Old building nondestructive testing method and system based on space-time intelligence

Abstract

The application discloses a method and system for nondestructive testing of old buildings based on space-time intelligence, and relates to the technical field of nondestructive testing, comprising: collecting a target mirror band weight image of a wall surface area to be tested; performing skeleton extraction and mask expansion processing on the orthogonal polarization image of the wall surface area to be tested to obtain a gray joint band mask image; performing pixel-by-pixel processing on the gray joint band mask image based on the target mirror band weight image to obtain a condensation joint mirror band evidence image and a mirror band influence trajectory atlas; obtaining a crack evolution evidence atlas based on posture data and sight distance data of the wall surface area to be tested; and generating a nondestructive testing result based on the mirror band influence trajectory atlas and the crack evolution evidence atlas; and the application improves the engineering application value of nondestructive testing of old buildings.

Description

Technical Field

[0001] This invention relates to the field of nondestructive testing technology, and in particular to a method and system for nondestructive testing of old buildings based on spatiotemporal intelligence. Background Technology

[0002] With the advancement of urban renewal, many older buildings still remain in use, especially those with brick-concrete and brick-wood structures. These buildings commonly suffer from problems such as aging mortar joints, surface weathering, water seepage, and localized cracks. In practical engineering, wall inspections of these buildings typically need to be conducted without damaging the structure. Therefore, non-destructive testing methods based on image acquisition and data analysis are increasingly being adopted. For example, multiple wall images are acquired and compared at different times to obtain information on crack development. Under the influence of changes in environmental humidity, condensation easily forms in the mortar joint areas of the wall surface, causing changes in surface reflection and thus affecting the brightness distribution and texture of the image.

[0003] In existing technologies, most image-based non-destructive testing methods directly perform differential analysis or edge extraction on wall images acquired at different times, without fully considering the impact of condensation on the reflection changes in the mortar joint area. In the presence of condensation, the mortar joint area will produce obvious brightness anomalies, which will change the original texture structure in the image and cause confusion between the cracked area and the non-cracked area in the image, thus making the subsequent crack identification results unstable. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies where grout seams can produce significant brightness anomalies, altering the original texture structure of images. This invention proposes a spatiotemporal intelligent non-destructive testing method and system for old buildings.

[0005] To address the problems existing in the prior art, the present invention adopts the following technical solution: Spatiotemporal intelligence-based non-destructive testing methods for old buildings include: S1. Acquire the target mirror zone weighted image of the wall area to be tested; S2. Perform skeleton extraction and mask expansion processing on the orthogonal polarization image of the wall area to be tested to obtain the mortar joint strip mask image; S3. Based on the target mirror strip weight image, the ash seam strip mask image is processed pixel by pixel to obtain the condensation ash seam mirror strip evidence image and the mirror strip influence trajectory map. S4. Based on the attitude data and line-of-sight data of the wall area to be tested, obtain the crack evolution evidence map; S5. Based on the mirror band influence trajectory map and crack evolution evidence map, generate non-destructive testing results.

[0006] Preferably, acquiring a target mirror-weighted image includes: A smart sensor is used to collect data on the same area of ​​the wall surface to be tested, resulting in parallel polarization images and orthogonal polarization images. Determine the lens band intensity image based on parallel polarization images and orthogonal polarization images; Linear normalization is performed on the mirror band intensity image to obtain the mirror band weight image; Smoothing filtering is performed on the mirror-weighted image to obtain the target mirror-weighted image.

[0007] Preferably, obtaining the mortar seam strip mask image includes: Determine the diffuse texture base map based on orthogonal polarization images; Perform a morphological opening operation on the diffuse texture base map to obtain the opening result image; Based on the diffuse texture base map and the opening operation result image, the grout seam line enhancement image is determined; Skeletonization processing is performed on the enhanced linear image of the grout joint to obtain the grout joint skeleton image; Dilation processing is performed on the mortar joint skeleton image to obtain a mortar joint strip mask image.

[0008] Preferably, the obtained condensation slit lens evidence image includes: The target mirror strip weight image is multiplied pixel by pixel with the gray seam strip mask image to obtain the condensation gray seam mirror strip evidence image; A robust texture replacement image is obtained by performing local median filtering on the diffuse texture base map. Linear normalization was performed on the condensation slit mirror image to obtain a fused weighted map; Based on the fusion weight image, the diffuse texture base map and the robust texture replacement image are fused pixel by pixel to obtain the de-mirror texture image; The target mirror band weighted image is multiplied pixel by pixel with the gray seam band mask image to obtain the condensation gray seam mirror band evidence image.

[0009] Preferably, the mirror band influence trajectory map is obtained, including: A coordinate transformation was performed on the condensation ash seam evidence images from multiple acquisition times to obtain a sequence of seam evidence images in a unified coordinate system. By performing time-dimensional overlay on the sequence of mirror-related evidence images in a unified coordinate system, a mirror-related influence trajectory map is obtained.

[0010] Preferably, the obtained crack evolution evidence map includes: Based on the fusion weight image, the diffuse texture base map and the robust texture replacement image are fused pixel by pixel to obtain the de-mirror texture image; Collect attitude data and line-of-sight data of the wall area to be tested; Based on pose data and line-of-sight data, cross-temporal phase registration is performed on the de-mirror textured images to obtain a sequence of de-mirror textured images in a unified coordinate system. Linear response extraction was performed on the de-textured image sequence under a unified coordinate system to obtain the crack response image sequence corresponding to each acquisition time. The crack response image sequence is overlaid over time to obtain a crack evolution evidence map.

[0011] Preferably, generating nondestructive testing results includes: The crack evolution evidence map and the mirror band influence trajectory map were aligned to the same coordinate system to obtain the aligned crack evolution evidence map and the aligned mirror band influence trajectory map. A non-destructive testing display layer is generated based on the aligned crack evolution evidence map; A non-destructive testing auxiliary layer is generated based on the aligned mirror band influence trajectory map; The non-destructive testing display layer and the non-destructive testing auxiliary layer are overlaid to create a non-destructive testing result image. The non-destructive testing result image is used as the non-destructive testing result of the wall area to be tested.

[0012] Preferably, the non-destructive testing result image is obtained, including: The dimensions of the non-destructive testing display layer and the non-destructive testing auxiliary layer are unified to obtain non-destructive testing display layer and non-destructive testing auxiliary layer with unified dimensions. Geometric registration is performed on a uniformly sized nondestructive testing display layer and a uniformly sized nondestructive testing auxiliary layer to obtain a registered nondestructive testing display layer and a registered nondestructive testing auxiliary layer. Perform transparency mapping on the registered non-destructive testing auxiliary layer to obtain a transparency layer; The registered nondestructive testing auxiliary layer and the registered nondestructive testing display layer are merged pixel-level based on the transparency layer to obtain the nondestructive testing result image.

[0013] To address the aforementioned problems, this invention also provides a spatiotemporal intelligence-based non-destructive testing system for old buildings, the system comprising: The mirror strip weight acquisition module acquires the target mirror strip weight image of the wall area to be tested. The grout joint mask generation module performs skeleton extraction and mask expansion processing on the orthogonal polarization image of the wall area to be tested, and obtains a grout joint strip mask image. The mirror band atlas generation module processes the gray seam strip mask image pixel by pixel based on the target mirror band weight image to obtain the condensation gray seam mirror band evidence image and the mirror band influence trajectory atlas. The crack atlas generation module generates a crack evolution evidence atlas based on the attitude data and line-of-sight data of the wall area to be tested. The test result generation module generates non-destructive testing results based on the mirror band influence trajectory map and the crack evolution evidence map.

[0014] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention introduces polarization imaging to obtain a weighted image of the mirror band and combines it with a slit strip mask image to extract the slit region pixel by pixel. This allows for the explicit separation of the abnormal reflection region caused by condensation at the image level, enabling a quantitative expression of the slit reflection changes. This effectively avoids the interference of condensation on the original texture information and improves the reliability of subsequent analysis data. 2. Based on the obtained image of the condensation crack mirror band evidence, this invention constructs a mirror band influence trajectory map and a crack evolution evidence map, and performs time superposition analysis from the dimensions of reflection change and structural change, respectively. This makes environmental factors and structural changes independent in data expression, thereby avoiding misjudging brightness changes caused by the environment as crack development and improving the stability and accuracy of crack identification results. 3. This invention further visualizes and overlays crack evolution information and mirror band influence information by unifying coordinate system alignment, layer generation, and transparency fusion processing. This allows the detection results to simultaneously include structural change information and environmental influence information, enhancing the interpretability and intuitiveness of the detection results, thereby improving the engineering application value of non-destructive testing of old buildings. Attached Figure Description

[0015] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating a spatiotemporal intelligence-based non-destructive testing method for old buildings, as provided in an embodiment of the present invention. Detailed Implementation

[0016] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0017] This embodiment provides a non-destructive testing method for old buildings based on spatiotemporal intelligence. (See also...) Figure 1 Specifically, including: S1. Acquire the target mirror zone weighted image of the wall area to be tested; In an embodiment of the present invention, acquiring a target mirror band weighted image includes: A smart sensor is used to collect data on the same area of ​​the wall surface to be tested, resulting in parallel polarization images and orthogonal polarization images. Determine the lens band intensity image based on parallel polarization images and orthogonal polarization images; Linear normalization is performed on the mirror band intensity image to obtain the mirror band weight image; Smoothing filtering is performed on the mirror-weighted image to obtain the target mirror-weighted image.

[0018] The smart sensor is fixed at the acquisition position aligned with the wall area to be tested. The acquisition field of view is set to cover the wall area to be tested and kept unchanged during acquisition. Under the same exposure and gain conditions, a first image with the polarization direction consistent with the set polarization direction is acquired and recorded as a parallel polarization image. Then, without changing the acquisition position and imaging parameters, the polarization direction is switched to be perpendicular to the set polarization direction and a second image is acquired and recorded as an orthogonal polarization image. Pixel grid consistency processing is performed on the parallel polarization image and the orthogonal polarization image. The pixel grid consistency processing includes converting the two images to the same resolution and the same pixel format, setting the pixel coordinate origin of the two images to the same position, aligning the number of pixel rows and columns of the two images, and outputting the aligned parallel polarization image and the aligned orthogonal polarization image. For each pixel location of the wall area to be tested, the pixel grayscale value of the parallel polarization image and the pixel grayscale value of the orthogonal polarization image are read. A pixel-by-pixel subtraction operation is performed according to the pixel position correspondence to obtain the lens intensity value at that pixel location. The lens intensity values ​​of all pixel locations are written into the output image in the original pixel grid order to obtain the lens intensity image. To avoid negative values ​​in the lens intensity image affecting subsequent processing, all pixel intensity values ​​of the lens intensity image are shifted to a non-negative range. This overall shift is achieved by calculating the minimum pixel intensity value of the lens intensity image and adding its negative value to each pixel intensity value of the lens intensity image. The shifted result is used as the input for subsequent linear normalization processing. In the translated mirror intensity image, all pixel positions are traversed to determine the minimum and maximum pixel intensity values. For each pixel position, the minimum pixel intensity value is subtracted from the pixel intensity value, and then divided by the difference between the maximum and minimum pixel intensity values ​​to obtain the normalized pixel weight value. All normalized pixel weight values ​​are written into the output image according to the original pixel grid to obtain the mirror weight image. The steps for performing smoothing filtering on the mirror weight image to obtain the target mirror weight image are as follows: a neighborhood window in pixels is determined and the window is slid across the mirror weight image in row and column order, with the center of the window covering each pixel position of the mirror weight image in turn. When the center of the window covers a certain pixel position, all pixel weight values ​​within the coverage area of ​​the window are read, the arithmetic mean of the pixel weight values ​​is calculated, and the arithmetic mean is written into the pixel position corresponding to the center of the window in the output image to complete the writing of the smoothing result for that pixel position. Repeat the sliding window reading and arithmetic average writing process for all pixel positions of the target mirror weight image to obtain a smoothed target mirror weight image. For pixel positions that are close to the image boundary and cause the window coverage to exceed the image boundary, a boundary extension method is used to form a complete window. The boundary extension method is to map the window position that exceeds the boundary to the nearest boundary pixel position and read the weight value of the boundary pixel to participate in the arithmetic average calculation, thereby ensuring that the target mirror weight image has a corresponding output result in the entire pixel range.

[0019] S2. Perform skeleton extraction and mask expansion processing on the orthogonal polarization image of the wall area to be tested to obtain the mortar joint strip mask image; In an embodiment of the present invention, obtaining a strip-shaped mask image of the grout seam includes: Determine the diffuse texture base map based on orthogonal polarization images; Perform a morphological opening operation on the diffuse texture base map to obtain the opening result image; Based on the diffuse texture base map and the opening operation result image, the grout seam line enhancement image is determined; Skeletonization processing is performed on the enhanced linear image of the grout joint to obtain the grout joint skeleton image; Dilation processing is performed on the mortar joint skeleton image to obtain a mortar joint strip mask image.

[0020] Read the orthogonal polarization image corresponding to the wall area to be tested, convert the orthogonal polarization image into a single-channel grayscale image while keeping the number of pixel rows and columns unchanged, and output the converted grayscale image as a diffuse texture base map. The shape of the structural element is determined to be linear, and multiple orientations are set for the linear structural element. The erosion operation is performed on the diffuse texture base map one by one using the linear structural element in each direction to obtain the erosion result in each direction. Then, the dilation operation is performed on the erosion result in each direction to obtain the opening operation result in each direction. The opening operation result in each direction is formed by taking the maximum value according to the pixel position to form the opening operation result image and output it. Read the pixel grayscale value of the diffuse texture base map at each pixel position and read the corresponding pixel grayscale value of the opening operation result image. Perform pixel-by-pixel subtraction operation according to the pixel position correspondence to obtain the enhanced grayscale value at that pixel position. Write all the enhanced grayscale values ​​into the output image according to the original pixel grid to obtain the gray seam line enhancement image. The grout seam enhancement image is converted into a binary image. The binary conversion uses the median gray level of the entire image as the segmentation benchmark, and pixels with a gray level higher than the median are set as foreground pixels and the rest are set as background pixels. Then, iterative thinning processing is performed on the binary image. In each iteration, all pixels are scanned and boundary foreground pixels that satisfy the endpoint preservation and connectivity preservation conditions are deleted according to the eight-neighbor connectivity relationship until no pixel deletion occurs in a complete scan. The grout seam skeleton image is then output. Starting with the foreground pixels of the mortar joint skeleton image, a circular neighborhood structure element is constructed in units of pixels. The radius of the structure element is calculated from the viewing distance data corresponding to the acquisition time. The calculation steps are as follows: first, determine the actual length of the wall corresponding to each pixel based on the imaging resolution of the smart sensor; then, determine the number of pixels corresponding to the radius of the structure element based on the actual width of the wall to be covered by the desired mask, and take the smallest integer not less than the number of pixels. The circular neighborhood structure element is used to perform dilation operation on the mortar joint skeleton image to generate a mortar joint strip mask image and output it.

[0021] S3. Based on the target mirror strip weight image, the ash seam strip mask image is processed pixel by pixel to obtain the condensation ash seam mirror strip evidence image and the mirror strip influence trajectory map. In an embodiment of the present invention, obtaining an image of condensation grout seam evidence includes: The target mirror strip weight image is multiplied pixel by pixel with the gray seam strip mask image to obtain the condensation gray seam mirror strip evidence image; A robust texture replacement image is obtained by performing local median filtering on the diffuse texture base map. Linear normalization was performed on the condensation slit mirror image to obtain a fused weighted map; Based on the fusion weight image, the diffuse texture base map and the robust texture replacement image are fused pixel by pixel to obtain the de-mirror texture image; The target mirror band weighted image is multiplied pixel by pixel with the gray seam band mask image to obtain the condensation gray seam mirror band evidence image.

[0022] The system reads the target lens weight image and the slit mask image, and performs size unification and pixel coordinate alignment processing on both. The size unification processing includes converting the two images to have the same number of pixel rows and columns and the same pixel format. The pixel coordinate alignment processing includes using the same pixel origin and the same row and column index order on the two images. At each pixel position, the system reads the pixel gray value of the target lens weight image and the pixel gray value of the slit mask image, performs a pixel-by-pixel multiplication operation according to the pixel position correspondence, writes the multiplication result to the corresponding pixel position of the output image, and after traversing all pixel positions, obtains the condensation slit lens evidence image and outputs it. A diffuse texture base map is read and a square sliding window in pixels is determined. The side length of the sliding window is determined by the number of rows and columns of pixels in the diffuse texture base map and is a minimum odd number of not less than three to ensure that the window has a center pixel position. The center of the window covers each pixel position of the diffuse texture base map in row and column order. When the center of the window covers a certain pixel position, all pixel gray values ​​within the coverage area of ​​the window are read and sorted by value. The median of the sorted values ​​is taken as the replacement gray value of the pixel position and written to the corresponding pixel position of the output image. After the window sliding and median writing are completed for all pixel positions, a robust texture replacement image is obtained and output. For pixel positions that are close to the image boundary and cause the window coverage area to exceed the image boundary, a boundary extension method is used to supplement the window. The boundary extension method is to map the position that exceeds the boundary to the nearest boundary pixel and read the gray value of the boundary pixel to participate in the sorting and take the median. By traversing all pixel positions of the condensation gray slit mirror evidence image, the minimum and maximum pixel gray values ​​of the image are determined. For each pixel position, the minimum pixel gray value is subtracted from the pixel gray value and divided by the difference between the maximum and minimum pixel gray values ​​to obtain the normalized pixel weight value, which is then written into the output image to form a fused weight map and output. At the same time, the minimum and maximum pixel gray values ​​are used as normalization parameters for this linear normalization process. Read the fusion weight map, diffuse texture base map, and robust texture replacement image. Perform size unification and pixel coordinate alignment to make the three correspond point by point on the pixel grid. At each pixel position, read the pixel weight value of the fusion weight map, the pixel gray value of the diffuse texture base map, and the pixel gray value of the robust texture replacement image. Perform linear weighted summation on the two gray values ​​according to the pixel weight value and write the weighted result to the corresponding pixel position of the output image. After traversing all pixel positions, the image with de-sanding and texture is obtained and output. The steps for obtaining the condensation grout lens band evidence image by multiplying the target lens band weight image and the grout band mask image pixel by pixel are the same as the aforementioned pixel-by-pixel multiplication steps. This is used to output the corresponding condensation grout lens band evidence image while generating the lens band texture image so as to form the lens band influence trajectory map in the future.

[0023] In an embodiment of the present invention, obtaining the mirror band influence trajectory map includes: A coordinate transformation was performed on the condensation ash seam evidence images from multiple acquisition times to obtain a sequence of seam evidence images in a unified coordinate system. By performing time-dimensional overlay on the sequence of mirror-related evidence images in a unified coordinate system, a mirror-related influence trajectory map is obtained.

[0024] Each image of the condensation seam lens evidence is read sequentially according to the acquisition time. At the same time, the attitude data and viewing distance data corresponding to the acquisition time are read and a one-to-one correspondence is established. The earliest image of the condensation seam lens evidence corresponding to the acquisition time is selected as the reference image and the pixel coordinates of the reference image are defined as a unified coordinate system. For each image of the condensation seam lens evidence other than the reference image, the geometric transformation parameters from the reference image are calculated. The geometric transformation parameters are determined by the attitude data and viewing distance data corresponding to the image. The determination process is to convert the change in viewing direction represented by the attitude data into the rotation amount of the image plane, convert the change in scale represented by the viewing distance data into the scaling amount of the image plane, and combine the rotation amount and the scaling amount into a planar similarity transformation. The planar similarity transformation is applied to each pixel coordinate of the condensation gray seam mirror evidence image. The transformed pixel coordinates are mapped to the pixel coordinate system of the reference image and resampling is performed. The resampling uses bilinear interpolation to calculate the gray value of the target pixel. The interpolation calculation is obtained by weighted summation of the gray values ​​of the four neighboring pixels into which the transformed coordinates fall according to the distance. For pixel positions that exceed the boundary of the reference image after transformation, a boundary clipping method is used to process them. The boundary clipping method is to output only pixels that fall within the effective pixel range of the reference image and keep the pixel row and column indices continuous. The above geometric transformation and resampling process is repeated for all the condensation gray seam mirror strip evidence images at all acquisition times to obtain a mirror strip evidence image sequence in a unified coordinate system and store it in the order of acquisition time. The sequence of mirror belt evidence images in a unified coordinate system is read sequentially according to the acquisition time. An accumulator image with the same size as the reference image is initialized, and all its pixel grayscale values ​​are set to zero. Each mirror belt evidence image in the sequence is accumulated pixel by pixel into the accumulator image. The accumulation method is to read the pixel grayscale value of the current mirror belt evidence image at each pixel position, add it to the pixel grayscale value of the same position in the accumulator image, and then write it back to the accumulator image. After completing the pixel-by-pixel accumulation of all mirror belt evidence images, the accumulator image is obtained. In order to make the accumulator image have comparable grayscale scales under different acquisition times, the grayscale value of each pixel in the accumulator image is divided by the number of images in the mirror belt evidence image sequence to obtain a time-dimensional average superimposed image. The time-dimensional average superimposed image is output as a mirror belt influence trajectory map.

[0025] S4. Based on the attitude data and line-of-sight data of the wall area to be tested, obtain the crack evolution evidence map; In an embodiment of the present invention, a crack evolution evidence map is obtained, including: Based on the fusion weight image, the diffuse texture base map and the robust texture replacement image are fused pixel by pixel to obtain the de-mirror texture image; Collect attitude data and line-of-sight data of the wall area to be tested; Based on pose data and line-of-sight data, cross-temporal phase registration is performed on the de-mirror textured images to obtain a sequence of de-mirror textured images in a unified coordinate system. Linear response extraction was performed on the de-textured image sequence under a unified coordinate system to obtain the crack response image sequence corresponding to each acquisition time. The crack response image sequence is overlaid over time to obtain a crack evolution evidence map.

[0026] Read the fused weighted image, diffuse texture base image, and robust texture replacement image. Perform size unification processing to ensure that the three images have the same number of pixel rows and columns and the same pixel format. Perform pixel coordinate alignment processing to ensure that the three images adopt the same pixel origin and the same row and column index order. At each pixel position, read the pixel weight value of the fused weighted image, the pixel gray value of the diffuse texture base image, and the pixel gray value of the robust texture replacement image. Calculate the weighted fused gray value according to the pixel weight value. The weighted fused gray value is equal to the product of the pixel weight value and the pixel gray value of the robust texture replacement image, plus the product of the complement of the pixel weight value and the pixel gray value of the diffuse texture base image. The complement of the pixel weight value is equal to one minus the pixel weight value. Write the weighted fused gray value to the corresponding pixel position of the output image. After traversing all pixel positions, obtain the de-textured image and output it. In the same acquisition of parallel polarization image and orthogonal polarization image, the attitude data output by attitude measurement unit and the line-of-sight data output by distance measurement unit are recorded synchronously. The attitude data and line-of-sight data are then correlated with the de-textured image acquired in this acquisition and stored according to the acquisition time. The image with de-textured textures is read sequentially from each acquisition time, along with the corresponding pose and viewing distance data. The image with de-textured textures acquired at the earliest acquisition time is selected as the reference image, and the pixel coordinates of the reference image are defined as a unified coordinate system. For each image with de-textured textures other than the reference image, the planar rotation amount relative to the reference image is determined based on its pose data, and the planar scaling amount relative to the reference image is determined based on its viewing distance data. The rotation amount and scaling amount are combined into a planar similarity transformation. The planar similarity transformation is applied to the pixel coordinates of the image with de-textured textures, and resampling is performed to generate a registered image in the unified coordinate system. The resampling uses bilinear interpolation, and boundary clipping is performed on pixel positions that exceed the boundary of the reference image. The resulting sequence of image with de-textured textures in the unified coordinate system is then stored sequentially from the acquisition time. For each registered, de-textured image in the sequence, the horizontal and vertical gradients are calculated to obtain a gradient magnitude image. The gradient magnitude image at each pixel position is obtained by summing the squares of the horizontal and vertical gradients and taking the square root. Directional enhancement processing is performed on the gradient magnitude image to highlight the slender linear structure. Directional enhancement processing is performed by using multi-directional linear structure elements to perform morphological top-hat operations on the gradient magnitude image and taking the maximum value of the result in each direction according to the pixel position to obtain the linear response image. The linear response images corresponding to each acquisition time are output in sequence as a crack response image sequence. Initialize a cumulative image with the same size as the reference image and set its pixel grayscale value to zero. Read the crack response images one by one in the order of acquisition time and accumulate them pixel by pixel into the cumulative image. After the accumulation of all crack response images is completed, divide the grayscale value of each pixel of the cumulative image by the number of crack response images to obtain the time-dimensional average superimposed image. Output the time-dimensional average superimposed image as a crack evolution evidence map.

[0027] S5. Based on the mirror band influence trajectory map and crack evolution evidence map, generate non-destructive testing results.

[0028] In embodiments of the present invention, generating nondestructive testing results includes: The crack evolution evidence map and the mirror band influence trajectory map were aligned to the same coordinate system to obtain the aligned crack evolution evidence map and the aligned mirror band influence trajectory map. A non-destructive testing display layer is generated based on the aligned crack evolution evidence map; A non-destructive testing auxiliary layer is generated based on the aligned mirror band influence trajectory map; The non-destructive testing display layer and the non-destructive testing auxiliary layer are overlaid to create a non-destructive testing result image. Read the crack evolution evidence map and the mirror band influence trajectory map, and convert the two maps to the same pixel format and the same number of pixel rows and columns. Select the crack evolution evidence map as the reference map and define its pixel coordinates as a unified coordinate system. Extract registration control point sets from the crack evolution evidence map and the mirror band influence trajectory map respectively. The extraction of registration control points is achieved by corner detection method. The corner detection method is to traverse the position of each pixel in the map and calculate the gray-level change matrix in the neighborhood of the pixel. Generate corner response map based on the response value of gray-level change matrix and sort it from large to small response value. Select corner pixels with non-repeating spatial positions as control points until traversal is completed, and obtain the control point set of crack evolution evidence map and mirror band influence trajectory map. Control point matching is performed on the two sets of control points. The control point matching is achieved by the correlation matching method based on local gray-scale blocks. The correlation matching method is to extract local gray-scale blocks of the same size in the corresponding map with each control point as the center. The normalized cross-correlation value is calculated for any local gray-scale block of the crack evolution evidence map and any local gray-scale block of the mirror zone influence trajectory map, and a similarity matrix is ​​formed. For each control point of the crack evolution evidence map, the control point of the mirror zone influence trajectory map with the largest cross-correlation value in the similarity matrix is ​​selected as the matching point and the matching pair is recorded to obtain the set of control point matching pairs. Based on the control point matching pair set estimation, the planar geometric transformation from the mirror band influence trajectory map to the crack evolution evidence map is performed. The planar geometric transformation adopts affine transformation, and the affine transformation parameters are solved by the least squares method. The solution process is to substitute the coordinates of each pair of matching points into the affine transformation equation to establish a system of linear equations and perform least squares to obtain the affine transformation matrix. The affine transformation matrix is ​​applied to the coordinates of each pixel in the mirror band influence trajectory map and resampling is performed to generate the aligned mirror band influence trajectory map. The resampling adopts bilinear interpolation and performs boundary clipping on the pixel positions that exceed the boundary of the reference map. The output is the aligned crack evolution evidence map and the aligned mirror band influence trajectory map. The aligned crack evolution evidence map is subjected to grayscale stretching to cover the entire display grayscale range. The grayscale stretching process is completed by traversing the entire image to obtain the minimum and maximum grayscale values ​​and generating the stretched grayscale value for each pixel by linear mapping. The stretched grayscale map is then output as a non-destructive testing display layer. The aligned mirror band influence trajectory map is subjected to grayscale normalization to obtain the transparency mapping base map. The grayscale normalization process is completed by traversing the entire image to obtain the minimum and maximum grayscale values ​​and generating normalized grayscale values ​​for each pixel by linear mapping. While keeping the pixel grid unchanged, the normalized grayscale image is output as a non-destructive testing auxiliary layer. This non-destructive testing auxiliary layer is then used to overlay the non-destructive testing display layer to generate the non-destructive testing result map.

[0029] In an embodiment of the present invention, a nondestructive testing result image is obtained, including: The dimensions of the non-destructive testing display layer and the non-destructive testing auxiliary layer are unified to obtain non-destructive testing display layer and non-destructive testing auxiliary layer with unified dimensions. Geometric registration is performed on a uniformly sized nondestructive testing display layer and a uniformly sized nondestructive testing auxiliary layer to obtain a registered nondestructive testing display layer and a registered nondestructive testing auxiliary layer. Perform transparency mapping on the registered non-destructive testing auxiliary layer to obtain a transparency layer; The registered nondestructive testing auxiliary layer and the registered nondestructive testing display layer are merged pixel-level based on the transparency layer to obtain the nondestructive testing result image.

[0030] The process reads the non-destructive testing display layer and the non-destructive testing auxiliary layer, obtaining the number of pixel rows and columns for each layer. The layer with the larger number of pixel rows and columns is selected as the target size. For layers with a smaller number of pixel rows or columns than the target size, resampling is performed to achieve the target size. The resampling uses bilinear interpolation, and the grayscale value is calculated pixel-by-pixel according to the target size. The interpolation calculation is obtained by reading the grayscale values ​​of the four neighboring pixels corresponding to the target pixel coordinates in the input layer and summing them according to the coordinate distance. For layers with a larger number of pixel rows or columns than the target size, cropping is performed to achieve the target size. The cropping method is to use the layer center as the cropping center and cut out a continuous pixel area according to the target size. The output non-destructive testing display layer and non-destructive testing auxiliary layer with uniform size are then executed. Using a uniformly sized nondestructive testing (NDT) display layer as a reference layer and defining its pixel coordinates as the registration coordinate system, corner point sets are extracted from both the uniformly sized NDT display layer and the uniformly sized NDT auxiliary layer. Corner point extraction is achieved by calculating the grayscale changes of each pixel's neighborhood, generating a corner point response map, and then selecting the corner points according to their response values. Matching is performed on the two sets of corner points. Matching is achieved by extracting local grayscale blocks of the same size from each corner point, calculating normalized cross-correlation values ​​to form a similarity matrix, and then selecting the corresponding corner point pairs with the highest cross-correlation values ​​to obtain the set of matched point pairs. Based on the set of matched point pairs, the affine transformation matrix from the uniformly sized NDT auxiliary layer to the uniformly sized NDT display layer is estimated. The affine transformation matrix is ​​obtained by substituting the coordinates of the matched point pairs into the affine transformation equations to establish a system of linear equations and solving them using the least squares method. An affine transformation matrix is ​​applied to the pixel coordinates of a uniformly sized non-destructive testing auxiliary layer, and resampling is performed to generate a registered non-destructive testing auxiliary layer. The resampling uses bilinear interpolation and performs boundary clipping on pixel positions that exceed the boundary of the reference layer. The registered non-destructive testing display layer and the registered non-destructive testing auxiliary layer are output. The minimum and maximum gray values ​​are obtained by traversing all pixel positions of the registered non-destructive testing auxiliary layer. For each pixel, the gray value is mapped to the transparency value by linear normalization and written to the corresponding pixel position of the transparency layer. The transparency value is equal to the pixel gray value minus the minimum gray value and divided by the difference between the maximum and minimum gray values. The transparency layer is then output. The algorithm reads the transparency layer, the registered non-destructive testing display layer, and the registered non-destructive testing auxiliary layer. It reads the transparency value at each pixel position according to the pixel location correspondence and reads the pixel grayscale values ​​of both layers. Based on the transparency value, it performs a linear weighted summation of the pixel grayscale values ​​of the two layers to obtain a blended grayscale value. The blended grayscale value is equal to the product of the transparency value and the pixel grayscale value of the non-destructive testing auxiliary layer, plus the product of the complement of the transparency value and the pixel grayscale value of the non-destructive testing display layer. The complement of the transparency value is equal to one minus the transparency value. The blended grayscale value is written to the corresponding pixel position in the output image. After traversing all pixel positions, the non-destructive testing result image is output.

[0031] The non-destructive testing result image is used as the non-destructive testing result of the wall area to be tested.

[0032] An embodiment of the present invention also provides a non-destructive testing system for old buildings based on spatiotemporal intelligence.

[0033] In this embodiment, the functions of each module / unit are as follows: The mirror strip weight acquisition module acquires the target mirror strip weight image of the wall area to be tested. The grout joint mask generation module performs skeleton extraction and mask expansion processing on the orthogonal polarization image of the wall area to be tested, and obtains a grout joint strip mask image. The mirror band atlas generation module processes the gray seam strip mask image pixel by pixel based on the target mirror band weight image to obtain the condensation gray seam mirror band evidence image and the mirror band influence trajectory atlas. The crack atlas generation module generates a crack evolution evidence atlas based on the attitude data and line-of-sight data of the wall area to be tested. The test result generation module generates non-destructive testing results based on the mirror band influence trajectory map and the crack evolution evidence map.

[0034] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A non-destructive testing method for old buildings based on spatiotemporal intelligence, characterized in that, Includes the following steps: S1. Acquire the target mirror zone weighted image of the wall area to be tested; S2. Perform skeleton extraction and mask expansion processing on the orthogonal polarization image of the wall area to be tested to obtain the mortar joint strip mask image; S3. Based on the target mirror strip weight image, the ash seam strip mask image is processed pixel by pixel to obtain the condensation ash seam mirror strip evidence image and the mirror strip influence trajectory map. S4. Based on the attitude data and line-of-sight data of the wall area to be tested, a crack evolution evidence map is obtained; S5. Based on the mirror band influence trajectory map and crack evolution evidence map, generate non-destructive testing results.

2. The non-destructive testing method for old buildings based on spatiotemporal intelligence according to claim 1, characterized in that, Acquire a weighted image of the target mirror band, including: A smart sensor is used to collect data on the same area of ​​the wall surface to be tested, resulting in parallel polarization images and orthogonal polarization images. Determine the lens band intensity image based on parallel polarization images and orthogonal polarization images; Linear normalization is performed on the mirror band intensity image to obtain the mirror band weight image; Smoothing filtering is performed on the mirror-weighted image to obtain the target mirror-weighted image.

3. The non-destructive testing method for old buildings based on spatiotemporal intelligence according to claim 2, characterized in that, Obtain the mortar seam strip mask image, including: Determine the diffuse texture base map based on orthogonal polarization images; Perform a morphological opening operation on the diffuse texture base map to obtain the opening result image; Based on the diffuse texture base map and the opening operation result image, the grout seam line enhancement image is determined; Skeletonization processing is performed on the enhanced linear image of the grout joint to obtain the grout joint skeleton image; Dilation processing is performed on the mortar joint skeleton image to obtain a mortar joint strip mask image.

4. The non-destructive testing method for old buildings based on spatiotemporal intelligence according to claim 3, characterized in that, Images of condensation slits obtained as evidence were included: The target mirror strip weight image is multiplied pixel by pixel with the gray seam strip mask image to obtain the condensation gray seam mirror strip evidence image; A robust texture replacement image is obtained by performing local median filtering on the diffuse texture base map. Linear normalization was performed on the condensation slit mirror image to obtain a fused weighted map; Based on the fusion weight image, the diffuse texture base map and the robust texture replacement image are fused pixel by pixel to obtain the de-mirror texture image; The target mirror band weighted image is multiplied pixel by pixel with the gray seam band mask image to obtain the condensation gray seam mirror band evidence image.

5. The non-destructive testing method for old buildings based on spatiotemporal intelligence according to claim 1, characterized in that, The influence trajectory map of the mirror band was obtained, including: A coordinate transformation was performed on the condensation ash seam evidence images from multiple acquisition times to obtain a sequence of seam evidence images in a unified coordinate system. By performing time-dimensional overlay on the sequence of mirror-related evidence images in a unified coordinate system, a mirror-related influence trajectory map is obtained.

6. The non-destructive testing method for old buildings based on spatiotemporal intelligence according to claim 4, characterized in that, The evidence map of crack evolution was obtained, including: Based on the fusion weight image, the diffuse texture base map and the robust texture replacement image are fused pixel by pixel to obtain the de-mirror texture image; Collect attitude data and line-of-sight data of the wall area to be tested; Based on pose data and line-of-sight data, cross-temporal phase registration is performed on the de-mirror textured images to obtain a sequence of de-mirror textured images in a unified coordinate system. Linear response extraction was performed on the de-textured image sequence under a unified coordinate system to obtain the crack response image sequence corresponding to each acquisition time. The crack response image sequence is overlaid over time to obtain a crack evolution evidence map.

7. The non-destructive testing method for old buildings based on spatiotemporal intelligence according to claim 1, characterized in that, Generate non-destructive testing results, including: The crack evolution evidence map and the mirror band influence trajectory map were aligned to the same coordinate system to obtain the aligned crack evolution evidence map and the aligned mirror band influence trajectory map. A non-destructive testing display layer is generated based on the aligned crack evolution evidence map; A non-destructive testing auxiliary layer is generated based on the aligned mirror band influence trajectory map; The non-destructive testing display layer and the non-destructive testing auxiliary layer are overlaid to create a non-destructive testing result image. The non-destructive testing result image is used as the non-destructive testing result of the wall area to be tested.

8. The non-destructive testing method for old buildings based on spatiotemporal intelligence according to claim 7, characterized in that, The non-destructive testing results are shown in the image, including: The dimensions of the non-destructive testing display layer and the non-destructive testing auxiliary layer are unified to obtain non-destructive testing display layer and non-destructive testing auxiliary layer with unified dimensions; Geometric registration is performed on a uniformly sized nondestructive testing display layer and a uniformly sized nondestructive testing auxiliary layer to obtain a registered nondestructive testing display layer and a registered nondestructive testing auxiliary layer. Perform transparency mapping on the registered non-destructive testing auxiliary layer to obtain a transparency layer; The registered nondestructive testing auxiliary layer and the registered nondestructive testing display layer are merged pixel-level based on the transparency layer to obtain the nondestructive testing result image.

9. A system applied to the spatiotemporal intelligence-based nondestructive testing method for old buildings as described in any one of claims 1-8, characterized in that, The system includes: The mirror strip weight acquisition module acquires the target mirror strip weight image of the wall area to be tested. The grout joint mask generation module performs skeleton extraction and mask expansion processing on the orthogonal polarization image of the wall area to be tested, and obtains a grout joint strip mask image. The mirror band atlas generation module processes the gray seam strip mask image pixel by pixel based on the target mirror band weight image to obtain the condensation gray seam mirror band evidence image and the mirror band influence trajectory atlas. The crack atlas generation module generates a crack evolution evidence atlas based on the attitude data and line-of-sight data of the wall area to be tested. The test result generation module generates non-destructive testing results based on the mirror band influence trajectory map and the crack evolution evidence map.

Images