Construction engineering quality detection and evaluation system based on machine vision

Abstract

The present invention relates to the field of image defect detection technology, specifically to a machine vision-based construction engineering quality inspection and assessment system, the system comprising an image acquisition module, a crack main axis identification module, a spatial encapsulation module, a structural overlap detection module, and a construction quality assessment module. In the present invention, a crack boundary layer with a spatial positioning function is constructed by scanning the image edge and marking the pixel position, pixel segments with strong continuity are screened to generate path lines, the continuity and representativeness of the crack trend are improved, rectangular blocks are encapsulated to enhance the organization and complete expression of cracks at the spatial level, coordinate comparison is performed with the structural stress area, a corresponding association is formed between the defect impact and the structural function, the spatial coverage relationship between the total amount of blocks and the number of overlaps is used to complete the determination of the impact of cracks on key areas, and a multi-dimensional building quality grade classification is established by combining crack morphology, position distribution and structural association information, thereby achieving a comprehensive assessment of building structural defects from the geometric to the structural level.

Description

Technical Field

[0001] The present invention relates to the technical field of image defect detection, and in particular to a construction engineering quality detection and evaluation system based on machine vision. Background Art

[0002] The field of image defect detection technology includes the use of computer vision and image processing methods to automatically identify and analyze defects on the surface of objects or inside structures. The core content is to accurately identify and locate defects such as cracks, deformations, holes, etc. in target objects by acquiring image data and performing processing methods such as feature extraction, image segmentation, edge detection and pattern recognition. It is widely used in industrial manufacturing, material analysis, safety testing, traffic monitoring, and building structure diagnosis. Its overall technical development relies on high-quality image acquisition methods, robust image processing algorithms and image understanding models, with the aim of achieving non-contact, efficient and automated detection of defects.

[0003] Among them, the construction project quality inspection and assessment system based on machine vision refers to the deployment of image acquisition equipment to capture images of the surface of the construction project structure, and the use of image analysis methods to identify, extract and classify defects such as concrete cracks, surface peeling, honeycombed surfaces, etc. presented in the image. It covers image acquisition, lighting control, image preprocessing, defect area extraction and defect type discrimination, and uses methods such as image edge feature enhancement, crack morphology feature extraction and regional grayscale distribution analysis to complete the identification and labeling of building surface defects, and complete the quality assessment level division according to the image geometric features of the defect area.

[0004] Existing construction project quality inspection and assessment processes rely on image feature extraction and regionalization of static image content. The recognition process primarily focuses on surface information within the image, lacking the ability to jointly represent the spatial distribution of the image and structure. This results in identification results that remain at the primary level of defect appearance. Defect recognition fails to consider directional trends and path connectivity, leading to issues such as identification breakage and morphological fragmentation when dealing with structural extension defects, making it difficult to capture the true spatial evolution of cracks. The lack of a spatial organization of encapsulated blocks makes it difficult to coordinate recognition results with key areas of the building structure, hindering effective data support for defect risk assessment in core areas subject to concentrated stresses. Quality assessment relies on single metrics such as image grayscale or area, ignoring the correlation between defects and structural layout. This can easily lead to assessment bias in complex spatial environments. For example, in areas of high stress concentration, even if continuous cracks appear, their high-risk nature is not recognized due to the lack of a mechanism to determine structural overlap, thus compromising the rigor and practicality of the overall quality assessment. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and propose a construction engineering quality detection and evaluation system based on machine vision.

[0006] In order to achieve the above-mentioned purpose, the present invention adopts the following technical solution: a construction engineering quality detection and evaluation system based on machine vision includes:

[0007] The image acquisition module acquires the building surface structure image, performs edge scanning on the non-background area in the image, extracts the boundary pixels of the crack area, and marks the position relationship of each pixel in the building image coordinate system to obtain the crack boundary coordinate layer;

[0008] The crack main axis identification module calculates the edge direction angle value of each crack boundary pixel point in the crack boundary coordinate layer according to the grayscale change direction of the neighboring pixels, sorts the pixel point sequence and classifies the direction angle change trend, constructs the main axis direction path line, and obtains the crack main axis direction line set;

[0009] The spatial encapsulation module divides the encapsulated rectangular areas aligned with the crack direction based on the spatial position relationship of the path lines in the crack main axis direction line set on the building image coordinates, extracts the boundary coordinates of each area, records the positioning block sequence number of each area, and obtains the crack path encapsulation block set;

[0010] The structural overlap detection module compares the spatial boundary coordinates of each block in the crack path encapsulation block set with the boundary data of the force-concentrated area in the building structure image, checks whether there are spatial segments with overlapping coordinates, summarizes the numbers of overlapping blocks, and obtains a structural block overlap record table.

[0011] As a further solution of the present invention, the crack boundary coordinate layer includes a boundary pixel distribution map, a pixel coordinate relationship table, and a crack image position index; the crack main axis direction line set includes a main axis path line sequence, a direction angle sequence set, and a direction line classification label; the crack path encapsulation block set includes encapsulated rectangular boundary data, path block positioning number, and block space direction information; the structural block overlap record table includes an overlapping block number table, a coincident space segment coordinate set, and a force area correspondence mapping result.

[0012] As a further solution of the present invention, the image acquisition module includes:

[0013] The image structure extraction submodule obtains the building surface structure image, performs grayscale conversion on the original image data, calculates and distinguishes the grayscale difference between the background area and the non-background area, selects the background grayscale threshold as the basis, classifies the area below the background grayscale threshold as the non-background area, and obtains the non-background pixel distribution map;

[0014] The edge pixel calibration submodule locates the edge contour of the non-background area in the image based on the non-background pixel distribution map, determines the gradient direction and grayscale change amplitude of adjacent pixels in the image, marks the crack boundary for pixels whose grayscale change amplitude is greater than the crack identification gradient threshold, identifies the coordinate information of all pixels within the boundary contour line, and obtains the crack edge pixel marking result;

[0015] The boundary coordinate generation submodule converts the positions of all pixel points within the crack boundary in the image matrix relative to the building image coordinate system based on the crack edge pixel marking results, establishes a horizontal and vertical coordinate system mapping relationship of the boundary points, and obtains the crack boundary coordinate layer.

[0016] As a further solution of the present invention, the crack main axis identification module includes:

[0017] The direction angle extraction submodule obtains the difference between the grayscale value of the adjacent pixels and the grayscale value of the target pixel based on each crack boundary pixel point in the crack boundary coordinate layer, calculates the distribution mean and change rate of the grayscale difference in the adjacent pixel direction, determines the direction of grayscale change, and performs gradient angle measurement and direction vector conversion on each crack boundary pixel point to obtain the edge direction angle value;

[0018] The sequence trend classification submodule analyzes the connectivity relationship of the crack boundary pixels based on the edge direction angle value, sorts the connected pixel sequences in adjacent order, calculates the direction angle change rate of each pixel and the difference between the direction angles of adjacent points, and classifies the direction angle change trend using the formula:

[0019] ;

[0020] Calculate and obtain the direction fluctuation trend value , filter the continuous pixel segments whose fluctuation amplitude is less than the direction fluctuation threshold to obtain a continuous stable direction sequence, where Indicates the The edge direction angle value of the crack boundary pixel point, Indicates the total number of pixels in the current connected pixel segment that participate in the calculation of the direction angle change. Indicates the The magnitude of the direction angle change in each direction segment, Indicates the The spatial spacing value of the pixels in the direction segment, Indicates the number of categorizable direction segments in the current crack. Indicates the The gradient value of the pixel point in the direction angle dimension, Indicates the number of participating points of the directional angle gradient in the current segment;

[0021] The path line construction submodule extracts the spatial coordinate information of all pixel segments in sequence according to the continuous stable direction sequence, constructs the path connection relationship based on the Euclidean distance and direction angle consistency value between adjacent coordinate points, and performs spatial path merging processing on the continuous segments to obtain the set of crack main axis direction lines.

[0022] As a further solution of the present invention, the space encapsulation module includes:

[0023] The path projection submodule extracts the spatial position point set of all path lines in the building image coordinate system based on the set of crack main axis direction lines. According to the endpoint coordinates, direction vector and the angle between the consecutive point pairs on each path line, the projection relationship between the direction line and the image reference coordinate axis is constructed to obtain the path space projection coordinate value.

[0024] The encapsulation area division submodule extracts the starting and ending coordinates of each path line according to the path space projection coordinate value and constructs the circumscribed rectangular area of ​​the line segment. It calculates the angle between the length direction of the circumscribed rectangle and the path direction, obtains the corrected angle value after direction alignment correction, and combines the corrected angle value with the boundary coordinates of the circumscribed area for joint calculation using the formula:

[0025] ;

[0026] Calculate the Crack Encapsulation area range value of the segment path , divide the package area according to the relative position of the package boundary in the image, establish the package rectangle with the same direction, and obtain the aligned package area set, where, 、 are the maximum and minimum horizontal coordinate values ​​of the path line endpoints, 、 are the maximum and minimum ordinate values, 、 Respectively The angle between the segment path line direction and the horizontal and vertical axes, is the total length of the path segments in the package area, Represents the total number of path segments;

[0027] The image block calibration submodule extracts the position block number and boundary information of each encapsulated rectangle in the building image coordinate system based on the aligned encapsulated area set, maps the number of each encapsulated block according to the image row and column index, records the coordinate index and path information corresponding to each encapsulated block, establishes a number list, and obtains the crack path encapsulated block set.

[0028] As a further solution of the present invention, the structure overlap detection module includes:

[0029] The boundary data extraction submodule extracts the upper left and lower right coordinate points of each encapsulated block based on the crack path encapsulation block set, calls the boundary coordinate set of the force concentration area in the building structure image, arranges all boundary coordinate values ​​in a coordinate matrix format, establishes a coordinate group after unifying the format, and obtains the structure and block boundary coordinate set;

[0030] The spatial comparison calculation submodule performs spatial coordinate intersection detection on the boundary area of ​​each encapsulated block and the building structure based on the structure and block boundary coordinate set, compares whether there is an overlapping relationship between the upper left and lower right coordinates, extracts the intersection boundary points, and calculates the ratio of the intersection area to the joint area using the formula:

[0031] ;

[0032] Calculate the The package block and Overlap judgment value of structure area , compare the judgment value with the spatial overlap range standard, identify all block numbers with overlap relationships, and obtain the overlapping block number set, where, represents the spatial overlap area, represents the area of ​​the spatial union, Indicates the The length of the line connecting the segment intersections, Indicates the length of the corresponding direction segment, Indicates the The coordinate difference, are the number of intersection segments and the number of difference points, respectively. Indicates the The crack path encapsulates the two-dimensional space boundary area of ​​the block in the image coordinate system. Indicates the The spatial boundary area of ​​the stress concentration area in the building structure image;

[0033] The overlapping number summary submodule extracts all the package numbers whose coordinates coincide with the force-bearing area according to the overlapping block number set and the package block sequence, injects the corresponding building image position index into the summary list, records the overlapping block number information, and obtains the structural block overlap record table.

[0034] As a further aspect of the present invention, the system further comprises a building quality assessment module;

[0035] The building quality assessment module determines whether cracks are concentrated in key areas of the building structure based on the spatial coverage relationship between the number of overlapping blocks in the structural block overlap record table and the total number of blocks in the crack path encapsulation block set, and classifies the cracks into different quality status categories to obtain a building structure quality assessment result;

[0036] The building structure quality assessment results include crack distribution concentration index, structural key area coverage, and quality grade classification results.

[0037] As a further solution of the present invention, the building quality assessment module includes:

[0038] The overlap ratio calculation submodule obtains the total number of encapsulated blocks and the number of blocks with overlapping relationships based on the structural block overlap record table and the crack path encapsulated block set, counts the distribution quantity values ​​of the two types of number sets in the building image coordinates, calculates the ratio of the overlapping quantity to the total encapsulated quantity, and obtains the spatial coverage value;

[0039] The distribution concentration determination submodule performs coordinate matching on the positions of all overlapping encapsulated blocks in the building image based on the spatial coverage value, and determines whether they are concentrated in the key areas by combining the key stress area numbers and position coordinate ranges identified in the building structure diagram, thereby obtaining key area concentration data.

[0040] The quality status classification submodule classifies the building structure quality status according to the key area concentration data and the degree of spatial overlap of the continuous block concentration value, generates each status level label and corresponding position index, and obtains the building structure quality assessment result.

[0041] Compared with the prior art, the advantages and positive effects of the present invention are:

[0042] In the present invention, a path line with stable and coherent characteristics is formed by directional analysis of the grayscale changes of crack boundary pixels and angular fluctuation constraints, thereby enhancing the connectivity and continuity of the crack direction description. The spatial position relationship of the path line in the building image coordinates is used to generate encapsulated rectangular blocks that closely correspond to the crack trend, thereby expanding the crack expression from linear paths to spatial blocks and improving the spatial organization expression capability. Each encapsulated block is then compared with the boundary information of the structural stress area at the coordinate level to form an overlapping expression between the defect impact area and the structural functional area, so that the defect identification result has the ability to determine the structural function association. Based on the quantitative coverage relationship between the total amount of encapsulated blocks and the overlap of the key structural areas, combined with the image spatial distribution, a quality grade classification mechanism is established to achieve a comprehensive expression of the crack spatial trend, geometric density and structural impact degree in the evaluation output, so that the defect expression moves from graphic information to structural semantics. At the same time, path continuity analysis and spatial encapsulation strategy are introduced to give the evaluation a multi-dimensional judgment capability, thereby improving the detection coverage depth, the richness of the evaluation level and the accuracy of the building quality judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 is a system flow chart of the present invention;

[0044] Figure 2 This is a flow chart of the image acquisition module of the present invention;

[0045] Figure 3 This is a flow chart of the crack main axis identification module of the present invention;

[0046] Figure 4 This is a flow chart of the space encapsulation module of the present invention;

[0047] Figure 5 This is a flow chart of the structure overlap detection module of the present invention;

[0048] Figure 6 This is a flow chart of the building quality assessment module of the present invention. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0050] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0051] See also Figure 1 , the construction engineering quality inspection and evaluation system based on machine vision includes:

[0052] The image acquisition module acquires images of building surface structures such as walls, beams, and columns, performs edge scanning on non-background areas in the image, extracts boundary pixels of the crack area, and marks the positional relationship of each pixel in the building image coordinate system to obtain a crack boundary coordinate layer.

[0053] The crack main axis identification module calculates the edge direction angle value of each crack boundary pixel point based on the grayscale change direction of the neighboring pixels in the crack boundary coordinate layer, sorts the pixel point sequence according to the crack connectivity relationship, classifies the direction angle change trend in the sequence, and selects continuous pixel segments with fluctuation amplitudes less than the fluctuation threshold to construct the main axis path line, thereby obtaining a set of crack main axis direction lines.

[0054] The spatial encapsulation module divides the crack path into encapsulated rectangular areas aligned with the crack direction based on the spatial position relationship of the path lines in the crack main axis direction line set on the building image coordinates. It extracts the boundary coordinates of each area and records the positioning block number of each area in the building image to obtain the crack path encapsulation block set.

[0055] The structural overlap detection module encapsulates the spatial boundary coordinates of each block in the crack path block set, compares them with the boundary data of the stress-concentrated area marked in the building structure image, checks whether there are spatial segments with overlapping coordinates, summarizes the numbers of all overlapping blocks, and obtains a structural block overlap record table;

[0056] The building quality assessment module determines whether cracks are concentrated in key areas of the building structure based on the spatial coverage relationship between the number of overlapping blocks in the structural block overlap record table and the total number of blocks in the crack path encapsulation block set, and divides the cracks into different quality status categories to obtain the building structure quality assessment results.

[0057] The crack boundary coordinate layer includes a boundary pixel distribution map, a pixel coordinate relationship table, and a crack image position index. The crack main axis direction line set includes a main axis path line sequence, a direction angle sequence set, and a direction line classification label. The crack path encapsulation block set includes encapsulated rectangular boundary data, path block positioning number, and block spatial direction information. The structural block overlap record table includes an overlapping block number table, a coincident spatial segment coordinate set, and a stress area correspondence mapping result. The building structure quality assessment results include a crack distribution concentration index, a structural key area coverage rate, and a quality grade classification result.

[0058] See also Figure 2 , the image acquisition module includes:

[0059] The image structure extraction submodule obtains the building surface structure image, performs grayscale conversion on the original image data, calculates and distinguishes the grayscale difference between the background area and the non-background area, selects the background grayscale threshold as the basis, classifies the area below the background grayscale threshold as the non-background area, and obtains the non-background pixel distribution map;

[0060] The process of acquiring building surface structure images usually uses drones or high-pixel image acquisition devices as preliminary equipment. During acquisition, it is necessary to ensure that the camera is perpendicular to the building surface and control the shooting distance within the range of 2 to 5 meters to reduce distortion errors. The acquired image must first be converted to a grayscale image to reduce the number of channels in the image processing process and improve the efficiency of boundary recognition. Grayscale conversion uses a weighted average method to process the RGB channel pixel values. The specific calculation method is:

[0061] The grayscale value of each pixel = 0.299×R+0.587×G+0.114×B. Taking the RGB values ​​of an image as (120, 100, 90) as an example, the grayscale value is: 0.299×120+0.587×100+0.114×90=104.84;

[0062] After the grayscale image is divided into regions, the pixel grayscale matrix is ​​read and the grayscale value of each pixel is compared with the set background grayscale threshold in turn. The background grayscale threshold is set based on the global grayscale histogram distribution of the image. The center value of the grayscale segment with the largest number of pixels is selected as the background reference grayscale value. This value is generally set in the range of 180~200 according to the image statistics. The grayscale value of 190 is selected as the background threshold (this setting refers to the difference in grayscale distribution between the background area and the non-background area in the reference image. In particular, the grayscale range of the concrete wall under natural light conditions is concentrated in the range of 180-210, while the grayscale corresponding to the cracks and holes in the wall is mostly in the range of 80-160, so 190 is selected as the background threshold). The background grayscale threshold is located in the median area of ​​the grayscale boundary segment mentioned above and can stably distinguish the concrete background area from the edge contour area. When the image brightness is increased or the shadow is enhanced, this value will linearly shift with the change of the overall pixel brightness mean value. It needs to be adjusted according to the grayscale mean μg of the current image. Usually, it is in the range of μg∈[170,200]. The threshold of 190 can meet the background filtering requirements of most images. The area with a grayscale value lower than 190 is identified as the non-background area. The smaller non-background dot matrix (less than 3×3 pixel blocks) is filtered out to avoid noise interference. Finally, a pixel distribution map of the non-background area is formed, which provides the pixel-based input for subsequent crack boundary identification, as shown in Table 1.

[0063] Table 1 Non-background area extraction sample table

[0064]

[0065] As shown in Table 1, pixels with grayscale values ​​less than 190 are identified as non-background areas. After performing this determination on all pixels in the image, a non-background pixel distribution map can be obtained.

[0066] The edge pixel calibration submodule locates the edge contour of the non-background area in the image based on the non-background pixel distribution map, determines the gradient direction and grayscale change amplitude of adjacent pixels in the image, marks the crack boundary for pixels whose grayscale change amplitude is greater than the crack identification gradient threshold, identifies the coordinate information of all pixels within the boundary contour line, and obtains the crack edge pixel marking results;

[0067] Based on the non-background pixel distribution map, its edge contour needs to be further identified. The process detects the grayscale change amplitude between pixels in an 8-neighborhood manner and calculates its directional gradient. The directional gradient is obtained by taking the maximum value of the grayscale difference between the current pixel and its right and bottom pixels as the edge gradient value. For example, if the grayscale value of point (25, 25) is 120, the right side (25, 26) is 160, and the bottom side (26, 25) is 110, then the gradient is max (|120-160|, |120-110|) = 40. The crack recognition gradient threshold is set to 35 (this value is obtained by analyzing multiple typical crack images. The gradient values ​​of the crack area are mostly between 30 and 60). ), so the pixel (25, 25) is marked as the crack boundary point. After executing this judgment process pixel by pixel, all candidate crack edge pixels are obtained. The candidate pixels are arranged in image coordinate order and their coordinate values ​​are recorded. Each coordinate point is stored in the form of (x, y), where x is the column coordinate and y is the row coordinate, forming preliminary crack contour marking data. On this basis, the part where the spacing between consecutive pixels exceeds 2 pixels is eliminated to exclude non-connected crack areas. In a certain 600×800 resolution image, a total of 3200 crack edge pixels are identified, accounting for 2.5% of the number of non-background pixels. The crack edge pixel marking result is thus established.

[0068] The boundary coordinate generation submodule converts the positions of all pixel points within the crack boundary in the image matrix relative to the building image coordinate system based on the crack edge pixel marking results, establishes the horizontal and vertical coordinate system mapping relationship of the boundary points, and obtains the crack boundary coordinate layer;

[0069] According to the pixel marking results of the crack edge, the (x, y) image coordinates recorded in the marking value are converted to the actual position coordinates in the building surface coordinate system. The pixel spacing during image acquisition and the pixel ratio of the building surface need to be introduced for conversion. Assuming that the image resolution is 800×600, the actual coverage width of the building is 4 meters and the height is 3 meters, the physical size corresponding to each pixel is the width: 4 / 800=0.005 meters, which is about 5 mm. All pixel coordinates of the crack edge are converted according to this ratio. For example, the pixel point (200, 300) is converted to (1.0 meters, 1.5 meters), which is the position of the crack in the building plane coordinate system. At the same time, a two-dimensional grid representation is established for all edge pixels to form a mesh coordinate set of the crack boundary line. This coordinate set is transferred to the building image registration system and anchored by comparing it with the actual building drawing projection view to complete the layer establishment. The result of this process is the crack boundary coordinate layer.

[0070] See also Figure 3 , the crack main axis identification module includes:

[0071] The direction angle extraction submodule obtains the difference between the grayscale value of the adjacent pixels and the grayscale value of the target pixel based on each crack boundary pixel in the crack boundary coordinate layer, calculates the distribution mean and change rate of the grayscale difference in the adjacent pixel direction, determines the direction of grayscale change, and performs gradient angle measurement and direction vector conversion on each crack boundary pixel to obtain the edge direction angle value;

[0072] Based on each crack boundary pixel in the crack boundary coordinate layer, the grayscale values ​​of the pixel to be identified and its 8 adjacent direction pixels in the image are obtained, and the direction is judged after performing the difference operation on each pair of grayscale values, where the pixel point is set as , and its eight adjacent pixels are recorded as , if the pixel The grayscale value of the adjacent points is 128, and the grayscale values ​​are 131, 129, 124, 120, 122, 130, 135, and 127 respectively. The direction change is mainly located in the direction area 2 to 5 with a grayscale decrease trend, indicating that there is an edge direction vector at this point. Then the gradient vector within the direction range is converted into an angle value, and the inverse tangent operation is used to convert the grayscale change direction into an angle range of 0° to 180°. In this example, the gradient change corresponding to the directions 2 to 5 is a negative value, and the angle after conversion is about 110°. This value is recorded as the direction angle of the pixel, and the above operations are performed on all crack boundary pixels in turn to form an angle map matrix. The direction angle value The range depends on the image resolution and sampling spacing, and is usually distributed between 5° and 175°. The data is saved in integer form. If there is discontinuous grayscale change in the transition area of ​​the image area boundary, the direction angle of the pixel with a difference less than 3 is temporarily determined to be invalid and eliminated in the next stage. In this process, the direction angle value range needs to be determined by interval judgment, and the grayscale difference threshold is set to ±5. If the grayscale difference of adjacent directions of a pixel point is less than this value, it is judged that the point has no direction angle information and is recorded as 0°. For example, the grayscale value of point B is 126, and the maximum difference value of its neighboring points is 3, then its direction angle is determined to be 0°. After completing the angle recognition of all boundary points, the edge direction angle value is generated.

[0073] The sequence trend classification submodule analyzes the connectivity of crack boundary pixels based on the edge direction angle value, sorts each connected pixel sequence in adjacent order, calculates the direction angle change rate of each pixel and the difference between the direction angles of adjacent points, and classifies the direction angle change trend using the formula:

[0074] ;

[0075] Calculate and obtain the direction fluctuation trend value , filter the continuous pixel segments whose fluctuation amplitude is less than the direction fluctuation threshold to obtain a continuous stable direction sequence, where Indicates the The edge direction angle value of the crack boundary pixel point, Indicates the total number of pixels in the current connected pixel segment that participate in the calculation of the direction angle change. Indicates the The magnitude of the direction angle change in each direction segment, Indicates the The spatial spacing value of the pixels in the direction segment, Indicates the number of categorizable direction segments in the current crack. Indicates the The gradient value of the pixel point in the direction angle dimension, Indicates the number of participating points of the directional angle gradient in the current segment;

[0076] According to the edge direction angle value, the connectivity relationship of the crack boundary pixels is called to extract the pixel sequence of each connected segment, and sort them according to the horizontal and vertical coordinates to form an ordered pixel chain. In a certain crack segment, assuming that the extracted pixel sequence is 15 points, each point has a corresponding angle value, such as 78°, 79°, 80°, 81°, 130°, 133°, 135°, 137°, 140°, 138°, 139°, 141°, 82°, 80°, 79°, then first calculate the direction angle difference between two adjacent points. If the angle difference in a certain segment does not exceed 5°, it is recorded as a stable direction segment. The angle difference in the segment is accumulated to calculate the total fluctuation amount, and multiplied by the pixel distance between the corresponding points, and the square sum operation is performed. Then, it is divided by the root of the sum of the absolute values ​​of the direction angle gradient values ​​in the segment to calculate the direction fluctuation trend value. The process uses the formula:

[0077] ;

[0078] Each segment of data is as follows:

[0079] : They are 78, 79, 80, 81, 130, 133, 135, 137, 140, 138, 139, 141, 82, 80, 79;

[0080] : are the angle differences of each segment, which are 1, 1, 50, 3, and 3 respectively;

[0081] : is the average pixel spacing within the segment, set to 1.2 pixels;

[0082] : Directional gradient values ​​such as 2, 2, 4, 3, 6, 5, 4, 4, 3, 2, 2, 3;

[0083] Substituting in:

[0084] ;

[0085] The calculated directional fluctuation trend value of this crack segment is 775.92, and the fluctuation threshold is set to 800 (the basis for this setting is that the directional angle variation range of common stable main axis segments in crack images is usually concentrated in an angle variation of no more than 8°, and the average gradient intensity is in the range of 40 to 60. Combined with the image sampling density (set to 1.2 pixels / mm in this image) and the crack length standard segment set between 12 and 20 pixels, the average directional fluctuation trend value calculated in multiple continuous segment samples is between 700 and 850. The fluctuation threshold of 800 is the 85th percentile corresponding value in 40 samples as the judgment standard, and this value shows a positive correlation with the increase of image resolution or the complexity of the crack. It needs to be dynamically adjusted in combination with pixel spacing, total crack length and angle variation). A value lower than this indicates that the segment is directional continuous and is recorded as a continuous stable directional sequence value.

[0086] The formula's operational logic is based on a comprehensive measurement of the pixel angular variation trend at the fracture boundary. The numerator reflects the cumulative fluctuation of the overall directional variation in the pixel sequence by summing the absolute values ​​of the angular differences between adjacent pixels. The numerator then multiplies the angular variation of each segment by its corresponding pixel spacing to represent the spatial weight of the directional fluctuation. This weight is then squared to emphasize the dominant role of large, continuous fluctuations in the overall trend. The denominator, on the other hand, sums the absolute values ​​of the directional gradient changes at all participating points and then takes the square root to form a normalized suppression term for the fluctuation trend. The logic is that if a fracture segment experiences a dramatic angular change accompanied by a high gradient strength, the influence of the directional fluctuation in the overall trend is weakened, thereby preventing a single point change from interfering with the overall trend. Ultimately, the formula constructs a fractional relationship using "total fluctuation amount + squared spatial influence weight" as the numerator and "total directional consistency gradient square root" as the denominator, forming a comprehensive evaluation index for the coherence of the main axis direction.

[0087] The path line construction submodule extracts the spatial coordinate information of all pixel segments according to the continuous stable direction sequence in sequence order, constructs the path connection relationship based on the Euclidean distance and direction angle consistency value between adjacent coordinate points, and performs spatial path merging processing on continuous segments to obtain the set of crack main axis direction lines;

[0088] According to the continuous stable direction sequence, the spatial coordinate value of each pixel point in each segment is selected. Assuming that a certain stable pixel direction sequence contains pixel points A (50, 80), B (51, 81), C (52, 82), D (53, 83), and E (54, 84), the distance between each point is extracted in turn, and the spatial distance between adjacent points is calculated using the Euclidean formula. After connecting them in sequence, a line segment set is formed. In this process, the direction angle consistency index is set. When the average difference in the direction angles of three adjacent points does not exceed the set value of 5°, they are merged into a path segment. Finally, multiple path segments are merged into a main axis set. The main axis set is stored in the form of an array, recording the starting point, end point and average path direction angle of each main axis. The structure example is as follows:

[0089] Table 2 Main axis direction line parameters

[0090]

[0091] As shown in Table 2, the path of principal axis number 1 consists of five consecutive coordinate points, with direction angles concentrated between 80° and 82° and a path length of 6.4 pixels. It is used to identify the main crack direction in subsequent crack structure identification and ultimately generate a set of crack principal axis direction lines.

[0092] See also Figure 4 , the space encapsulation module includes:

[0093] The path projection submodule extracts the spatial position point set of all path lines in the building image coordinate system based on the set of crack main axis direction lines. According to the endpoint coordinates, direction vector and the angle between the consecutive point pairs on each path line, the projection relationship between the direction line and the image reference coordinate axis is constructed to obtain the path space projection coordinate value.

[0094] Based on the spatial position relationship of the path lines in the crack main axis direction line set on the building image coordinates, first extract the coordinates of all endpoints of each path in the path line set and the horizontal and vertical distribution coordinate values ​​of the continuous points of the path. For example, a path line segment identified in an image is selected, whose starting coordinates are point A (150, 240) and the ending coordinates are point B (320, 410). Then the path line can be defined as a vector set, which consists of several pixel pairs. Then, according to the horizontal coordinate range of each path line, the vector set is obtained. And vertical coordinate range , calculate its projection boundary in the image, record each pixel point on the path line by scanning point by point, and record its horizontal and vertical index values ​​in the image matrix. During the processing, in order to ensure the accuracy of coordinate projection, take the building image with a pixel resolution of 0.25mm / px as an example. If the horizontal span of the path line is 680px, then its corresponding projection length is 170mm. At the same time, combined with the direction vector of each path line, such as the direction vector of the path line is , the angle between it and the horizontal axis of the building image can be calculated by inverse trigonometric function Finally, the angle is used to determine the projection relationship between the path direction and the image reference coordinates, extract the direction angle of each crack path line, and adjust the coordinate boundary range accordingly. At the same time, the path line number and its corresponding coordinate value are bound and encoded and output to obtain the path space projection coordinate value.

[0095] The encapsulation area division submodule extracts the coordinates of the beginning and end of each path line according to the path space projection coordinate value and constructs the circumscribed rectangular area of ​​the line segment. It calculates the angle between the length direction of the circumscribed rectangle and the path direction, obtains the corrected angle value after direction alignment correction, and combines the corrected angle value with the boundary coordinates of the circumscribed area for joint calculation using the formula:

[0096] ;

[0097] Calculate the Crack Encapsulation area range value of the segment path , divide the package area according to the relative position of the package boundary in the image, establish the package rectangle with the same direction, and obtain the aligned package area set, where, 、 are the maximum and minimum horizontal coordinate values ​​of the path line endpoints, 、 are the maximum and minimum ordinate values, 、 Respectively The angle between the segment path line direction and the horizontal and vertical axes, is the total length of the path segments in the package area, Represents the total number of path segments;

[0098] The relationship between cracks and paths is one of inclusion and composition. Each crack is composed of multiple path segments, and a path is the spatial representation of the crack. When extracting cracks from an image, the crack is segmented into several path segments. Each path segment has clear start and end coordinates and a direction angle, and these are used as the basis for constructing the encapsulation region.

[0099] According to the path space projection coordinate value, the coordinate boundary of the first and last endpoints of each path segment in the building image is extracted, the coordinate range of the circumscribed rectangle boundary required to encapsulate the path is calculated, and the angle alignment between the rectangle and the path line direction is performed. The angle correction method is used to reset the rectangle boundary direction. For example, the minimum and maximum horizontal coordinates of the path line in the image are 、 , the minimum and maximum ordinates are 、 , then its width and height are 170px and 170px respectively, combined with the angle between the path line and the horizontal axis , the vertical axis angle , the number of direction segments is 3, and the total path length is , substitute into the formula:

[0100] ;

[0101] The angle terms are , , take the cosine and sine values ​​respectively:

[0102] ;

[0103] ;

[0104] ;

[0105] Substitute the above data into:

[0106] ;

[0107] Table 3 Crack path line packaging parameters:

[0108]

[0109] As shown in Table 3, the package area range value is used to calibrate the comprehensive index of the package boundary, which is evaluated in combination with the direction projection angle, boundary coordinate range and total path length.

[0110] The operation logic of this formula is to comprehensively measure the spatial range and directional consistency of the path encapsulation area. Specifically, the formula It reflects the sum of the horizontal and vertical dimensions of the encapsulation rectangle in the image coordinate system, that is, the overall scale of the encapsulation boundary, which serves as the measurement basis for the degree of path space occupation; while the molecular part It means adding the cosine and sine values ​​of the angle between the path line direction and the coordinate axis to comprehensively evaluate the coordination degree of the biaxial distribution of the path direction in the image coordinate system, reflecting the matching degree between the path line direction and the image packaging direction. The greater the matching degree of the direction, the better the alignment of the packaging path. Used to normalize the encapsulation index of the entire path, is the total length of the path segments. Its square root is processed to suppress the weight influence of extreme length segments. The +1 operation avoids the abnormal situation of the denominator being 0, and at the same time serves to buffer the ratio amplification caused by length. The overall logic is based on the actual geometric dimensions of the path encapsulation area, supplemented by the contribution of the path direction component and the total length of the path. The obtained encapsulation area range value achieves a balanced representation between spatial scale, directional alignment and length control.

[0111] The image block calibration submodule extracts the position block number and boundary information of each encapsulated rectangle in the building image coordinate system based on the aligned encapsulated area set, maps the number of each encapsulated block according to the image row and column index, records the coordinate index and path information corresponding to each encapsulated block, establishes a number list, and obtains the crack path encapsulated block set;

[0112] According to the aligned encapsulated area set, the rectangular boundary coordinates of each encapsulated area are extracted, and the image block number of each encapsulated area is located according to its horizontal and vertical index values ​​in the building image matrix. The image partition range is numbered and mapped. If the image size is 1920×1080px and is divided into blocks of 320×270px, the entire image is divided into 24 blocks of 6×4. For example, if the boundary of a certain encapsulated rectangular area is in the horizontal coordinate range of 190–290 and the vertical coordinate range of 300–520, the block number is calculated as the first block on the horizontal axis and the second block on the vertical axis, and the block number is (1, 2). The corresponding number ID can be block number ID = 1×4+2=6. Similar mapping records are performed on all encapsulated areas. Finally, a building image block number index table corresponding to each encapsulated path is established, and a mapping relationship is established. The corresponding encapsulated path, boundary coordinates and block number are output, that is, the crack path encapsulated block set is obtained.

[0113] See also Figure 5 , the structural overlap detection module includes:

[0114] The boundary data extraction submodule encapsulates the block set based on the crack path, extracts the upper left and lower right coordinate points of each encapsulated block, calls the boundary coordinate set of the force concentration area in the building structure image, arranges all boundary coordinate values ​​in the coordinate matrix format, establishes a coordinate group after unifying the format, and obtains the structure and block boundary coordinate set;

[0115] Based on the crack path encapsulation block set, the coordinate points of the upper left corner and lower right corner of each encapsulated block are extracted, the coordinate values ​​in the direction of the image coordinate axis are normalized, and the pixel units are unified into millimeter units to facilitate the spatial overlap judgment with consistent scale. The boundary coordinates of the stress concentration area are extracted from the building structure image. The boundary information is expressed by linear annotations or regional frame lines in the structural drawing. In the example, the upper left coordinates of the encapsulated block are (120mm, 80mm), and the lower right coordinates are (180mm, 140mm). The upper left coordinates of the structural area boundary are (160mm, 100mm), and the lower right coordinates are (220mm, 160mm). The two sets of boundary values ​​are unified into arrays. and , detect whether there is an intersection relationship between the two sets of boundaries, and make an interval intersection judgment of the boundary range. First, determine whether there is an intersection in the x-axis direction, that is, the maximum left boundary is smaller than the minimum right boundary. If true, it means there is an overlapping area in the x direction, and the y direction is judged This is also true, so there is an intersection area between the two in space, and the intersection coordinates are recorded as (160, 100) to (180, 140). This area can be called by subsequent modules for overlapping area calculation, and finally the structure and block boundary coordinate set is established.

[0116] The spatial contrast calculation submodule performs spatial coordinate intersection detection on the boundary area of ​​each encapsulated block and the building structure based on the structure and block boundary coordinate set, compares whether there is an overlapping relationship between the upper left and lower right coordinates, extracts the intersection boundary points, and calculates the ratio of the intersection area to the joint area using the formula:

[0117] ;

[0118] Calculate the The package block and Overlap judgment value of structure area , compare the judgment value with the spatial overlap range standard, identify all block numbers with overlap relationships, and obtain the overlapping block number set, where, represents the spatial overlap area, represents the area of ​​the spatial union, Indicates the The length of the line connecting the segment intersections, Indicates the length of the corresponding direction segment, Indicates the The coordinate difference, are the number of intersection segments and the number of difference points, respectively. Indicates the The crack path encapsulates the two-dimensional space boundary area of ​​the block in the image coordinate system. Indicates the The spatial boundary area of ​​the stress concentration area in the building structure image;

[0119] According to the structure and block boundary coordinate sets, the intersection coordinates corresponding to the encapsulated block and the structure area are extracted, and the area calculation and joint area comparison are performed. The overlapping area is the product of the width and height of the intersection rectangle. In the above example, the width of the intersection rectangle is , the height is , so the overlapping area is The total area of ​​the package block is The structural area is , the joint area is , substitute the above values ​​into the formula, where the length of the line connecting the boundary intersections is , corresponding to the length of the direction segment , coordinate difference , after substituting:

[0120] ;

[0121] The result shows that the degree of overlap is about 0.5, and the ratio threshold is set at 0.3 in the overlap judgment standard (the basis for this setting is that the average proportion of the stress concentration area in the building structure image to the total pixel area of ​​the image is between 24.6% and 32.8%. In order to consider the maximum redundant overlap effect of the path encapsulation block under extreme overlap, the overlap rate is set higher than the threshold near the upper limit, which may have an interference effect on the structure. Therefore, 0.3 is used as the risk identification limit value. This value fluctuates with the expansion degree of the stress area boundary in the structure annotation map and the image resolution. When the boundary range expands or the resolution decreases, the overlapping area will increase, thereby increasing the overlap ratio. Therefore, the threshold should be adjusted synchronously according to the image physical scale standard and the structural area division granularity). If the overlap judgment value exceeds this standard, it is identified as a structural overlap block, then the current result meets the conditions and should be included in the overlap record range.

[0122] Table 4 Calculation table of overlap between package area and structure area

[0123]

[0124] Table 4 lists the actual calculation process and results of spatial overlap, verifying that the formula can be used to identify structural overlap areas.

[0125] The operational logic of the formula is to quantitatively measure the spatial relationship between the overlap of the crack encapsulation block and the structural area in the image coordinate system to comprehensively reflect the spatial overlap intensity and geometric feature coupling. The numerator of the formula consists of two terms. The first term is the overlapping area. , which directly reflects the spatial overlap range between the packaging area and the structural area. The second term is the sum of the products of the lengths of all overlapping boundary segments and their direction segments, reflecting the consistency and continuity of the overlapping contours along the direction, and reflecting the geometric similarity between the crack path and the structural boundary. The sum of the two gives the total strength index of the structural overlap. The first term in the denominator is the joint area of ​​the packaging area and the structural area. , used to normalize the overlap intensity. The second term is the square root sum of the absolute values ​​of the coordinate differences of all pixels in the overlapping area, which is intended to reflect the degree of geometric distortion in the boundary area. The square root operation enhances the sensitivity to small boundary changes, and finally forms a composite ratio with boundary overlap degree, geometric consistency and normalization control, which can present the overlap relationship of the encapsulated block to the structural area in multiple dimensions.

[0126] The overlapping number summary submodule extracts all package numbers whose coordinates coincide with the force-bearing area according to the overlapping block number set and the package block sequence, inserts the corresponding building image position index into the summary list, records the overlapping block number information, and obtains the structural block overlap record table;

[0127] According to the overlapping block number set, all block numbers that meet the overlap judgment value greater than the standard value (such as 0.3) are screened, their numbers, positions and image row and column index information are sorted, and an index mapping table is established. In the example, A1 that meets the conditions is used as a valid record item. Its upper left corner coordinates in the image are (120, 80). When the image resolution is 1mm / pixel, the corresponding row and column numbers are row 80 and column 120, so the record number is , the structural region number is , the position index is (120, 80), and the final output is a tabular structure block overlap record set, which can provide a spatial corresponding reference for structural response analysis or image risk detection, and finally obtain the structure block overlap record table.

[0128] See also Figure 6 , the building quality assessment modules include:

[0129] The overlap ratio calculation submodule obtains the total number of encapsulated blocks and the number of overlapping blocks based on the structural block overlap record table and the crack path encapsulation block set. It then counts the distribution values ​​of the two types of number sets in the building image coordinates, calculates the ratio of the overlapping number to the total encapsulation number, and obtains the spatial coverage value.

[0130] Based on the structural block overlap record table and the crack path encapsulation block set, first extract the marked block numbers in the structural block overlap record table and store them as a number array. , and extract all block numbers from the encapsulated block set to form the full set , count the numbers in the two sets accordingly, count the total number of overlapping numbers and the total number of encapsulated blocks, and obtain the overlap ratio = overlapping encapsulated blocks / total number of encapsulated blocks. If the total number of encapsulated blocks extracted in the actual project is 50, and 12 encapsulated blocks are detected to overlap with the structural blocks, then the overlap ratio = 12 ÷ 50 = 0.24. This is, then, for each overlapping block, extract its encapsulated area. The encapsulated area coordinates are based on the upper left vertex and the lower right vertex Expressed as follows, using the area formula Calculate the area of ​​each overlapping block, if The coordinates are (100, 120) to (180, 200), then A = (180-100) × (200-120) = 80 × 80 = 6400 pixels². At the same time, the corresponding area of ​​the entire set of encapsulated blocks is also calculated. For example, another block The area of ​​all overlapping blocks and the encapsulated full set is (50, 60) to (130, 140), with an area of ​​6400 pixels². The areas of all overlapping blocks and the encapsulated full set are ΣX and ΣY respectively. In the example, if ΣX = 76800 pixels² and ΣY = 256000 pixels², the area ratio is 76800 ÷ 256000 = 0.3. A spatial offset statistical set is constructed, and the center coordinates of the overlapping blocks are further identified on the structural diagram. The distance difference between the overlapping blocks and the geometric center point of the corresponding structural stress zone is calculated and statistically analyzed. All the above data are combined to perform a comprehensive comparison to obtain the spatial coverage of the building crack influence range. The process is shown in Table 5:

[0131] Table 5 Statistics of building encapsulation blocks

[0132]

[0133] As shown in Table 5, some numbers in the package block have spatial overlap. The spatial coverage value can be directly obtained by calculating the ratio of the overlapping area to the total package area and comparing the numbers.

[0134] The distribution concentration determination submodule matches the coordinates of all overlapping encapsulated blocks in the building image based on the spatial coverage value. It then combines the key stress area numbers and position coordinate ranges identified in the building structure diagram to determine whether they are concentrated in the key area and obtain the key area concentration data.

[0135] According to the spatial coverage value, each overlapping encapsulated block is positioned as a point set {( , ), ,…, }, and synchronously call the key structural area position set set in the building structure image { , ,…, }, by judging the ownership relationship of the coordinate points, check one by one whether the packaging area falls into the above-mentioned key structure area enclosing frame. For example, if the packaging block number The center coordinates are (140, 160), and the key area The boundary of is (120, 140) to (200, 220), then Those falling into the critical structural area are marked as "concentrated". In specific application scenarios, for example, for a five-story office building structure, the critical stress area is the intersection of the load-bearing column concentration position and the shear wall. In the third and fourth-story images, it is detected that the crack encapsulation blocks are densely concentrated at the boundary of this area. The system further counts the number of concentrated encapsulation areas and their proportion to the total number of overlapping blocks. If there are 10 overlapping blocks, 7 of which are located in the critical structural area, the concentration value is 0.7. According to the actual project settings, a concentration ≥ 0.6 is judged as "concentrated distribution", and a concentration < 0.3 is judged as "non-concentrated". This classification standard is set with reference to the importance level of the building structure and the safety assessment index. A mapping set is formed between the concentration value and the key area number, and then the key area concentration data is obtained.

[0136] The quality status classification submodule classifies the building structure quality status according to the key area concentration data, based on the continuous block concentration value and the degree of spatial overlap, generates each status level label and corresponding position index, and obtains the building structure quality assessment results;

[0137] According to the concentration data of key areas, the value is compared with the preset building quality status grade benchmark value, and the building structure quality grade is set to be divided into five categories, and the corresponding concentration value ranges are: Class I [0, 0.2), Class II [0.2, 0.4), Class III [0.4, 0.6), Class IV [0.6, 0.8), Class V [0.8, 1]. For example, if the concentration calculated above is 0.7, the building structure assessment grade is Class IV, and its position range and distribution number are marked. A two-dimensional mapping structure is constructed in the system output to show the image. The image position coordinates are used as indexes and the quality level is used as the value to form a spatial distribution quality assessment matrix. In addition, for those cracks with unclear boundary overlap but located in important functional areas of the structure (such as beam end interfaces, stair connections, etc.), additional judgments are also required. If the crack sequence has continuous overlapping blocks exceeding the set lower limit (such as ≥3), it can be forcibly classified as concentrated to ensure the comprehensiveness of the assessment judgment. The final classification level and its corresponding encapsulated block number, spatial coordinates, concentration degree and other elements are recorded in the output structure table as the basis for subsequent structural safety analysis to obtain the building structure quality assessment results.

[0138] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. The construction engineering quality detection and evaluation system based on machine vision is characterized by: The system comprises: The image acquisition module acquires the building surface structure image, performs edge scanning on the non-background area in the image, extracts the boundary pixels of the crack area, and marks the position relationship of each pixel in the building image coordinate system to obtain the crack boundary coordinate layer; The crack main axis identification module calculates the edge direction angle value of each crack boundary pixel point in the crack boundary coordinate layer according to the grayscale change direction of the neighboring pixels, sorts the pixel point sequence and classifies the direction angle change trend, constructs the main axis direction path line, and obtains the crack main axis direction line set; The spatial encapsulation module divides the encapsulated rectangular areas aligned with the crack direction based on the spatial position relationship of the path lines in the crack main axis direction line set on the building image coordinates, extracts the boundary coordinates of each area, records the positioning block sequence number of each area, and obtains the crack path encapsulation block set; The structural overlap detection module compares the spatial boundary coordinates of each block in the crack path encapsulation block set with the boundary data of the force-concentrated area in the building structure image, checks whether there are spatial segments with overlapping coordinates, summarizes the numbers of overlapping blocks, and obtains a structural block overlap record table.

2. The machine vision-based construction engineering quality detection and evaluation system according to claim 1 is characterized in that: The crack boundary coordinate layer includes a boundary pixel distribution map, a pixel coordinate relationship table, and a crack image position index; the crack main axis direction line set includes a main axis path line sequence, a direction angle sequence set, and a direction line classification label; the crack path encapsulation block set includes encapsulated rectangular boundary data, path block positioning number, and block space direction information; the structural block overlap record table includes an overlapping block number table, a coincident space segment coordinate set, and a force area correspondence mapping result.

3. The machine vision-based construction engineering quality detection and evaluation system according to claim 1 is characterized in that: The image acquisition module includes: The image structure extraction submodule obtains the building surface structure image, performs grayscale conversion on the original image data, calculates and distinguishes the grayscale difference between the background area and the non-background area, selects the background grayscale threshold as the basis, classifies the area below the background grayscale threshold as the non-background area, and obtains the non-background pixel distribution map; The edge pixel calibration submodule locates the edge contour of the non-background area in the image based on the non-background pixel distribution map, determines the gradient direction and grayscale change amplitude of adjacent pixels in the image, marks the crack boundary for pixels whose grayscale change amplitude is greater than the crack identification gradient threshold, identifies the coordinate information of all pixels within the boundary contour line, and obtains the crack edge pixel marking result; The boundary coordinate generation submodule converts the positions of all pixel points within the crack boundary in the image matrix relative to the building image coordinate system based on the crack edge pixel marking results, establishes a horizontal and vertical coordinate system mapping relationship of the boundary points, and obtains the crack boundary coordinate layer.

4. The machine vision-based construction engineering quality detection and evaluation system according to claim 1 is characterized in that: The crack main axis identification module includes: The direction angle extraction submodule obtains the difference between the grayscale value of the adjacent pixels and the grayscale value of the target pixel based on each crack boundary pixel point in the crack boundary coordinate layer, calculates the distribution mean and change rate of the grayscale difference in the adjacent pixel direction, determines the direction of grayscale change, and performs gradient angle measurement and direction vector conversion on each crack boundary pixel point to obtain the edge direction angle value; The sequence trend classification submodule analyzes the connectivity relationship of the crack boundary pixels based on the edge direction angle value, sorts the connected pixel sequences in adjacent order, calculates the direction angle change rate of each pixel and the difference between the direction angles of adjacent points, and classifies the direction angle change trend using the formula: ； Calculate and obtain the direction fluctuation trend value , filter the continuous pixel segments whose fluctuation amplitude is less than the direction fluctuation threshold to obtain a continuous stable direction sequence, where Indicates the The edge direction angle value of the crack boundary pixel point, Indicates the total number of pixels in the current connected pixel segment that participate in the calculation of the direction angle change. Indicates the The magnitude of the direction angle change in each direction segment, Indicates the The spatial spacing value of the pixels in the direction segment, Indicates the number of categorizable direction segments in the current crack. Indicates the The gradient value of the pixel point in the direction angle dimension, Indicates the number of participating points of the directional angle gradient in the current segment; The path line construction submodule extracts the spatial coordinate information of all pixel segments in sequence according to the continuous stable direction sequence, constructs the path connection relationship based on the Euclidean distance and direction angle consistency value between adjacent coordinate points, and performs spatial path merging processing on the continuous segments to obtain the set of crack main axis direction lines.

5. The machine vision-based construction engineering quality detection and evaluation system according to claim 1 is characterized in that: The space encapsulation module includes: The path projection submodule extracts the spatial position point set of all path lines in the building image coordinate system based on the set of crack main axis direction lines. According to the endpoint coordinates, direction vector and the angle between the consecutive point pairs on each path line, the projection relationship between the direction line and the image reference coordinate axis is constructed to obtain the path space projection coordinate value. The encapsulation area division submodule extracts the starting and ending coordinates of each path line according to the path space projection coordinate value and constructs the circumscribed rectangular area of ​​the line segment. It calculates the angle between the length direction of the circumscribed rectangle and the path direction, obtains the corrected angle value after direction alignment correction, and combines the corrected angle value with the boundary coordinates of the circumscribed area for joint calculation using the formula: ； Calculate the Crack Encapsulation area range value of the segment path , divide the package area according to the relative position of the package boundary in the image, establish the package rectangle with the same direction, and obtain the aligned package area set, where, 、 are the maximum and minimum horizontal coordinate values ​​of the path line endpoints, 、 are the maximum and minimum ordinate values, 、 Respectively The angle between the segment path line direction and the horizontal and vertical axes, is the total length of the path segments in the package area, Represents the total number of path segments; The image block calibration submodule extracts the position block number and boundary information of each encapsulated rectangle in the building image coordinate system based on the aligned encapsulated area set, maps the number of each encapsulated block according to the image row and column index, records the coordinate index and path information corresponding to each encapsulated block, establishes a number list, and obtains the crack path encapsulated block set.

6. The machine vision-based construction engineering quality detection and evaluation system according to claim 1 is characterized in that: The structure overlap detection module includes: The boundary data extraction submodule extracts the upper left and lower right coordinate points of each encapsulated block based on the crack path encapsulation block set, calls the boundary coordinate set of the force concentration area in the building structure image, arranges all boundary coordinate values ​​in a coordinate matrix format, establishes a coordinate group after unifying the format, and obtains the structure and block boundary coordinate set; The spatial comparison calculation submodule performs spatial coordinate intersection detection on the boundary area of ​​each encapsulated block and the building structure based on the structure and block boundary coordinate set, compares whether there is an overlapping relationship between the upper left and lower right coordinates, extracts the intersection boundary points, and calculates the ratio of the intersection area to the joint area using the formula: ； Calculate the The package block and Overlap judgment value of structure area , compare the judgment value with the spatial overlap range standard, identify all block numbers with overlap relationships, and obtain the overlapping block number set, where, represents the spatial overlap area, represents the area of ​​the spatial union, Indicates the The length of the line connecting the segment intersections, Indicates the length of the corresponding direction segment, Indicates the The coordinate difference, are the number of intersection segments and the number of difference points, respectively. Indicates the The crack path encapsulates the two-dimensional space boundary area of ​​the block in the image coordinate system. Indicates the The spatial boundary area of ​​the stress concentration area in the building structure image; The overlapping number summary submodule extracts all the package numbers whose coordinates coincide with the force-bearing area according to the overlapping block number set and the package block sequence, injects the corresponding building image position index into the summary list, records the overlapping block number information, and obtains the structural block overlap record table.

7. The machine vision-based construction engineering quality detection and evaluation system according to claim 1 is characterized in that: The system also includes a building quality assessment module; The building quality assessment module determines whether cracks are concentrated in key areas of the building structure based on the spatial coverage relationship between the number of overlapping blocks in the structural block overlap record table and the total number of blocks in the crack path encapsulation block set, and classifies the cracks into different quality status categories to obtain a building structure quality assessment result; The building structure quality assessment results include crack distribution concentration index, structural key area coverage, and quality grade classification results.

8. The machine vision-based construction engineering quality detection and evaluation system according to claim 7 is characterized in that: The building quality assessment module includes: The overlap ratio calculation submodule obtains the total number of encapsulated blocks and the number of blocks with overlapping relationships based on the structural block overlap record table and the crack path encapsulated block set, counts the distribution quantity values ​​of the two types of number sets in the building image coordinates, calculates the ratio of the overlapping quantity to the total encapsulated quantity, and obtains the spatial coverage value; The distribution concentration determination submodule performs coordinate matching on the positions of all overlapping encapsulated blocks in the building image based on the spatial coverage value, and determines whether they are concentrated in the key areas by combining the key stress area numbers and position coordinate ranges identified in the building structure diagram, thereby obtaining key area concentration data. The quality status classification submodule classifies the building structure quality status according to the key area concentration data and the degree of spatial overlap of the continuous block concentration value, generates each status level label and corresponding position index, and obtains the building structure quality assessment result.

Images