An image processing-based exterior wall defect detection and temperature analysis system

Abstract

This invention discloses an image processing-based system for detecting defects and analyzing temperature in exterior walls, comprising: a defect detection module, which identifies defect areas based on visible light images and generates a defect mask. M d ( x , y The defect detection module calculates the geometric quantization parameters of the defect region, including at least one of the actual area, actual total length, actual width, and actual perimeter of the defect region. The infrared detection module establishes a temperature field based on the mapping relationship between the pseudo-color of the infrared image and temperature. T ( x , y ); Coupling module, the coupling module is used to mask defects. M d ( x , y Registered to the infrared image coordinate system to form an infrared mask. M d t ( x , y ), and combined with infrared mask M d t ( x , y and temperature field T ( x , y Calculate the temperature difference in the defect area. T d The coupling module is able to quantify the geometric parameters and temperature difference of the defect region. T d Assess the risk index of defective areas E .

Description

Technical Field

[0001] This invention relates to the field of building maintenance technology, and in particular to an image processing-based system for detecting exterior wall defects and analyzing temperature. Background Technology

[0002] As a primary structural element of a building, the exterior walls bear crucial functions such as thermal insulation, rain and sun protection, typically accounting for over 60% of the building's total area. However, constantly exposed to the natural environment, exterior walls are susceptible to various defects such as peeling, hollowing, and cracking due to temperature stress, freeze-thaw cycles, structural settlement, and exposure to wind and sun. These defects not only affect the building's appearance and functionality but, over time, can also lead to safety hazards such as tile detachment and insulation layer failure, even endangering personal safety and public property. Therefore, regular and accurate inspection and evaluation of exterior walls has become a vital aspect of building operation and maintenance management, and is of practical significance for ensuring building structural safety and extending its service life.

[0003] Currently, defect detection in building exterior walls primarily relies on two technologies: visible light detection and infrared detection. Visible light analysis can visually reveal surface anomalies such as cracks and damage; infrared thermal imaging can capture the temperature field distribution on the wall surface, detecting hidden temperature anomalies such as hollow areas and leaks. However, in these technologies, visible light detection and infrared detection are independent of each other, requiring inspectors to switch between multiple systems. This is not only inefficient but also makes it impossible to correlate appearance anomalies with temperature anomalies, or to quantitatively assess the overall risk of exterior wall defects, thus failing to meet increasingly sophisticated inspection needs. Summary of the Invention

[0004] One objective of this invention is to provide an image processing-based system for detecting exterior wall defects and analyzing temperature, thereby improving the accuracy and reliability of defect detection on building exterior walls.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: an image processing-based external wall defect detection and temperature analysis system, comprising: a defect detection module, wherein the defect detection module is used to identify defect areas based on visible light images and generate a defect mask. M d ( x , y The defect detection module calculates the geometric quantization parameters of the defect region, which include at least one of the actual area, actual total length, actual width, and actual perimeter of the defect region; the infrared detection module is used to establish a temperature field based on the mapping relationship between the pseudo-color of the infrared image and temperature. T ( x , y); Coupling module, the coupling module being used to connect the defect mask M d ( x , y Registered to the infrared image coordinate system to form an infrared mask. M d t ( x , y ), and in combination with the infrared mask M d t ( x , y and the temperature field T ( x , y Calculate the temperature difference in the defect area. T d The coupling module can be based on the geometric quantization parameters of the defect region and the temperature difference. T d Assess the risk index of defective areas E .

[0006] As a preferred embodiment, the defect detection module includes a preprocessing unit for converting the visible light image into a grayscale image. G ( x , y ), for the grayscale image G ( x , y Perform threshold segmentation to obtain a binary mask. M ( x , y ).

[0007] As a preferred embodiment, the defect detection module further includes a block detection unit, which is based on the binary mask. M ( x , y Identify blocky defect regions; the blocky detection unit is set with a minimum area threshold. A min and the maximum area ratio threshold ρ max The block detection unit can screen those that meet the requirements. A min ≤ A j,pix < A max Effective blocky defect area D j Among them, the maximum area threshold Amax = ρ max · W · H , W The width of the visible light image. H The length of the visible light image. A j,pix For the j-th effective blocky defect region D j The pixel outline area.

[0008] Preferably, the block detection unit is able to calculate the effective block defect region. D j The geometric quantization parameters of the blocky defects, wherein the geometric quantization parameters of the blocky defects include at least one of the following: the effective blocky defect region. D j actual area A j = A j,pix · k A The effective blocky defect region D j Area percentage in visible light images r A,j =( A j,pix / ( W · H ))×100%; the effective blocky defect area D j actual perimeter P j = P j,pix · k L The effective blocky defect region D j Circularity C j =(4π A j,pix ) / ( P j,pix 2 The effective blocky defect region D j equivalent diameter Φ j =2 R j,pix · k L ;in, k A The pixel area coefficient. Pj,pix For the j-th effective blocky defect region D j The pixel outline perimeter, R j,pix The effective blocky defect region D j The minimum outer radius of the pixel's concentric circle. k L This is the pixel length coefficient.

[0009] As a preferred embodiment, the defect detection module includes a crack detection unit, which is based on the binary mask. M ( x , y Crack mask extraction is performed using inverse threshold segmentation. M c ( x , y ), and based on the crack mask M c ( x , y The crack detection unit identifies the crack defect area and extracts the centerline of the crack defect area. s ( x , y ), to obtain the skeleton point set of the crack defect region. S ={( x , y )| s ( x , y )=1}.

[0010] As a preferred embodiment, the crack detection unit calculates the geometric quantization parameters of the crack defect region, wherein the geometric quantization parameters include at least one of the following: the actual total length of the skeleton of the crack defect region. L sum =| S |· k L The actual longest path in the crack defect region. L max =| P max |· k L The pixel half-width of the crack defect region The maximum pixel width of the crack defect region The actual maximum width of the crack defect region. The principal axis direction angle of the crack defect region θ ;in, k LThe pixel length coefficient. P max For the skeleton point set S The longest continuous path in the, M c For the crack mask M c ( x , y The boundary set of ) p For the skeleton point set S Skeletal points, q Boundary set M c Any pixel in, θ For the skeleton point set S The angle between the longest side of the smallest bounding rectangle and the horizontal direction of the visible light image.

[0011] As a preferred embodiment, the infrared detection module includes a channel difference unit, which selects a difference formula based on the pseudo-color mode of the infrared image and calculates the channel difference corresponding to each pixel in the infrared image. d ( x , y ).

[0012] As a preferred embodiment, the infrared detection module further includes a temperature field generation unit, which has at least one of a first mode and a second mode: in the first mode, the infrared detection module is used to generate a temperature field on the infrared image. n c (i) The actual temperature at each reference point T i Quantified in t℃ T’ i And construct the fitting function. f ( d To generate a temperature field T ( x , y )= f ( d ( x , y ), where the fitting function f ( d (Regarding) T’ i and d The function; in the second mode, the infrared detection module can be based on channel difference. d ( x , y )% quantile ea and (100-a)% quantile e 100-a Channel difference d ( x , y Truncation and normalization are used to obtain temperature characteristic values. The value of 'a' ranges from [0,1], and 'clip(·,0,1)' is the cutoff function. The infrared detection module is set with a minimum reference temperature. T min and maximum reference temperature T max To generate a temperature field .

[0013] As a preferred embodiment, the infrared detection module further includes a visualization unit, which is used to identify thermal anomaly regions and generate a thermal anomaly mask. M hot ( x , y )= I ( T ( x , y )≥ T th The visualization unit calculates the thermal anomaly geometric quantization parameters of the thermal anomaly region, which include at least one of the following: the area of ​​the thermal anomaly region. The area ratio of the thermal anomaly region in the infrared image. ;in, T th The thermal threshold, k A The pixel area coefficient. A real This represents the actual area corresponding to the visible light image.

[0014] As a preferred embodiment, the coupling module calculates the average temperature of the defect region. The average temperature of the background area To obtain the defect temperature difference The coupling module calculates the significance of thermal anomalies. In conjunction with the geometric quantization parameters of the defective area, a risk index is obtained. E =λ· N vis +(1-λ)· (| Z d |); where the background mask M b t =1- Md t , σ T Let λ be the standard deviation of the temperature field and λ be the weighting coefficient. N vis Obtained by normalization of geometric quantization parameters. (·) is a monotonically increasing mapping function.

[0015] As a preferred embodiment, the image processing-based exterior wall defect detection and temperature analysis system further includes a basic calibration module. This module is used to establish a mapping relationship between the pixel size of the visible light image and the actual size, based on the actual area corresponding to the visible light image, to obtain the pixel area coefficient. k A and pixel length coefficient k L .

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: This application enables the visible light defect mask to be coupled through a coupling module. M d ( x , y Registered to the infrared image coordinate system and correlated with the temperature field T ( x , y Coupling, and then assessing the risk index of the defective area. E This method enables the coupled evaluation of geometric quantification parameters and thermal anomaly intensity in defect areas, overcoming the shortcomings of single detection methods such as visible light detection or infrared detection alone, which cannot comprehensively evaluate hidden defects. Attached Figure Description

[0017] Figure 1 This is a flowchart of an image processing-based exterior wall defect detection and temperature analysis system according to some embodiments of this application.

[0018] Figure 2 This is a schematic diagram of the interface of a defect detection module according to some embodiments of this application.

[0019] Figure 3 This is a schematic diagram of the interface of an infrared detection module according to some embodiments of this application.

[0020] Figure 4 These are visible light images of crack defects according to some embodiments of this application.

[0021] Figure 5 This is a schematic diagram of converting a visible light image of a crack defect into a grayscale image according to some embodiments of this application.

[0022] Figure 6This is a schematic diagram of a visible light image of a crack defect after denoising processing, according to some embodiments of this application.

[0023] Figure 7 This is a schematic diagram of converting a grayscale image of a crack defect into a binary mask according to some embodiments of this application.

[0024] Figure 8 These are visible light images of blocky defects according to some embodiments of this application.

[0025] Figure 9 yes Figure 8 A diagram illustrating the conversion to grayscale.

[0026] Figure 10 Yes Figure 9 A schematic diagram after contrast processing.

[0027] Figure 11 yes Figure 10 A schematic diagram of conversion to a binary mask.

[0028] Figure 12 The block detection unit according to some embodiments of this application is in Figure 11 Based on the extraction of blocky defect regions, superimposed on Figure 8 A schematic diagram.

[0029] Figure 13 This is a schematic diagram showing the block defect geometric quantization parameters of the block defect region calculated by the block detection unit according to some embodiments of this application.

[0030] Figure 14 These are visible light images of crack defects according to other embodiments of this application.

[0031] Figure 15 yes Figure 14 A diagram illustrating the conversion to grayscale.

[0032] Figure 16 Yes Figure 15 A schematic diagram after contrast processing.

[0033] Figure 17 yes Figure 16 A schematic diagram of conversion to a binary mask.

[0034] Figure 18 The crack detection unit according to some embodiments of this application is in Figure 17 On the basis of retain N r A schematic diagram after connecting regions.

[0035] Figure 19 The crack detection unit according to some embodiments of this application is in Figure 18 Based on the identification of blocky defect regions, superimposed on Figure 14 A schematic diagram.

[0036] Figure 20 This is a schematic diagram showing the calculation of the geometric quantization parameters of the crack defect region by the block detection unit according to some embodiments of this application.

[0037] Figure 21 These are infrared images according to some embodiments of this application.

[0038] Figure 22 The infrared detection unit according to some embodiments of this application is in Figure 21 The temperature field image is generated based on this.

[0039] Figure 23 This is a schematic diagram of the infrared detection unit calculating the geometric quantization parameters of thermal anomalies according to some embodiments of this application.

[0040] Figure 24 The risk index is calculated using the coupling module according to some embodiments of this application. E A schematic diagram. Detailed Implementation

[0041] The present invention will now be further described in conjunction with specific embodiments. It should be noted that, without conflict, the various embodiments or technical features described below can be arbitrarily combined to form new embodiments.

[0042] It should be noted that the terms "first," "second," etc., in the specification and claims of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0043] It should be noted that, as used in this application, the terms “basically,” “approximately,” and similar terms are used to indicate approximation rather than degree, and are intended to describe inherent deviations in measured or calculated values ​​that would be recognized by a person skilled in the art.

[0044] An image processing-based system for detecting exterior wall defects and analyzing temperature, such as Figures 1-3 As shown, it includes: a defect detection module, an infrared detection module, and a coupling module. The defect detection module is used to identify defect regions based on visible light images and generate a defect mask. M d ( x , yThe defect detection module calculates the geometric quantization parameters of the defect region, including at least one of the actual area, actual total length, actual width, and actual perimeter of the defect region. The infrared detection module establishes a temperature field based on the mapping relationship between the pseudo-color of the infrared image and temperature. T ( x , y The coupling module is used to mask defects. M d ( x , y Registered to the infrared image coordinate system to form an infrared mask. M d t ( x , y ), and combined with infrared mask M d t ( x , y and temperature field T ( x , y Calculate the temperature difference in the defect area. T d The coupling module is able to quantify the geometric parameters and temperature difference of the defect region. T d Assess the risk index of defective areas E .

[0045] It should be understood that, through the coupling module, visible light defect masks can be... M d ( x , y Registered to the infrared image coordinate system and correlated with the temperature field T ( x , y Coupling, and then assessing the risk index of the defective area. E This method enables the coupled evaluation of geometric quantification parameters and thermal anomaly intensity in defect areas, overcoming the shortcomings of single detection methods such as visible light detection or infrared detection alone, which cannot comprehensively evaluate hidden defects. This, in turn, improves the accuracy and reliability of defect detection on building exterior walls.

[0046] In some embodiments, the image processing-based exterior wall defect detection and temperature analysis system includes a basic calibration module. This module is used to establish a mapping relationship between the pixel size of the visible light image and the actual size, based on the actual area corresponding to the visible light image, and to obtain the pixel area coefficient. k A and pixel length coefficient k LThis allows for the rapid quantification of the actual area of ​​the defect region, providing a risk index for the defect region. E The assessment provides reliable data support.

[0047] In at least one instance, the user only needs to input the actual area corresponding to the visible light image, and the basic calibration module will automatically calculate the actual physical area corresponding to a single pixel based on the image resolution. It should be understood that this setting can simplify the operation process, ensure the logical consistency and universality of actual area quantification, and help to be compatible with more building exterior wall scenarios, such as scenarios where there are no obvious calibration objects.

[0048] In at least one embodiment, the user can select a calibration object in a visible light image and input the actual size of the calibration object. The basic calibration module can also automatically calculate the actual physical area corresponding to a single pixel based on the image resolution. Those skilled in the art can adjust "input the actual area corresponding to the visible light image" or "input the actual area corresponding to a part of the visible light image" according to actual needs. Such adjustments are all within the protection scope of this application.

[0049] In one specific embodiment, in the basic calibration module, the pixel size of the visible light image is set to be... W · H ,in, W The width of the visible light image. H The visible light image is the length of the image, and the total number of pixels in the visible light image is... N pix = W · H The actual area corresponding to the visible light image is A real Pixel area coefficient Among them, pixel area coefficient k A This refers to the actual physical area corresponding to a single pixel.

[0050] Furthermore, an equivalent square approximation is used; in other words, the visible light image is approximated as a square to obtain the pixel length coefficient. It should be understood that by approximating with an equivalent square, the actual physical length of a single pixel's side length can be obtained even when the actual dimensions of the visible light image's length and width are missing. This not only improves the system's versatility but also helps simplify the system's operation.

[0051] It should be understandable that the pixel length coefficient k L Alternatively, the actual size corresponding to the visible light image length can be compared with the visible light image length. HThe ratio is calculated; or, it is obtained by comparing the actual size corresponding to the width of the visible light image with the width of the visible light image. W The ratio is calculated to further improve the length coefficient. k L The accuracy of such adjustments is ensured, and all such adjustments fall within the scope of protection of this application.

[0052] In some embodiments, the defect detection module includes a preprocessing unit, such as... Figures 4-11 ,as well as Figures 14-18 As shown, the preprocessing unit is used to convert the visible light image into a grayscale image. G ( x , y For grayscale images G ( x , y Perform threshold segmentation to obtain a binary mask. M ( x , y It should be understood that grayscale processing allows a color visible light image to be converted into a single-channel grayscale image. G ( x , y ),like Figure 4 and Figure 5 As shown, this reduces the computational load of subsequent system processing, simplifies the data processing flow, and improves data processing efficiency. Furthermore, through binarization, grayscale images can be... G ( x , y Convert to a binary mask M ( x , y ),like Figure 5 and Figure 7 As shown, having a clearly distinguishable target area and background helps to achieve subsequent defect area identification and contour extraction. The target area includes block defect areas and crack defect areas.

[0053] In some embodiments, a binary mask ,in, t For the segmentation threshold, specifically, the segmentation threshold t The segmentation threshold can be set by the user or generated adaptively by the system. This application specifies the segmentation threshold. t The method of generating the threshold is not specifically limited. In at least one embodiment, the segmentation threshold is... t The optimal threshold is obtained through the Otsu thresholding method (maximum inter-class variance method), and then further optimized by user-inputted fixed values. This improves the system's operability and interactivity while reducing operational difficulty. Specifically, when segmenting images with a bright foreground and a dark background, the segmentation threshold... tThe value range can be 120~180; when segmenting images with a dark foreground and a light-colored background, the segmentation threshold... t The value can range from 50 to 100.

[0054] In some embodiments, the preprocessing unit is further configured to perform noise reduction processing on the visible light image, such as... Figure 4 and Figure 6 As shown. Specifically, the preprocessing unit integrates at least two of the following filtering algorithms: Gaussian filtering, median filtering, mean filtering, and bilateral filtering. Users can then select one or more of these algorithms to specifically eliminate interference such as normal distribution noise, impulse noise, and random noise generated during the visible light image acquisition process, thereby improving the quality of the visible light image and enhancing the accuracy and reliability of defect detection on building exterior walls.

[0055] In some embodiments, the preprocessing unit is further configured to process grayscale images. G ( x , y Perform contrast processing, such as... Figure 9 and Figure 10 ,as well as Figure 15 and Figure 16 As shown. Specifically, the preprocessing unit integrates at least two of the following equalization methods: HE (Histogram Equalization), AHE (Adaptive Histogram Equalization), CLAHE (Contrast Limited Adaptive Histogram Equalization), BBHE (Brightness Preserving Bi-Histogram Equalization), DSIHE (Dualistic Sub-Image Histogram Equalization), RMSHE (Recursive Mean-Separate Histogram Equalization), and DHE (Dynamic Histogram Equalization). Users can then select one or more of these methods to improve the grayscale image. G ( x , y Increase the overall or local contrast, enhance the difference between blocky defect areas and crack defect areas and the background, and improve the visibility of each defect area.

[0056] In some embodiments, the preprocessing unit is further configured to process the binary mask. M ( x , y Optimization processing is performed. Specifically, the binary mask is optimized. M ( x , y Performing opening and closing operations suppresses small noise points and fills in voids to obtain an optimized mask. ,in, V This is a structural element. It should be understood that the optimized mask generated by the preprocessing unit... M’ ( x , y It can replace a binary mask. M ( x , y It is used in defect detection modules, infrared detection modules, and coupling modules to further improve the accuracy and reliability of defect detection on building exterior walls.

[0057] Exterior wall defects are mainly divided into block defects and crack defects. The main difference between the two lies in their "shape and outline" and "extension characteristics." Specifically, the overall shape of a block defect area is approximately clump-like, sheet-like, or irregularly patchy, with relatively regular boundaries and a clear demarcation from the background; furthermore, the width and length of the block defect area are similar, with no obvious direction of extension; typical block defects include wall peeling and hollow areas. The overall shape of a crack defect area is approximately long and thin, or linear, with relatively blurred boundaries; furthermore, the width of the crack defect area is much smaller than its length, with a clear direction of extension; typical crack defects include bifurcated linear cracks, alligator cracks, and network cracks.

[0058] In some embodiments, the defect detection module integrates at least two external wall defect detection algorithms to identify and locate the binary mask. M ( x , y ) and optimized mask M’ ( x , y The defect detection module identifies both block-shaped and crack-shaped defect regions within the image. In one specific embodiment, the defect detection module integrates at least two of the following: the Canny edge detection algorithm, a morphological processing algorithm, and an image segmentation algorithm. The Canny edge detection algorithm, through adaptive threshold control, can detect binary masks. M ( x , y ) and optimized mask M’ ( x , yThe morphological processing algorithm optimizes the edge contour of the defect region in the binary mask, making it particularly suitable for identifying subtle defects, such as cracks. Through morphological operations such as dilation, erosion, opening, and closing, the algorithm can optimize the contour of the defect region, effectively highlighting the binary mask. M ( x , y ) and optimized mask M’ ( x , y Image segmentation algorithms can enhance structural features in defective regions. By integrating region growing and watershed algorithms, image segmentation algorithms can achieve accurate segmentation of complex images, facilitating targeted feature extraction and quantitative analysis of defective regions.

[0059] In at least one embodiment, the defect detection module integrates a Canny edge detection algorithm, a morphological processing algorithm, and an image segmentation algorithm. The Canny edge detection algorithm generates an initial contour of the defect region, the morphological processing algorithm optimizes this initial contour, and the image segmentation algorithm improves the segmentation accuracy of the defect region, thereby contributing to the improved numerical accuracy of the subsequently generated geometric quantization parameters. In at least one embodiment, the user can select the algorithm according to their needs.

[0060] It is worth mentioning that those skilled in the art can adjust the algorithm type integrated in the defect detection module according to actual needs, or can integrate different algorithm types into the block detection unit and crack detection unit described below respectively. Such adjustments are all within the protection scope of this application.

[0061] In some embodiments, the defect detection module further includes a block detection unit, which is based on a binary mask. M ( x , y Identify blocky defect regions. It should be understood that blocky detection units can also be based on optimized masks. M’ ( x , y The system identifies blocky defect regions. Furthermore, the blocky detection unit is set with a minimum area threshold. A min and the maximum area ratio threshold ρ max The block-shaped detection unit can be used to screen and meet the requirements. A min ≤ A j,pix < A max Effective blocky defect area D j Among them, the maximum area threshold A max = ρ max · W · H , W The width of the visible light image. H The length of the visible light image. A j,pix For the j-th effective blocky defect region D j The pixel outline area.

[0062] In at least one embodiment, the block detection unit is used to optimize the mask. M’ ( x , y In this process, the contours of blocky defect regions are used as the basic geometric units to extract the set of connected regions {Ω} and / or the set of contours {Ω}. C j Blocky defect areas at the edges of the contact image are removed, and those that meet the criteria are selected. A min ≤ A j,pix < A max The blocky defect region is thus obtained, thereby acquiring the effective blocky defect region. D j .like Figure 8 and Figure 12 As shown, the block inspection unit extracts the effective block defect area. D j This image is then overlaid on the visible light image to facilitate user observation of defects and identification of the extracted effective blocky defect areas. D j Is this reasonable? It's understandable. This setting can reduce recognition errors caused by shadows at the edges of building walls and background clipping, and it also helps to avoid noise areas affecting subsequent geometric quantification calculations of blocky defect areas.

[0063] In at least one embodiment, the block detection unit is capable of setting a minimum area threshold. A min and maximum area ratio threshold ρ max Provides default preset values. In a specific instance, the minimum area threshold... A min The default preset value can be taken from 50 pixels to 500 pixels, which is the maximum area ratio threshold. ρ max The default preset value can be taken from 10% to 50%.

[0064] In at least one embodiment, the minimum area threshold A min and maximum area ratio threshold ρ max The system allows for user-defined settings, enhancing its operability. Alternatively, users can adjust default preset values, simplifying operation and facilitating quick learning. Specifically, users can determine the minimum area threshold based on factors such as noise filtering levels, the presence of identification errors, and whether it matches the actual distribution patterns of external wall defects. A min and maximum area ratio threshold ρ max Whether it is compatible with the resolution of visible light images and the actual size of defects.

[0065] In some embodiments, such as Figure 13 As shown, the block detection unit can calculate the effective block defect area. D j The geometric quantization parameters of the blocky defects include at least one of the following: Effective blocky defect area D j actual area A j = A j,pix · k A ; Effective blocky defect area D j Area percentage in visible light images r A,j =( A j,pix / ( W · H ))×100%=( A j / A real )×100%; Effective blocky defect area D j actual perimeter P j = P j,pix · k L ;in, P j,pix For the j-th effective blocky defect region D j The pixel outline perimeter; Effective blocky defect area D j Circularity C j =(4π A j,pix) / ( P j,pix 2 ); among which, roundness C j The closer it is to 1, the larger the effective blocky defect area. D j The closer to a circle; the more circular. C j The smaller the value, the closer it is to 0, the larger the effective blocky defect area. D j The closer it is to an elongated shape; Effective blocky defect area D j equivalent diameter Φ j =2 R j,pix · k L ;in, R j,pix For effective blocky defect areas D j The minimum outer radius of the pixel.

[0066] It should be understood that by providing effective blocky defect areas D j actual area A j and actual perimeter P j This helps users more intuitively determine the size and severity of exterior wall defects corresponding to block-shaped defect areas. Furthermore, by providing effective block-shaped defect areas... D j Circularity C j and equivalent diameter Φ j This helps users determine the development trend of blocky defect areas. For example, when the roundness... C j The closer the value is to 1, the more uniform the extension of the external wall defect corresponding to the blocky defect area is in all directions, and it tends to peel off as a whole under the action of temperature stress or wind pressure; when the roundness is... C j When the value is low, it indicates that the external wall defect corresponding to the block defect area has a sharp protrusion. Stress concentration is likely to occur at the "sharp corner" position, and the extension direction of the sharp corner needs to be paid special attention to.

[0067] In addition, when dealing with exterior wall defects such as peeling, loosening, or hollowing, it is necessary not only to repair the visible defects on the exterior wall surface, but also to reinforce the wall by providing effective blocky defect areas. D j CircularityC j and equivalent diameter Φ j This helps users determine the size of the circular area requiring drilling and grouting reinforcement to ensure anchorage to a solid base layer. In other words, the block detection unit can calculate the effective block defect area. D j The geometric quantification parameters of block defects provide a reliable basis for the assessment and subsequent maintenance of exterior wall defects.

[0068] In some embodiments, the defect detection module includes a crack detection unit, which is based on a binary mask. M ( x , y Crack mask extraction is performed using inverse threshold segmentation. M c ( x , y ), and based on the crack mask M c ( x , y Identify crack and defect areas. For example... Figure 14 and Figure 19 As shown, the crack inspection unit overlays the identified crack defect area onto the visible light image, allowing the user to observe the defect and determine whether the identified crack defect area is reasonable. It should be understood that the crack detection unit can also be based on an optimized mask. M’ ( x , y Crack mask extraction is performed using inverse threshold segmentation. M c ( x , y ).

[0069] In at least one instance, a crack mask Specifically, because the crack defect region is within the binary mask... M ( x , y ) and optimized mask M’ ( x , y In the image, it appears as a thin, dark line, hence the inverse threshold is used. t c Crack mask generated as a reference M c ( x , y To meet the subsequent processing condition of "the crack defect area is white and the background is black", such as Figure 17 As shown.

[0070] In at least one embodiment, the crack detection unit first passes an inverse threshold.t c The crack defect area and background are separated, and then preserved. N r Connected regions are used to reduce noise interference, such as Figure 17 and Figure 18 As shown, to obtain a refined crack mask. M c ( x , y ),in, N r This represents the maximum number of connected regions that need to be retained. It should be understood that this processing effectively eliminates isolated noise points caused by uneven lighting or fine textures, improving the crack mask while preserving the overall integrity of the crack defect. M c ( x , y The signal-to-noise ratio (SNR) provides a high-quality foundation for the subsequent calculation of geometric quantification parameters of crack defects. In one specific embodiment, N r The value ranges from 1 to 3. It should be understood that those skilled in the art can adjust this value based on the actual branching pattern of the crack defect. N r The range of values ​​for which such adjustments are made is within the scope of protection of this application.

[0071] Furthermore, such as Figure 19 As shown, the crack detection unit is able to extract the centerline of the crack defect region. s ( x , y ), to obtain the skeleton point set of the crack defect region. S ={( x , y )| s ( x , y It can be understood that by using skeletonization, the width attribute of the crack defect region can be separated from the length attribute, making the length characteristics of the crack defect region closer to its true geometric direction, and providing an accurate geometric basis for the subsequent calculation of crack defect geometric quantification parameters, thereby improving the reliability of the system's quantitative analysis.

[0072] In one embodiment, the crack detection unit obtains the total pixel length of the crack defect region. It should be understood that the total skeleton length of the crack defect region is approximately equal to the total number of skeleton points. Therefore, the total pixel length of the crack defect region is... L sum,pix =| S The crack detection unit can also search the skeleton point set. S Continuous paths in Pand obtain the skeleton point set. S Longest continuous path in .

[0073] In some embodiments, such as Figure 20 As shown, the crack detection unit can calculate the geometric quantization parameters of the crack defect region, which include at least one of the following: The actual total length of the skeleton in the crack defect area L sum = L sum,pix · k L =| S |· k L ; The actual longest path in the crack defect region L max =| P max |· k L ; The pixel half-width of the crack defect area ;in, M c For crack masking M c ( x , y The boundary set of ) p For the skeleton point set S Skeletal points, q Boundary set M c Any pixel in the array.

[0074] Maximum pixel width of the crack defect area ; The actual maximum width of the crack defect area ; Principal axis orientation angle of the crack defect region θ .

[0075] It should be understandable that by masking the cracks... M c ( x , y Perform Euclidean distance transformation HThe (·) operator obtains the nearest Euclidean distance from any skeleton point p to the boundary of the crack defect region, which is approximately equivalent to obtaining the local pixel half-width at skeleton point p. This process allows the width attribute of the crack defect region to be removed, which helps to provide a more accurate calculation basis for subsequent calculations of the maximum pixel width and actual maximum width of the crack defect region, thereby improving the reliability of the system's quantitative analysis.

[0076] In at least one instance, θ For skeleton point set S The angle between the longest side of the smallest bounding rectangle and the horizontal direction of the visible light image. Specifically, the skeleton point set. S In the minimum bounding rectangle, the longer side is defined as the "principal axis" to characterize the extension direction of the crack defect region, and the shorter side is defined as the "secondary axis" to characterize the width direction of the crack defect region. The crack detection unit calculates the angle between the "principal axis" and the positive direction of the horizontal axis of the visible light image and normalizes it to [-90°, 90°] to obtain the principal axis direction angle. θ .

[0077] Furthermore, the crack detection unit can be based on the principal axis direction angle. θ Classify the types of external wall defects corresponding to crack defect areas, for example: when | θ When | < 30°, it is judged as a horizontal crack; when 60° < | θ When |≤90°, it is judged as a vertical crack; when 30°≤| θ When the angle is ≤60°, it is determined to be a diagonal crack. It should be understood that those skilled in the art can adjust the threshold according to the actual situation, and such adjustments are all within the scope of protection of this application.

[0078] It is worth mentioning that by providing the actual total length of the skeleton in the crack defect area... L sum and the actual longest path L max This allows for a direct and intuitive reflection of the overall extent of the external wall defect corresponding to the crack area, helping users to intuitively assess the scale and severity of the defect. Furthermore, by providing the actual maximum width of the crack area... w max This helps users assess the extent of damage and leakage risk of external wall defects corresponding to cracked areas. For example, when the actual maximum width w max When the crack reaches a certain value, it indicates that the crack defect has posed a significant threat to the integrity of the wall's waterproof layer or structure, requiring sealing or reinforcement measures; when the actual maximum width w max When the damage is relatively small, it may only require surface treatment of the exterior walls or regular monitoring.

[0079] In addition, by calculating the principal axis orientation angle of the crack defect region θ This helps users determine the relationship between the direction of the external wall defect corresponding to the crack and the direction of the building structure's stress. For example, when the crack direction is perpendicular to or at a large angle to the main stress direction of the wall, it indicates that the crack may be in an unstable propagation state; when the crack direction is parallel to the main stress direction of the wall, it may be a shrinkage crack or a structural crack, with a relatively low risk of propagation. Furthermore, this can be combined with the actual longest path of the crack defect area. L max and actual maximum width w max It can help users determine the depth and range of grouting reinforcement required, or plan the layout of monitoring points according to the direction of cracks.

[0080] It is worth mentioning that the defect detection module integrates block detection units and crack detection units, enabling more comprehensive coverage of the main damage types of building exterior walls and helping to avoid omissions and misjudgments caused by single detection modes. Furthermore, the system of this application can process and quantify block defects and crack defects in parallel, which helps improve the detection and evaluation efficiency of exterior wall defects.

[0081] In some embodiments, the defect detection module includes a correction unit, which provides defect marking and editing functions. Specifically, the correction unit includes a line drawing tool, allowing users to draw the outline of blocky defect areas, the path of crack defect areas, and other defect markings using a mouse, touch screen, or other means, thereby helping to compensate for the shortcomings of automated algorithms in identifying complex or irregular defects.

[0082] Furthermore, the correction unit also includes a segmentation tool, providing one or more segmentation modes such as rectangle, circle, polygon, and free selection, which helps users accurately select the defect area of ​​interest and complete the segmentation, enabling targeted processing of specific areas. Specifically, the specific area extracted by the segmentation tool can form an independent processing object. For example, when a segment of a crack defect area is extracted to form a specific area, the crack detection unit can generate the geometric quantization parameters of the crack defect in that specific area.

[0083] Furthermore, the correction unit also includes editing tools. Specifically, editing tools include, but are not limited to, eraser, undo, and modify. Among them, the eraser is used to erase defect marks made by the tracing tool; the undo is used to undo the operation steps of the tracing tool and the segmentation tool; and the modify is used to perform operations such as scaling, rotating, dragging, and merging on defect marks.

[0084] Furthermore, the correction unit also includes a batch processing tool. Specifically, the batch processing tool is used to traverse all visible light image files that meet the format requirements within the target folder, and in conjunction with the defect detection module, identify defect areas in the visible light images and generate geometric quantization parameters for the defect areas, thereby significantly improving the detection efficiency of exterior wall defects. The target folder can be specified by the user as one or more. It should be understood that the batch processing tool can also be used to traverse multiple visible light image files selected by the user; this application does not impose specific limitations on this.

[0085] It's worth mentioning that when a group of visible light images in the target folder meet the same shooting conditions, the user can generate the pixel area coefficient using only one of the visible light images. k A and pixel length coefficient k L The system can convert the pixel area coefficient k A and pixel length coefficient k L This can be used for other visible light images within the same group, thereby reducing the overall computational load of the system and simplifying user operations.

[0086] Furthermore, the correction unit also includes a history tracking tool. The history tracking tool is used to record every operation in the system, including but not limited to the operation type (e.g., obtaining pixel area coefficients through the basic calibration module). k A and pixel length coefficient k L Grayscale images are generated through a preprocessing unit. G ( x , y ) and binary mask M ( x , y ), for binary masks M ( x , y The noise reduction, contrast enhancement, and optimization processes are performed, and the effective block defect area is calculated using a block detection unit. D j The geometric quantification parameters of block defects are calculated by the crack detection unit, and various operations performed by the correction unit are corrected, including operation parameters and operation time, so that users can view historical records and perform one-click undo and redo operations, thereby greatly reducing the risk of misoperation.

[0087] In some embodiments, the infrared detection module includes a channel difference unit, which selects a difference formula based on the pseudo-color mode of the infrared image and calculates the channel difference corresponding to each pixel in the infrared image.d ( x , y It is understandable that directly using RGB values ​​for linear conversion might destroy the original temperature gradient information, leading to blurred isothermal regions or loss of temperature abrupt change points. Using channel interpolation, however, is more sensitive in capturing temperature changes on the building's exterior surface, thereby improving the temperature field generated subsequently. T ( x , y The reliability of ).

[0088] In at least one embodiment, the red-green channel difference corresponding to each pixel in the infrared image is calculated. d ( x , y )= R ( x , y )- G ( x , y ),in R ( x , y ) is the coordinate ( x , y Pixel at ) R value, G ( x , y ) is the coordinate ( x , y Pixel at ) G The value. It should be understood that by using the difference between the red and green channels, the overall brightness shift caused by changes in ambient lighting or camera gain fluctuations can be offset. In other words, common-mode interference is removed through differential operations, resulting in a better correlation between the extracted features and temperature, thus providing a basis for the subsequently generated temperature field. T ( x , y This provides a reliable basis for quantitative analysis. It is worth mentioning that those skilled in the art can also use red-blue channel difference or blue-green channel difference to match the corresponding pseudo-color mode according to actual needs; such adjustments are all within the scope of protection of this application.

[0089] In some embodiments, the infrared detection module further includes a temperature field generation unit, which has at least one of a first mode and a second mode. Specifically, in the first mode, the infrared detection module is used to generate a temperature field on an infrared image. n c (i) The actual temperature at each reference point T i Quantified in t℃ T’ i And construct the fitting function. f (d To generate a temperature field T ( x , y )= f ( d ( x , y ), where the fitting function f ( d (Regarding) T’ i and d The function calculates the channel difference between each pixel in the aforementioned infrared image. d ( x , y Substitute into the fitting function f ( d The temperature field can then be obtained. T ( x , y ).

[0090] It should be understood that the first mode is suitable for cases with a reference point, where the pixel position of the reference point ( x i , y i The actual temperature of the reference point can be entered by the user or manually selected from the infrared image. T i The pixel color value of the reference point can be read and input by the user from the field. R i , G i , B i The temperature field can be read from the infrared image by the temperature field generation unit. Preferably, the number of reference points... n c (i) There should be 2-5 [element / components], covering the temperature range from low to high as much as possible, thereby improving the temperature field. T ( x , y It is closer to the actual temperature distribution on the exterior wall surface of the building.

[0091] In at least one embodiment, the infrared detection module... n c (i) The actual temperature at each reference point T i Quantified in units of 0.5℃ T’ i , specifically Furthermore, the red-green channel difference at each reference point is calculated. d i =R i - G i To obtain calibration sample pairs Furthermore, the infrared detection module can detect the difference between the red and green channels. d i Sort the data and construct a fitting function. f ( d ).

[0092] In one specific embodiment, when the number of reference points is small, the infrared detection module can calculate the fitting function through piecewise linear interpolation. f ( d Specifically, , It should be understandable that piecewise linear interpolation is used to calculate the fitted function. f ( d This reduces the overall computational load of the system, thereby improving computational efficiency; and it strictly passes through all reference points, which helps avoid oscillations that may be caused by higher-order interpolation. It is worth mentioning that those skilled in the art can enable the infrared detection module to provide one or more other fitting methods according to actual needs, including but not limited to: Lagrange interpolation, cubic spline interpolation, least squares linear fitting, and polynomial fitting; such adjustments are all within the scope of protection of this application.

[0093] In the second mode, the infrared detection module can be based on channel difference. d ( x , y )% quantile e a and (100-a)% quantile e 100-a Channel difference d ( x , y Truncation and normalization are used to obtain temperature characteristic values. The value of 'a' ranges from [0,1]. `clip(·,0,1)` is a cutoff function used to restrict the calculation result within the parentheses to the range [0,1], taking 0 if less than 0 and 1 if greater than 1. It should be understood that the second mode is suitable for situations where a reference point is difficult to obtain, thus expanding the applicability of the system.

[0094] In at least one embodiment, the value of 'a' is 2; in other words, the infrared detection module can be based on the channel difference. d ( x , y 2% quantile e 2 and 98th percentile e 98 Channel differenced ( x , y Truncation and normalization are used to obtain temperature characteristic values. It should be understandable that the 2% quantile is used. e 2 and 98th percentile e 98 This removes 2% of the extreme values ​​at both ends of the data distribution, effectively filtering out noise points and improving the temperature field generated subsequently. T ( x , y The reliability of ) is ensured. It is worth mentioning that those skilled in the art can adjust the value of 'a' according to actual needs, and such adjustments fall within the scope of protection of this application. Preferably, the value of 'a' ranges from 1 to 5.

[0095] Furthermore, in the second mode, the infrared detection module is set with a minimum reference temperature. T min and maximum reference temperature T max To generate a temperature field In one specific embodiment, the lowest reference temperature T min The highest reference temperature is 20.0℃. T max The maximum temperature is 70.0℃. It's understandable that under extreme conditions of sunny, windless weather with direct sunlight in summer, the highest temperature on the exterior wall surface of ordinary buildings is typically 50℃~60℃, while dark-colored, west-facing exterior insulation layers may approach 70℃. Therefore, the maximum reference temperature is set at 70.0℃. T max Setting it to 70.0℃ allows it to cover most of the temperature range of building exterior surfaces.

[0096] It is worth mentioning that those skilled in the art can determine the minimum reference temperature based on the specific environment of the building's exterior wall, such as geographical location, season, weather, and sunlight. T min and maximum reference temperature T max The values ​​and precision are adjusted to ensure the generated temperature field... T ( x , y These adjustments are more in line with the actual situation and are all within the scope of protection of this application.

[0097] It is understandable that while related technologies can generate infrared images of building exterior surfaces to display temperature distribution using infrared thermal imaging technology, the resolution of infrared images is low, making it difficult to combine them with defect areas in visible light images. Furthermore, infrared thermal imaging technology relies on specialized calibration equipment, making it difficult to achieve accurate "color-temperature" mapping in scenarios such as civil buildings and construction projects. In this embodiment, however, the first and second modes of the temperature field generation unit can generate a more refined temperature field. T ( x , y This helps to achieve more accurate color-temperature mapping and provides a reliable data foundation for the fusion of infrared and visible light images. It is also beneficial for conducting correlation analysis between appearance anomalies and temperature anomalies and for quantitatively assessing the comprehensive risks of exterior wall defects.

[0098] In some embodiments, the infrared detection module further includes a visualization unit for identifying thermal anomaly regions and generating a thermal anomaly mask. M hot ( x , y )= I ( T ( x , y )≥ T th ),in, T th This refers to the thermal threshold. Specifically, the temperature field. T ( x , y The temperature range is [T’ min , T’ max Isotherm intervals are T Therefore, the isothermal level τ can be obtained. k = T’ min + k T , k =1,…, K , It should be understandable that, in T ( x , y In the case of the first mode generated by the infrared detection module, T’ min and T’ max Can be read by the visualization unit T ( x , yThe minimum and maximum temperature values ​​are obtained in [the following context]. T ( x , y In the case of generation by the second mode of the infrared detection module, T’ min = T min , T’ max = T max It is worth mentioning that those skilled in the art can also manually input the information according to actual needs. T’ min and T’ max or generated in other ways T’ min and T’ max Such adjustments are all within the scope of protection of this application.

[0099] Furthermore, a threshold region Ω is constructed for each level. k ={( x , y )| T ( x , y )≥τ k}, to make the continuous temperature field T ( x , y It is discretized into multiple temperature rise levels, allowing users to perform tiered assessments of defect severity using multi-level isotherms, facilitating the analysis of temperature rise gradients and trends, such as... Figure 21 and Figure 22 As shown. Furthermore, to highlight areas of thermal anomaly, users can input a thermal threshold into the visualization unit. T th This allows the generation of a thermal anomaly mask. M hot ( x , y This is understandable, as it helps to achieve objective quantification and automatic segmentation of thermal anomaly areas, and helps to avoid subjective differences in human interpretation.

[0100] In some embodiments, such as Figure 23 As shown, the visualization unit can calculate the thermal anomaly geometric quantization parameters of the thermal anomaly region, which include at least one of the following: Area of ​​thermal anomaly region ; Area percentage of thermal anomaly regions in infrared images .

[0101] It should be understandable that by providing the area of ​​the thermal anomaly region... A hot This can intuitively reflect the extent of potential defects such as hollow areas and leaks in the exterior walls, helping users quickly determine the severity of exterior wall defects corresponding to areas of thermal anomaly. By providing the area percentage... r hot This helps eliminate the scale effects caused by different shooting distances or image resolutions, objectively presenting the severity of exterior wall defects corresponding to thermal anomaly areas in the detection field of view in a relatively proportional manner. For example, when the area of ​​the thermal anomaly region... A hot or area percentage r hot When the temperature exceeds a certain threshold, it indicates that the external wall defects corresponding to the thermal anomaly area have developed into regional damage, which may affect the structural safety of the wall and requires detailed investigation or overall repair; when the area of ​​the thermal anomaly area... A hot or area percentage r hot If the defect is small, it indicates that the defect is still in a local or early stage and can be treated by surface treatment or regular monitoring.

[0102] It is worth mentioning that when a pair of visible light images and infrared images in the target folder meet the same shooting conditions, the user can generate the pixel area coefficient solely from the visible light image. k A and pixel length coefficient k L The system can output the actual area corresponding to the visible light image. A real Pixel area coefficient k A and pixel length coefficient k L This is used to generate the corresponding infrared image, thereby reducing the overall computational load of the system and simplifying user operations. Those skilled in the art can further generate the actual area corresponding to the infrared image based on the infrared image. A’ real Pixel area coefficient k’ A and pixel length coefficient k’ L Such adjustments are all within the scope of protection of this application.

[0103] In some embodiments, such as Figure 24 As shown, the coupling module calculates the average temperature of the defect region. The average temperature of the background area To obtain the defect temperature difference The coupling module calculates the significance of thermal anomalies. And by combining the geometric quantification parameters of the defect area, a risk index is obtained. E =λ· N vis +(1-λ)· (| Z d |); where the background mask M b t =1- M d t , σ T Let λ be the standard deviation of the temperature field and λ be the weighting coefficient. N vis Obtained by normalization of geometric quantization parameters. (·) is a monotonically increasing mapping function.

[0104] It is understandable that this setup facilitates a comprehensive coupled assessment of the geometric quantification parameters and thermal anomaly intensity of the defect area, enabling a complete reflection of the damage state of external wall defects from both geometric and thermal perspectives. Specifically, the defect detection module can provide feedback on the geometric scale and extent of blocky and crack defects, while the infrared detection module can provide feedback on temperature anomalies and potential hazards. Furthermore, the coupling module, by introducing a weighting coefficient λ, can flexibly adjust the contribution of geometric and thermal features, thereby adjusting the risk index. E It can comprehensively consider the size, severity, and potential hazards of the defective area, making the results of exterior wall defect detection more objective and accurate. For example, when the risk index... E A high risk index indicates that the external wall defect corresponding to the defect area not only has significant geometric extension but is also accompanied by strong thermal anomalies, requiring reinforcement or repair measures; when the risk index is high... E At lower levels, regular monitoring or surface treatment can be implemented. Therefore, the coupling module provides a more reliable basis for the inspection and maintenance of building exterior walls.

[0105] The basic principles, main features, and advantages of this invention have been described above. Those skilled in the art should understand that this invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made without departing from the spirit and scope of the invention, and all such changes and modifications fall within the scope of the invention as claimed. The scope of protection claimed by this invention is defined by the appended claims and their equivalents.

Claims

1. A system for detecting exterior wall defects and analyzing temperature based on image processing, characterized in that, include: A defect detection module is used to identify defect regions based on visible light images and generate a defect mask. M d ( x , y The defect detection module can calculate the geometric quantization parameters of the defect region, which include at least one of the actual area, actual total length, actual width, and actual perimeter of the defect region. An infrared detection module is used to establish a temperature field based on the mapping relationship between the pseudo-color of an infrared image and temperature. T ( x , y ); The coupling module is used to connect the defect mask. M d ( x , y Registered to the infrared image coordinate system to form an infrared mask. M d t ( x , y ), and in combination with the infrared mask M d t ( x , y and the temperature field T ( x , y Calculate the temperature difference in the defect area. T d The coupling module can be based on the geometric quantization parameters of the defect region and the temperature difference. T d Assess the risk index of defective areas E .

2. The image processing-based external wall defect detection and temperature analysis system according to claim 1, characterized in that, The defect detection module includes a preprocessing unit, which is used to convert visible light images into grayscale images. G ( x , y ), for the grayscale image G ( x , y Perform threshold segmentation to obtain a binary mask. M ( x , y ).

3. The image processing-based external wall defect detection and temperature analysis system according to claim 2, characterized in that, The defect detection module further includes a block detection unit, which is based on the binary mask. M ( x , y Identify blocky defect regions; The block detection unit is set with a minimum area threshold. A min and the maximum area ratio threshold ρ max The block detection unit can screen those that meet the requirements. A min ≤ A j,pix < A max Effective blocky defect area D j ; Among them, the maximum area threshold A max = ρ max · W · H , W The width of the visible light image. H The length of the visible light image. A j,pix For the j-th effective blocky defect region D j The pixel outline area.

4. The image processing-based external wall defect detection and temperature analysis system according to claim 3, characterized in that, The block detection unit calculates the effective block defect region. D j The geometric quantization parameters of the blocky defects, wherein the geometric quantization parameters of the blocky defects include at least one of the following: The effective blocky defect region D j actual area A j = A j,pix · k A The effective blocky defect region D j Area percentage in visible light images r A,j =( A j,pix / ( W · H ))×100%; the effective blocky defect area D j actual perimeter P j = P j,pix · k L The effective blocky defect region D j Circularity C j =(4π A j,pix ) / ( P j,pix 2 The effective blocky defect region D j equivalent diameter Φ j =2 R j,pix · k L ; in, k A The pixel area coefficient. P j,pix For the j-th effective blocky defect region D j The pixel outline perimeter, R j,pix The effective blocky defect region D j The minimum outer radius of the pixel's concentric circle. k L This is the pixel length coefficient.

5. The image processing-based external wall defect detection and temperature analysis system according to claim 2, characterized in that, The defect detection module includes a crack detection unit, which is based on the binary mask. M ( x , y Crack mask extraction is performed using inverse thresholding. M c ( x , y ), and based on the crack mask M c ( x , y The crack detection unit identifies the crack defect area and extracts the centerline of the crack defect area. s ( x , y ), to obtain the skeleton point set of the crack defect region. S ={( x , y )| s ( x , y )=1}.

6. The image processing-based external wall defect detection and temperature analysis system according to claim 5, characterized in that, The crack detection unit can calculate the geometric quantization parameters of the crack defect region, and the geometric quantization parameters of the crack defect include at least one of the following: The actual total length of the skeleton in the crack defect region L sum =| S |· k L The actual longest path in the crack defect region. L max =| P max |· k L The pixel half-width of the crack defect region The maximum pixel width of the crack defect region The actual maximum width of the crack defect region. The principal axis direction angle of the crack defect region θ ; in, k L The pixel length coefficient. P max For the skeleton point set S The longest continuous path in the, M c For the crack mask M c ( x , y The boundary set of ) p For the skeleton point set S Skeletal points, q Boundary set M c Any pixel in, θ For the skeleton point set S The angle between the longest side of the smallest bounding rectangle and the horizontal direction of the visible light image.

7. The image processing-based external wall defect detection and temperature analysis system according to any one of claims 1-6, characterized in that, The infrared detection module includes a channel difference unit, which selects a difference formula based on the pseudo-color mode of the infrared image and calculates the channel difference corresponding to each pixel in the infrared image. d ( x , y ).

8. The image processing-based external wall defect detection and temperature analysis system according to claim 7, characterized in that, The infrared detection module further includes a temperature field generation unit, which has at least one of a first mode and a second mode: In the first mode, the infrared detection module is used to detect infrared images. n c (i) The actual temperature at each reference point T i Quantified in t℃ T’ i And construct the fitting function. f ( d To generate a temperature field T ( x , y )= f ( d ( x , y ), where the fitting function f ( d (Regarding) T’ i and d The function; In the second mode, the infrared detection module can be based on channel difference. d ( x , y )% quantile e a and (100-a)% quantile e 100-a Channel difference d ( x , y Truncating normalization to obtain temperature characteristic values The value of 'a' ranges from [0,1], and clip(·,0,1) is the cutoff function. The infrared detection module is set with a minimum reference temperature. T min and maximum reference temperature T max To generate a temperature field .

9. The image processing-based external wall defect detection and temperature analysis system according to claim 7, characterized in that, The infrared detection module also includes a visualization unit, which is used to identify thermal anomaly regions and generate a thermal anomaly mask. M hot ( x , y )= I ( T ( x , y )≥ T th The visualization unit can calculate the thermal anomaly geometric quantization parameters of the thermal anomaly region, which include at least one of the following: The area of ​​the thermal anomaly region The area ratio of the thermal anomaly region in the infrared image. ; in, T th The thermal threshold, k A The pixel area coefficient. A real This represents the actual area corresponding to the visible light image.

10. The image processing-based external wall defect detection and temperature analysis system according to any one of claims 1-6, characterized in that, The coupling module calculates the average temperature of the defect region. The average temperature of the background area To obtain the defect temperature difference ; The coupling module calculates the significance of thermal anomalies. In conjunction with the geometric quantization parameters of the defective area, a risk index is obtained. E =λ· N vis +(1-λ)· (| Z d |); Among them, background mask M b t =1- M d t , σ T Let λ be the standard deviation of the temperature field and λ be the weighting coefficient. N vis Obtained by normalization of geometric quantization parameters. (·) is a monotonically increasing mapping function.

11. The image processing-based external wall defect detection and temperature analysis system according to any one of claims 1-6, characterized in that, It also includes a basic calibration module, which is used to establish a mapping relationship between the pixel size of the visible light image and the actual size, based on the actual area corresponding to the visible light image, and to obtain the pixel area coefficient. k A and pixel length coefficient k L .

Images