperature measurement sub-region.

[0068] The thermal conductivity coefficients (extracted from historical data) and steady-state temperature data of each sub-region are used as material parameters and boundary conditions of the finite element model, respectively. A discrete finite element model is constructed in combination with the basic principle of heat conduction, and a set of linear equations about the intensity of the internal heat source is formed based on the thermal conduction equilibrium relationship of each element.

[0069] Here, the internal heat source intensity inversion is a post-processing calculation, not an unknown quantity to be solved. The inverted internal heat source intensity is used as an input parameter for solving the subsequent heat conduction equation to improve the consistency between the theoretical temperature reference field and the actual heat distribution. For a two-dimensional plane with no internal heat source in steady state or with a known internal heat source distribution, the heat conduction equation is used to perform physical consistency verification and smooth interpolation on the infrared thermometry results to generate the theoretical temperature reference field.

[0070] The temperature distribution pattern, heat conduction characteristics, and internal heat source intensity distribution range of each sub-region are statistically analyzed. The average, maximum, minimum, and standard deviation of historical temperature data for each sub-region are calculated to obtain the temperature distribution pattern. The range and probability distribution of the values ​​of the heat conduction coefficient and internal heat source intensity are statistically analyzed.

[0071] The least squares method was used to fit the correspondence between the detector output gray value and the actual temperature to obtain the detector radiation response function. Blackbody radiation data at different temperatures were collected, and 100 frames of infrared thermal images were collected at each temperature point. The average gray value of each frame of thermal images was taken as the gray value corresponding to that temperature point. With the actual temperature as the independent variable and the average gray value as the dependent variable, a quadratic polynomial was used for least squares fitting, and the goodness of fit R² was not less than 0.999.

[0072] Extract the spatial adjacency relationships and heat transfer paths of historical temperature measurement zones, construct a temperature measurement spatial topology graph, with each temperature measurement sub-region as a node and the adjacency relationship between sub-regions as edges. The weight of the edge is the thermal resistance between two adjacent sub-regions. The thermal resistance is calculated based on the thermal conductivity coefficient and contact area of ​​the sub-region. The heat transfer path is determined by the shortest path algorithm of the topology graph.

[0073] The process of solving the theoretical temperature reference field for each thermometric sub-region based on the heat conduction differential equation includes: A two-dimensional steady-state heat conduction differential equation is established for the target region. For a homogeneous medium without internal heat sources in steady state, the heat flowing into the control volume per unit time is equal to the heat flowing out of the control volume. When internal heat sources exist, the sum of the heat flowing in and the heat generated by the internal heat sources is equal to the heat flowing out. By performing heat balance analysis on the infinitesimal control volume in the two-dimensional plane, the following equation can be obtained: In the formula, , These are the thermal conductivity coefficients in the x and y directions, respectively. For sub-region temperature, The intensity of the internal heat source.

[0074] The equation is based on the assumption that the thickness of the temperature measurement target is much smaller than its length and width (thickness / length-to-width ratio < 1 / 10), heat is mainly transferred within a two-dimensional plane, the target is in a steady-state operation, the internal temperature does not change with time, and the effects of thermal radiation and convection on internal heat conduction are ignored. This assumption is applicable to scenarios where internal heat conduction is dominant, such as metal equipment and thick-walled electrical equipment; for thin-walled equipment with rapid heat dissipation, the surface convection heat transfer coefficient needs to be added as a third type of boundary condition.

[0075] Combining boundary conditions and historical thermal conductivity data, the two-dimensional steady-state thermal conductivity differential equation is solved discretically using the finite difference method. The boundary conditions include first-type boundary conditions and second-type boundary conditions. The first-type boundary condition is the temperature value on the known boundary, which is directly measured by an infrared thermal imager. The second-type boundary condition is the heat flux density on the known boundary, which is measured by a heat flux sensor.

[0076] The target region is divided into grids corresponding to the temperature measurement sub-regions. Each grid node corresponds to the geometric center of a temperature measurement sub-region. The grid division adopts a rule that perfectly matches the temperature measurement sub-regions, that is, one temperature measurement sub-region corresponds to one grid cell, and the coordinates of the grid nodes are completely consistent with the coordinates of the geometric center of the corresponding temperature measurement sub-region, ensuring that the grid division and the spatial distribution of the temperature measurement sub-regions are completely synchronized. The central difference scheme is used to discretize the heat conduction differential equation. The core is to transform the partial derivatives in the equation into the ratio of the temperature difference between adjacent grid nodes. The specific discretization process and corresponding formulas are as follows: For any grid node (i,j), its corresponding partial derivative in the x-direction Discretized using the central difference scheme as y-direction partial derivative Discretized using the central difference scheme as .

[0077] in, , These represent the grid step size in the x and y directions, respectively, which are equal to the actual physical dimensions of the corresponding thermometric sub-region in the x and y directions. , Let be the heat transfer coefficients in the x-direction between node (i,j) and its adjacent nodes (i+1,j) and (i-1,j), respectively. The average of the heat transfer coefficients of two adjacent grid cells is taken. , Let be the thermal conductivity coefficients in the y-direction between node (i,j) and its adjacent nodes (i,j+1) and (i,j-1), respectively. Similarly, the average of the thermal conductivity coefficients of two adjacent grid cells is taken. Let (i,j) be the temperature value of the grid node. , , , Let be the temperature values ​​of the four adjacent grid nodes above, below, left, and right of node (i,j). Substituting the discretized partial derivative expression into the two-dimensional steady-state heat conduction differential equation, and rearranging, we obtain a system of linear equations concerning the temperature values ​​of each grid node, expressed as: In the formula, , , , , These are all coefficient terms, calculated from the grid step size and thermal conductivity. The constant term is determined by the intensity of the internal heat source. The boundary conditions are determined. For the first type of boundary condition, the temperature values ​​of the boundary nodes are known and can be directly substituted into the equations. For the second type of boundary condition, the discrete equations for the boundary nodes are derived through the relationship between heat flux density and temperature gradient. The purpose of solving this system of equations is to smooth and interpolate the discrete temperature points obtained from infrared thermometry under physical constraints, generating a continuous theoretical temperature reference field.

[0078] After the linear equation system is constructed, Gaussian elimination is used to solve it. First, the linear equation system is transformed into an upper triangular matrix form, and then the temperature value of each grid node is solved sequentially through back substitution. During the solution process, the calculation error is controlled to be less than 1%. The obtained temperature value of each grid node is the theoretical temperature reference value of the corresponding thermometric sub-region.

[0079] An environmental background radiation temperature correction term is superimposed to generate the final theoretical temperature reference field. The environmental background radiation temperature is obtained by acquiring background infrared thermal images around the target and calculating the average temperature of the background region. The correction term is converted into a temperature increment after calculation in the radiance domain. The formula for calculating the correction term is as follows: in For the target surface emissivity, The ambient background radiance, Atmospheric transmittance, This is the Planck inverse function. The Planck inverse function is realized using piecewise polynomial fitting, with a fitting error of less than 0.05℃ in the 8μm to 14μm band, ranging from -20℃ to 150℃. The theoretical temperature reference value is added to the correction term to obtain the final theoretical temperature reference field.

[0080] The initial map generation module is used to configure differentiated temperature confidence intervals for each sub-region based on temperature measurement accuracy requirements and environmental fluctuation characteristics, and generate an initial three-dimensional temperature field feature map. The process includes: A three-dimensional coordinate system is constructed with spatial coordinates (x,y) as the horizontal axis and temperature T as the vertical axis. The spatial coordinates (x,y) are consistent with the coordinate system of the three-dimensional point cloud data of the lidar, and the coordinate range covers the entire spatial range and temperature range of the temperature measurement target.

[0081] The theoretical temperature reference values ​​of each temperature measurement sub-region are mapped to a three-dimensional coordinate system. A continuous temperature reference surface is generated using bilinear interpolation. The theoretical temperature reference value of each sub-region is assigned to the three-dimensional coordinate point corresponding to its geometric center. Then, bilinear interpolation is performed between four adjacent coordinate points to calculate the temperature value at the interpolation point. The interpolation step size is set to 1 pixel to ensure that the generated temperature reference surface has the same pixel resolution as the infrared thermal image. For sub-regions at the image edges, a boundary processing method of copying the edge is used to avoid interpolation distortion in edge regions.

[0082] Based on the temperature measurement accuracy requirements of each sub-region, differentiated temperature confidence intervals are configured for the temperature reference surface. For critical temperature measurement sub-regions such as electronic device chip areas and power equipment connector areas, the required temperature measurement accuracy is ±0.5℃, and the corresponding temperature confidence interval is set to the theoretical temperature reference value ±0.5℃. For non-critical temperature measurement sub-regions such as equipment casing areas, the required temperature measurement accuracy is ±1℃, and the corresponding temperature confidence interval is set to the theoretical temperature reference value ±1℃. Simultaneously, the confidence interval width is adjusted according to environmental fluctuation characteristics. When the environmental temperature fluctuation exceeds ±2℃, the confidence interval width for all sub-regions is increased by 0.5℃.

[0083] The upper and lower boundaries of the confidence interval are marked in the three-dimensional coordinate system to form an initial three-dimensional temperature field feature map. The upper and lower boundaries of the confidence interval are generated using the same interpolation method as the temperature reference surface and are distinguished by different colors.

[0084] The dynamic temperature field feature map module is used to collect data on the emissivity distribution of the target surface, the atmospheric radiation attenuation coefficient, and the dark current noise of the detector. It constructs a multi-field coupled adaptive correction tensor and performs pixel-by-pixel correction on the initial three-dimensional temperature field feature map to obtain the dynamic temperature field feature map.

[0085] The process of constructing a multi-field coupled adaptive correction tensor includes: Based on the emissivity distribution data of the target surface, an emissivity correction factor is calculated, which is applied to the radiance domain. At the same temperature, the radiance of an actual object equals the blackbody radiance multiplied by the object's emissivity. Therefore, when there is a deviation between the actual emissivity and the standard emissivity, compensation is needed using a correction factor. The calculation formula is as follows: In the formula, This is the emissivity correction factor. The target surface emissivity spectrum is derived from multi-source environmental calibration data to measure the surface emissivity. The standard emissivity is used. When the deviation between the measured emissivity and the standard emissivity exceeds ±5%, emissivity correction is initiated. The emissivity correction coefficient is limited to a range of 0.8 to 1.2.

[0086] Based on atmospheric temperature and humidity data, an atmospheric radiation correction factor is calculated, which is applied to the radiance domain. This is based on the standard radiative transfer equation: The true radiance of the target is obtained by deformation: In the formula, The radiance measured by the detector. Radiation along the upward path of the atmosphere. This is downward atmospheric radiation. Atmospheric transmittance, The target emission rate.

[0087] Atmospheric radiation correction factor is defined as The ratio is used only when the temperature fluctuation range is less than ±10℃ and the result is approximately linear; when the temperature fluctuation exceeds ±10℃, the above complete radiative transfer equation is used directly for pixel-by-pixel calculation without using a correction factor.

[0088] Based on the detector dark current noise data, a noise correction coefficient is calculated. The detector dark current noise is obtained by continuously acquiring 100 frames of infrared thermal images under the condition that the detector shutter is closed and there is no radiation input, calculating the average grayscale value of all pixels, and then converting it into the corresponding radiance. Dark current is additive noise, which is corrected by direct subtraction.

[0089] The emissivity correction coefficient, atmospheric radiation correction coefficient, and noise correction coefficient are arranged according to spatial dimensions to construct a 3×N-dimensional multi-field coupled adaptive correction tensor, where N is the number of temperature measurement sub-regions. The first row of the tensor corresponds to the emissivity correction coefficient, the second row corresponds to the atmospheric radiation correction coefficient, and the third row corresponds to the noise correction coefficient. Each column corresponds to a temperature measurement sub-region, and the three correction coefficients in that column are uniformly applied to all pixels in that sub-region.

[0090] The process of pixel-by-pixel correction of the initial three-dimensional temperature field feature map includes: The correction factor is applied to the radiance domain and then converted to a temperature value. For each pixel, its corresponding radiance is calculated. Then multiply by the emissivity correction factor and the atmospheric radiation correction factor, and subtract the radiance corresponding to dark current noise to obtain the corrected radiance. : Using Planck's inverse function Convert to corrected temperature value .

[0091] Based on the noise level of each sub-region, the width of the temperature confidence interval is dynamically adjusted. The noise level is obtained by calculating the gray-level variance of all pixels in the sub-region. When the gray-level variance increases by 10%, the width of the confidence interval for that sub-region is increased by 10%. When the gray-level variance decreases by 10%, the width of the confidence interval for that sub-region is decreased by 10%. The adjustment range of the confidence interval width does not exceed ±50% of the initial width.

[0092] A correction coefficient thermal layer is overlaid on the three-dimensional temperature field feature map to visually display the degree of correction in each region. The color of the thermal map is gradient from blue to red. Blue indicates that the correction coefficient is close to 1 and the degree of correction is small, while red indicates that the correction coefficient deviates more from 1 and the degree of correction is large. The transparency of the thermal map is set to 0.5.

[0093] A real-time data interface is established with multiple source sensors, including infrared thermal imagers, lidar, emissivity meters, weather stations, and heat flux sensors. The data interface uses the TCP / IP protocol, supports real-time data transmission, and has a transmission rate of no less than 10 Mbps. Under normal circumstances, the multi-field coupled adaptive correction tensor is updated once every preset time interval. The preset time interval is set to 1 to 10 seconds according to the rate of environmental change. When the weather station detects a sudden change in ambient temperature (more than ±3℃ within 1 minute) or a sudden change in humidity (more than ±10% within 1 minute), the correction tensor is immediately updated.

[0094] The reconstruction and localization module is used to acquire real-time infrared thermal image frame data and map it to the spatial coordinate system of the dynamic temperature field feature map. It calculates the mean temperature deviation and spatial structure similarity of the corresponding sub-regions to complete temperature field reconstruction and abnormal thermal zone localization. The process includes: Real-time infrared thermal image frame data is converted into grayscale temperature maps and mapped to the spatial coordinate system of the dynamic temperature field feature map. The same preprocessing operations as the data preprocessing module are performed on the real-time infrared thermal image frames, including semantic segmentation, contour extraction, spatial projection correction, and sub-region division, to obtain real-time temperature measurement sub-regions. A one-to-one correspondence is established between the real-time temperature measurement sub-regions and the sub-regions in the dynamic temperature field feature map using spatial identifiers. If there is a positional deviation between the real-time sub-region and the reference sub-region (deviation less than 3 pixels), nearest neighbor interpolation is used for registration; if the deviation exceeds 3 pixels, a new reference sub-region is generated and the theoretical temperature reference field is updated.

[0095] Calculate the deviation between the actual mean temperature and the theoretical baseline temperature for each temperature measurement sub-region. The actual mean temperature is the arithmetic mean of the real-time temperature measurement sub-region after removing 3σ outlier pixels (noisy pixels whose grayscale values ​​deviate from the sub-region mean by more than 3 standard deviations). The formula for calculating the mean temperature deviation is as follows: ,in This represents the temperature mean deviation. This is the average actual temperature. The theoretical temperature reference value is derived from the dynamic temperature field characteristic spectrum.

[0096] The spatial structural similarity index is used to calculate the similarity between the feature maps of real-time and dynamic temperature fields. This index measures the similarity between two images from three aspects: brightness, contrast, and structure. The comprehensive structural similarity index is obtained by calculating brightness similarity, contrast similarity, and structural similarity separately, and then multiplying the three together. The calculation formula is as follows: In the formula, SSIM represents the structural similarity index. , These represent the pixel mean values ​​of the corresponding regions in the real-time temperature field and the dynamic spectrum, respectively. , These represent the pixel variances of the corresponding regions in the real-time temperature field and the dynamic spectrum, respectively. This represents the pixel covariance between two regions. , This is a constant used to avoid a denominator of zero. Before calculating SSIM, the temperature values ​​of the real-time temperature field and dynamic spectrum need to be normalized so that their dynamic range is mapped to the [0,1] interval. At this point, we take... , ,Right now , The structural similarity index ranges from 0 to 1, with values ​​closer to 1 indicating greater similarity in the spatial structures of the two temperature fields.

[0097] When the average temperature deviation exceeds a preset threshold or the structural similarity is lower than a preset threshold, the sub-region is marked as an abnormal hot zone and an alarm message is output. The preset threshold for the average temperature deviation is set according to the importance of the sub-region: 2℃ for critical sub-regions (such as power connectors and chip core areas) and 5℃ for non-critical sub-regions (such as equipment casings and brackets); the preset threshold for structural similarity is 0.8. A level one alarm is triggered when both the average temperature deviation exceeds the preset threshold and the structural similarity is lower than the preset threshold; a level two alarm is triggered when only one condition is met. The alarm message includes the location, number, actual temperature, theoretical temperature, temperature deviation, and structural similarity of the abnormal hot zone, and the abnormal hot zone is highlighted in red on the three-dimensional temperature field feature map.

[0098] In summary, this embodiment provides an infrared thermal imaging temperature field reconstruction system based on spectral analysis. Through simultaneous preprocessing of multi-source data and pixel-level semantic segmentation, it accurately divides the temperature measurement sub-regions and extracts radiation features, constructing a structured dataset and significantly improving the effectiveness and standardization of the original data. Relying on historical operating condition data mining and solving the thermal conduction differential equation, a theoretical temperature reference field that closely matches the actual thermal characteristics of the target is generated, ensuring the theoretical accuracy of the temperature field reconstruction from the source. The use of a three-dimensional temperature field feature map combined with differentiated confidence interval settings takes into account the temperature measurement accuracy requirements of different regions, making the temperature field presentation more intuitive and adaptable. Through multi-field coupling adaptive correction tensors, pixel-by-pixel dynamic correction is achieved, effectively offsetting interference factors such as emissivity, atmospheric radiation, and detector noise, significantly improving temperature measurement accuracy and environmental adaptability. Ultimately, through real-time data mapping and joint judgment of deviation and similarity, the system automatically completes accurate reconstruction of the temperature field and rapid location of abnormal hot areas without human intervention. While improving temperature measurement accuracy and reconstruction effect, it also realizes intelligent monitoring and fault early warning of the thermal state of complex targets, providing stable and reliable technical support for non-contact thermal detection in industries such as industry and power.

[0099] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0100] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An infrared thermal image temperature field reconstruction system based on spectral analysis, characterized in that, include: The data preprocessing module is used to simultaneously preprocess the raw infrared thermal image frame sequence and multi-source environmental calibration data, divide the target temperature measurement sub-regions, fuse calibration parameters to extract the radiation feature vectors of each sub-region, and construct a structured temperature measurement partition dataset. The reference field solution module is used to mine historical temperature measurement datasets under the same operating conditions, extract the regional thermal conductivity coefficient, detector radiation response function, and internal heat source intensity and the topological relationship with the temperature measurement space, and solve the theoretical temperature reference field of each temperature measurement sub-region based on the thermal conductivity differential equation. The initial map generation module is used to configure differentiated temperature confidence intervals for each sub-region based on temperature measurement accuracy requirements and environmental fluctuation characteristics, and generate an initial three-dimensional temperature field feature map. The dynamic temperature field feature map module is used to collect target surface emissivity distribution, atmospheric radiation attenuation coefficient and detector dark current noise data, construct multi-field coupled adaptive correction tensor, and perform pixel-by-pixel correction on the initial three-dimensional temperature field feature map to obtain the dynamic temperature field feature map. The reconstruction and positioning module is used to acquire real-time infrared thermal image frame data and map it to the spatial coordinate system of dynamic temperature field feature map. It calculates the temperature mean deviation and spatial structure similarity of the corresponding sub-regions to complete temperature field reconstruction and abnormal hot zone positioning.

2. The infrared thermal imaging temperature field reconstruction system based on spectral analysis according to claim 1, characterized in that, The raw infrared thermal image frame sequence includes a grayscale thermal image stream, frame synchronization timestamps, and detector gain configuration files; the multi-source environmental calibration data includes a blackbody calibration coefficient matrix, an atmospheric transmittance parameter table, a target surface emissivity map, and lidar three-dimensional point cloud data.

3. The infrared thermal imaging temperature field reconstruction system based on spectral analysis according to claim 2, characterized in that, The process of dividing the target temperature measurement sub-region by the data preprocessing module includes: The U-Net network is used to perform pixel-level semantic segmentation on infrared grayscale thermal images to identify the target subject, background and interference thermal areas; The Canny edge detection algorithm is used to extract the contour features of the target subject, and spatial projection correction is performed by combining it with the 3D point cloud data of LiDAR. Based on contour features, the target body is divided into several continuous temperature measurement sub-regions, generating an initial temperature measurement partition sequence; The target temperature measurement sub-region is obtained by removing non-effective temperature measurement sub-regions in the initial sequence whose area is smaller than a preset threshold and whose gray-level variance is lower than the noise level.

4. The infrared thermal imaging temperature field reconstruction system based on spectral analysis according to claim 1, characterized in that, The process by which the data preprocessing module constructs the structured temperature measurement zoning dataset includes: Based on the pixel coordinates of the sub-regions and the point cloud data of the lidar, the actual physical size and temperature measurement distance of each sub-region are calculated. By combining the blackbody calibration coefficient matrix, the gray values ​​of the sub-regions are converted into the original radiance values, and then into radiance values. By introducing atmospheric transmittance parameters, the standard radiative transfer model is used to correct the radiance value for atmospheric attenuation. Calculate the radiation flux density and gray-level gradient variance of each sub-region to generate a radiation feature vector. The radiation feature vector includes the average gray value of the sub-region, the gray-level gradient variance, the radiation flux density, and the geometric center coordinates of the sub-region. Each temperature measurement sub-region is assigned a unique spatial identifier, and its spatial coordinates, radiation feature vector, and calibration parameters are encapsulated into structured data units, which are then arranged according to spatial topological relationships to form a structured temperature measurement partition dataset.

5. The infrared thermal imaging temperature field reconstruction system based on spectral analysis according to claim 1, characterized in that, The process by which the reference field solving module extracts the relationship between the regional thermal conductivity coefficient, the intensity of the internal heat source, the detector radiation response function, and the topological relationship of the temperature measurement space includes: Select historical temperature measurement records that match the material, structure, and environmental conditions of the current temperature measurement target to ensure consistency between historical data and current temperature measurement conditions; Analyze the historical structured temperature measurement zoning dataset, extract the heat conduction parameters of each sub-region, combine the temperature change data in the historical temperature measurement records, calculate the internal heat source intensity data of each sub-region through the heat conduction equation, and statistically analyze the temperature distribution pattern, heat conduction characteristics and internal heat source intensity distribution range of each sub-region. The detector's radiation response function is obtained by fitting the correspondence between the detector's output grayscale value and the actual temperature using the least squares method. Extract the spatial adjacency relationships and heat flow transfer paths of historical temperature measurement zones to construct a temperature measurement spatial topology map.

6. The infrared thermal imaging temperature field reconstruction system based on spectral analysis according to claim 1, characterized in that, The process by which the reference field solving module solves for the theoretical temperature reference field of each thermometric sub-region based on the heat conduction differential equation includes: Establish the two-dimensional steady-state heat conduction differential equation for the target region; By combining boundary conditions and historical heat transfer coefficient data, the two-dimensional steady-state heat transfer differential equation is solved discretically using the finite difference method to obtain the theoretical temperature reference value for each sub-region. By superimposing the ambient background radiation temperature correction term, the final theoretical temperature reference field is generated.

7. The infrared thermal imaging temperature field reconstruction system based on spectral analysis according to claim 1, characterized in that, The process by which the initial map generation module generates the initial three-dimensional temperature field feature map includes: Construct a three-dimensional coordinate system with spatial coordinates (x, y) as the horizontal axis and temperature T as the vertical axis; The theoretical temperature reference values ​​of each temperature measurement sub-region are mapped to a three-dimensional coordinate system, and a continuous temperature reference surface is generated by bilinear interpolation. Based on the temperature measurement accuracy requirements of each sub-region, differentiated temperature confidence intervals are configured for the temperature reference surface; Mark the upper and lower boundaries of the confidence interval in the three-dimensional coordinate system to form an initial three-dimensional temperature field feature map.

8. The infrared thermal imaging temperature field reconstruction system based on spectral analysis according to claim 1, characterized in that, The process of constructing a multi-field coupled adaptive correction tensor through dynamic graph correction includes: Calculate the emissivity correction coefficient based on the emissivity distribution data of the target surface; Calculate the atmospheric radiation correction coefficient based on atmospheric temperature and humidity data; Calculate the radiance value corresponding to the dark current noise based on the detector's dark current noise data; The emissivity correction coefficient, atmospheric radiation correction coefficient, and dark current noise radiance value are arranged according to spatial dimensions to construct a 3×N-dimensional multi-field coupled adaptive correction tensor, where N is the number of thermometric sub-regions.

9. The infrared thermal imaging temperature field reconstruction system based on spectral analysis according to claim 8, characterized in that, The dynamic map correction process for pixel-by-pixel correction of the initial three-dimensional temperature field feature map includes: The temperature values ​​of the initial three-dimensional temperature field feature map are converted into radiance values, and then pixel-wise operations are performed with the multi-field coupled adaptive correction tensor to obtain the corrected radiance values. The corrected radiance values ​​are converted into temperature values ​​using the Planck inverse function, resulting in the corrected temperature reference surface. The width of the temperature confidence interval is dynamically adjusted based on the noise level of each sub-region; A correction coefficient thermodynamic layer is overlaid on the three-dimensional temperature field feature map to show the degree of correction in each region; The multi-field coupled adaptive correction tensor is updated once every preset time interval. When a sudden change in ambient temperature or humidity is detected, the correction tensor update is triggered.

10. The infrared thermal imaging temperature field reconstruction system based on spectral analysis according to claim 1, characterized in that, The process by which the reconstruction and positioning module completes temperature field reconstruction and abnormal hot zone location includes: Real-time infrared thermal image frame data is converted into grayscale temperature maps and mapped to the spatial coordinate system of dynamic temperature field feature map; Calculate the deviation between the actual mean temperature and the theoretical reference temperature value for each temperature measurement sub-region; The spatial structural similarity between the feature maps of the real-time temperature field and the dynamic temperature field is calculated using the structural similarity index. When the average temperature deviation exceeds a preset threshold or the structural similarity is lower than a preset threshold, the sub-region is marked as an abnormal hot zone and an alarm message is output.

Images