Method and system for detecting thermal defects of glass curtain wall

Abstract

The present application relates to a kind of glass curtain wall heat defect detection method and system.By collecting curtain wall multi-modal sensing data and spatial attitude information and parsing to obtain effective thermal radiation component, based on multispectral visible light image constructs pixel level curtain wall area radiation characteristic distribution and combines spatial attitude information to establish environmental radiation mapping relationship, three types of data are used to carry out radiation inverse operation jointly restore single visual angle curtain wall real temperature field data, to the global smooth temperature field that multi-angle temperature field data is integrated, according to heat defect determination standard, complete temperature field analysis and identification and output standardization curtain wall heat defect detection result, it is helpful to peel off the measurement deviation caused by the reflection radiation of multiple sources environment, adapt the emissivity spatial difference problem brought by curtain wall facade glass material mixed, weaken the negative influence of reflection difference on fusion result for different shooting angles, provide stable and reliable technical support for building facade all-weather nondestructive heat defect census, curtain wall long-term energy saving operation and maintenance evaluation.

Description

Technical Field

[0001] This invention relates to the field of building curtain wall health detection and radiation physics modeling technology, and more specifically, to a method and system for detecting thermal defects in glass curtain walls. Background Technology

[0002] Glass curtain walls are widely used in modern building facades and are a key building envelope system of concern in the fields of building energy conservation and building operation and maintenance. Infrared thermal imaging detection methods, with their advantages of non-contact and long-distance scanning, are gradually being applied to the investigation of curtain wall thermal performance and the inspection of damage defects. Conventional infrared detection methods can rely on infrared detectors to collect target radiation signals and calculate the apparent temperature of components based on the laws of blackbody radiation, enabling a rapid survey of the thermal condition of building facades.

[0003] In the process of realizing this invention, the inventors discovered that existing infrared thermal defect detection methods for glass curtain walls have insurmountable core defects: the emissivity range of glass curtain wall materials is large and the spatial distribution is uneven. At the same time, the facade is easily affected by multi-angle environmental reflection and radiation interference. Conventional infrared detection is difficult to remove the interference of various environmental reflection components, cannot restore the true temperature field of curtain wall components, has a large temperature measurement deviation, and the subtle thermal defect features are easily masked by reflected noise, which greatly reduces the reliability of curtain wall thermal defect identification. Summary of the Invention

[0004] Based on this, and to address the aforementioned problems, this invention provides a method and system for detecting thermal defects in glass curtain walls. It acquires multimodal sensing data and spatial attitude information of the curtain wall, including polarized infrared and multispectral visible light images, and analyzes this data to obtain effective thermal radiation components. Based on the multispectral visible light images, it constructs a pixel-level distribution of the curtain wall's regional radiation characteristics and establishes an environmental radiation mapping relationship by combining spatial attitude information. It uses a combination of these three types of data to perform radiation inverse calculations to reconstruct the true temperature field data of the curtain wall from a single perspective. It integrates the multi-view temperature field data to obtain a globally smoothed temperature field. Based on thermal defect judgment standards, it completes temperature field analysis and identification and outputs standardized curtain wall thermal defect detection results. This helps to eliminate measurement deviations caused by multi-source environmental reflected radiation, adapts to the spatial differences in emissivity caused by the mixture of glass materials on the curtain wall facade, weakens the negative impact of reflection differences from different shooting angles on the fusion results, improves the detection capability of weak thermal defects such as hollow air filling failure and coating aging, expands the applicable time period for curtain wall infrared detection, and provides stable and reliable technical support for all-weather non-destructive thermal defect surveys of building facades and long-term energy-saving operation and maintenance assessments of curtain walls.

[0005] In a first aspect, the present invention provides a method for detecting thermal defects in glass curtain walls, comprising: acquiring multimodal sensing data and spatial attitude information of the curtain wall including polarized infrared images and multispectral visible light images; performing polarization feature analysis on the polarized infrared images to obtain effective thermal radiation components; generating a radiation characteristic distribution of the curtain wall area based on the multispectral visible light images; combining the environmental radiation components obtained from the spatial attitude information to construct an environmental radiation mapping relationship; performing radiation inverse calculation using the effective thermal radiation components, the radiation characteristic distribution of the curtain wall area, and the environmental radiation mapping relationship as joint inputs to reconstruct the true temperature field data of the curtain wall from a single perspective; acquiring multiple sets of true temperature field data of the curtain wall from different shooting perspectives and integrating them to generate a globally smooth temperature field; analyzing and identifying the globally smooth temperature field according to a preset thermal defect judgment standard, and outputting standardized curtain wall thermal defect detection results.

[0006] Optionally, multimodal sensing data and spatial attitude information of the curtain wall, including polarized infrared images and multispectral visible light images, are acquired. Polarization feature analysis is performed on the polarized infrared images to obtain the effective thermal radiation component. This includes: using a synchronous sensing device to perform external field scanning and acquisition of the curtain wall, obtaining polarized infrared images, multispectral visible light images, and synchronously recorded spatial attitude information; solving for polarization quantization indices on the polarized infrared images to obtain the radiation source classification results for each pixel; filtering the classification results to remove interfering pixels caused by mirror reflections, obtaining pixel data of the thermal radiation of the glass curtain wall itself, and outputting the effective thermal radiation component. By synchronously acquiring polarized infrared images, multispectral visible light images, and associated spatial attitude information using a synchronous sensing device, solving for polarization quantization indices on the polarized infrared images to classify pixel radiation sources, and removing interfering pixels from mirror reflections to obtain the effective thermal radiation component, it is possible to distinguish between the thermal radiation of the glass itself and the polarized reflected radiation from the environment, eliminate invalid pixel data corresponding to high reflectivity noise, reduce radiation measurement deviations caused by sunlight and mirror reflections from surrounding buildings, expand the applicable time period for curtain wall infrared detection, and provide clean, reliable, and effective radiation input data for subsequent applications.

[0007] Optionally, the radiation characteristic distribution of the curtain wall area is generated based on the multispectral visible light image, and the environmental radiation components are deduced by combining spatial attitude information to construct an environmental radiation mapping relationship. This includes: extracting reflection feature data inside the multispectral visible light image; retrieving a pre-stored material radiation association model to perform matching calculations on the reflection feature data to generate a pixel-level curtain wall area radiation characteristic distribution; deducing the sky, sunlight, and building radiation components according to the spatial attitude information to obtain various environmental components and shooting orientation information; and performing correlation processing on the pixel-level curtain wall area radiation characteristic distribution, environmental radiation components, and shooting orientation information to obtain an environmental radiation mapping relationship. By extracting the reflection feature data of multispectral visible light images, and calling the pre-stored material radiation association model to generate the pixel-level radiation characteristic distribution of the curtain wall area, the system relies on spatial attitude information to infer the radiation components of various environments such as sky, sunlight, and buildings, and combines the shooting orientation information to complete the association processing to construct the environmental radiation mapping relationship. It can match the emissivity parameters corresponding to different glass materials pixel by pixel, fully cover the multi-source environmental reflection radiation in urban scenes, eliminate the uniform emissivity assumption error caused by the mixing of curtain wall glass types, accurately quantify the contribution of environmental reflection radiation in the whole scene, and provide a modeling foundation that fits the real working conditions of the curtain wall for subsequent radiation inverse calculation.

[0008] The above implementation process, obtaining the environmental radiation mapping relationship, also includes: pre-storing radiation correlation models for multiple types of glass materials; retrieving preset material parameters of the curtain wall and performing matching calculations with the radiation correlation models to obtain the pixel-level curtain wall area radiation characteristic distribution; extrapolating the radiation components of the sky, sunlight, and buildings based on spatial attitude information to obtain various environmental radiation components and shooting orientation information; and performing correlation processing on the pixel-level curtain wall area radiation characteristic distribution, various environmental radiation components, and shooting orientation information to obtain the environmental radiation mapping relationship. By pre-storing radiation correlation models corresponding to multiple types of glass materials, when multispectral visible light images are missing, the preset material parameters of the curtain wall are retrieved and matched with the radiation correlation models to generate the pixel-level curtain wall area radiation characteristic distribution. Based on spatial attitude information, various environmental radiation components are extrapolated and combined with shooting orientation information to construct the environmental radiation mapping relationship. This allows for curtain wall radiation characteristic and environmental radiation modeling even in multispectral acquisition failure scenarios, ensuring the complete operation of the temperature reconstruction process under data loss conditions and improving the environmental adaptability and operational stability of the entire detection solution.

[0009] Optionally, radiation inverse calculation is performed using the effective thermal radiation component, the radiation characteristic distribution of the curtain wall area, and the environmental radiation mapping relationship as joint inputs to restore the true temperature field data of the curtain wall from a single viewpoint. This includes: listing the effective thermal radiation component, the radiation characteristic distribution of the curtain wall area, and the environmental radiation mapping relationship as unified calculation inputs; sequentially performing atmospheric compensation, reflection stripping, and blackbody temperature inverse conversion on all input data to obtain an intermediate corrected image; and performing pixel filtering processing on the intermediate corrected image to remove distorted pixels, thereby obtaining the true temperature field data of the curtain wall from a single viewpoint. By unifying the effective thermal radiation component, the distribution of radiation characteristics of the pixel-level curtain wall area, and the environmental radiation mapping relationship as joint inputs, atmospheric compensation, reflection stripping, and blackbody temperature inverse conversion are performed sequentially to obtain intermediate corrected images and filter out distorted pixels to restore the true temperature field data of the curtain wall from a single viewpoint. This can simultaneously eliminate multiple interferences from atmospheric path radiation, material emissivity inhomogeneity, and multi-source environmental reflection, accurately stripping various interference radiation signals, restoring the true thermodynamic temperature of the glass curtain wall at the pixel level, significantly reducing the error of the temperature measurement system for coated glass curtain walls, and providing high-precision temperature basis data for multi-view fusion and weak thermal defect identification.

[0010] In the above implementation process, atmospheric compensation, reflection stripping, and blackbody temperature inverse conversion operations are sequentially performed on all input data, including: atmospheric compensation calculation based on the collected atmospheric transmission parameters and combined with the effective thermal radiation component to obtain the surface radiance after removing atmospheric path radiation; reflection stripping calculation based on the surface radiance, the radiation characteristic distribution of the curtain wall area, and the environmental radiation mapping relationship to obtain the pure target blackbody radiance; and blackbody temperature inverse conversion of the pure target blackbody radiance based on the following formula to obtain the single pixel temperature value: in, This represents the actual thermodynamic temperature value of a single-pixel glass curtain wall. A pre-configured blackbody radiation constant; Preset the long-wave infrared equivalent working wavelength for infrared imaging equipment; To obtain the pure target blackbody radiance after removing atmospheric and environmental reflections, all single-pixel temperature values ​​are integrated according to image spatial arrangement rules to generate an intermediate correction image. Atmospheric compensation calculations are performed based on field-collected atmospheric transmission parameters to eliminate atmospheric path radiation interference. Environmental reflection signals are separated by reflection stripping operations, combined with the radiation characteristic distribution of the curtain wall area and the environmental radiation mapping relationship. Then, the temperature of the pure target blackbody radiance is calculated pixel by pixel using Planck's inverse operation formula and integrated to generate an intermediate correction image. This step-by-step, layer-by-layer elimination of multi-layer measurement errors caused by atmospheric attenuation and multi-source environmental reflections accurately calculates the true thermodynamic temperature of glass curtain wall pixels, significantly reducing daytime temperature measurement errors of low-emissivity coated glass such as Low-E and improving the accuracy of single-view temperature field reconstruction.

[0011] Optionally, multiple sets of single-view real temperature field data of the curtain wall from different shooting angles are acquired and integrated to generate a globally smooth temperature field. This includes: collecting single-view real temperature field data of the curtain wall from different shooting angles on the same facade; classifying the interference levels of the single-view real temperature field data of the curtain wall from each angle according to the optical reflection variation law, and obtaining the fusion weights for each angle; identifying the angle data corresponding to the high interference level and completing the removal process; performing weighted calculations on the remaining angle data to obtain a globally smooth temperature field. By collecting single-view real temperature field data of the same curtain wall facade from multiple angles, classifying the interference levels of each angle according to the optical reflection variation law and matching the corresponding fusion weights, and removing the high interference angle data, the remaining valid data is used to generate a globally smooth temperature field through weighted calculations. This can distinguish the data quality differences caused by the difference in the degree of reflection under different shooting angles, avoid interference from the fusion results of high reflection distortion angle data, weaken the temperature value abrupt changes caused by single-view reflection noise, reduce the discrete error of the overall temperature field reconstruction of the curtain wall, and further improve the reliability of subsequent defect identification.

[0012] Optionally, the global smooth temperature field is analyzed and identified according to preset thermal defect judgment criteria, and standardized curtain wall thermal defect detection results are output. This includes: retrieving differentiated screening methods adapted to different structural parts of the curtain wall; performing multi-dimensional feature analysis on each region of the global smooth temperature field to obtain information on all thermal anomaly areas; and summarizing and organizing all thermal anomaly area information to obtain standardized curtain wall thermal defect detection results. By retrieving differentiated screening methods adapted to different structural parts of the curtain wall, performing multi-dimensional feature analysis on each region of the global smooth temperature field to extract information on thermal anomaly areas, and summarizing and organizing to obtain standardized curtain wall thermal defect detection results, this method can adapt to different thermal defect manifestations corresponding to various curtain wall components. It captures weak abnormal features in the temperature field from multiple dimensions, reduces the problems of missed detection and false detection caused by single-dimensional identification, effectively identifies hidden thermal defects with small amplitudes such as hollow air filling failure and coating aging, improves the standardization of defect output, and provides a clear and complete defect judgment basis for curtain wall operation and maintenance.

[0013] In the above implementation process, after outputting standardized curtain wall thermal defect detection results, the following steps are also included: statistically calculating the temperature of all pixels within the globally smoothed temperature field to obtain curtain wall thermal insulation-related indicators; and binding and archiving the thermal insulation indicators with thermal anomaly area information to obtain a comprehensive curtain wall evaluation file. By statistically calculating the temperature of all pixels within the globally smoothed temperature field to obtain curtain wall thermal insulation-related indicators, and binding and archiving the thermal insulation indicators with the thermal anomaly area information obtained through analysis and identification to form a comprehensive curtain wall evaluation file, the curtain wall temperature distribution, thermal defect location, and curtain wall energy-saving performance data can be simultaneously stored. This achieves integrated recording of defects and thermal insulation performance, providing complete and traceable data support for long-term operation and maintenance and major renovation decisions of the curtain wall.

[0014] Secondly, the present invention also provides a thermal defect detection system for glass curtain walls. This system includes: a data acquisition and analysis module, used to acquire multimodal sensing data and spatial attitude information of the curtain wall, including polarized infrared images and multispectral visible light images, and to perform polarization feature analysis on the polarized infrared images to obtain effective thermal radiation components; a radiation mapping construction module, used to generate a radiation characteristic distribution of the curtain wall area based on the multispectral visible light images, and to construct an environmental radiation mapping relationship by combining the environmental radiation components derived from the spatial attitude information; a temperature field restoration module, used to perform radiation inverse calculation using the effective thermal radiation components, the radiation characteristic distribution of the curtain wall area, and the environmental radiation mapping relationship as joint inputs to restore the true temperature field data of the curtain wall from a single perspective; a temperature field fusion module, used to acquire multiple sets of true temperature field data of the curtain wall from multiple shooting angles and integrate them to generate a globally smooth temperature field; and a defect identification and output module, used to analyze and identify the globally smooth temperature field according to a preset thermal defect judgment standard, and output standardized curtain wall thermal defect detection results.

[0015] According to the technical solution of the present invention, by acquiring multimodal sensing data and spatial attitude information of the curtain wall, including polarized infrared images and multispectral visible light images, and analyzing them to obtain effective thermal radiation components, a pixel-level distribution of the radiation characteristics of the curtain wall area is constructed based on the multispectral visible light images, and an environmental radiation mapping relationship is established by combining the spatial attitude information. The three types of data are used to jointly carry out radiation inverse calculation to restore the true temperature field data of the curtain wall from a single perspective. The multi-view temperature field data are integrated to obtain a globally smooth temperature field. According to the thermal defect judgment standard, the temperature field is analyzed and identified, and standardized curtain wall thermal defect detection results are output. This helps to eliminate measurement deviations caused by multi-source environmental reflected radiation, adapt to the problem of spatial differences in emissivity caused by the mixture of glass materials on the curtain wall facade, weaken the negative impact of reflection differences from different shooting angles on the fusion results, improve the detection capability of weak thermal defects such as hollow air filling failure and coating aging, expand the applicable time period of curtain wall infrared detection, and provide stable and reliable technical support for all-weather non-destructive thermal defect survey of building facades and long-term energy-saving operation and maintenance assessment of curtain walls. Attached Figure Description

[0016] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of a method for detecting thermal defects in glass curtain walls provided in an embodiment of the present invention; Figure 2 This is a flowchart of the multimodal data acquisition and parsing method provided in the embodiments of the present invention; Figure 3 This is a flowchart of the radiation mapping construction method provided in the embodiments of the present invention; Figure 4 This is a flowchart of the radiation mapping substitution construction method provided in the embodiments of the present invention; Figure 5 This is a flowchart of the temperature field restoration method provided in the embodiments of the present invention; Figure 6 This is a flowchart of the stepwise correction method for radiation signals provided in the embodiments of the present invention; Figure 7 This is a flowchart of the multi-view temperature field fusion method provided in the embodiments of the present invention; Figure 8 This is a flowchart of the thermal defect analysis and identification method provided in the embodiments of the present invention; Figure 9 This is a flowchart of the curtain wall evaluation file generation method provided in the embodiments of the present invention; Figure 10 This is a schematic diagram of the thermal defect detection system for glass curtain walls provided in an embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will now be described with reference to the accompanying drawings. For example, the flowcharts and block diagrams in the drawings illustrate the architecture, functions, and operations of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagram may represent a module, program segment, or part of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and combinations of blocks in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or can be implemented using a combination of dedicated hardware and computer instructions. In addition, the functional modules in the various embodiments of the present invention may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0019] Please refer to Figure 1 , Figure 1 This is a flowchart of a method for detecting thermal defects in glass curtain walls provided in an embodiment of the present invention. The method for detecting thermal defects in glass curtain walls includes: Step S100: Collect multimodal sensing data and spatial attitude information of the curtain wall, including polarized infrared images and multispectral visible light images, and perform polarization feature analysis on the polarized infrared images to obtain the effective thermal radiation components.

[0020] In step S100 above, a synchronous sensing device equipped with a long-wave infrared polarization camera and a multispectral visible light camera with an integrated micro-polarization array performs a field scanning operation on the curtain wall, simultaneously acquiring two types of image data: polarized infrared images and multispectral visible light images, and simultaneously recording the spatial attitude information corresponding to the acquisition time. The long-wave infrared polarization camera operates in a band covering 8~12μm and can simultaneously acquire radiance maps in four directions: 0°, 45°, 90°, and 135°, without rotating the polarizer. The acquired four-directional polarized infrared raw images are preprocessed with non-uniformity correction and bad pixel replacement to eliminate sensor response differences and obtain corrected four-directional radiance maps. Subsequently, based on the corrected four-directional radiance maps, the Stokes parameters are calculated using the following formula: Where I0 is the pixel value of the radiance image corresponding to the 0° polarization direction acquired by the polarization camera; I 45 The pixel values ​​of the radiance image corresponding to the 45° polarization direction acquired by the polarization camera; I 90 The pixel values ​​of the radiance image corresponding to the 90° polarization direction acquired by the polarization camera; I 135 S1 represents the pixel value of the radiance image corresponding to the 135° polarization direction acquired by the polarization camera; S2 represents the first Stokes parameter, which represents the total radiance; S3 represents the second Stokes parameter, which represents the difference between the horizontal and vertical linear polarization components; and S4 represents the third Stokes parameter, which represents the difference between the 45° and 135° linear polarization components.

[0021] Furthermore, the polarization quantization index, linear polarization degree, is calculated based on the following formula: The system uses linear polarization to classify the radiation sources of each pixel. Physically, glass mirror reflection exhibits strong polarization characteristics; for example, the polarization effect of mirror reflection is significant around a Brewster angle of approximately 56°. Meanwhile, the thermal radiation from the glass itself is unpolarized radiation, thus distinguishing between the two types of radiation signals. Pixel filtering is performed on the classification results. Different types of glass, coatings, incident angles, and environmental reflection conditions result in different polarization behaviors. For example, a threshold DOP < 0.35 can be set to identify areas dominated by the thermal radiation of the glass itself and retain them, while DOP ≥ 0.35 can be identified as areas dominated by the environmental reflection of the mirror and filtered out as interfering pixels. Only pixel data corresponding to the thermal radiation of the glass curtain wall itself are retained, and the effective thermal radiation components are output to provide clean, reliable, and effective radiation input data for subsequent radiation modeling and temperature field reconstruction.

[0022] Step S200: Generate the radiation characteristic distribution of the curtain wall area based on the multispectral visible light image, and construct the environmental radiation mapping relationship by combining the environmental radiation components obtained by spatial attitude information.

[0023] In step S200 above, the multispectral visible light image obtained from the acquisition operation is read. The multispectral visible light camera covers the 400~1000nm wavelength band and includes no less than 8 spectral bands. The reflection feature data inside the image is extracted, and the pre-stored radiation correlation models corresponding to multiple types of glass materials are retrieved to perform matching calculations on the reflection feature data. For the known glass type and coating type actually used in the curtain wall, an empirical mapping between near-infrared reflectivity and long-wave infrared emissivity is pre-established through offline calibration samples to construct the mapping relationship. ,in, The long-wave infrared emissivity of the glass at the pixel position corresponding to the image coordinates (x, y); These are the visible light spectral reflectances of the pixel in the 700nm and 850nm wavelength bands, respectively. The pre-established radiation correlation mapping function covers six types of glass: ordinary float glass, single silver Low-E, double silver Low-E, heat-reflective, body-colored, and glaze-printed glass. It generates the radiation characteristic distribution of the curtain wall area at the pixel level. In the specific implementation process, when the emissivity of the same material area changes by more than 0.15, it is marked as a candidate area for coating degradation. The emissivity estimation error can be controlled within 0.05.

[0024] Furthermore, the spatial attitude information recorded synchronously during the acquisition phase is read, and the three types of radiation components—sky, sunlight, and buildings—are extrapolated based on this spatial attitude information: the diffuse radiance of the sky can be solved using the MODTRAN atmospheric radiative transfer model combined with airborne meteorological sensors to collect data on temperature, humidity, visibility, and aerosol type; the radiance of surrounding buildings is extrapolated based on the 360° panoramic visible light image of the UAV combined with the equivalent temperature of the building materials; the direct solar radiance uses the NRELSPA algorithm to calculate the solar azimuth and elevation angles, and only includes the solar component when the sun is within ±3° solid angle of the glass mirror reflection; the various environmental radiation components and shooting orientation information are obtained by comprehensive solution; the radiation characteristic distribution of the pixel-level curtain wall area, various environmental radiation components, and shooting orientation information are correlated and processed to quantify the multi-source reflection radiance corresponding to each pixel, obtain the environmental radiation mapping relationship, and solve the temperature measurement deviation problem caused by the simplified modeling of single sky temperature.

[0025] Step S300: Using the effective thermal radiation component, the radiation characteristic distribution of the curtain wall area, and the environmental radiation mapping relationship as joint inputs, perform radiation inverse calculation to restore the true temperature field data of the single-view curtain wall.

[0026] In step S300 above, the effective thermal radiation component, the radiation characteristic distribution of the pixel-level curtain wall area, and the environmental radiation mapping relationship are uniformly listed as the calculation input; atmospheric compensation operation, reflection stripping operation, and blackbody temperature inverse conversion operation are sequentially performed on all input data to obtain the intermediate corrected image; atmospheric transmission parameters are collected by the airborne meteorological sensor equipment in the field, and the atmospheric compensation calculation formula can be expressed as: , where L s The surface radiance of the glass curtain wall after eliminating atmospheric path radiation (surface radiance); L c The measured radiance is obtained from the raw data acquired by the infrared camera; Atmospheric transmittance (atmospheric transmission parameter); L a The atmospheric path radiation radiance is given. Atmospheric compensation calculations are performed by combining atmospheric transmission parameters and effective thermal radiation components to eliminate atmospheric path radiation interference, resulting in the surface radiance after removing atmospheric path radiation. Reflection stripping calculations are then performed based on the surface radiance, the radiation characteristic distribution of the curtain wall area, and the environmental radiation mapping relationship, which can be expressed as: , where L BB This refers to the pure target blackbody radiance obtained after removing atmospheric and environmental reflection interference. Lr represents the long-wave infrared emissivity of the glass at the corresponding pixel location; Lr represents the ambient reflected radiance received by the curtain wall surface; by separating the multi-source ambient reflected radiation, the pure target blackbody radiance is obtained; then, based on Planck's inverse operation formula, the single-pixel temperature value is calculated pixel by pixel for the pure target blackbody radiance. The formula can be expressed as: ,in, This represents the actual thermodynamic temperature value of a single-pixel glass curtain wall. A pre-configured blackbody radiation constant; The infrared imaging device is pre-set with a long-wave infrared equivalent operating wavelength; all single-pixel temperature values ​​are integrated according to the image spatial arrangement rules to generate an intermediate corrected image; simultaneously, three effective reconstruction constraints can be set for synchronous determination: emissivity. >0.15, linear polarization degree DOP<0.35, shooting incident angle Pixels with a temperature <60° that do not meet the constraint are marked as distorted pixels; the intermediate corrected image is subjected to pixel filtering processing to remove distorted pixels, and the true temperature field data of the single-view curtain wall is obtained.

[0027] Step S400: Obtain multiple sets of single-view curtain wall real temperature field data from different shooting angles, and integrate them to generate a global smooth temperature field.

[0028] In step S400 above, single-view real temperature field data of the curtain wall corresponding to different shooting angles under the same curtain wall facade are collected; the interference level can be divided into different levels according to the Fresnel optical reflection variation law, and the Fresnel reflectance calculation formula can be expressed as: ,in, The angle of incidence is Fresnel reflectivity of the glass surface; The angle of incidence of the camera's light relative to the normal to the glass surface; The refractive index of the glass medium is 1.5, for example (suitable for the 8~12μm long-wave infrared band); effective equivalent weight of the viewing angle. The effective equivalent emissivity satisfies ; The fusion weights are the temperature field data corresponding to the current perspective. The effective equivalent emissivity of the glass at the corresponding incident angle; The measurement variance of the temperature field is reconstructed from this perspective, characterizing the noise level of the single-view data; The glass material has an inherent normal emissivity (emissivity in the normal direction and under perpendicular incidence). The glass refractive index is 1.5 for example, with an effective equivalent weight. For example, when the incident angle is greater than 60°, it is determined to be a high-interference viewpoint. The viewpoint data corresponding to the high-interference level is identified and removed. Only the low-interference effective viewpoint data is retained for weighted fusion calculation to generate a global smooth temperature field.

[0029] Step S500: Analyze and identify the global smooth temperature field according to the preset thermal defect judgment criteria, and output standardized curtain wall thermal defect detection results.

[0030] In step S500 above, differentiated screening methods adapted to different structural parts of the curtain wall are retrieved. Multi-dimensional judgment rules can be configured for four types of curtain wall-specific thermal defects: For sealant aging identification, an abnormality is marked if the temperature difference between the curtain wall edge area and the center of the panel exceeds 2.5℃; for insulated glass inflation failure detection, the temperature field power spectrum is extracted using a two-dimensional Fourier transform, and a failure is judged when the ratio of the average energy in the 0.1~0.50 / cm frequency band to the average low-frequency background energy exceeds 3.0; for structural thermal bridge identification, the temperature field is matched with the BIM keel position, and the temperature difference between the keel area and the glass panel surface is considered. A value greater than 4℃ is used to mark thermal bridges; coating degradation is based on pixel-level emissivity distribution maps, and a change in emissivity greater than 0.15 in the same type of glass area is used to mark degradation candidate areas; multi-dimensional feature analysis is performed on the global smooth temperature field region by region to extract information on all thermal anomaly areas; all thermal anomaly area information is summarized and standardized, and the output format of defect location, defect type, temperature difference amplitude, and risk level is unified to obtain standardized curtain wall thermal defect detection results; it can accurately identify weak thermal defects such as hollow air filling failure and sealing aging that cannot be distinguished by traditional infrared detection, and significantly improve the detection rate of hidden thermal defects in curtain walls.

[0031] Please refer to Figure 2 , Figure 2 The flowchart below shows a multimodal data acquisition and parsing method provided in an embodiment of the present invention; the multimodal data acquisition and parsing method includes: Step S20: Carry out external field scanning and acquisition of curtain wall data by equipping synchronous sensing devices, and acquire two types of image data: polarized infrared image and multispectral visible light image, as well as synchronously recorded spatial attitude information.

[0032] In step S20 above, a drone is used as the platform. The airborne synchronous sensing equipment includes a long-wave infrared polarization camera, a multispectral visible light camera, and an inertial navigation and positioning unit. The long-wave infrared polarization camera operates in the 8~12μm band and integrates a micro-polarization array structure, eliminating the need for a rotating polarizer. It can simultaneously acquire raw radiance images in four polarization directions: 0°, 45°, 90°, and 135°. The multispectral visible light camera covers the 400~1000nm band and has at least eight independent spectral acquisition channels. During the field inspection, the drone is controlled to scan the curtain wall facade at a uniform speed. The sensing equipment maintains synchronous acquisition with timestamps. While outputting two types of image data, it simultaneously records the spatial position, shooting posture, and other spatial attitude information corresponding to the acquisition time of each frame of image, realizing a one-to-one binding of image data and pose data to ensure subsequent pixel-level spatial matching calculations.

[0033] Step S21: Solve the polarization quantization index of the polarized infrared image to obtain the radiation source division results of each pixel.

[0034] In step S21 above, four sets of radiance images with different polarization angles are read and acquired, the Stokes polarization parameters are solved pixel by pixel, and the polarization quantification index linear polarization degree is calculated based on the three sets of Stokes parameters. From the perspective of physical radiation mechanism, the specular environment reflected radiation generated by the glass surface has significant polarization characteristics, and the polarization effect reaches its strongest near Brewster angle of 56°. The thermal radiation emitted by the glass itself is unpolarized radiation. Based on the above physical characteristics, the radiation source of the pixel signal can be distinguished according to the linear polarization degree value of each pixel, and the preliminary division of the radiation source of the whole image pixels is completed.

[0035] Step S22: Filter the division results to remove interfering pixels formed by the reflection of the mirror environment, obtain the pixel data of the thermal radiation of the glass curtain wall body, and output the effective thermal radiation component.

[0036] In step S22 above, the preset linear polarization degree screening threshold can be set to 0.35. The pixel radiation source classification result is judged pixel by pixel: if the pixel satisfies DOP<0.35, it is determined that the pixel signal is mainly due to the thermal radiation of the glass curtain wall body and is retained; if the pixel satisfies DOP≥0.35, it is determined that the pixel signal is mainly due to the reflection radiation of the mirror environment, and is marked as an interference pixel and filtered out; after the filtering is completed, the radiation data corresponding to all retained pixels are summarized to form a pure glass curtain wall body thermal radiation pixel dataset, which is output as the effective thermal radiation component; by completing the separation of emission component and reflection component through polarization feature, the interference of mirror reflection caused by daytime sunlight and surrounding buildings can be greatly reduced, and effective daytime detection of curtain wall within a suitable viewing angle range on sunny days can be performed, providing a low-noise input data source for subsequent radiation characteristic modeling and temperature field inverse calculation.

[0037] Please refer to Figure 3 , Figure 3 This is a flowchart of a radiation map construction method provided in an embodiment of the present invention; the radiation map construction method includes: Step S30: Extract the reflection feature data inside the multispectral visible light image; retrieve the pre-stored material radiation correlation model to perform matching calculations on the reflection feature data, and generate the pixel-level curtain wall area radiation characteristic distribution.

[0038] In step S30 above, the UAV is simultaneously equipped with a multispectral visible light camera covering no less than 8 spectral bands and 400~1000nm wavelength to complete image acquisition; the reflectance feature data of the 700nm and 850nm bands of the multispectral image are extracted as input; according to Kirchhoff's radiation law, the near-infrared 850nm reflectance and long-wave infrared emissivity of Low-E coated glass have a strong negative correlation, with an example correlation coefficient of 0.89; a material radiation correlation mapping model covering six types of glass, including ordinary float glass, single silver Low-E glass, double silver Low-E glass, heat reflective glass, body-colored glass, and glaze-printed glass, is pre-established; the extracted pixel reflectance features are matched and calculated pixel by pixel with the pre-stored material radiation correlation model to output the pixel-level radiation characteristic distribution map of the entire curtain wall; if the estimated emissivity value of the same type of glass in the same curtain wall area changes abruptly by more than 0.15, the area is simultaneously marked as a candidate area for coating degradation; the overall emissivity estimation error can be controlled within 0.05, solving the temperature measurement error problem caused by the unified emissivity modeling of multiple types of glass in the curtain wall.

[0039] Step S31: Based on the spatial attitude information, extrapolate the sky, sunlight, and building radiation components to obtain various environmental components and shooting location information.

[0040] In step S31 above, the spatial attitude information synchronously output by the UAV's onboard IMU / GPS is read, and combined with the temperature, humidity, visibility, and aerosol type data collected by the meteorological sensor, the MODTRAN atmospheric radiative transfer model is called to calculate the diffuse radiation of the sky; the NRELSPA algorithm is used to input the solar azimuth and elevation angles at the time of shooting to solve for the direct solar radiation component; the surrounding building materials are matched based on the 360° panoramic visible light image of the UAV, and the radiation components of the surrounding buildings are estimated by looking up the table based on the material's equivalent temperature; the direct solar radiation is only included in the reflection component when the solar azimuth falls within the ±3° solid angle range of the glass mirror reflection; the specular reflection solid angle corresponding to each pixel is solved by combining the glass surface normal vector corresponding to the camera's shooting attitude, and the shooting azimuth information corresponding to each pixel is output synchronously, thus fully quantifying the contributions of the three types of environmental radiation in densely populated urban areas: sky, sunlight, and surrounding buildings.

[0041] Step S32: Correlate the distribution of radiation characteristics of the pixel-level curtain wall area, the environmental radiation component, and the shooting orientation information to obtain the environmental radiation mapping relationship.

[0042] In step S32 above, the radiation characteristic distribution of the curtain wall area at the pixel level, the three types of environmental radiation components (sky, sunlight, and building) and the shooting orientation information corresponding to the pixel are coupled and correlated on a single pixel basis. The reflection radiance is calculated pixel by pixel based on the glass Fresnel specular reflection model to construct a complete pixel-level environmental reflection radiation mapping relationship. Unlike the traditional simplified modeling method that only uses a single sky temperature, this method completes multi-source radiation field modeling by integrating three types of environmental radiation sources. This helps to reduce the environmental reflection correction error in densely populated urban areas and provides accurate environmental radiation input parameters for subsequent radiation inverse calculation and reflection interference removal.

[0043] Please refer to Figure 4 , Figure 4 This is a flowchart of a radiation mapping substitution construction method provided in an embodiment of the present invention; the radiation mapping substitution construction method includes: Step S40: Pre-store radiation correlation models for multiple types of glass materials.

[0044] In step S40 above, an offline database of near-infrared reflectance-long-wave infrared emissivity mapping for six types of architectural glass is established and stored as a radiation correlation model. The reflectance ranges of 700nm and 850nm bands and the corresponding long-wave infrared emissivity benchmark values ​​for each type of glass are recorded. Simultaneously, the material library of the curtain wall BIM building information model is connected to store the preset glass material parameters for each section of the curtain wall for alternative estimation under the condition of multispectral image acquisition failure.

[0045] Step S41: Retrieve the preset material parameters of the curtain wall and perform matching calculations with the radiation correlation model to obtain the radiation characteristic distribution of the curtain wall area at the pixel level.

[0046] In step S41 above, when multispectral visible light image acquisition fails and reflection feature data cannot be extracted, the preset glass material parameters in the curtain wall BIM information database are retrieved, and the material parameters are matched and calculated with the pre-stored radiation correlation models of multiple types of glass materials. The pixel-level curtain wall area radiation characteristic distribution map is generated by assigning values ​​to each grid. This alternative construction method can continuously output emissivity distribution data under conditions such as multispectral acquisition failure, occlusion, and strong light overexposure, ensuring that the entire detection process is not interrupted.

[0047] Step S42: Based on the spatial attitude information, extrapolate the radiation components of the sky, sunlight, and buildings to obtain various environmental radiation components and shooting orientation information.

[0048] In step S42 above, the spatial attitude information synchronously recorded by the airborne machine is read, and the three types of environmental radiation components of sky, sunshine and surrounding buildings are calculated by using the MODTRAN atmospheric model, NREL solar azimuth algorithm and panoramic visible light map respectively. The shooting azimuth information corresponding to each pixel is solved synchronously to fully quantify the contribution of multi-source reflection radiation of urban environment.

[0049] Step S43: Correlate the distribution of radiation characteristics of the pixel-level curtain wall area, various environmental radiation components, and shooting orientation information to obtain the environmental radiation mapping relationship.

[0050] In step S43 above, the pixel-level radiation characteristic distribution, three types of environmental radiation components, and shooting orientation information generated by replacing material parameters are coupled and correlated pixel by pixel. The specular reflection radiance corresponding to each pixel is calculated, and a complete environmental radiation mapping relationship is constructed. This ensures that the reflection radiation modeling results adapted to the actual working conditions of the curtain wall can still be output even in the case of multispectral acquisition failure, thereby improving the operational stability and environmental adaptability of the entire detection system.

[0051] Please refer to Figure 5 , Figure 5 This is a flowchart of a temperature field restoration method provided in an embodiment of the present invention; the temperature field restoration method includes: Step S50: The effective thermal radiation component, the radiation characteristic distribution of the curtain wall area, and the environmental radiation mapping relationship are uniformly listed as calculation inputs.

[0052] In step S50 above, the effective thermal radiation component, the radiation characteristic distribution of the pixel-level curtain wall area, and the environmental radiation mapping relationship are read. The three sets of data are uniformly registered to the same pixel spatial coordinate system to remove the data mismatch problem caused by temporal misalignment and pixel offset. They are integrated into a complete set of radiation inverse calculation input datasets, and the atmospheric transmission parameters collected by the airborne meteorological sensor are synchronously brought in as auxiliary inputs.

[0053] Step S51: List the effective thermal radiation components, the radiation characteristic distribution of the curtain wall area, and the environmental radiation mapping relationship as unified calculation inputs.

[0054] In step S51 above, atmospheric compensation is first performed based on atmospheric transmission parameters to eliminate atmospheric path radiation interference. Then, reflection stripping is performed based on pixel-level emissivity and environmental radiation mapping. Finally, the thermodynamic temperature is solved pixel by pixel using Planck's inverse formula. All pixel temperature values ​​are integrated to generate an unconstrained intermediate correction image, which eliminates atmospheric attenuation, material emissivity inhomogeneity, and multiple radiation interferences from multi-source environmental reflections layer by layer.

[0055] Step S52: Perform pixel filtering on the intermediate corrected image to remove distorted pixels and obtain the true temperature field data of the single-view curtain wall.

[0056] In step S52 above, a triple effective reconstruction constraint is set to synchronously verify each pixel: emissivity. >0.15, linear polarization degree DOP<0.35, shooting incident angle <60°; if any constraint is not met, the pixel is determined to be a distorted pixel and is filtered out; after filtering out distorted pixels, the effective pixel temperature data is retained to form complete single-view curtain wall real temperature field data.

[0057] Furthermore, it is possible to address the issue of emissivity that satisfies the aforementioned triple constraints. Pixel regions within the range of 0.15 to 0.25 (inclusive) are separately marked as low-confidence temperature zones; reflection stripping operations within this range... The denominator is small, making environmental reflection estimation errors easily amplified, resulting in lower temperature inversion reliability compared to normal emissivity regions. For this region, a confidence weight label is added to the temperature field output data. Note that the temperature value in this region is not used as an independent basis for subsequent defect judgment; instead, its fusion weight is reduced during the multi-view temperature field fusion stage, or a polarization enhancement correction process is triggered. By fusing multiple sets of radiance data with different polarization angles, the accuracy of reflection component estimation is improved, thereby enhancing the reliability of temperature inversion in low emissivity regions. Pixels with emissivity below 0.15 are directly discarded as distorted pixels and do not enter the low-confidence processing flow.

[0058] Please refer to Figure 6 , Figure 6 The flowchart below shows a stepwise radiation signal correction method provided in an embodiment of the present invention; the stepwise radiation signal correction method includes: Step S60: Based on the collected atmospheric transmission parameters, perform atmospheric compensation calculations in conjunction with the effective thermal radiation components to obtain the surface radiance after removing atmospheric path radiation.

[0059] In step S60 above, air temperature, humidity, visibility and aerosol type data are collected by airborne meteorological sensors. The atmospheric transmission parameters at the corresponding flight altitude can be solved by the MODTRAN / MODTRAN atmospheric radiative transfer model to eliminate measurement errors caused by atmospheric medium self-radiation and infrared radiation air attenuation, and output the surface radiance after eliminating atmospheric path radiation interference.

[0060] Step S61: Perform reflection stripping calculation based on surface radiance, radiation characteristic distribution of the curtain wall area, and environmental radiation mapping relationship to obtain the pure target blackbody radiance.

[0061] In step S61 above, the pixel-level emissivity distribution map and pixel-by-pixel reflectance are retrieved, and the environmental reflected radiation component is separated using the reflectance stripping formula; a lower limit constraint is applied to the emissivity value. ≥0.15, to avoid distortion in pixel calculations with extremely low emissivity, and to obtain the pure target blackbody radiance dominated only by the true temperature of the glass.

[0062] Step S62: Based on the following formula, perform inverse blackbody temperature conversion on the pure target blackbody radiance to obtain the single pixel temperature value: in, This represents the actual thermodynamic temperature value of a single-pixel glass curtain wall. A pre-configured blackbody radiation constant; Preset the long-wave infrared equivalent working wavelength for infrared imaging equipment; This represents the radiance of the pure target blackbody obtained after removing atmospheric and environmental reflections.

[0063] In step S62 above, the pure target blackbody radiance is substituted pixel by pixel into the Planck inverse function to obtain the thermodynamic temperature value corresponding to each pixel; at the same time, reasonable upper and lower limits can be set for the output temperature value to eliminate abnormal temperature values ​​that exceed the physical range of 200K~350K, so as to prevent radiation calculation distortion from producing extreme temperatures without physical meaning.

[0064] Step S63: Integrate all single-pixel temperature values ​​according to the image space arrangement rules to generate an intermediate correction image.

[0065] In step S63 above, the real temperature values ​​obtained by solving pixel by pixel are filled one by one according to the row and column space coordinates of the original polarized infrared image, the linear background gradient interference is removed, and an intermediate corrected image of the complete two-dimensional pixel temperature distribution is generated, which serves as the basic temperature data for subsequent distorted pixel screening and multi-view fusion.

[0066] Please refer to Figure 7 , Figure 7 This is a flowchart of a multi-view temperature field fusion method provided in an embodiment of the present invention; the multi-view temperature field fusion method includes: Step S70: Collect single-view real temperature field data of the curtain wall from different shooting angles on the same facade.

[0067] In step S70 above, the UAV is controlled to complete multi-view scanning and acquisition along multiple parallel routes on the curtain wall facade. Each route corresponds to a set of independent polarized infrared and multispectral imaging data. After temperature reconstruction, multiple single-view real temperature fields covering the same curtain wall facade with different shooting incident angles are output. During the flight process, the incident angle of the camera relative to the glass surface corresponding to each frame image is recorded synchronously and stored in conjunction with the temperature field pixels to provide the incident angle basis for subsequent interference classification.

[0068] Step S71: Based on the optical reflection variation law, the interference level of the single-view curtain wall real temperature field data from each viewpoint is divided to obtain the fusion weight of each viewpoint.

[0069] In step S71 above, the viewing angle interference levels are divided according to the Fresnel optical reflection variation law; the weights are assigned according to the shooting incident angle, and the larger the incident angle and the stronger the Fresnel reflection, the lower the weight is assigned, thus quantifying the differences in data quality from different viewing angles.

[0070] Step S72: Identify the viewpoint data corresponding to the high interference level and complete the removal process. Perform weighted calculation on the remaining viewpoint data to obtain the global smooth temperature field.

[0071] In step S72 above, the preset incident angle threshold is 60°. When the shooting incident angle is greater than 60°, it is determined to be a high interference level, and all temperature field data of this group of viewpoints are directly removed. Only the effective viewpoint data with low incident angle and low Fresnel reflection interference are retained. Pixel-level weighted average fusion calculation is carried out with the matching weight of each viewpoint to eliminate the temperature change caused by single-view specular reflection noise and generate a global smooth temperature field.

[0072] Please refer to Figure 8 , Figure 8 This is a flowchart of a multi-view temperature field fusion method provided in an embodiment of the present invention; the multi-view temperature field fusion method includes: Step S80: Retrieve the differentiated screening method for different structural parts of the curtain wall.

[0073] In step S80 above, differentiated defect screening rules are pre-configured for different types of curtain wall components, covering four types of curtain wall-specific defects: sealant aging identification rules; insulated glass inflation failure convection ripple detection rules; structural thermal bridge identification rules; coating degradation identification rules; and simultaneously accessing the curtain wall BIM model to read the coordinates of the segmented sealant and metal keel positions, achieving precise matching of component positions and temperature field pixels, and adapting to the differentiated thermal feature identification needs of different structural parts.

[0074] Step S81: Perform multi-dimensional feature analysis on the global smooth temperature field region by region to obtain information on all thermal anomaly regions.

[0075] In step S81 above, multi-dimensional feature analysis is performed on the grid-by-grid area of ​​the global smooth temperature field: the aging criterion for sealant is that the temperature difference between the sealing area at the edge of the curtain wall and the center of the panel is greater than 2.5°C; for hollow air filling failure, the power spectrum of the temperature field is solved by two-dimensional Fourier transform, and failure is marked if the ratio of the average energy in the 0.1~0.5 / cm spatial frequency band to the low-frequency background energy is greater than 3.0; structural thermal bridges are matched with BIM keel coordinates, and thermal bridges are judged if the temperature difference between the keel area and the glass panel is greater than 4°C; abrupt changes in the emissivity distribution map of the same glass area greater than 0.15 are marked as coating degradation; the results of all four types of defect identification are integrated, and the information of all thermal anomaly areas including defect location, defect type, temperature difference amplitude, and risk level is output, which can detect small-amplitude temperature difference air filling failure defects that cannot be identified by traditional infrared methods.

[0076] Step S82: Summarize and organize all thermal anomaly area information to obtain standardized curtain wall thermal defect detection results.

[0077] In step S82 above, the output format of all thermal anomaly areas is standardized, including the plane coordinates of the defect, the floor to which it belongs, the defect type, the temperature difference range, the risk level, and the repair suggestions; the multi-type defect areas are summarized and statistically analyzed in layers to generate a standardized curtain wall thermal defect detection result document, which supports the use of curtain wall overhaul and energy-saving renovation decisions for commercial complexes and high-rise office buildings.

[0078] Please refer to Figure 9 , Figure 9 This is a flowchart of a multi-view temperature field fusion method provided in an embodiment of the present invention; the multi-view temperature field fusion method includes: Step S90: Perform statistical calculations on the temperature of all pixels within the global smooth temperature field to obtain relevant indicators for curtain wall thermal insulation.

[0079] In step S90 above, the temperature of all pixels in the global smooth temperature field is statistically analyzed by region. The heat transfer coefficient U value of the curtain wall is calculated in batches by combining the emissivity distribution map of each glass pixel, which serves as the core indicator of energy saving and heat insulation of the curtain wall. The difference between the U value and the design value in different regions is statistically analyzed to quantify the degree of heat insulation performance degradation caused by coating degradation and air filling failure, and the heat insulation performance distribution map of the entire curtain wall pixel is output.

[0080] Step S91: Bind and archive the thermal insulation index and thermal anomaly area information to obtain the comprehensive evaluation file of the curtain wall.

[0081] In step S91 above, the global temperature field distribution map, the distribution maps of the four types of thermal anomalies and defects, the U-value index of the curtain wall zoning insulation, the flight acquisition parameters, the original imaging data, and the defect identification confidence level are all bound, stored and archived to form a complete and traceable comprehensive assessment file for the long-term energy-saving operation and maintenance of the curtain wall; a single drone flight can simultaneously output defect reports and curtain wall energy-saving performance assessment drawings, providing complete data support for the periodic inspection, major renovation and reconstruction, and energy consumption optimization of building curtain walls.

[0082] Please refer to Figure 10 , Figure 10This is a schematic diagram of the thermal defect detection system for glass curtain walls provided in an embodiment of the present invention. The system includes a data acquisition and analysis module 10, a radiation mapping construction module 20, a temperature field restoration module 30, a temperature field fusion module 40, and a defect identification output module 50. The data acquisition and analysis module 10 is used to acquire multimodal sensing data and spatial attitude information of the curtain wall, including polarized infrared images and multispectral visible light images, and to perform polarization feature analysis on the polarized infrared images to obtain the effective thermal radiation component. The radiation mapping construction module 20 is used to generate the radiation characteristic distribution of the curtain wall area based on the multispectral visible light images, and to construct an environmental radiation mapping relationship by combining the environmental radiation component obtained from the spatial attitude information. The temperature field restoration module 30 is used to perform radiation inverse calculation using the effective thermal radiation component, the radiation characteristic distribution of the curtain wall area, and the environmental radiation mapping relationship as joint inputs to restore the true temperature field data of the curtain wall from a single perspective. The temperature field fusion module 40 is used to acquire multiple sets of true temperature field data of the curtain wall from different shooting angles and integrate them to generate a globally smooth temperature field. The defect identification output module 50 is used to analyze and identify the globally smooth temperature field according to a preset thermal defect judgment standard and output standardized curtain wall thermal defect detection results.

[0083] The above description is merely an embodiment of the present invention and is not intended to limit the scope of protection of the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for detecting thermal defects in glass curtain walls, characterized in that, include: Multimodal sensing data and spatial attitude information of the curtain wall, including polarized infrared images and multispectral visible light images, are acquired, and polarization feature analysis is performed on the polarized infrared images to obtain the effective thermal radiation components. The radiation characteristic distribution of the curtain wall area is generated based on the multispectral visible light image, and the environmental radiation components are deduced by combining the spatial attitude information to construct the environmental radiation mapping relationship. Using the effective thermal radiation component, the radiation characteristic distribution of the curtain wall area, and the environmental radiation mapping relationship as joint inputs, radiation inverse calculation is performed to restore the true temperature field data of the curtain wall from a single perspective. Acquire multiple sets of single-view curtain wall real temperature field data from different shooting angles, and integrate them to generate a global smooth temperature field; The global smooth temperature field is analyzed and identified according to the preset thermal defect judgment criteria, and standardized curtain wall thermal defect detection results are output.

2. The method according to claim 1, characterized in that, The acquisition includes multimodal sensing data and spatial attitude information of the curtain wall, comprising polarized infrared images and multispectral visible light images. Polarization feature analysis is performed on the polarized infrared images to obtain the effective thermal radiation components, including: Equipped with synchronous sensing devices, the curtain wall external field scanning and acquisition operation is carried out to obtain two types of image data: polarized infrared images and multispectral visible light images, as well as synchronously recorded spatial attitude information. The polarization quantization index of the polarized infrared image is solved to obtain the radiation source classification results of each pixel; The division results are filtered to remove interfering pixels caused by reflections from the mirror environment, and the pixel data of the thermal radiation of the glass curtain wall body is obtained, and the effective thermal radiation component is output.

3. The method according to claim 1, characterized in that, The process of generating the radiation characteristic distribution of the curtain wall area based on the multispectral visible light image, and constructing an environmental radiation mapping relationship by combining the environmental radiation components derived from spatial attitude information, includes: Extract reflection feature data from the multispectral visible light image; retrieve a pre-stored material radiation correlation model to perform matching operations on the reflection feature data, and generate a pixel-level distribution of radiation characteristics of the curtain wall area; Based on the spatial attitude information, the components of sky, sunlight, and building radiation were extrapolated to obtain various environmental components and shooting location information; The radiation characteristic distribution, environmental radiation component, and shooting orientation information of the pixel-level curtain wall area are correlated to obtain the environmental radiation mapping relationship.

4. The method according to claim 3, characterized in that, The obtained environmental radiation mapping relationship also includes: Pre-store radiation correlation models for multiple types of glass materials; The preset material parameters of the curtain wall are retrieved and matched with the radiation correlation model to obtain the radiation characteristic distribution of the curtain wall area at the pixel level. Based on the spatial attitude information, the radiation components of the sky, sunlight, and buildings are deduced to obtain various environmental radiation components and shooting orientation information; The radiation characteristic distribution of the pixel-level curtain wall area, various environmental radiation components, and the shooting orientation information are correlated to obtain the environmental radiation mapping relationship.

5. The method according to claim 1, characterized in that, The process of performing radiation inverse calculation using the effective thermal radiation components, the radiation characteristic distribution of the curtain wall area, and the environmental radiation mapping relationship as joint inputs to reconstruct the true temperature field data of the curtain wall from a single viewpoint includes: The effective thermal radiation components, the radiation characteristic distribution of the curtain wall area, and the environmental radiation mapping relationship are uniformly listed as calculation inputs; Atmospheric compensation, reflection stripping, and blackbody temperature inverse conversion are performed sequentially on all input data to obtain intermediate corrected images. The intermediate corrected image is subjected to pixel filtering to remove distorted pixels, thereby obtaining the true temperature field data of the curtain wall from a single viewpoint.

6. The method according to claim 5, characterized in that, The sequential execution of atmospheric compensation, reflection stripping, and blackbody temperature inverse conversion operations on all input data includes: Based on the collected atmospheric transmission parameters, atmospheric compensation calculations are performed using the effective thermal radiation components to obtain the surface radiance after removing atmospheric path radiation. Based on the surface radiance, the distribution of radiation characteristics of the curtain wall area, and the environmental radiation mapping relationship, a reflection stripping calculation is performed to obtain the pure target blackbody radiance. The blackbody temperature is inversely converted from the pure target blackbody radiance based on the following formula to obtain the single-pixel temperature value: in, This represents the actual thermodynamic temperature value of a single-pixel glass curtain wall. A pre-configured blackbody radiation constant; Preset the long-wave infrared equivalent working wavelength for infrared imaging equipment; The radiance of the pure target blackbody obtained after removing atmospheric and environmental reflections; All the single-pixel temperature values ​​are integrated according to the image space arrangement rules to generate an intermediate correction image.

7. The method according to claim 1, characterized in that, The process of acquiring multiple sets of single-view curtain wall real temperature field data from different shooting angles and integrating them to generate a globally smooth temperature field includes: Collect single-view real temperature field data of the curtain wall from different shooting angles on the same facade; Based on the optical reflection variation law, the interference level of the single-view curtain wall real temperature field data from each viewpoint is divided to obtain the fusion weight of each viewpoint. The viewpoint data corresponding to the high interference level is identified and removed. The remaining viewpoint data is then weighted to obtain a globally smooth temperature field.

8. The method according to claim 1, characterized in that, The step of analyzing and identifying the global smooth temperature field according to a preset thermal defect judgment standard, and outputting standardized curtain wall thermal defect detection results, includes: Retrieve differentiated screening methods suitable for different structural parts of the curtain wall; Multi-dimensional feature analysis is performed on the global smooth temperature field region by region to obtain information on all thermal anomaly regions. Information on all thermal anomaly areas is summarized and organized to obtain standardized results for curtain wall thermal defect detection.

9. The method according to claim 8, characterized in that, After outputting the standardized curtain wall thermal defect detection results, the following is also included: Statistical calculations are performed on the temperature of all pixels within the global smooth temperature field to obtain relevant indicators for curtain wall thermal insulation. The thermal insulation index and thermal anomaly area information are bound and archived to obtain a comprehensive evaluation file for the curtain wall.

10. A thermal defect detection system for glass curtain walls, characterized in that, include: The data acquisition and analysis module is used to acquire multimodal sensing data and spatial attitude information of the curtain wall, including polarized infrared images and multispectral visible light images, and to perform polarization feature analysis on the polarized infrared images to obtain the effective thermal radiation components. The radiation mapping construction module is used to generate the radiation characteristic distribution of the curtain wall area based on the multispectral visible light image, and construct the environmental radiation mapping relationship by combining the environmental radiation components obtained by spatial attitude information. The temperature field restoration module is used to perform radiation inverse calculations using the effective thermal radiation components, the radiation characteristic distribution of the curtain wall area, and the environmental radiation mapping relationship as joint inputs to restore the real temperature field data of the curtain wall from a single perspective. The temperature field fusion module is used to acquire multiple sets of real temperature field data of the single-view curtain wall from different shooting angles and integrate them to generate a global smooth temperature field. The defect identification output module is used to analyze and identify the global smooth temperature field according to the preset thermal defect judgment criteria, and output standardized curtain wall thermal defect detection results.

Images