A method for diagnosing hollowing of building outer wall based on infrared thermal imaging

Abstract

This invention relates to the field of non-destructive testing of building structures, specifically to a method for diagnosing hollow areas in building exterior walls based on infrared thermal imaging. The method includes dividing the target building exterior wall into independent wall units and collecting thermal property parameters; using a drone to perform pre-sampling with a very limited time sequence and extracting suspected hollow areas; establishing a one-dimensional unsteady-state heat conduction model of the wall and inverting it to obtain accurate thermal property parameters of the wall surface; defining a hollow detection index (DI); adaptively predicting the optimal acquisition window for each wall unit based on weather forecast data; planning the drone inspection route based on the optimal window and completing image acquisition; and finally achieving accurate identification and quantitative diagnosis of hollow defects. This invention solves the problems of invalid data and high rework rates caused by relying on manual timing in existing technologies. It eliminates the need for continuous, intensive data collection, is compatible with large-scale drone inspections, and significantly improves the data validity and diagnostic accuracy of exterior wall hollow detection.

Description

Technical Field

[0001] This invention belongs to the field of non-destructive testing of building structures, specifically relating to a method for diagnosing hollow areas in building exterior walls based on infrared thermal imaging. Background Technology

[0002] Hollow areas in building exterior walls are defects caused by the separation of the exterior wall finish layer, insulation layer, and base wall, resulting in air layers. This is a core cause of exterior wall detachment and falling objects from heights, and is a mandatory inspection item in building project acceptance and existing building safety inspections.

[0003] Existing technologies for detecting hollow exterior walls are divided into two categories: contact and non-contact. Contact detection methods, such as manual tapping, ultrasonic rebound, and pull-out testing, have drawbacks including low detection efficiency, high risk of working at heights, reliance on human experience for detection results, and inability to achieve full coverage of building exterior walls. They can only be used for sampling inspections and cannot meet the large-scale inspection needs of super high-rise buildings and large residential communities.

[0004] Non-contact inspection, centered on infrared thermal imaging, utilizes the difference in thermal conductivity between the air gap and building materials such as concrete, tiles, and insulation boards. Under thermal excitation, a detectable temperature difference forms between the air gap area and the normally bonded area on the wall surface. By acquiring the infrared thermal radiation signal of the wall surface with an infrared thermal imager and converting it into a surface temperature field image, the location and extent of the air gap defect can be identified. This method does not require contact with the wall surface and can be carried out by drones to achieve rapid, full-coverage scanning of exterior walls from high altitudes. It is currently the mainstream technical solution for large-scale inspection of air gaps in building exterior walls.

[0005] Existing passive infrared thermal imaging detection technology has the following core defects:

[0006] First, the infrared detectability of hollow areas is not an inherent property of the defect, but rather a transient result that dynamically changes with the time of data collection, the angle of the sun, the ambient temperature, and the heat exchange conditions of the wall. Existing inspection solutions rely on manual experience to select a uniform data collection time, which cannot adapt to the thermal response characteristics of walls with different orientations, materials, and floors. This results in a data collection efficiency of only 30%-50%, a high rework rate, and poor adaptability to large-scale projects.

[0007] Second, existing hollow drum detection algorithms have obvious limitations. One type of algorithm relies on dense time-series data acquisition throughout the entire period, which cannot be adapted to the endurance of drones and the needs of engineering implementation. Another type of algorithm only performs static processing on a single frame of infrared image, which cannot eliminate the interference of dynamic environmental factors, resulting in a significant increase in false detection rate and false negative rate under complex on-site conditions.

[0008] Third, existing technologies can only perform post-process defect identification on acquired infrared images. They cannot quantitatively assess the detectability of voids in advance, adaptively predict the optimal acquisition window, or guide the planning of UAV inspection routes and acquisition timing. The inspection process is highly blind and cannot solve the core problem of invalid data from the source. Summary of the Invention

[0009] The purpose of this invention is to provide a method for diagnosing hollow areas in building exterior walls based on infrared thermal imaging, so as to solve the problems mentioned in the background art.

[0010] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0011] A method for diagnosing hollow areas in building exterior walls based on infrared thermal imaging includes the following steps:

[0012] S1 divides the exterior wall of the target building into several independent wall units and collects the thermal properties of each wall unit.

[0013] S2 collects on-site environmental parameters and controls the drone to complete a full-coverage scan of each wall unit to obtain time-series infrared thermal images.

[0014] S3 preprocesses the time-series infrared thermal image to obtain the time-series surface temperature field of the wall and extracts areas with abnormal temperature change rates as suspected hollow areas.

[0015] S4 establishes a one-dimensional unsteady-state heat conduction model of the wall, and uses the time-series surface temperature field as the observation value to invert and obtain the accurate heat conduction model of each wall unit;

[0016] S5 defines the hollow detection index DI, quantifies the hollow detection capability, and combines weather forecast data to predict the optimal acquisition time window for each wall unit;

[0017] The S6 plans the drone inspection route based on the optimal acquisition time window to complete the acquisition of high-resolution infrared images of the exterior wall and the diagnosis of hollow defects.

[0018] Furthermore, in S1, the facade model is the BIM model of the target building, construction drawings, or a three-dimensional point cloud model and facade texture map generated by oblique photography of UAV.

[0019] Furthermore, in S1, the wall unit is divided into three dimensions: wall orientation, finishing layer material, and floor section, and each wall unit is configured with a unique identification code; in S1, the thermal property parameters include the material, thickness, density, specific heat capacity, thermal conductivity, and solar absorptivity of the finishing layer, insulation layer, and base wall.

[0020] Furthermore, the on-site environmental parameters include ambient air temperature, wind speed, total solar radiation intensity, solar altitude angle, and solar azimuth angle, with a sampling frequency of no less than 1 time per minute.

[0021] Furthermore, in S3, the preprocessing sequentially includes non-uniformity correction, bad pixel removal, image registration and radiometric calibration, converting the infrared image grayscale value into the wall surface temperature value.

[0022] Furthermore, in step S3, the temperature change rate of each pixel is calculated by time-series temperature difference. Areas where the difference between the temperature change rate and the overall mean of the wall exceeds 2 times the standard deviation are marked as temperature abnormal areas. After removing small noise areas through connected component analysis, suspected hollow areas are obtained.

[0023] Furthermore, in S4, the one-dimensional unsteady-state heat conduction model of the wall is divided into a finishing layer, a hollow air layer, an insulation layer, and a base wall along the thickness direction, and its general governing equation is: ; In the formula, x is the coordinate along the thickness of the wall, ρ(x), c(x), and λ(x) are the density, specific heat capacity, and thermal conductivity of the material at position x, respectively, and T(x,t) is the temperature value at position x at time t.

[0024] Furthermore, in step S4, a genetic algorithm-Levenberger-Marquardt hybrid optimization algorithm is used for parameter inversion. The inversion parameters include the actual thermal conductivity of the finishing layer, the actual thermal conductivity of the insulation layer, the thickness of the air gap layer, and the equivalent thermal conductivity.

[0025] Furthermore, according to the method of claim 1, the formula for calculating the detectability index DI of the void in step S5 is: ; In the formula, T def T represents the surface temperature of the hollow area. norm NETD represents the surface temperature of adjacent normal areas and the noise equivalent temperature difference of the infrared thermal imager. When DI≥3, the hollow area in the corresponding time period is determined to be stable and detectable.

[0026] This application also discloses an electronic device, including:

[0027] At least one processor; and

[0028] A memory communicatively connected to the at least one processor; wherein,

[0029] The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the above-described method for diagnosing hollow building exterior walls based on infrared thermal imaging according to the present invention.

[0030] Beneficial effects: This invention achieves adaptive and personalized prediction of the optimal collection time window for different wall units through the technical path of "extremely limited presampling-model inversion-window prediction". It gets rid of the dependence on human experience, can effectively improve the data of UAV inspection, avoid the problem of rework and reflying, shorten the inspection cycle of large-scale projects, and reduce inspection costs.

[0031] This invention can complete the construction of a wall heat conduction model and parameter inversion through extremely limited time-series pre-sampling, without the need for continuous and intensive data collection on the wall. It is fully compatible with the endurance of drones and the engineering requirements of large-scale inspections, and can be applied in multi-building and large-area inspection projects such as high-rise buildings and large residential communities.

[0032] This invention defines a quantifiable hollowness detectability index (DI). By inverting the parameters of a one-dimensional unsteady-state heat conduction model, the interference of dynamic environmental factors is effectively eliminated, and the intrinsic thermophysical parameters of hollowness defects are obtained. This solves the problem that existing technologies cannot assess the detectability of hollowness in advance, and provides a scientific and quantitative basis for predicting the optimal acquisition window.

[0033] This invention enables the optimal acquisition window prediction results to guide the planning of UAV inspection routes and acquisition timing. At the same time, it sets up a real-time dynamic correction mechanism to cope with sudden weather changes, forming a fully closed-loop automated inspection process of "pre-sampling-model inversion-window prediction-route planning-precise acquisition-defect diagnosis", which greatly improves the intelligence level of large-scale UAV exterior wall inspection.

[0034] In the image processing stage, this invention combines a pre-constructed precise heat conduction model of the wall surface to compensate for environmental interference in infrared images, effectively eliminating temperature field distortions caused by differences in wall corners, direct sunlight angles, and wind speed changes. Combined with an optimized deep learning model, the accuracy of hollow area identification can be improved, and the thickness of hollow areas can be quantitatively inverted, providing accurate data support for risk level assessment of external wall defects. Attached Figure Description

[0035] Figure 1 This is a flowchart of a method for diagnosing hollow areas in building exterior walls based on infrared thermal imaging, according to the present invention.

[0036] Figure 2 is a flowchart of the building exterior wall unit division and thermal property parameter acquisition in an embodiment of the present invention;

[0037] Figure 3 is a flowchart of the on-site environment acquisition and UAV time-series infrared pre-sampling in an embodiment of the present invention;

[0038] Figure 4 is a flowchart of the pre-sampling infrared image preprocessing and initial extraction of suspected hollow areas in an embodiment of the present invention;

[0039] Figure 5 is a flowchart of the construction of the wall heat conduction model and the inversion of thermal property parameters in an embodiment of the present invention;

[0040] Figure 6 is a flowchart of the detectionability assessment and optimal acquisition window prediction of the hollow drum in an embodiment of the present invention;

[0041] Figure 7 is a flowchart of the drone inspection planning and accurate diagnosis of hollow exterior walls in an embodiment of the present invention;

[0042] Figure 8 is a time series curve of the inversion accuracy verification of the one-dimensional unsteady heat conduction model of the wall in an embodiment of the present invention;

[0043] Figure 9 shows the time-series variation of DI value in a typical hollow area and the coordinate diagram of the optimal acquisition window verification in an embodiment of the present invention. Detailed Implementation

[0044] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0045] This invention provides a method for diagnosing hollow areas in building exterior walls based on infrared thermal imaging, such as... Figure 1 As shown, the steps include:

[0046] S1 divides the exterior wall of the target building into several independent wall units and collects the thermal properties of each wall unit.

[0047] S2 collects on-site environmental parameters and controls the drone to complete a full-coverage scan of each wall unit to obtain time-series infrared thermal images.

[0048] S3 preprocesses the time-series infrared thermal image to obtain the time-series surface temperature field of the wall and extracts areas with abnormal temperature change rates as suspected hollow areas.

[0049] S4 establishes a one-dimensional unsteady-state heat conduction model of the wall, and uses the time-series surface temperature field as the observation value to invert and obtain the accurate heat conduction model of each wall unit;

[0050] S5 defines the hollow detection index DI, quantifies the hollow detection capability, and combines weather forecast data to predict the optimal acquisition time window for each wall unit;

[0051] The S6 plans the drone inspection route based on the optimal acquisition time window to complete the acquisition of high-resolution infrared images of the exterior wall and the diagnosis of hollow defects.

[0052] The invention will be further described in detail below with reference to the accompanying drawings and specific engineering embodiments. The testing object in this embodiment is an 18-story shear wall structure residential building in a provincial capital city. This building was completed in 2018, with a building height of 54m. The exterior walls use a composite structure of "8mm thick ceramic tile finish + 30mm thick extruded polystyrene insulation layer + 200mm thick concrete base". The building has complete facades on all four sides (east, south, west, and north). The floors are divided into a low-rise section (floors 1-6), a mid-rise section (floors 7-12), and a high-rise section (floors 13-18). The total exterior wall testing area is approximately 12,000 square meters. The specific implementation steps are as follows:

[0053] Step S1 involves preprocessing and dividing the wall into units to obtain the facade model of the target building, such as... Figure 2 As shown, in this embodiment, the target building does not have a complete BIM model and construction drawings. Therefore, a DJI M300RTK drone equipped with a P1 full-frame camera was used to collect oblique photogrammetric data of the target building. The flight altitude was 60m, and the drone flew in 5 loops with a heading overlap rate of 80% and a lateral overlap rate of 70%. After the data collection was completed, software was used to generate a high-precision 3D point cloud model and facade texture map of the target building. The model's planar accuracy was better than 5cm, and the facade texture resolution was better than 3mm / pixel.

[0054] Based on the facade model of the target building, the exterior walls of the target building are divided into several independent wall units according to three core dimensions: wall orientation, finishing material, and floor section. First, the building is divided into four primary zones according to its orientation: east facade, south facade, west facade, and north facade. Each primary zone is further divided into three secondary zones according to the floor section: low-rise, mid-rise, and high-rise. Since the finishing material of all facades in this project is ceramic tile with no material difference, the exterior walls of the entire building are ultimately divided into 12 independent wall units.

[0055] Each wall unit is assigned a unique identification code, with the coding rule being "Orientation-Segment-Serial Number". The specific codes are as follows: East facade low-rise DD-01, East facade mid-rise DZ-01, East facade high-rise DG-01, South facade low-rise ND-01, South facade mid-rise NZ-01, South facade high-rise NG-01, West facade low-rise XD-01, West facade mid-rise XZ-01, West facade high-rise XG-01, North facade low-rise BD-01, North facade mid-rise BZ-01, North facade high-rise BG-01.

[0056] The thermal properties of each wall unit were collected, and the specific parameters are as follows:

[0057] Ceramic tile finish: nominal thickness 8mm, density 2300kg / m³, specific heat capacity 840J / ( ), nominal thermal conductivity 1.8 W / ( The solar absorptivity is 0.7.

[0058] Extruded polystyrene (XPS) insulation layer: nominal thickness 30mm, density 35kg / m³, specific heat capacity 1500J / (m³). ), nominal thermal conductivity 0.030 W / ( );

[0059] Concrete base layer: 200mm thickness, 2400kg / m³ density, 920J / (m³ specific heat capacity). Thermal conductivity 1.74 W / ( );

[0060] Indoor environmental parameters: Indoor ambient temperature (T) in Taking the standard value of 25℃, the indoor convective heat transfer coefficient h in Take 8.7W / ( ).

[0061] Step S2: On-site environmental parameter acquisition and extremely limited time-series presampling, such as... Figure 3 As shown, a portable automatic weather station is deployed at an unobstructed location at the center of the roof of the target building to collect on-site environmental parameters in real time, including ambient air temperature, wind speed, wind direction, and total solar radiation intensity, with a sampling frequency of no less than once per minute; the solar altitude angle and solar azimuth angle are calculated in real time through a GPS positioning module, with a sampling frequency of once per minute.

[0062] This inspection used a DJI M300 RTK drone, equipped with an H20T infrared thermal imager and a visible light camera. The core parameters of the infrared thermal imager are: infrared resolution of 640×512, temperature measurement range of -20℃ to 150℃, temperature measurement accuracy of ±2℃, noise equivalent temperature difference NETD≤50mK, and response wavelength of 8-14μm; the visible light camera has a resolution of 20 megapixels and supports RTK centimeter-level positioning.

[0063] The two pre-sampling time points were set to 1 hour and 2 hours after local sunrise, with a 60-minute interval between the two points, meeting the 30-60 minute interval requirement. Pre-sampling employed a low-resolution fast scanning mode, with the infrared image resolution set to 320×256. The flight speed for each wall unit was set to 5 m / s, the flight distance to the wall was 20 m, the forward overlap was 60%, and the lateral overlap was 50%. The acquisition time for a single wall unit was controlled within 35 seconds, and the total pre-sampling time for 12 wall units was 8 minutes, meeting the requirement that the total pre-sampling time not exceed 10 minutes. Finally, the temporal infrared thermal image and corresponding visible light image of each wall unit at two time points were obtained.

[0064] Step S3: Preprocessing of presampled data and initial extraction of features from suspected hollow areas, such as... Figure 4 As shown, the time-series infrared thermal images of the 12 pre-sampled wall units were sequentially standardized and preprocessed. The specific process is as follows:

[0065] Non-uniformity correction: Using the two-point correction parameters of the infrared camera at the factory, the non-uniformity of the pixel response in the infrared thermal image is corrected, and the non-uniformity of the corrected image is less than 0.5%;

[0066] Defect removal: A 3×3 window median filtering algorithm is used to traverse all pixels in the image, identify defective pixels whose gray values ​​differ from those of surrounding pixels by more than 5 times the standard deviation, and replace them with the median of the surrounding pixels to eliminate dead pixels, flashes and other noise in the infrared image.

[0067] Image registration: The SIFT feature matching algorithm is used to extract and match feature points from the infrared thermal images of two time points of the same wall unit. Combined with the random sampling consensus algorithm, erroneous matching points are eliminated. The registration accuracy is controlled within 0.8 pixels, ensuring that pixels at the same spatial location correspond one-to-one in the time series images.

[0068] Radiometric calibration: Based on the radiometric calibration coefficient of the infrared camera, the grayscale values ​​of the corrected infrared image are converted into radiance values, and then converted into wall surface temperature values ​​through Planck's law, thus obtaining the time-series surface temperature field data of each wall unit.

[0069] After preprocessing, the registered temporal temperature field is subjected to differential processing to calculate the temporal temperature change rate of each pixel. The calculation formula is as follows: ; in, This represents the pixel temperature value at the first time point. This represents the pixel temperature value at the second time point. For the time interval between two nodes, generate a temperature change rate distribution map for each wall unit.

[0070] The mean μ and standard deviation σ of the overall temperature change rate of the wall unit were calculated. Areas where the difference between the temperature change rate and the mean exceeded twice the standard deviation were marked as temperature anomaly areas. Through connected component analysis, tiny noise areas with an area less than 50 cm² were removed, and the remaining areas were identified as suspected hollow areas. The location, area, centroid coordinates, and time-series temperature change curve of each suspected hollow area were recorded. In this embodiment, a total of 129 suspected hollow areas were identified in 12 wall units.

[0071] Step S4: Construction of a one-dimensional unsteady-state heat conduction model of the wall and inversion of intrinsic parameters, as follows: Figure 5As shown, for each wall unit, a one-dimensional unsteady heat conduction model of the wall is established based on the thermal property parameters collected in step S1. The model is divided into four structural layers from the outside to the inside along the wall thickness direction: finishing layer, air layer, insulation layer, and base wall.

[0072] Construct the general governing equations for one-dimensional unsteady heat conduction: Equation (1): ; In the formula, x is the coordinate along the thickness of the wall, x=0 is the outer surface of the finishing layer, x=δ is the inner surface of the base wall; ρ(x) is the density of the material at position x, c(x) is the specific heat capacity of the material at position x, λ(x) is the thermal conductivity of the material at position x, and T(x,t) is the temperature value at position x at time t.

[0073] The general control equation of equation (1) is discretized using the finite difference method: the wall is divided into 50 uniform grid units along the thickness direction, the time step is set to 10s, and the unconditionally stable implicit difference scheme is used for discretization. The discretized linear equations are solved by the chasing method to obtain the temperature distribution of the wall along the thickness direction with time.

[0074] Set the initial and boundary conditions for the model:

[0075] Initial conditions: The ambient temperature at the first pre-sampling time point is taken as the initial external temperature. Combined with the indoor ambient temperature, the initial temperature distribution of the wall along the thickness direction is obtained through steady-state heat conduction calculation.

[0076] External surface boundary conditions: Third-type boundary conditions are adopted, comprehensively considering solar radiation absorption, convective heat transfer, and long-wave radiation heat transfer. The boundary condition equations are as follows: ; In the formula, α s Let I(t) be the solar absorptivity of the surface layer, I(t) be the solar radiation intensity at time t, and h be the solar absorptivity of the surface layer. c T is the convective heat transfer coefficient. air (t) represents the ambient air temperature, h r T is the radiative heat transfer coefficient. sky (t) represents the effective sky temperature; where the convective heat transfer coefficient is calculated using the empirical formula h. c =2.8 + 3.0 × v, where v is the actual measured wind speed on site, and the effective sky temperature is calculated using the formula T. sky =T air Calculations are performed using -6, with the radiation heat transfer coefficient taken as 5.0 W / ( );

[0077] Inner surface boundary conditions: Third type boundary conditions are used, considering indoor convective heat transfer. The boundary condition equations are: ; In the formula, h in T is the indoor convective heat transfer coefficient. in This refers to the indoor ambient temperature.

[0078] Using the pre-sampled time-series surface temperature field data obtained in step S3 as the observed values, the unknown parameters of the heat conduction model are inverted using a hybrid optimization algorithm of genetic algorithm-Leuwenburg-Marquardt (GA-LM). The specific process is as follows:

[0079] Set the unknown parameter to be inverted: the actual thermal conductivity λ of the finish layer. s Actual thermal conductivity λ of the insulation layer ins , thickness d of the air gap, and equivalent thermal conductivity λ of the air gap air ;

[0080] Set the parameter search range: λ s ∈[1.0,2.5]W / ( ), λ ins ∈[0.025,0.05]W / ( ), d∈[0.5,20]mm, λ air ∈[0.024,0.1]W / ( );

[0081] Global search using genetic algorithm: The population size of the genetic algorithm is set to 50, the crossover probability to 0.7, the mutation probability to 0.05, and the number of generations to 100. The optimization objective is to minimize the mean square error between the surface temperature value calculated by the model and the measured temperature value, and to obtain the optimal initial values ​​of the parameters.

[0082] Levenberg-Marquardt algorithm for local fine-tuning: Starting with the initial values ​​obtained from the genetic algorithm, the Levenberg-Marquardt algorithm is used for local fine-tuning. The convergence condition is set to the rate of change of the objective function being less than 10. -6 If the iteration count exceeds 50, stop the iteration and output the optimal inversion parameter values.

[0083] In this embodiment, the mean square error between the surface temperature value calculated by the inverted model and the pre-sampled measured value is 0.21℃, which is less than the temperature measurement accuracy of the infrared camera ±2℃. The model accuracy meets the requirements, and finally, the accurate heat conduction model of each wall unit and the intrinsic thermophysical parameters of all suspected hollow areas are obtained.

[0084] Step S5: Quantitative evaluation of the detectability of hollow drum and prediction of the optimal acquisition time window, such as... Figure 6 As shown, the detectability index DI for hollow areas is defined as the core parameter for quantitatively evaluating the detectability of hollow area defects. The formula for calculating DI is: ; In the formula, T def T represents the surface temperature of the wall in the hollow area. norm NETD represents the surface temperature of the wall in the adjacent normal area, and NETD represents the noise equivalent temperature difference of the infrared thermal imager. According to the infrared detection industry standard, when DI≥3, it is determined that the hollow defect is stable and detectable in the corresponding time period and can be clearly identified by the infrared thermal imager.

[0085] Based on the accurate heat conduction model of the wall obtained in step S4, the on-site environmental parameters at the current moment are input, and the DI value of the suspected hollow area in each wall unit is calculated to complete the pre-quantitative assessment of the detectability of hollow areas.

[0086] By using the professional meteorological data interface of the local meteorological bureau, high-precision hourly meteorological forecast data for the target area for the next 72 hours is obtained, including hourly solar radiation intensity, solar altitude angle, solar azimuth angle, ambient air temperature, wind speed, and precipitation probability. First, periods with a precipitation probability of more than 30% are removed. In this embodiment, there is no precipitation in the next 3 days, so there is no need to remove periods.

[0087] By substituting hourly weather forecast data into the precise heat conduction model of each wall unit, the surface temperature of the suspected hollow area and the surface temperature of the normal area within each wall unit are calculated hourly, resulting in hourly temperature difference and DI value, and generating a continuous curve of the DI value of each wall unit changing over time.

[0088] Time periods with a DI ≥ 3 and a continuous duration ≥ 15 minutes were selected as candidate acquisition windows for the wall unit. A multi-factor weighted ranking method was used to prioritize the candidate acquisition windows. The ranking weight factors included: average DI (40%), continuous window duration (30%), time matching degree of windows facing the same direction (20%), and meteorological parameter stability (10%). Finally, the optimal acquisition time window for each wall unit was obtained: 8:00-9:30 AM for the three wall units on the east facade, 10:00-11:30 AM for the three wall units on the south facade, 2:30-4:00 PM for the three wall units on the west facade, and 12:00-1:30 PM for the three wall units on the north facade.

[0089] Step S6: Drone inspection route planning, image acquisition, and precise diagnosis of voids, such as... Figure 7 As shown, based on the optimal acquisition time window of each wall unit, and combined with the spatial location, orientation, and size of the wall unit, the ant colony optimization algorithm is used to generate the global inspection route of the UAV. The goal of route optimization is to minimize the total flight distance and the total inspection time, while ensuring that the acquisition time of each wall unit falls entirely within its optimal acquisition window.

[0090] The constraints for flight path planning are as follows: the maximum flight time of a single UAV is 40 minutes, the flight distance of a single UAV is no more than 10km; the safe distance between the flight and the wall is no less than 15m, the flight altitude range is 5m-60m; the obstacle avoidance distance is no less than 2m; the forward overlap rate of high-resolution data acquisition is 80%, and the lateral overlap rate is 70%.

[0091] For each wall unit, high-resolution acquisition parameters were set: infrared image resolution was set to 640×512, flight distance to the wall was 15m, flight speed was 2m / s, and the corresponding ground sampling distance was 2.3mm / pixel, meeting the accuracy requirements for void detection. A real-time dynamic correction mechanism was also set. When the deviation between the real-time solar radiation intensity and the predicted value exceeded 20%, or the deviation between the wind speed and the predicted value exceeded 3m / s and lasted for more than 10 minutes, the correction mechanism was immediately triggered, and steps S4 to S6 were re-executed to update the optimal acquisition window and inspection route. In this embodiment, the UAV flew four sorties, with a total flight time of 128 minutes, completing the acquisition of high-resolution infrared and visible light images of the entire 12,000㎡ exterior wall of the building. The acquisition time for all wall units fell within the optimal acquisition window.

[0092] The high-resolution infrared thermal images were standardized and preprocessed to obtain the surface temperature field image of each wall unit. Based on the accurate heat conduction model of the wall unit obtained in step S4, the theoretical temperature distribution of each pixel was calculated. Environmental interference compensation was performed on the measured temperature field image to eliminate temperature field distortion caused by differences in the angle of direct sunlight, wind speed changes, and wall structure, so as to obtain the pure temperature field distribution caused only by defects in the internal structure of the wall.

[0093] An improved YOLOv8 deep learning model was used to perform target detection and instance segmentation of hollow areas in compensated infrared thermal images. The model was based on pre-trained weights from the COCO dataset and fine-tuned using 10,000 labeled infrared images of hollow building exterior walls. The training set samples included hollow defect samples with different finishing layer materials, different orientations, and different environmental conditions. The model's average detection accuracy mAP@0.5 reached 98.2%.

[0094] Based on the segmented hollow areas, combined with UAV RTK positioning data and camera internal and external parameters, the world coordinates of each hollow area are calculated using photogrammetry principles, corresponding to the specific floor and axis position of the building facade. At the same time, the actual area, length, and width of the hollow area are calculated. Combined with the hollow air layer thickness obtained from step S4, the hollow risk level is classified: hollow area less than 0.5㎡ is considered minor hollow, 0.5-2㎡ is considered general hollow, and greater than 2㎡ is considered severe hollow.

[0095] The final product is a standardized diagnostic report for hollow areas in building exterior walls. The report includes a project overview, testing basis, testing methods, a summary of testing results, detailed information on each hollow defect (location, size, risk level), corresponding infrared images, visible light images, temperature field distribution maps, and targeted rectification suggestions.

[0096] like Figure 8 As shown, based on the building exterior wall conditions of the embodiment, the inversion accuracy of the one-dimensional unsteady-state heat conduction model is verified. The main plot covers the full sunshine period from 6:00 to 18:00, showing the temporal changes of the temperature measured by UAV infrared and the temperature calculated by the GA-LM hybrid algorithm inversion model. The two curves highly overlap, synchronously responding to the dynamic changes of solar radiation and ambient temperature, with no systematic bias. The subplot quantitatively displays the absolute error over the entire time period, with a root mean square error (RMSE) of only 0.21℃ and a maximum absolute error of 0.38℃, far lower than the ±2℃ temperature measurement accuracy of infrared thermal imagers, fully meeting the model accuracy requirements of this application. The results prove that this application can obtain an accurate wall heat conduction model through extremely limited temporal presampling, accurately reconstructing the dynamic law of the wall surface temperature field, providing reliable support for subsequent quantitative assessment of the detectability of voids and prediction of the optimal acquisition window, verifying the scientific nature and engineering practicality of the method in this application.

[0097] like Figure 9 As shown in the figure, covering the entire time period from 6:00 to 18:00 on a single day, the DI time-series curves of the hollow areas in the four directions of east, south, west, and north are plotted. The stable detectable thresholds of NETD=50mK and DI≥3 for the infrared spectrometer are marked, along with the optimal acquisition windows for each direction predicted by the method of this application. The curve trends perfectly match the solar radiation patterns and the thermal response characteristics of the wall orientation. Within the optimal window for each orientation, the DI value is consistently higher than the detectable threshold of 3, which is completely consistent with the adaptive prediction results of this application, verifying the accuracy of the quantitative assessment method for the detectability of hollow areas. This figure visually demonstrates that the method of this application can accurately predict the optimal detection time for different walls, fundamentally solving the blindness of traditional manual timing selection and ensuring the effectiveness and diagnostic accuracy of UAV inspection data.

[0098] This application also provides an embodiment of an electronic device. The electronic device is manifested in the form of a general-purpose computing device. The components of the electronic device may include, but are not limited to: one or more processors or processing units, memory, and buses connecting different components (including memory and processing units).

[0099] A bus refers to one or more of several bus architectures, including memory buses or memory controllers, peripheral buses, graphics acceleration ports, processors, or local buses using any of the various bus architectures. Examples of these architectures include, but are not limited to, Industry Standard Architecture (ISA) buses, Micro Channel Architecture (MCA) buses, Enhanced ISA buses, Video Electronics Standards Association (VESA) local buses, and Peripheral Component Interconnect (PCI) buses.

[0100] Electronic devices typically include a variety of computer-readable media. These media can be any available media that can be accessed by the electronic device, including volatile and non-volatile media, and removable and non-removable media.

[0101] The memory may include computer-readable media in the form of volatile memory, such as random access memory (RAM) and / or cache memory. Electronic devices may further include other removable / non-removable, volatile / non-volatile computer device storage media. By way of example only, the storage system may be used to read and write non-removable, non-volatile magnetic media.

[0102] The electronic device can also communicate with one or more external devices (e.g., keyboard, pointing device, camera, etc.), may include a display, and may communicate with one or more devices that enable a user to interact with the electronic device, and / or with any device that enables the electronic device to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed via an input / output (I / O) interface. Furthermore, the electronic device can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN)) and / or public networks, such as the Internet) via a network adapter. The network adapter communicates with other modules of the electronic device via a bus. The processor executes various functional applications and data processing by running programs stored in memory, such as implementing the infrared thermal imaging-based method for diagnosing hollow areas in building exterior walls provided in the above embodiments of the present invention.

[0103] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for diagnosing a hollowing of an external wall of a building based on infrared thermography, characterized in that, Includes the following steps: S1 divides the exterior wall of the target building into several independent wall units and collects the thermal properties of each wall unit. S2 collects on-site environmental parameters and controls the drone to complete a full-coverage scan of each wall unit to obtain time-series infrared thermal images. S3 preprocesses the time-series infrared thermal image to obtain the time-series surface temperature field of the wall and extracts areas with abnormal temperature change rates as suspected hollow areas. S4 establishes a one-dimensional unsteady-state heat conduction model of the wall, and uses the time-series surface temperature field as the observation value to invert and obtain the accurate heat conduction model of each wall unit; S5 defines the hollow detection index DI, quantifies the hollow detection capability, and combines weather forecast data to predict the optimal acquisition time window for each wall unit; The S6 plans the drone inspection route based on the optimal acquisition time window to complete the acquisition of high-resolution infrared images of the exterior wall and the diagnosis of hollow defects.

2. The method for diagnosing hollow areas in building exterior walls based on infrared thermal imaging according to claim 1, characterized in that, In S1, the facade model is the BIM model of the target building, construction drawings, or a three-dimensional point cloud model and facade texture map generated by oblique photography of UAV.

3. The method for diagnosing hollow areas in building exterior walls based on infrared thermal imaging according to claim 1, characterized in that, In S1, the wall unit is divided into three dimensions: wall orientation, finishing layer material, and floor section. Each wall unit is configured with a unique identification code. In S1, the thermal property parameters include the material, thickness, density, specific heat capacity, thermal conductivity, and solar absorptivity of the finishing layer, insulation layer, and base wall.

4. The method for diagnosing hollow areas in building exterior walls based on infrared thermal imaging according to claim 1, characterized in that, The on-site environmental parameters include ambient air temperature, wind speed, total solar radiation intensity, solar altitude angle and solar azimuth angle, and the sampling frequency is no less than once per minute.

5. The method for diagnosing hollow areas in building exterior walls based on infrared thermal imaging according to claim 1, characterized in that, In step S3, the preprocessing includes non-uniformity correction, bad pixel removal, image registration and radiometric calibration, converting the infrared image grayscale value into the wall surface temperature value.

6. The method for diagnosing hollow areas in building exterior walls based on infrared thermal imaging according to claim 1, characterized in that, In S3, the temperature change rate of each pixel is calculated by time-series temperature difference. Areas where the difference between the temperature change rate and the overall mean of the wall exceeds 2 times the standard deviation are marked as temperature abnormal areas. After removing small noise areas through connected component analysis, suspected hollow areas are obtained.

7. The method for diagnosing hollow areas in building exterior walls based on infrared thermal imaging according to claim 1, characterized in that, In S4, the one-dimensional unsteady-state heat conduction model of the wall is divided into a finishing layer, a hollow air layer, an insulation layer, and a base wall along the thickness direction. Its general governing equation is: ； In the formula, x is the coordinate along the thickness of the wall, ρ(x), c(x), and λ(x) are the density, specific heat capacity, and thermal conductivity of the material at position x, respectively, and T(x,t) is the temperature value at position x at time t.

8. The method for diagnosing hollow areas in building exterior walls based on infrared thermal imaging according to claim 1, characterized in that, In step S4, a genetic algorithm-Levenberger-Marquardt hybrid optimization algorithm is used for parameter inversion. The inversion parameters include the actual thermal conductivity of the finishing layer, the actual thermal conductivity of the insulation layer, the thickness of the air gap layer, and the equivalent thermal conductivity.

9. The method for diagnosing hollow areas in building exterior walls based on infrared thermal imaging according to claim 1, characterized in that, In step S5, the calculation formula for the detectability index DI of the hollow drum is as follows: ； where T def is the surface temperature of the hollow area, T norm is the surface temperature of the adjacent normal area, NETD is the noise equivalent temperature difference of the infrared thermal imager; when DI≥3, it is determined that the hollow area in the corresponding period has stable detectability.

Images