External wall tile hollowing detection method based on thermal image sub-region and double threshold segmentation

Abstract

The present invention provides a method for detecting hollowing in exterior wall tiles based on thermal image sub-regions and dual-threshold segmentation, belonging to the field of infrared nondestructive testing technology for building structures. The method divides the infrared image into several sub-regions based on the grid-like distribution of brick joints, and only performs subsequent testing on the tile areas, thereby reducing the interference of brick joints in the infrared image on the quantitative detection of hollowing. Furthermore, the neighborhood dual-threshold image segmentation method combines two parameters, minimum hollow size and minimum temperature difference between the hollow area and the background area, to determine each pixel point. This method solves the under-segmentation or over-segmentation problems of traditional image segmentation methods and improves the quantitative detection accuracy of hollowing.

Description

Technical Field

[0001] The present invention belongs to the technical field of infrared non-destructive testing of building structures, and in particular relates to a method for detecting hollowing of exterior wall tiles based on thermal image sub-regions and dual-threshold segmentation. Background Art

[0002] In the field of building inspection, the main causes of hollowing include substandard construction quality, aging, and environmental influences. These factors gradually reduce the bond strength between tiles and the wall substrate over time, leading to localized debonding and the formation of hollows. If not promptly reinforced, the area of ​​the hollows will gradually expand, eventually leading to complete debonding of the tiles, which can in turn cause accidents such as falling objects. Therefore, there is an urgent need to develop efficient and accurate inspection technologies to comprehensively survey and rapidly diagnose defects in building exterior wall tiles.

[0003] Related technologies for detecting hollow ceramic tiles on building exterior walls primarily include tapping, pulling, and infrared thermography. While the tapping method is simple to use, it is inefficient, carries high risks, and lacks objectivity. The pulling method is destructive to the tiles and unsuitable for large-scale inspections. Infrared thermography, by contrast, can scan large wall surfaces from a distance, offering advantages such as high efficiency and low cost.

[0004] The detection principle of infrared thermography is that, under the stimulation of sunlight, heat flow is generated inside the ceramic tile. Due to the influence of discontinuous hollowing on heat conduction, different temperature distributions are formed on the surface of the ceramic tile. Infrared detection and diagnosis technology measures this infrared radiation energy to measure the temperature and temperature distribution of the object surface, and then judge whether there are hollows inside. However, in this detection method, because the brick joint area is affected by the thermal conductivity and emissivity differences of the material and the local thermal bridge effect, its surface will show localized high temperature phenomenon similar to hollowing, which interferes with the quantitative detection of hollowing defects. Secondly, traditional image segmentation methods may lead to over-segmentation, missing the target area, or under-segmentation, leaving a large amount of background area, making it difficult to accurately identify hollows.

[0005] Therefore, it is necessary to provide a method for detecting hollowing of exterior wall tiles based on thermal image sub-regions and dual threshold segmentation to solve the above problems. Summary of the Invention

[0006] The present invention provides a method for detecting hollowing of exterior wall tiles based on thermal image sub-regions and dual-threshold segmentation. By dividing the tiles into regions and adopting an image segmentation method based on neighborhood dual-thresholds, the method overcomes the shortcomings of traditional detection methods caused by brick joint interference, image noise and local brightness fluctuations, and achieves accurate identification and quantitative detection of hollowing areas, thereby solving at least one technical problem involved in the background technology.

[0007] In order to solve the above-mentioned technical problems, the present invention is achieved as follows:

[0008] A method for detecting hollowing of exterior wall tiles based on thermal image sub-regions and dual threshold segmentation includes the following steps:

[0009] Step S1, collecting infrared images of exterior wall tiles;

[0010] Step S2: Setting a strip mask to cover the brick joints in the infrared image, dividing the infrared image into several sub-regions by the crisscross strip mask, and using the neighborhood double-threshold image segmentation method to determine the hollow points in each sub-region, resetting the pixel values ​​of the hollow points to 1 and the pixel values ​​of the non-hollow points to 0, outputting a binary image, and removing small areas by opening operation to obtain the hollow areas;

[0011] Step S3: establishing a correlation between the actual area of ​​the hollow region and the image area, and calculating the actual area of ​​the hollow region.

[0012] As a preferred improvement, the following steps are also included before step S2:

[0013] The acquired infrared images are subjected to noise reduction and enhancement processing, wherein homomorphic filtering is used for noise reduction and contrast-limited adaptive histogram equalization is used for enhancement.

[0014] As a preferred improvement, the process of determining hollow drum points using the neighborhood double-threshold image segmentation method specifically includes the following steps:

[0015] Step S21: For any pixel point a(i, j) in the sub-region, a neighborhood matrix with 2n×2n pixels is constructed with it as the center, and the grayscale values ​​of all pixels in the neighborhood matrix are collected to form a grayscale value set;

[0016] Step S22: Set the hollow point determination condition. If the neighborhood matrix of pixel point a(i, j) satisfies both of the following conditions, pixel point a(i, j) is determined to be a hollow point; otherwise, pixel point a(i, j) is determined to be a non-hollow point:

[0017] (1) The grayscale difference of pixel a(i,j) is not less than the grayscale threshold x. The grayscale difference represents the pixel value u of pixel a(i,j). a and the minimum value u in the gray value set min The difference, that is, u a -u min ≥x;

[0018] (2) In the neighborhood matrix of pixel a(i,j), the number of pixels whose grayscale difference is not less than the grayscale threshold x is N1 ≥ n × n;

[0019] In step S23 , the pixel values ​​of the hollow points are reset to 1, and the pixel values ​​of the non-hollow points are reset to 0, a binary image is output, and the hollow areas are obtained by removing the small areas through an opening operation.

[0020] As a preferred improvement, in the initial stage of hollowing determination, the grayscale threshold x is set to an initial value based on an empirical value; in the subsequent hollowing determination stage, the value of the grayscale threshold x is determined based on the grayscale difference between the minimum hollowing area and the background area, expressed as:

[0021] x=I 空鼓 -I 背景 ;

[0022] Where, I 空鼓 Indicates the average grayscale of the minimum hollow area; I 背景 Represents the average grayscale of the background area.

[0023] As a preferred improvement, the size of the neighborhood matrix 2n×2n is determined according to the size a×a of the inscribed square of the minimum hollow area:

[0024] ;

[0025] Where, L 、 B Represent the pixel length and width of the target object respectively; l 、 b They represent the actual length and width of the target object respectively; a represents the side length of the inscribed square of the minimum hollow area.

[0026] As a preferred improvement, the specific steps of step S3 are as follows:

[0027] Step S31: Count the number of non-zero pixels in the binary image k ;

[0028] Step S32: Calculate the actual area of ​​the area corresponding to the infrared image S , expressed as:

[0029] ;

[0030] Where, 、 Respectively represent the length and width of the area corresponding to the infrared image; H Indicates the distance between the lens and the area corresponding to the infrared image; P_L 、 P_W They represent the lens CMOS size; F represents the camera focal length;

[0031] Step S33, calculate the actual area of ​​the hollow area , the calculation process is expressed as:

[0032] ;

[0033] Where M and N are the resolutions of the infrared image.

[0034] The beneficial effects of the present invention are:

[0035] (1) The infrared image is divided into several sub-areas by the brick joints distributed in a grid pattern, and only the tile area is subsequently detected, which reduces the interference of brick joints in the infrared image on the quantitative detection of hollows;

[0036] (2) The neighborhood double-threshold image segmentation method combines the two parameters of minimum hollow size and minimum temperature difference between the hollow area and the background area to judge each pixel, solving the under-segmentation or over-segmentation problems of traditional image segmentation methods and improving the quantitative detection accuracy of hollows. DETAILED DESCRIPTION

[0037] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0038] This embodiment provides a method for detecting hollowing of exterior wall tiles based on thermal image sub-regions and dual threshold segmentation, comprising the following steps:

[0039] Step S1, collecting infrared images of exterior wall tiles.

[0040] The acquisition method involves using a drone equipped with a dual-light camera, which is positioned according to the drone's movement to capture images of exterior wall tiles at different locations. The captured infrared images require noise reduction and enhancement.

[0041] In this implementation, noise reduction is performed using homomorphic filtering. Homomorphic filtering is a frequency-domain image processing method that combines image frequency filtering with grayscale transformation to adjust the image's grayscale range, highlight image details, enhance contrast, and reduce uneven lighting.

[0042] After noise reduction, images may still have low contrast and blurred details. In such cases, image enhancement is necessary to increase contrast and detail, improve visual quality, and enhance the accuracy of subsequent research. This implementation utilizes contrast-limited adaptive histogram equalization (CLAHE) for image enhancement. The basic idea is to limit noise amplification and mitigate over-enhancement of local contrast by limiting the height of the local histogram.

[0043] Step S2, setting a strip mask to cover the brick joints in the infrared image, dividing the infrared image into several sub-areas by the criss-cross strip masks, and using the neighborhood double-threshold based image segmentation method to determine the hollow points in each sub-area, resetting the pixel values ​​of the hollow points to 1, and resetting the pixel values ​​of the non-hollow points to 0, outputting a binary image, and removing small areas by opening operation to obtain the hollow areas.

[0044] On the exterior walls of buildings, the brick joints form a regular, crisscross grid pattern. Therefore, strip masks can be installed to cover the gaps according to their width and spacing. Specifically, the width of the strip masks is the same as the width of the brick joints, and the spacing between two adjacent strip masks is the same as the spacing between two adjacent brick joints.

[0045] The setting of the strip mask can effectively shield the brick joints and only retain the image of the tile itself as the sub-area for detection. On the one hand, it reduces the calculation amount of hollow detection, and on the other hand, it can also avoid the interference of brick joints on the quantitative detection of hollows.

[0046] The process of determining hollow drum points using the neighborhood double-threshold image segmentation method specifically includes the following steps:

[0047] Step S21: For any pixel point a(i, j) in the sub-region, a neighborhood matrix with 2n×2n pixels is constructed with it as the center, and the grayscale values ​​of all pixels in the neighborhood matrix are collected to form a grayscale value set;

[0048] Step S22: Set the hollow point determination condition. If the neighborhood matrix of pixel point a(i, j) satisfies both of the following conditions, pixel point a(i, j) is determined to be a hollow point; otherwise, pixel point a(i, j) is determined to be a non-hollow point:

[0049] (1) The grayscale difference of pixel a(i,j) is not less than the grayscale threshold x. The grayscale difference represents the pixel value u of pixel a(i,j). a and the minimum value u in the gray value set min The difference, that is, u a -u min ≥x;

[0050] (2) In the neighborhood matrix of pixel a(i,j), the number of pixels whose grayscale difference is not less than the grayscale threshold x is N1 ≥ n × n;

[0051] In step S23 , the pixel values ​​of the hollow points are reset to 1, and the pixel values ​​of the non-hollow points are reset to 0, a binary image is output, and the hollow areas are obtained by removing the small areas through an opening operation.

[0052] Under sunlight stimulation, the normal area is in good contact with the wall base (such as cement mortar), and heat is quickly transferred to the base. Therefore, the surface temperature drops quickly, and it appears as a low-temperature area in the infrared image, and the grayscale value of the pixel is small. However, the hollow area has an air layer in the middle, and the thermal conductivity of the air is poor, so the heat cannot be quickly transferred, resulting in the surface temperature of the hollow area dropping slowly. Therefore, it will appear as a high-temperature area in the infrared image, and the grayscale value of the pixel is large. The present invention indicates the relative temperature of the current area by calculating the grayscale difference of the current pixel. Generally speaking, the pixel with the smallest grayscale value is the normal area. When the difference between the grayscale value of any pixel and the grayscale value of the pixel with the smallest grayscale value is greater than or equal to the grayscale threshold, it indicates that the relative temperature of the two exceeds the preset threshold, and it can be determined as a hollow point. By traversing all pixels, the hollowing determination of the entire sub-area is achieved.

[0053] For any hollow region with n×n pixels, to ensure that all points in the hollow region are judged as hollow points, all pixels in the hollow region must meet the above hollow judgment conditions (1) and (2). In order to ensure that the neighborhood matrix formed by any point in the hollow region can include the entire n×n hollow region, and at the same time the matrix must be as small as possible, the neighborhood matrix size is selected to be 2n×2n. The method provided by the present invention achieves full coverage judgment of the n×n hollow region when the neighborhood matrix size is as small as possible, avoids the possibility of missed detection and false detection, and improves the accuracy of hollow detection.

[0054] The value of the grayscale threshold x is determined according to the grayscale difference between the minimum hollow area and the background area, which can be expressed as:

[0055] x=I 空鼓 -I 背景 ;

[0056] Where, I 空鼓 Indicates the average grayscale of the minimum hollow area; I 背景 Represents the average grayscale of the background area.

[0057] In the initial stage of the determination, no hollow areas are identified in the sub-regions. At this time, the initial value of x is set according to the empirical value. As the hollow areas are gradually identified, the minimum hollow area is continuously updated, the average grayscale of the minimum hollow area is also constantly changing, and the grayscale threshold x is also iteratively updated accordingly, realizing dynamic adjustment of the grayscale threshold x and improving the accuracy of hollow determination. It can be understood that as the determination area increases, the minimum hollow area gradually stabilizes, the average grayscale of the minimum hollow area also tends to be stable, and the value of the grayscale threshold x also gradually tends to be stable.

[0058] The size of the neighborhood matrix 2n×2n is determined by the size a×a of the inscribed square of the minimum hollow area:

[0059] ;

[0060] Where, L 、 B Represent the pixel length and width of the target object respectively; l 、 b They represent the actual length and width of the target object respectively; a represents the side length of the inscribed square of the minimum hollow area.

[0061] For the minimum hollow area of ​​any shape, an inscribed square can be determined, and the inscribed square can roughly reflect the morphology of the hollow area. Therefore, the present invention uses the inscribed square to approximately replace the hollow area, ignores the area outside the inscribed square, and uses the inscribed square as the area for hollow judgment.

[0062] The image segmentation method based on neighborhood double threshold combines two parameters, minimum hollow size and minimum temperature difference between the hollow area and the background area, to judge each pixel in the sub-area, aiming to maximize the recognition of complete hollow areas while effectively reducing the erroneous segmentation of the background area.

[0063] Step S3: establishing a correlation between the actual area of ​​the hollow region and the image area, and calculating the actual area of ​​the hollow region.

[0064] The specific steps of step S3 are as follows:

[0065] Step S31: Count the number of non-zero pixels in the binary image k ;

[0066] Step S32: Calculate the actual area of ​​the area corresponding to the infrared image S , expressed as:

[0067] ;

[0068] Where, 、 Respectively represent the length and width of the area corresponding to the infrared image; H Indicates the distance between the lens and the area corresponding to the infrared image; P_L 、 P_W They represent the lens CMOS size; F represents the camera focal length;

[0069] Step S33, calculate the actual area of ​​the hollow area , the calculation process is expressed as:

[0070] ;

[0071] Where M and N are the resolutions of the infrared image.

[0072] The above describes the embodiments of the present invention, but the present invention is not limited to the above specific implementation methods. The above specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are protected by the present invention.

Claims

1. A method for detecting hollowing of exterior wall tiles based on thermal image sub-regions and dual threshold segmentation, characterized in that: The steps include: Step S1, collecting infrared images of exterior wall tiles; Step S2: Setting a strip mask to cover the brick joints in the infrared image, dividing the infrared image into several sub-regions by the crisscross strip mask, and using the neighborhood double-threshold image segmentation method to determine the hollow points in each sub-region, resetting the pixel values ​​of the hollow points to 1 and the pixel values ​​of the non-hollow points to 0, outputting a binary image, and removing small areas by opening operation to obtain the hollow areas; Step S3, establishing a correlation between the actual area of ​​the hollow region and the image area, and calculating the actual area of ​​the hollow region; The process of determining hollow drum points using the neighborhood double-threshold image segmentation method specifically includes the following steps: Step S21: For any pixel point a(i, j) in the sub-region, a neighborhood matrix with 2n×2n pixels is constructed with it as the center, and the grayscale values ​​of all pixels in the neighborhood matrix are collected to form a grayscale value set; Step S22: Set the hollow point determination condition. If the neighborhood matrix of pixel point a(i, j) satisfies both of the following conditions, pixel point a(i, j) is determined to be a hollow point; otherwise, pixel point a(i, j) is determined to be a non-hollow point: (1) The grayscale difference of pixel a(i,j) is not less than the grayscale threshold x. The grayscale difference represents the pixel value u of pixel a(i,j). a and the minimum value u in the gray value set min The difference, that is, u a -u min ≥x; (2) In the neighborhood matrix of pixel a(i,j), the number of pixels whose grayscale difference is not less than the grayscale threshold x is N1 ≥ n × n; Step S23, resetting the pixel value of the hollow point to 1, resetting the pixel value of the non-hollow point to 0, outputting a binary image, and removing small areas by opening operation to obtain the hollow area; The specific steps of step S3 are as follows: Step S31: Count the number of non-zero pixels in the binary image k ; Step S32: Calculate the actual area of ​​the area corresponding to the infrared image S , expressed as: ； Where, 、 Respectively represent the length and width of the area corresponding to the infrared image; H Indicates the distance between the lens and the area corresponding to the infrared image; P_L 、 P_W They represent the lens CMOS size; F represents the camera focal length; Step S33, calculate the actual area of ​​the hollow area , the calculation process is expressed as: ； Where M and N are the resolutions of the infrared image.

2. The exterior wall tile hollowing detection method based on thermal image sub-region and double threshold segmentation according to claim 1 is characterized in that: Before step S2, the following steps are also included: The acquired infrared images are subjected to noise reduction and enhancement processing, wherein homomorphic filtering is used for noise reduction and contrast-limited adaptive histogram equalization is used for enhancement.

3. The method for detecting hollowing of exterior wall tiles based on thermal image sub-regions and dual threshold segmentation according to claim 1 is characterized in that: In the initial stage of hollowing determination, the grayscale threshold x is set to an initial value based on empirical values; in the subsequent hollowing determination stage, the value of the grayscale threshold x is determined based on the grayscale difference between the minimum hollowing area and the background area, expressed as: x=I 空鼓 -I 背景 ; Where, I 空鼓 Indicates the average grayscale of the smallest hollow area; I 背景 Represents the average grayscale of the background area.

4. The method for detecting hollowing of exterior wall tiles based on thermal image sub-regions and dual threshold segmentation according to claim 1 is characterized in that: The size of the neighborhood matrix 2n×2n is determined by the size a×a of the inscribed square of the minimum hollow area: ； Where, L 、 B Represent the pixel length and width of the target object respectively; l 、 b They represent the actual length and width of the target object respectively; a represents the side length of the inscribed square of the minimum hollow area.

Images