Infrared thermal image detection positioning method and device based on building outer wall insulation layer hollow defect

Abstract

The application discloses an infrared thermal image detection positioning method and device based on building outer wall insulation layer hollowing defects, and relates to the technical field of building outer wall defect detection. The method comprises the following steps: step one: dividing a to-be-detected outer wall facade into multiple rectangular detection units according to floor height and horizontal span; in a continuous detection period containing a complete temperature rising section and a complete temperature falling section, infrared thermal images are collected at preset time intervals, and ambient temperature and near-wall surface wind speed are recorded synchronously to obtain surface temperature field data of each detection unit at each collection time; the infrared thermal image sequence of the outer wall is continuously collected in the complete temperature rising and falling period, then the thermal abnormal area is extracted frame by frame and the dynamic change of the area with time is tracked, then the characteristics of the hollowing area are refined from the area change, the hollowing depth is judged combined with the time evolution law of the boundary temperature gradient, and finally the risk classification and positioning distribution map are given by comprehensively considering the three dimensions of area, depth and height.

Description

Technical Field

[0001] This invention belongs to the field of building exterior wall defect detection technology, specifically an infrared thermographic detection and positioning method and device for hollow defects in building exterior wall insulation layer. Background Technology

[0002] During long-term use, building exterior wall insulation layers are susceptible to delamination and hollowing defects due to factors such as temperature cycling, wind loads, and material aging. If not detected and repaired promptly, this can lead to insulation layer detachment and even injuries. Currently, infrared thermography is widely used in engineering projects for detecting hollow areas in exterior walls. The basic principle is to utilize the thermal resistance difference caused by air gaps within the hollow area, creating temperature anomalies on the wall surface that can be captured by an infrared thermal imager. Suspected hollow areas are then located through manual interpretation or simple temperature difference threshold comparison.

[0003] However, existing technologies rely solely on surface temperature differences at a single moment for judgment, failing to effectively distinguish between hollow defects and non-hollow factors such as thermal bridging and abnormal moisture content, which also cause temperature anomalies on the wall surface, resulting in a high misjudgment rate. Furthermore, the lack of quantitative analysis on the dynamic changes in area of ​​thermally abnormal regions during temperature rise and fall limits the accuracy of detection results. Based on this, an infrared thermal imaging detection and positioning method and device for hollow defects in building exterior wall insulation layers is proposed. Summary of the Invention

[0004] The purpose of this invention is to provide an infrared thermal imaging detection and positioning method and device for hollow defects in building exterior wall insulation layers, so as to solve the problems mentioned in the background art.

[0005] Infrared thermography detection and localization method based on hollow defects in building exterior wall insulation layer includes: Step 1: Divide the exterior wall facade to be inspected into multiple rectangular inspection units according to the floor height and horizontal span. In a continuous inspection cycle that includes a complete heating section and a complete cooling section, collect infrared thermal images at preset time intervals and simultaneously record the ambient temperature and wind speed near the wall to obtain the surface temperature field data of each inspection unit at each collection time. Step 2: Calculate the local temperature difference field for each frame of thermal image within each detection unit, identify thermal anomaly pixels, extract thermal anomaly regions through connected component analysis, track the same thermal anomaly region across frames based on geometric center offset and area change rate constraints, and obtain the area time series of each thermal anomaly region. Step 3: Separate the area sequence of the heating segment and the area sequence of the cooling segment from the area time series, calculate the area expansion rate and the area contraction rate respectively, and use the ratio of the area expansion rate to the area contraction rate as the hot spot breathing asymmetry coefficient. At the same time, calculate the center temperature decay lag time. Step 4: Extract the average temperature difference between the boundary pixels of each thermal anomaly region and the adjacent non-anomaly pixels as the boundary temperature gradient, construct the boundary temperature gradient time series, and calculate the time difference between the peak time of the boundary gradient and the peak time of the area as the boundary gradient lead. Step 5: Use the thermal spot breathing asymmetry coefficient and area expansion rate as the first-level judgment conditions to verify the authenticity of the void. Use the boundary gradient lead amount and center temperature decay lag time as the second-level judgment conditions to subdivide the void depth level and output the void defect location distribution map and risk level table.

[0006] As a further aspect of the present invention, the specific method for calculating the local temperature difference field and identifying thermal anomaly pixels is as follows: The global median of the surface temperature values ​​of all pixels within the detection unit in the thermal image frame is calculated as the local background temperature. The difference between the surface temperature value of each pixel and the local background temperature is taken as the local temperature difference value of that pixel. The mean and standard deviation of the absolute values ​​of the local temperature differences of all pixels are calculated. The mean plus a preset multiple of the standard deviation is taken as the thermal anomaly judgment threshold. Pixels with an absolute value of local temperature difference greater than the thermal anomaly judgment threshold are marked as thermal anomaly pixels. Eight-neighbor connected component analysis is performed on the thermal anomaly pixels to remove connected components with an area smaller than a preset area threshold, and the effective thermal anomaly region is output.

[0007] As a further aspect of the present invention, the specific method for tracking the same thermal anomaly region across frames is as follows: Record the geometric center coordinates of each thermal anomaly region in the previous frame. In the next frame, search for the nearest connected component whose geometric center coordinates are closest. When the geometric center offset distance of the nearest connected component is less than a preset offset threshold and the area change rate is less than a preset change rate threshold, it is confirmed as a continuation of the same thermal anomaly region and the area value of that frame is added to the area time series. When no matching connected component is found in a frame, the area of ​​that frame is recorded as zero and the tracking continues.

[0008] As a further aspect of the present invention, the specific method for calculating the area expansion rate and area contraction rate is as follows: In the area sequence of the heating phase, the frame where the area first exceeds a preset area threshold is taken as the expansion start frame, and the frame where the area reaches the maximum value is taken as the expansion end frame. The area expansion rate is obtained by dividing the difference between the area of ​​the expansion end frame and the area of ​​the expansion start frame by the time interval between the two frames. In the area sequence of the cooling phase, the frame where the area continues to decrease from the maximum value is taken as the contraction start frame, and the frame where the area falls back to below a preset proportion of the maximum value is taken as the contraction end frame. The area contraction rate is obtained by dividing the difference between the area of ​​the contraction start frame and the area of ​​the contraction end frame by the time interval between the two frames.

[0009] As a further aspect of the present invention, the specific method for calculating the center temperature decay hysteresis time is as follows: Extract the temperature time series of the pixel at the geometric center of the thermal anomaly region throughout the entire detection cycle. Find the moment when the temperature reaches its peak during the heating phase and record it as the peak moment. In the cooling phase, find the moment when the pixel temperature drops back to the preset ratio of the difference between the peak temperature and the end temperature of the cooling phase and record it as the half-life moment. Use the time difference between the half-life moment and the peak moment as the center temperature decay lag time.

[0010] As a further aspect of the present invention, the specific method for extracting the boundary temperature gradient is as follows: In each frame of thermal image, pixels belonging to the thermal anomaly region and having at least one eight-neighbor pixel not belonging to the thermal anomaly region are designated as boundary pixels. The absolute value of the difference between the temperature value of each boundary pixel and the temperature value of its adjacent non-thermal anomaly pixel is calculated, and the mean of the absolute values ​​of the temperature differences of all boundary pixels is taken as the boundary temperature gradient of that frame. The boundary temperature gradients of all frames are arranged by time to obtain the boundary temperature gradient time series, and the time corresponding to the maximum value of the series is taken as the peak time of the boundary gradient.

[0011] As a further aspect of the present invention, the specific method of the hierarchical decision-making rule is as follows: First layer of authenticity verification: When the thermal spot breathing asymmetry coefficient is greater than the preset breathing threshold and the area expansion rate is greater than the preset expansion threshold, it is judged as a real cavitation defect and enters the second layer of judgment; otherwise, it is judged as a non-cavitation temperature abnormality and marked for exclusion. Second layer of depth level subdivision: When the boundary gradient lead is greater than the preset lead threshold and the center temperature decay lag time is less than the preset lag threshold, it is judged as a shallow cavitation; when the boundary gradient lead is less than or equal to the preset lead threshold and the center temperature decay lag time is greater than or equal to the preset lag threshold, it is judged as a deep cavitation; when the boundary gradient lead is greater than the preset lead threshold and the center temperature decay lag time is greater than or equal to the preset lag threshold, it is judged as a compound cavitation.

[0012] As a further aspect of the present invention, the specific method for outputting the risk level table is as follows: For each actual hollow defect area, an area risk factor, a depth risk factor, and a height risk factor are calculated. The area risk factor is determined by the proportion of the hollow area to the area of ​​the detection unit. The depth risk factor is assigned a value based on the hollow depth level, with deeper hollows assigned a higher value than shallow hollows. The height risk factor is determined by the proportion of the floor where the hollow is located to the total number of floors in the building. The three risk factors are weighted and summed according to preset weights to obtain a comprehensive risk index. Based on the comprehensive risk index, hollow defects are divided into three levels: high risk, medium risk, and low risk. Each risk level area is marked with a different color on the hollow defect location distribution map.

[0013] An infrared thermal imaging detection and positioning device for hollow defects in building exterior wall insulation layers is provided. The device is equipped with a computer program. When the computer program runs, it causes the device to execute the aforementioned disclosed infrared thermal imaging detection and positioning method for hollow defects in building exterior wall insulation layers, causing the device to perform the following operations: The thermal imaging sequence acquisition module is used to divide the exterior wall facade to be inspected into multiple rectangular detection units according to the floor height and horizontal span. In a continuous detection cycle that includes a complete heating segment and a complete cooling segment, infrared thermal images are acquired at preset time intervals and ambient temperature and near-wall wind speed are recorded simultaneously to obtain the surface temperature field data of each detection unit at each acquisition time. The thermal anomaly region tracking module is used to calculate the local temperature difference field for each frame of thermal image in each detection unit, identify thermal anomaly pixels and extract thermal anomaly regions through connected component analysis, and track the same thermal anomaly region across frames based on geometric center offset and area change rate constraints to obtain the area time series of each thermal anomaly region. The feature extraction module is used to separate the area sequence of the heating segment and the area sequence of the cooling segment from the area time series, calculate the area expansion rate and the area contraction rate respectively, use the ratio of the area expansion rate to the area contraction rate as the hot spot breathing asymmetry coefficient, and calculate the center temperature decay lag time. The boundary gradient analysis module is used to extract the average temperature difference between the boundary pixels of each thermal anomaly region and the adjacent non-anomaly pixels as the boundary temperature gradient, construct the boundary temperature gradient time series, and calculate the time difference between the peak time of the boundary gradient and the peak time of the area as the boundary gradient lead. The hollow blemish determination and location module is used to verify the authenticity of hollow blemishes by using the thermal spot breathing asymmetry coefficient and area expansion rate as the first-level determination conditions, and to further subdivide the hollow blemish depth level by using the boundary gradient lead amount and the center temperature decay lag time as the second-level determination conditions, and output a hollow blemish defect location distribution map and risk level table.

[0014] Compared with the prior art, the beneficial effects of the present invention are: This invention first continuously acquires infrared thermal image sequences of the exterior wall during a complete temperature rise and fall cycle, then extracts thermal anomaly areas frame by frame and tracks their dynamic changes in area over time. Next, it extracts the unique rapid expansion and slow contraction breathing asymmetry characteristics and the lag in center temperature decay characteristics of the hollow area from the area changes. Then, it combines the time evolution law of the boundary temperature gradient to determine the depth of the hollow area. Finally, it provides a risk classification and location distribution map by combining the three dimensions of area, depth and height. By using the new feature quantity of thermal spot breathing asymmetry coefficient, it fundamentally distinguishes hollow areas from thermal bridges and moisture content anomalies, thus improving the accuracy of detection results. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the system framework structure of the present invention; Figure 2 This is a schematic diagram of the infrared thermal imaging detection and positioning device of the present invention. Detailed Implementation

[0016] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Example 1: Please refer to Figure 1 This application provides an infrared thermal imaging detection and localization method for hollow defects in building exterior wall insulation layers, including: Step 1: Divide the exterior wall facade to be inspected into multiple rectangular inspection units according to the floor height and horizontal span. In a continuous inspection cycle that includes a complete heating section and a complete cooling section, collect an infrared thermal image once at the same preset time interval, and simultaneously record the ambient temperature and wind speed near the wall to obtain the surface temperature field data of each inspection unit at each collection time. Here, the complete heating period refers to the time from when the wall begins to be heated by solar radiation after sunrise until the wall surface temperature reaches its peak for the day. The complete cooling period refers to the time from when the wall surface temperature reaches its peak until after sunset when the wall temperature drops to within 2 degrees Celsius of the pre-sunrise temperature. The reason for choosing a complete heating and cooling cycle is that there is a fundamental difference in the thermal anomaly area change behavior between the hollow area and the normally bonded area during the heating and cooling processes. This difference cannot be fully captured in a single moment or a single heating period. The rectangular detection unit is divided into units along the building height direction with each floor as a unit, and along the horizontal direction with 3-meter units. For example, in a 6-story residential building, each floor is 2 meters high. The wall is 0.8 meters high and 18 meters wide. It is divided into 6 units along the height and 6 units along the horizontal, for a total of 36 detection units. The physical dimensions of each detection unit are approximately 2.8 meters by 3 meters. The preset interval for infrared thermal image acquisition is 5 minutes. The entire detection cycle lasts from 6:00 AM to 8:00 PM, a total of 14 hours, corresponding to 168 frames. The first 84 frames cover the heating phase, and the last 84 frames cover the cooling phase. The wind speed near the wall is collected synchronously by an ultrasonic anemometer placed 0.5 meters in front of the wall surface, and the ambient temperature is collected synchronously by a temperature sensor placed in a shaded area. The acquisition time is strictly aligned with the thermal image, and the clock error does not exceed 3 seconds.

[0018] It should be noted that dividing the exterior facade into inspection units aims to ensure that the orientation, height, sunlight conditions, and building structure within the same inspection unit are basically consistent. All subsequent analyses will be compared relatively within the same inspection unit to eliminate the influence of environmental differences at different locations on the overall deviation of surface temperature. The reason for choosing to take pictures during the complete heating and cooling cycle rather than at a single moment is that, due to the debonding, air gaps are formed in the hollow areas. During the heating phase, heat accumulates on the surface of the finish, causing the area of ​​the thermal anomaly to continuously expand outward. During the cooling phase, the heat accumulated in the cavity is slowly released through radiation, causing the area of ​​the thermal anomaly to shrink significantly with a lag. This dynamic behavior of the thermal anomaly area expanding and contracting over time is a unique physical phenomenon of hollow defects. Thermal bridge areas and areas with abnormal moisture content do not have this characteristic. Subsequent steps will utilize this characteristic to achieve accurate identification of hollow defects.

[0019] The exterior wall was divided into small rectangular blocks according to floor height and horizontal span. Infrared thermal images were taken every few minutes from before sunrise to after sunset, while simultaneously recording the ambient temperature and wind speed near the wall. The advantage of this method is that the orientation and sunlight conditions within each small cell are essentially the same, allowing for comparisons within the same cell during subsequent analysis and eliminating interference from environmental differences at different locations. Furthermore, capturing the entire temperature rise and fall cycle is crucial to fully capturing the entire process of expansion and contraction of the hollow areas; this dynamic change cannot be observed by capturing only a single moment. The entire building's exterior wall was divided into blocks, and thermal images were continuously taken throughout a complete daytime cycle.

[0020] Step 2: Perform local temperature difference field calculation and thermal anomaly region identification on each frame of thermal image within each detection unit, track the area change of the same thermal anomaly region in consecutive frames, and obtain the area time series of each thermal anomaly region. The specific method is as follows: For a single frame thermal image within a single detection unit, the surface temperature values ​​of all pixels within that detection unit are extracted, and the median global temperature of that frame is calculated as the local background temperature of that frame. The median is used instead of the mean because when there is a large area of ​​voids or thermal bridges within the detection unit, the mean will be skewed by the abnormal area, while the median is not sensitive to extreme values ​​and can more robustly reflect the temperature level of normal areas. The difference between the surface temperature value and the local background temperature of each pixel within the detection unit is calculated to obtain the local temperature difference value corresponding to that pixel. A positive value indicates that the temperature is higher than the background, and a negative value indicates that the temperature is lower than the background.

[0021] The specific method for identifying thermal anomalies in pixels based on local temperature differences is as follows: The absolute values ​​of local temperature differences for all pixels within the detection unit are statistically analyzed. The mean and standard deviation of these absolute values ​​are calculated. The thermal anomaly threshold is set to the mean plus twice the standard deviation. Pixels with absolute local temperature differences greater than this threshold are marked as thermal anomaly pixels. A thermal anomaly pixel marking matrix is ​​established, with rows and columns corresponding to the spatial positions of pixels within the detection unit. Initially, all elements are 0. The element values ​​corresponding to the thermal anomaly pixels are set to 1. Eight-neighbor connected component analysis is performed. Starting from the top left corner of the marking matrix, the first unvisited pixel with an element value of 1 is found as a seed point. A new connected component number is created, and the seed point is added to this connected component. The eight neighboring positions of the seed point are checked: above, below, left, right, upper left, upper right, lower left, and so on. In the lower right corner, if a neighboring element has a value of 1 and has not been visited, it is added to the current connected component and marked as visited. At the same time, this neighboring element is used as a new seed point to recursively process its eight neighbors until all eight neighbors of the abnormal pixels that have been added to the connected component have been visited or have an element value of 0. Then, the marking matrix is ​​returned to continue scanning to find the next unvisited abnormal pixel and establish the next connected component. After all abnormal pixels have been visited, all connected components are output. For each connected component, the number of pixels contained is calculated and multiplied by the area represented by a single pixel to obtain the area of ​​the thermal anomaly region in the frame. The area represented by a single pixel is equal to the square of the spatial resolution of the infrared thermal imager. For example, if the spatial resolution is 5 mm, then the area represented by a single pixel is 25 square millimeters. Connected components with an area less than the preset area threshold of 0.03 square meters are removed because thermal anomaly regions that are too small may be caused by noise or local stains.

[0022] The specific method for tracking the same thermal anomaly region across frames is as follows: For each thermal anomaly region identified in the first frame, its geometric center coordinates are recorded. In the second frame, the nearest connected component with the geometric center coordinates is searched as a continuation of the same thermal anomaly region. If the geometric center offset distance of the nearest connected component is less than the preset offset threshold of 0.3 meters and the area change rate is less than 50%, it is confirmed as a continuation of the same region, and the area value of that frame is added to the area time series of that region. The tracking continues frame by frame until the end of the detection cycle. If no matching connected component is found in a certain frame, it means that the area of ​​the thermal anomaly region has shrunk to below the elimination threshold at that moment. The area of ​​that frame is recorded as 0 and the tracking continues.

[0023] Output the area time series of each tracked thermal anomaly region within each detection unit, including region number, area value of each frame, and geometric center coordinates of each frame.

[0024] By independently identifying thermal anomaly regions in each frame of thermal image and tracking them across frames, the complete dynamic process of thermal anomaly area changing over time is obtained, providing a data foundation for subsequently using the differences in area expansion and contraction behavior to distinguish hollow areas from other temperature anomalies. Using the median as the local background temperature avoids the anomaly region from pulling the baseline, and the dual constraints of area change rate and geometric center offset ensure the accuracy of cross-frame tracking.

[0025] The process involves first calculating the median temperature within each cell as the normal background temperature. Then, the difference between each pixel's temperature and the background temperature is calculated. Pixels with particularly large differences are marked as outliers. Connected component analysis is then used to group adjacent outliers into a single region. Finally, the geometric center position and area change are compared frame by frame to connect the same regions into an area time series. The median, rather than the average, is used as the background temperature because if there happens to be a large hollow area in a cell, the average will be skewed, while the median is unaffected. During cross-frame tracking, both center offset and area change rate are constrained simultaneously, preventing two different outlier regions from being mistakenly identified as the same. This allows for the identification of temperature-abnormal regions in each frame's thermal image and the tracking of how the area of ​​the same outlier region changes from morning to night.

[0026] Step 3: Perform segmented analysis of temperature rise and fall on the area time series of each thermal anomaly region, calculate the area expansion rate and area contraction rate, and thus obtain the hot spot respiration asymmetry coefficient. Simultaneously, calculate the center temperature decay lag time. The specific method is as follows: For a single thermal anomaly region, the area sequence of the heating segment and the area sequence of the cooling segment are separated from its area time series. The area sequence of the heating segment corresponds to the first 84 frames, and the area sequence of the cooling segment corresponds to the last 84 frames. For the area sequence of the heating segment, the frame where the area first exceeds 0.03 square meters is identified as the expansion start frame, and the frame where the area reaches its maximum value is identified as the expansion end frame. The difference between the area of ​​the expansion end frame and the area of ​​the expansion start frame is divided by the time interval between the two frames to obtain the area expansion rate, which is in square meters per minute. Here, the area expansion rate reflects how quickly the thermal anomaly region expands outward during the heating process.

[0027] For the area sequence of the cooling segment, the frame where the area begins to continuously decrease from the maximum value is identified as the shrinkage start frame, and the frame where the area falls back to less than 30% of the maximum value is identified as the shrinkage end frame. The area shrinkage rate is obtained by dividing the difference between the area of ​​the shrinkage start frame and the area of ​​the shrinkage end frame by the time interval between the two frames. The unit is square meters per minute. Here, the area shrinkage rate reflects the speed at which the thermal anomaly area shrinks inward during the cooling process.

[0028] The thermal spot breathing asymmetry coefficient is calculated by dividing the area expansion rate by the area contraction rate to obtain the thermal spot breathing asymmetry coefficient corresponding to the thermal anomaly region. Here, the thermal spot breathing asymmetry coefficient refers to the ratio of the expansion rate of the thermal anomaly region during the heating stage to the contraction rate during the cooling stage. This coefficient is the core characteristic quantity that distinguishes cavitation defects from other temperature anomalies.

[0029] It should be noted that the physical mechanism of the hollow area determines that its thermal breathing asymmetry coefficient is significantly greater than 1, exhibiting rapid expansion and slow contraction: during the heating phase, solar radiation heats the air temperature inside the cavity rapidly, and the thermal expansion of the air further increases the thermal resistance. The heat is trapped on the surface of the finishing layer and diffuses rapidly laterally, causing the thermal anomaly area to expand rapidly. During the cooling phase, although the surface dissipates heat quickly, the heat accumulated inside the cavity is continuously replenished to the surface through thermal radiation, causing the surface temperature to decay slowly, and the contraction of the thermal anomaly area to lag significantly. In contrast, in thermal bridge areas such as concrete beams and columns, the temperature anomaly range is determined by the structural geometry. The anomaly area remains basically constant during heating and cooling, with both the area expansion rate and contraction rate approaching zero, and the thermal breathing asymmetry coefficient is close to 1 or meaningless. In areas with high moisture content, the surface temperature is lower due to the heat absorption of moisture evaporation, forming cold spots rather than hot spots. Furthermore, evaporation accelerates during heating, causing the cold spot area to expand, while evaporation slows down during cooling, causing the cold spot area to shrink. The direction of area change is opposite to that of the hollow area.

[0030] The specific method for calculating the center temperature decay lag time is as follows: extract the temperature time series of the pixel at the geometric center of the thermal anomaly region throughout the entire detection cycle, find the moment when the temperature reaches its peak during the heating phase and record it as t_peak, find the moment when the temperature of the pixel drops back to 50% of the difference between the peak temperature and the end temperature of the cooling phase during the cooling phase and record it as t_half, and subtract t_peak from t_half as the center temperature decay lag time, in minutes. Here, the center temperature decay lag time reflects the speed of heat dissipation at the center point of the thermal anomaly region. In the hollow area, due to the release of heat stored in the cavity, this time is significantly longer than that in the normal area.

[0031] Output parameters such as the thermal spot breathing asymmetry coefficient, area expansion rate, area contraction rate, and center temperature decay hysteresis time for each thermal anomaly region.

[0032] By leveraging the characteristic of the thermal spot breathing asymmetry coefficient, the unique physical behavior of rapid expansion and slow contraction in the hollow region was utilized to achieve an essential distinction between hollow defects and non-hollow temperature anomalies such as thermal bridges and abnormal moisture content. This breaks through the limitations of traditional infrared thermography, which relies solely on temperature difference or heating rate at a single moment. The center temperature decay lag time further characterizes the physical characteristics of cavity heat storage and release from a time dimension, providing a basis for subsequent hollow depth determination.

[0033] By dividing the area sequence into heating and cooling segments, the rates of area expansion and contraction are calculated respectively. Dividing these two rates yields the thermal spot breathing asymmetry coefficient. The time required for the temperature at the center of the abnormal region to decrease from its peak to half is also calculated. In the hollow region, due to the internal air gap, heat is trapped on the surface and rapidly diffuses outward during heating, causing a sharp increase in area. During cooling, the heat stored in the cavity slowly escapes, causing the area to shrink very slowly. Therefore, the breathing asymmetry coefficient is significantly greater than 1. In contrast, the abnormal area of ​​the thermal bridge region remains basically unchanged, and the area change direction of the moisture content abnormal region is even reversed. This coefficient distinguishes the hollow region from other temperature anomalies.

[0034] Step 4: Calculate the temporal evolution characteristics of the boundary temperature gradient for each thermal anomaly region. Obtain the time difference between the peak time of the boundary gradient and the peak time of the area gradient as the boundary gradient lead. Combined with the hot spot breathing asymmetry coefficient and the center temperature decay lag time, determine the authenticity and depth level of the void based on the hierarchical decision rule. The specific method is as follows: For a single thermal anomaly region, the set of boundary pixels of the region is extracted in each frame of the thermal image. Boundary pixels refer to pixels that belong to the thermal anomaly region and have at least one eight-neighbor pixel that does not belong to the thermal anomaly region. For each boundary pixel, the absolute value of the difference between its temperature value and the temperature value of its adjacent non-thermal anomaly pixel is calculated. The average of the absolute values ​​of the temperature differences of all boundary pixels is taken as the boundary temperature gradient of the frame. Here, the boundary temperature gradient reflects the steepness of the temperature transition between the thermal anomaly region and the surrounding normal region. The boundary temperature gradients of the thermal anomaly region in all frames are arranged by time to obtain the boundary temperature gradient time series. The time corresponding to the frame in the series where the boundary temperature gradient reaches its maximum value is recorded as the boundary gradient peak time.

[0035] The boundary gradient lead is calculated by subtracting the peak time of the boundary gradient from the time corresponding to the frame when the area of ​​the thermal anomaly region reaches its maximum value in step three. The time difference is positive when the peak time of the boundary gradient is earlier than the peak time of the area, and is defined as the boundary gradient lead in minutes. Here, the boundary gradient lead reflects the temporal relationship between the maximum temperature difference at the boundary of the thermal anomaly region and the maximum area.

[0036] It should be noted that different delamination depths lead to significant differences in boundary gradient advance: shallow delamination refers to cases where the debonding location is less than 20 mm from the outer surface. Due to the proximity of the cavity to the surface, heat rapidly accumulates within the thin decorative layer during the initial heating phase, causing a sharp increase in the boundary temperature difference. However, the hot spot area is still small at this stage, and the area only gradually increases as the heat continues to diffuse laterally. Therefore, the boundary gradient peak value significantly precedes the area peak value, resulting in a large boundary gradient advance. Deep delamination refers to cases where the debonding location is more than 40 mm from the outer surface. Heat needs to penetrate a thicker insulation layer to reach the cavity interface. The boundary temperature difference increases slowly and is almost synchronous with the area expansion. The boundary gradient peak value is close to the area peak value, resulting in a small boundary gradient advance.

[0037] The specific method for making judgments based on hierarchical decision-making rules is as follows: The first layer of judgment is for authenticity verification: when the thermal spot breathing asymmetry coefficient is greater than the preset threshold of 1.2 and the area expansion rate is greater than the preset threshold of 0.0002 square meters per minute, the thermal anomaly area is judged to be a real hollow defect and enters the second layer of judgment; when the thermal spot breathing asymmetry coefficient is less than or equal to 1.2 or the area expansion rate is less than or equal to 0.0002 square meters per minute, it is judged to be a non-hollow temperature anomaly and is excluded. The judgment basis is that the thermal anomaly area of ​​the real hollow area shows a significant rapid expansion and slow contraction characteristic during the temperature rise and fall process and the expansion rate reaches the observable level, while the area of ​​the thermal bridge area remains almost unchanged and the area of ​​the moisture content anomaly area changes in the opposite direction.

[0038] The second layer of classification is the depth of the void: when the boundary gradient lead is greater than a preset threshold of 40 minutes and the center temperature decay lag time is less than a preset threshold of 200 minutes, it is classified as a shallow void. The basis for this classification is that shallow cavities are close to the surface, resulting in a rapid establishment of boundary temperature difference but less heat storage and faster heat dissipation. When the boundary gradient lead is less than or equal to a preset threshold of 40 minutes and the center temperature decay lag time is greater than or equal to a preset threshold of 200 minutes, it is classified as a deep void. The basis for this classification is that deep cavities have a long heat conduction path, resulting in a slow establishment of boundary temperature difference but greater heat storage and slower heat dissipation. When the boundary gradient lead is greater than 40 minutes and the center temperature decay lag time is greater than or equal to 200 minutes, it is classified as a composite void. The basis for this classification is that there may be multiple layers of debonding or large-area cavities, resulting in the simultaneous appearance of shallow and deep features.

[0039] Output the determination result for each thermal anomaly region, including parameters such as region number, whether it is a true void, void depth level, hot spot breathing asymmetry coefficient, area expansion rate, area contraction rate, center temperature decay lag time, and boundary gradient lead.

[0040] By combining the temporal evolution characteristics of the boundary temperature gradient with the thermal spot breathing characteristics, and utilizing the physical mechanism difference that shallow hollow boundary temperature difference establishes quickly but stores little heat, while deep hollow boundary temperature difference establishes slowly but stores a lot of heat, a non-contact determination of hollow depth is achieved. This overcomes the technical bottleneck of traditional infrared thermography, which can only locate the hollow position but cannot determine the hollow depth. In the hierarchical decision rule, the first layer uses the breathing asymmetry coefficient and expansion rate as dual conditions to verify the authenticity of the hollow, and the second layer uses the combination of boundary gradient lead and center temperature decay lag time to distinguish the hollow depth. The decision parameters of each layer are strongly correlated with the physical mechanism of the hollow, avoiding the decision conflicts that may be caused by parallel decision-making under multiple conditions.

[0041] By calculating the steepness of temperature change at the boundary of each abnormal area over time, the time difference between the moment of maximum boundary temperature difference and the moment of maximum area is identified as the boundary gradient lead. Then, a two-layer judgment is made: the first layer uses the breathing asymmetry coefficient and expansion rate to verify the presence of voids; the second layer uses the boundary gradient lead and the center temperature decay lag time to determine the depth of the void. Shallow voids are close to the surface, so the boundary temperature difference is established quickly, but heat storage is low and heat dissipation is fast. Deep voids are the opposite; the heat conduction path is long, resulting in a slower establishment of the boundary temperature difference, but more heat storage and slower heat dissipation. Using the combination of these two sets of parameters, the depth of the void can be determined without contacting the wall, thus determining whether a void is real or false and how deep it is.

[0042] Step 5: Calculate the comprehensive risk index based on the hollow area, hollow depth level, and floor location of each actual hollow defect area. Then, classify the risk according to the comprehensive risk index and output a distribution map of external wall hollow defects and a risk level table. The specific method is as follows: For each actual hollow defect area, three dimensions of risk factors are calculated: The first dimension is the area risk factor, which is the area of ​​the hollow area (i.e., the maximum value of the area in the time series) divided by the area of ​​the detection unit where the area is located to obtain the area ratio. The area ratio is multiplied by 100 to obtain the area risk factor, and the maximum value of the area risk factor is limited to 10. The second dimension is the depth risk factor, with shallow hollow areas assigned a value of 4, deep hollow areas assigned a value of 7, and composite hollow areas assigned a value of 9. The reason why deep hollow areas have a higher risk than shallow hollow areas is that deep debonding means that the bond between the insulation board and the base wall has failed, and there is a safety hazard of the entire insulation board falling off. Shallow debonding usually only involves local bulging of the finishing layer, and the hazard of falling off is relatively small. The third dimension is the height risk factor, which is the height ratio obtained by dividing the number of floors where the hollow area is located by the total number of floors in the building. The height ratio is multiplied by 10 to obtain the height risk factor. The height risk factor reflects the engineering reality that the risk of falling objects is greater when the hollow defect is located at a high place.

[0043] The comprehensive risk index for the hollow defect area is calculated by multiplying the area risk factor by a weight of 0.30, the depth risk factor by a weight of 0.45, and the height risk factor by a weight of 0.25. The sum of the three weights is 1. The reason why the depth risk factor has the largest weight is that the hollow depth directly determines the detachment mode and the degree of safety hazard. The harm of deep whole board detachment is far greater than that of shallow local bulging, and it is the most critical dimension in risk assessment.

[0044] When the comprehensive risk index is greater than or equal to 7.0, it is judged as a high-risk hollow defect; when the comprehensive risk index is greater than or equal to 4.0 and less than 7.0, it is judged as a medium-risk hollow defect; when the comprehensive risk index is less than 4.0, it is judged as a low-risk hollow defect.

[0045] Based on the geometric center coordinates and area of ​​all actual hollow defect areas, a distribution map of hollow defects is drawn on the exterior wall elevation drawing. Different colors are used to mark different risk levels: red for high-risk areas, yellow for medium-risk areas, and blue for low-risk areas. The hollow depth level and hot spot breathing asymmetry coefficient are also noted next to each marked area. A risk level table is generated, including the number of each hollow defect area, its location in the detection unit, hollow area, hollow depth level, hot spot breathing asymmetry coefficient, center temperature decay lag time, boundary gradient lead, comprehensive risk index, risk level, and treatment recommendations. For high-risk areas, it is recommended to immediately erect protective netting and anchor it for reinforcement or partially dismantle and rebuild. For medium-risk areas, it is recommended to prioritize treatment during the next exterior facade maintenance cycle and establish a ground warning zone during this period. For low-risk areas, it is recommended to include them in the annual infrared re-testing plan for continuous monitoring to see if the hollow area expands.

[0046] The comprehensive risk index is calculated by weighting risk factors from three dimensions: area, depth, and height. Among them, the depth risk factor has the largest weight, which is consistent with the actual engineering situation where the damage of deep delamination and whole board detachment is far greater than that of shallow local bulging. The hollow defect location distribution map simultaneously marks the location, area, depth, and breathing characteristic parameters, providing complete defect characteristic data for the formulation of repair plans. This allows maintenance personnel to select targeted repair processes according to the depth of the hollow. Shallow hollows can be filled with grouting, while deep hollows require the removal and re-bonding of the insulation board. By scoring based on three dimensions—area size, depth of hollow areas, and floor level—deep hollow areas are given the highest weight because they could potentially cause the entire insulation board to detach. Hollow areas at higher floors are also weighted appropriately due to the greater risk of falling objects. The three weighted scores are summed to categorize the areas into high, medium, and low risk levels. Finally, these levels are marked on the exterior wall elevation drawing using red, yellow, and blue colors, along with suggested treatments. This allows maintenance personnel to immediately identify which areas require immediate attention and which can be addressed in the next maintenance cycle. Shallow hollow areas can be treated with grouting, while deep hollow areas require complete removal and redoing of the insulation board. This eliminates the need for guesswork based on experience. All confirmed hollow defects are prioritized according to their degree of danger, making it easier for maintenance personnel to decide where to repair first.

[0047] Infrared thermal image sequences of the exterior wall are continuously acquired during the complete temperature rise and fall cycle. Then, thermal anomaly areas are extracted frame by frame and their area changes dynamically over time. Next, the characteristics of the hollow areas are extracted from the area changes. The depth of the hollow areas is determined by combining the time evolution law of the boundary temperature gradient. Finally, the risk classification and location distribution map are given by combining the three dimensions of area, depth and height.

[0048] Example 2: Please refer to Figure 2 As shown, this embodiment also provides an infrared thermal imaging detection and positioning device based on hollow defects in building exterior wall insulation layers. This device is equipped with a computer program. When the computer program runs, it causes the device to execute the aforementioned disclosed infrared thermal imaging detection and positioning method based on hollow defects in building exterior wall insulation layers, causing the device to perform the following operations: The thermal imaging sequence acquisition module is used to divide the exterior wall facade to be inspected into multiple rectangular detection units according to the floor height and horizontal span. In a continuous detection cycle that includes a complete heating segment and a complete cooling segment, infrared thermal images are acquired at preset time intervals and ambient temperature and near-wall wind speed are recorded simultaneously to obtain the surface temperature field data of each detection unit at each acquisition time. The thermal anomaly region tracking module is used to calculate the local temperature difference field for each frame of thermal image in each detection unit, identify thermal anomaly pixels and extract thermal anomaly regions through connected component analysis, and track the same thermal anomaly region across frames based on geometric center offset and area change rate constraints to obtain the area time series of each thermal anomaly region. The feature extraction module is used to separate the area sequence of the heating segment and the area sequence of the cooling segment from the area time series, calculate the area expansion rate and the area contraction rate respectively, use the ratio of the area expansion rate to the area contraction rate as the hot spot breathing asymmetry coefficient, and calculate the center temperature decay lag time. The boundary gradient analysis module is used to extract the average temperature difference between the boundary pixels of each thermal anomaly region and the adjacent non-anomaly pixels as the boundary temperature gradient, construct the boundary temperature gradient time series, and calculate the time difference between the peak time of the boundary gradient and the peak time of the area as the boundary gradient lead. The hollow blemish determination and location module is used to verify the authenticity of hollow blemishes by using the thermal spot breathing asymmetry coefficient and area expansion rate as the first-level determination conditions, and to further subdivide the hollow blemish depth level by using the boundary gradient lead amount and the center temperature decay lag time as the second-level determination conditions, and output a hollow blemish defect location distribution map and risk level table.

[0049] By dividing the exterior wall facade into rectangular detection units and acquiring infrared thermal image sequences and synchronous environmental parameters at preset intervals within a continuous detection cycle containing complete heating and cooling phases, the surface temperature field data of each detection unit at each acquisition moment is obtained. Subsequently, the local temperature difference field is calculated for each frame of thermal image, and thermal anomaly pixels are identified. Thermal anomaly regions are extracted through connected component analysis. Based on geometric center offset and area change rate constraints, the same thermal anomaly region is tracked across frames to obtain the area time series of each thermal anomaly region. Then, the heating and cooling phases are separated from the area time series, and the area expansion rate and area contraction rate are calculated respectively. The ratio of the two is used as the thermal spot breathing asymmetry coefficient, and the center temperature decay lag time is calculated synchronously. Thus, the unique thermal characteristics of the rapid expansion and slow contraction of the hollow area are utilized. The response behavior fundamentally distinguishes hollow defects from non-hollow temperature anomalies such as thermal bridging and abnormal moisture content. Finally, it extracts the boundary temperature gradient time evolution characteristics to calculate the boundary gradient lead. The thermal spot breathing asymmetry coefficient and area expansion rate are used as the first-level conditions to verify the authenticity of hollow defects. The boundary gradient lead and center temperature decay lag time are used as the second-level conditions to further subdivide the hollow depth level. Combined with the three dimensions of area, depth and height, a comprehensive risk index is calculated to output a hollow defect location distribution map and risk level table. This not only enables accurate identification of hollow defects and non-contact determination of depth, but also provides complete defect feature data support for maintenance process selection, improves the accuracy of infrared thermal imaging detection, and avoids the high misjudgment rate problem of single-moment temperature difference determination methods.

[0050] Example 3: As Example 3 of the present invention, in specific implementation, compared with Example 1 and Example 2, the technical solution of this example is to combine the solutions of Example 1 and Example 2.

[0051] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0052] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An infrared thermal imaging detection and localization method for hollow defects in building exterior wall insulation layers, characterized in that, include: Step 1: Divide the exterior wall facade to be inspected into multiple rectangular inspection units according to the floor height and horizontal span. In a continuous inspection cycle that includes a complete heating section and a complete cooling section, collect infrared thermal images at preset time intervals and simultaneously record the ambient temperature and wind speed near the wall to obtain the surface temperature field data of each inspection unit at each collection time. Step 2: Calculate the local temperature difference field for each frame of thermal image within each detection unit, identify thermal anomaly pixels, extract thermal anomaly regions through connected component analysis, track the same thermal anomaly region across frames based on geometric center offset and area change rate constraints, and obtain the area time series of each thermal anomaly region. Step 3: Separate the area sequence of the heating segment and the area sequence of the cooling segment from the area time series, calculate the area expansion rate and the area contraction rate respectively, and use the ratio of the area expansion rate to the area contraction rate as the hot spot breathing asymmetry coefficient. At the same time, calculate the center temperature decay lag time. Step 4: Extract the average temperature difference between the boundary pixels of each thermal anomaly region and the adjacent non-anomaly pixels as the boundary temperature gradient, construct the boundary temperature gradient time series, and calculate the time difference between the peak time of the boundary gradient and the peak time of the area as the boundary gradient lead. Step 5: Use the thermal spot breathing asymmetry coefficient and area expansion rate as the first-level judgment conditions to verify the authenticity of the void. Use the boundary gradient lead amount and center temperature decay lag time as the second-level judgment conditions to subdivide the void depth level and output the void defect location distribution map and risk level table.

2. The infrared thermal imaging detection and location method for hollow defects in building exterior wall insulation layer according to claim 1, characterized in that, The specific method for calculating the local temperature difference field and identifying thermal anomaly pixels is as follows: The global median of the surface temperature values ​​of all pixels in the detection unit in the thermal image of the frame is calculated as the local background temperature. The difference between the surface temperature value of each pixel and the local background temperature is taken as the local temperature difference value of that pixel. The mean and standard deviation of the absolute values ​​of the local temperature difference values ​​of all pixels are calculated. The mean plus a preset multiple of the standard deviation is taken as the thermal anomaly judgment threshold. Pixels with a local temperature difference value greater than the thermal anomaly judgment threshold are marked as thermal anomaly pixels. Perform eight-neighbor connected component analysis on thermally anomalous pixels, remove connected components with an area smaller than a preset area threshold, and output the effective thermally anomalous region.

3. The infrared thermal imaging detection and localization method for hollow defects in building exterior wall insulation layer according to claim 2, characterized in that, The specific method for tracking the same thermal anomaly region across frames is as follows: Record the geometric center coordinates of each thermal anomaly region in the previous frame. In the next frame, search for the nearest connected component whose geometric center coordinates are closest. When the geometric center offset distance of the nearest connected component is less than a preset offset threshold and the area change rate is less than a preset change rate threshold, it is confirmed as a continuation of the same thermal anomaly region and the area value of that frame is added to the area time series. If no matching connected component is found in a frame, the area of ​​that frame is recorded as zero and the process continues to track the next frame.

4. The infrared thermal imaging detection and location method for hollow defects in building exterior wall insulation layer according to claim 1, characterized in that, The specific method for calculating the rate of area expansion and the rate of area contraction is as follows: In the area sequence of the heating phase, the frame where the area first exceeds a preset area threshold is taken as the expansion start frame, and the frame where the area reaches the maximum value is taken as the expansion end frame. The area expansion rate is obtained by dividing the difference between the area of ​​the expansion end frame and the area of ​​the expansion start frame by the time interval between the two frames. In the area sequence of the cooling phase, the frame where the area continues to decrease from the maximum value is taken as the contraction start frame, and the frame where the area falls back to below a preset proportion of the maximum value is taken as the contraction end frame. The area contraction rate is obtained by dividing the difference between the area of ​​the contraction start frame and the area of ​​the contraction end frame by the time interval between the two frames.

5. The infrared thermal imaging detection and positioning method for hollow defects in building exterior wall insulation layer according to claim 4, characterized in that, The specific method for calculating the center temperature decay hysteresis time is as follows: Extract the temperature time series of the pixel at the geometric center of the thermal anomaly region throughout the entire detection cycle. Find the moment when the temperature reaches its peak during the heating phase and record it as the peak moment. In the cooling phase, find the moment when the pixel temperature drops back to the preset ratio of the difference between the peak temperature and the end temperature of the cooling phase and record it as the half-life moment. Use the time difference between the half-life moment and the peak moment as the center temperature decay lag time.

6. The infrared thermal imaging detection and location method for hollow defects in building exterior wall insulation layer according to claim 1, characterized in that, The specific method for extracting the boundary temperature gradient is as follows: In each frame of thermal image, pixels belonging to the thermal anomaly region and having at least one eight-neighbor pixel not belonging to the thermal anomaly region are designated as boundary pixels. The absolute value of the difference between the temperature value of each boundary pixel and the temperature value of its adjacent non-thermal anomaly pixel is calculated, and the mean of the absolute values ​​of the temperature differences of all boundary pixels is taken as the boundary temperature gradient of that frame. The boundary temperature gradients of all frames are arranged by time to obtain the boundary temperature gradient time series, and the time corresponding to the maximum value of the series is taken as the peak time of the boundary gradient.

7. The infrared thermal imaging detection and location method for hollow defects in building exterior wall insulation layer according to claim 1, characterized in that, The specific method of hierarchical decision-making rules is as follows: First layer of authenticity verification: When the hot spot breathing asymmetry coefficient is greater than the preset breathing threshold and the area expansion rate is greater than the preset expansion threshold, it is judged as a real hollow defect and enters the second layer of judgment; otherwise, it is judged as a non-hollow temperature abnormality and marked for exclusion. The second layer of depth level subdivision is as follows: when the boundary gradient lead is greater than the preset lead threshold and the center temperature decay lag time is less than the preset lag threshold, it is determined to be a shallow void; when the boundary gradient lead is less than or equal to the preset lead threshold and the center temperature decay lag time is greater than or equal to the preset lag threshold, it is determined to be a deep void; when the boundary gradient lead is greater than the preset lead threshold and the center temperature decay lag time is greater than or equal to the preset lag threshold, it is determined to be a composite void.

8. The infrared thermal imaging detection and location method for hollow defects in building exterior wall insulation layer according to claim 7, characterized in that, The specific method for outputting the risk level table is as follows: For each actual hollow defect area, an area risk factor, a depth risk factor, and a height risk factor are calculated. The area risk factor is determined by the proportion of the hollow area to the area of ​​the detection unit. The depth risk factor is assigned a value based on the hollow depth level, with deeper hollows assigned a higher value than shallow hollows. The height risk factor is determined by the proportion of the floor where the hollow is located to the total number of floors in the building. The three risk factors are weighted and summed according to preset weights to obtain a comprehensive risk index. Based on the comprehensive risk index, hollow defects are divided into three levels: high risk, medium risk, and low risk. Each risk level area is marked with a different color on the hollow defect location distribution map.

9. An infrared thermal imaging detection and positioning device for hollow defects in building exterior wall insulation layers, characterized in that, The device is equipped with a computer program. When the computer program runs, it causes the device to execute the aforementioned disclosed infrared thermal imaging detection and location method based on the hollow defects in the building's exterior wall insulation layer, and causes the device to perform the following operations: The thermal imaging sequence acquisition module is used to divide the exterior wall facade to be inspected into multiple rectangular detection units according to the floor height and horizontal span. In a continuous detection cycle that includes a complete heating segment and a complete cooling segment, infrared thermal images are acquired at preset time intervals and ambient temperature and near-wall wind speed are recorded simultaneously to obtain the surface temperature field data of each detection unit at each acquisition time. The thermal anomaly region tracking module is used to calculate the local temperature difference field for each frame of thermal image in each detection unit, identify thermal anomaly pixels and extract thermal anomaly regions through connected component analysis, and track the same thermal anomaly region across frames based on geometric center offset and area change rate constraints to obtain the area time series of each thermal anomaly region. The feature extraction module is used to separate the area sequence of the heating segment and the area sequence of the cooling segment from the area time series, calculate the area expansion rate and the area contraction rate respectively, use the ratio of the area expansion rate to the area contraction rate as the hot spot breathing asymmetry coefficient, and calculate the center temperature decay lag time. The boundary gradient analysis module is used to extract the average temperature difference between the boundary pixels of each thermal anomaly region and the adjacent non-anomaly pixels as the boundary temperature gradient, construct the boundary temperature gradient time series, and calculate the time difference between the peak time of the boundary gradient and the peak time of the area as the boundary gradient lead. The hollow blemish determination and location module is used to verify the authenticity of hollow blemishes by using the thermal spot breathing asymmetry coefficient and area expansion rate as the first-level determination conditions, and to further subdivide the hollow blemish depth level by using the boundary gradient lead amount and the center temperature decay lag time as the second-level determination conditions, and output a hollow blemish defect location distribution map and risk level table.

Images