Construction quality evaluation method based on prefabricated component surface pore area distribution analysis

Abstract

The present application relates to the technical field of image data processing, and especially relates to a construction quality evaluation method based on precast component surface pore area distribution analysis, which comprises the following steps: obtaining a precast component gray image; expanding an initial neighborhood window of each pixel point according to a preset step length, and determining a target neighborhood window when a termination condition of expansion is met; quantifying the geometric symmetry of pixel distribution in the target neighborhood window to obtain a symmetry score; fusing gradient consistency, direction difference, symmetry score and gray standard deviation to calculate the probability that the window belongs to a pore defect area; and evaluating whether the precast component is qualified according to the total area of the pore defect. In the identification of the pore defect area, the present application considers the texture characteristics, direction mutation and geometric symmetry through multi-dimensional feature fusion, thereby improving the identification accuracy of the pore defect area.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image data processing, and in particular to a construction quality evaluation method based on surface pore area distribution analysis of prefabricated components. BACKGROUND

[0002] As the core component of building industrialization, prefabricated components are applied to the field of fabricated buildings, and their production quality is directly related to the safety and performance of building structures. The pore defect area on the surface of prefabricated components is a common quality defect in the production of prefabricated components. If not accurately detected and controlled, it will damage the integrity of the component surface and reduce the aesthetics of the building. In addition, as a surface depression, the pore defect area will become a point of accumulation for moisture and corrosive media, accelerating the aging and erosion of the surface material of the component and greatly weakening its protective performance. Therefore, it is particularly necessary to detect the pore defect area on the surface of prefabricated components.

[0003] The traditional detection method for the pore defect area on the surface of prefabricated components is mainly based on visual inspection, which has obvious limitations in modern large-scale industrial production. This method mainly relies on manual visual inspection and touch, and is prone to miss small diameter fine pore defect areas, and cannot quantify the distribution density of the defect area.

[0004] In recent years, image processing-based detection methods for pore defect areas on the surface of prefabricated components have gradually attracted attention. However, related technologies have significant bottlenecks in practical applications: the pore defect area on the surface of prefabricated components and the artificially processed or naturally formed pitted area have some similarities in gray scale distribution. Related technologies rely on single gray scale features or simple texture analysis and fail to deeply explore the essential differences between the two, resulting in frequent confusion between pore defect areas and pitted areas, high misjudgment rate of detection results, and inability to provide reliable basis for prefabricated component quality evaluation. SUMMARY

[0005] To solve the technical problem of easily confusing pore defect areas and pitted areas and causing misjudgment, the present application provides a construction quality evaluation method based on surface pore area distribution analysis of prefabricated components, which includes the following steps:

[0006] A surface image of the precast component is acquired and preprocessed to obtain a grayscale image of the precast component. For the grayscale image of the precast component, the initial neighborhood window of each pixel is expanded by a preset step size. The gradient consistency and directional difference of the expanded neighborhood window during the expansion process are calculated. When the expansion termination condition is met, the expanded neighborhood window at the termination is the target neighborhood window. Taking the center pixel of the target neighborhood window as the center of symmetry, several preset symmetry axes are used to calculate the geometric symmetry of the pixel distribution within the target neighborhood window, which is recorded as the symmetry score of the target neighborhood window. The gradient consistency, directional difference, symmetry score, and grayscale standard deviation of the target neighborhood window are fused to calculate the probability that the target neighborhood window belongs to a porosity defect region. When the probability is greater than a preset threshold, the target neighborhood window is marked as a porosity defect region. The precast component is evaluated for qualification based on the ratio of the total area of ​​the porosity defect region to the total area of ​​the surface image of the precast component.

[0007] This invention utilizes a multi-feature fusion and adaptive window mechanism to detect and assess the quality of porosity defects on the surface of precast components. First, an initial neighborhood window is defined for each pixel. Local features of defects at different scales are captured by calculating gradient consistency and directional differences. Then, the window is dynamically expanded based on a preset step size, and the defect area is adaptively covered based on feature stability, effectively solving the problem of a single window being unable to adapt to diverse defect sizes. Simultaneously, symmetry scores and grayscale standard deviations are introduced. The probability that the target neighborhood window belongs to a porosity defect area is calculated through multi-dimensional feature fusion, taking into account texture characteristics, directional abrupt changes, and geometric symmetry, significantly improving the accuracy of distinguishing between porosity areas, rough surfaces, and background areas. Finally, the degree of defect is quantified by the ratio of the total area of ​​the porosity defect area to the total area of ​​the precast component image. This indicator comprehensively reflects the quantity and size impact of porosity defects, providing a quantitative basis for determining whether the surface quality of the precast component meets standards, thereby achieving objective and standardized quality judgment.

[0008] Preferably, the gradient consistency of the extended neighborhood window is the ratio of the magnitude of the vector sum of the gradient vectors of all pixels within the extended neighborhood window to the sum of the magnitudes.

[0009] This invention effectively quantifies the directional uniformity of texture in local regions by calculating the ratio of the magnitude of the sum of pixel gradient vectors to the sum of their magnitudes within a neighborhood window. For background regions with uniform texture, the magnitude of the sum of gradient vectors is close to the sum of their magnitudes due to the concentrated gradient directions, resulting in a ratio close to 1, which can accurately identify uniform background regions. For regions with disordered gradient directions, such as areas with pores or pitted surfaces, the gradient vectors cancel each other out, causing the magnitude of the sum of gradient vectors to be much smaller than the sum of their magnitudes, resulting in a significantly lower ratio, which can distinguish defective regions from background regions. This metric considers both gradient direction and intensity characteristics, providing a reliable quantitative basis for texture consistency for subsequent neighborhood window expansion and defect classification, effectively improving the stability and accuracy of feature extraction.

[0010] Preferably, the calculation of the gradient consistency of the extended neighborhood window includes: constructing the structure tensor of the extended neighborhood window; calculating two eigenvalues ​​of the structure tensor; and calculating the consistency index of the two eigenvalues ​​of the structure tensor, denoted as the gradient consistency of the extended neighborhood window.

[0011] Preferably, the calculation of the symmetry score of the target neighborhood window includes: constructing several symmetry axes with the center pixel of the target neighborhood window as the symmetry center; calculating the grayscale difference between the pixel on the several symmetry axes and its symmetric point, and recording it as the symmetry score of the target neighborhood window.

[0012] This invention constructs multiple axes of symmetry with the center of the neighborhood window as the center of symmetry. Symmetry is quantified and scored by quantifying the grayscale difference between pixels on the symmetry axis and their symmetrical points. For pores that are circular or elliptical in shape, due to their significant symmetry features and small grayscale differences, the symmetry score approaches 1, allowing for accurate identification. This design effectively amplifies the symmetry feature differences between pore defect areas, rough areas, and background areas, providing key quantitative basis for distinguishing pore defect areas from other areas and improving the targeting and accuracy of defect identification.

[0013] Preferably, the probability that the target neighborhood window belongs to a porosity defect region satisfies the expression... ;in, It is the first The probability that a target neighborhood window belongs to a porosity defect region. It is the first Normalized values ​​of directional differences among target neighborhood windows It is the first Gradient consistency of each target neighborhood window It is the first The normalized value of the gray standard deviation of each target neighborhood window. It is the first Symmetrical scores for each target neighborhood window , , They are , , The weight.

[0014] This invention presents a probabilistic calculation model for porosity defect regions. By integrating multi-dimensional features such as directional differences, gradient consistency, grayscale standard deviation, and symmetry scores, and dynamically adjusting the contribution of each feature using weights, it achieves accurate probabilistic discrimination. For porosity defect regions, small directional differences, weak gradient consistency, significant grayscale fluctuations, and symmetry features lead to a high probability value in the model's comprehensive calculation. This method overcomes the limitations of single features, characterizing the essential differences between defective and non-defective regions from multiple dimensions. This lays a solid foundation for subsequent defect classification and quality assessment, significantly improving detection accuracy and reliability, and adapting to the needs of complex engineering scenarios.

[0015] Preferably, the , , It was calculated using the Analytic Hierarchy Process (AHP).

[0016] Preferably, the expansion termination condition includes: when the absolute difference in gradient consistency and the absolute difference in orientation difference between two adjacent expanded neighborhood windows are both less than their respective preset thresholds.

[0017] For pore defect regions of varying sizes, when the window expands to cover the entire defect, the gradient consistency and directional differences between adjacent windows tend to stabilize, which can lock the target neighborhood window that fits the defect contour, laying a reliable foundation for subsequent multi-feature fusion detection.

[0018] Preferably, the directional difference of the extended neighborhood window is the result of summing the gradient direction angle differences between all pixels in the extended neighborhood window and their neighboring pixels and then normalizing them.

[0019] Preferably, the calculation of the gradient direction angle difference between all pixels in the extended neighborhood window and their adjacent pixels includes: taking any pixel as the target pixel, calculating the sum of the absolute differences in the gradient direction angle between the target pixel and its immediate right-hand and top-hand pixels, and recording it as the gradient direction angle difference between the target pixel and its adjacent pixels.

[0020] Preferably, the directional difference of the extended neighborhood window is calculated based on the gradient orientation histogram of the extended neighborhood window.

[0021] The beneficial effects of this invention are as follows: This invention utilizes a multi-feature fusion and adaptive windowing mechanism to achieve the detection and quality assessment of porosity defects on the surface of precast components. For pores of varying sizes, when the window expands to cover the entire defect, the gradient consistency and directional differences between adjacent windows tend to stabilize. This allows for locking onto a target neighborhood window that conforms to the defect contour, providing a reliable window foundation for subsequent multi-feature fusion detection. Simultaneously, by fusing multi-dimensional features such as directional differences, gradient consistency, grayscale standard deviation, and symmetry scores, and dynamically adjusting the contribution of each feature using weights, accurate probabilistic discrimination is achieved. Finally, the degree of defect is quantified by the ratio of the total area of ​​the porosity defect region to the total area of ​​the precast component image. This indicator comprehensively reflects the quantity and size impact of the porosity defect region, providing a quantitative basis for determining whether the surface quality of the precast component meets standards, thereby achieving objective and standardized quality judgment. Attached Figure Description

[0022] Figure 1 A flowchart of a construction quality assessment method based on the analysis of the distribution of pore areas on the surface of prefabricated components, provided in an embodiment of the present invention. Detailed Implementation

[0023] This invention provides a construction quality assessment method based on the analysis of the porosity distribution on the surface of precast components, such as... Figure 1 As shown, the method includes steps S100-S500:

[0024] Step S100: Acquire surface images of precast components and preprocess them to obtain grayscale images of the precast components.

[0025] It should be noted that when evaluating the porosity defect area on the surface of precast components, it is first necessary to acquire images of the component surface using photographic equipment, and then preprocess them to lay the foundation for subsequent analysis. The preprocessing steps include noise reduction and grayscale conversion, and the order of these operations must be arranged appropriately.

[0026] Specifically, firstly, a high-resolution image of the surface of the precast component is captured using a photographic device; then, the color image is converted to grayscale, transforming the three-dimensional color information into a single-channel grayscale value to reduce subsequent computation; finally, a filter such as a Gaussian filter is used to denoise the grayscale image, reducing random noise interference in the image.

[0027] At this point, grayscale images of the prefabricated components have been obtained.

[0028] Step S200: For the grayscale image of the precast component, expand the initial neighborhood window of each pixel by a preset step size, calculate the gradient consistency and orientation difference of the expanded neighborhood window during the expansion process of the initial neighborhood window, and when the expansion termination condition is met, the expanded neighborhood window at the termination is the target neighborhood window.

[0029] It should be noted that, in order to adapt to the pore defect areas and pitted areas of different sizes on the surface of the precast components, an initial neighborhood window needs to be set first to provide a basis for subsequent dynamic expansion.

[0030] Specifically, an initial neighborhood window is defined for each pixel in the grayscale image of the prefabricated component. The shape of the initial neighborhood window can be flexibly set; it can be circular or square, or the implementer can customize the shape according to actual inspection needs to adapt to different types of surface defects. Regarding the initial neighborhood window size, if a circle is used, the radius can be preset to 5 pixels. In practical applications, it is also necessary to dynamically adjust the size based on the image resolution, size, and the radius of the largest possible porosity defect area on the surface of the prefabricated component. This ensures that the initial neighborhood window size covers the feature range of the porosity defect area without being too large and including too much irrelevant background information, laying the foundation for accurate calculation of subsequent gradient and texture features.

[0031] It should be noted that due to the varying dimensions of porosity and pitted areas on the surface of precast components, a single, fixed-size initial neighborhood window is insufficient to adapt to all scenarios. Therefore, this invention proposes a dynamic expansion mechanism based on gradient consistency and directional differences. The specific calculation methods for gradient consistency and directional differences will be explained later.

[0032] Specifically, the initial neighborhood window of each pixel is expanded one by one according to a preset step size. The gradient consistency and orientation difference of the expanded neighborhood windows are calculated during the expansion process. The gradient consistency and orientation difference of two adjacent expanded neighborhood windows are compared. Expansion stops when the difference in these two features between adjacent expanded neighborhood windows is less than their respective preset thresholds. The expanded neighborhood window at the point of termination is the target neighborhood window. Furthermore, a maximum number of expansions needs to be set to prevent continuous calculations from causing the target neighborhood window to become too large. The maximum number of expansions can be set to 20, or adjusted according to requirements. The preset step size is 1 pixel by default, but can also be adjusted as needed.

[0033] definition For the first The absolute difference in gradient consistency between two adjacent expanded neighborhood windows corresponding to an initial neighborhood window. For the first The absolute difference in orientation between two adjacent expanded neighbor windows corresponding to an initial neighbor window. The preset threshold for gradient consistency is set to 0.05, and the threshold for orientation difference is set to 15. Both thresholds can be set according to requirements.

[0034] With preset step size For example, combining the preset radius of the initial neighborhood window mentioned earlier... The radius of the expanded neighborhood window after the initial expansion is [number] pixels. Subsequent expansions will increase by this step size each time. and When the gradient and orientation features within the expanded neighborhood window have stabilized, it can be determined that the current expanded neighborhood window completely covers the rough surface area or the porosity defect area. At this point, the expanded neighborhood window is the target neighborhood window. If either of the two features is not less than its corresponding threshold and the preset maximum number of expansions has not been reached, the neighborhood window continues to expand by a preset step size. Expansion stops when both feature threshold conditions are met simultaneously or the preset maximum number of expansions is reached. When expansion stops, the expanded neighborhood window is the target neighborhood window.

[0035] The previous section introduced how to obtain the target neighborhood window. The following section explains how to calculate the gradient consistency and orientation difference of the expanded neighborhood window during the expansion process.

[0036] It should be noted that a major difference between the pitted surface area, the porosity defect area, and the background area of ​​a precast component is that the background area and the pitted surface area show greater differences in all directions, while the porosity defect area shows less difference in all directions. To quantify this characteristic, a gradient consistency index is introduced for calculation, and this index is used to distinguish between the pitted surface area, the porosity defect area, and the background area.

[0037] Specifically, the gradient consistency calculation in this invention provides two implementation methods: the first is to quantize by the ratio of the magnitude of the vector sum of the gradient vectors of all pixels in the extended neighborhood window to the sum of the magnitudes of each gradient vector; the second is to quantize based on the eigenvalues ​​of the structure tensor of the extended neighborhood window. This method uses the ability of the structure tensor to describe local texture and reflects the concentration of gradient directions through the difference in eigenvalues. This is an existing technology and will not be elaborated on here.

[0038] Both approaches offer gradient consistency quantization paths from the perspectives of vector composition and tensor analysis, respectively, allowing you to choose the appropriate solution based on your specific scenario. This section primarily focuses on the first approach.

[0039] It should be noted that the background area on the surface of precast components typically has the strongest texture, followed by the rough surface area, while the texture of the pore defect area is the weakest. To effectively distinguish between background areas, rough surface areas, and pore defect areas, this invention uses texture characteristics as the core judgment criterion. Specifically, each pixel within an extended neighborhood window has its own gradient direction angle: if the gradient directions of most pixels within the extended neighborhood window tend to be consistent, it indicates that the area has strong texture and is more likely to be a background area; conversely, if the gradient directions are chaotic, it is more likely to be a rough surface area or a pore defect area.

[0040] To quantify the aforementioned gradient consistency, this invention introduces the mathematical principle of vector synthesis: In a two-dimensional plane, the magnitude of the sum of two vectors is less than or equal to the sum of the magnitudes of the two vectors, and equality holds if and only if the two vectors have the same direction. This conclusion also applies to multiple vectors. The ratio of the magnitude of the sum of the gradient vectors of all pixels within the extended neighborhood window to the sum of the magnitudes of all individual vectors is calculated. The closer this ratio is to 1, the more similar the directions of most gradient vectors are, indicating stronger texture; the closer it is to 0, the more chaotic the gradient directions are, indicating weaker texture. This quantitative indicator provides a precise basis for distinguishing the texture differences between background areas, rough areas, and areas with porosity or defects.

[0041] Based on the above logic, the gradient consistency of the expanded neighborhood window satisfies the following relationship:

[0042] ;

[0043] in, It is the first Gradient consistency of extended neighborhood windows, , They are the first Within the first extended neighborhood window 1 pixel , Gradient in direction; It is the first The total number of pixels within an extended neighborhood window.

[0044] This formula quantifies the consistency of gradient directions within the extended neighborhood window. The closer it is to 1, the more uniform the gradient direction of the pixels within the extended neighborhood window, the stronger the texture of the region, and the more likely it is to be a background region. The closer it is to 0, the more chaotic the gradient direction of the pixels within the extended neighborhood window, and the more likely it is to be a pore defect region. A value close to 0.5 indicates that the expanded neighborhood window has some directionality but is generally cluttered, and is more likely to be a pitted area.

[0045] The previous section explained how to obtain the gradient consistency of the extended neighborhood window; the following section explains how to obtain the directional differences of the extended neighborhood window.

[0046] It's important to note that directional difference excels at identifying regions where gradient direction changes abruptly, such as the edges of porosity defects or local breaks in rough surfaces. Gradient consistency, on the other hand, tends to classify areas with uniform local texture as background regions. Relying solely on GVC (Gradient Value Capture) makes it difficult to accurately identify regions with abrupt gradient direction changes. Therefore, directional difference needs to be used to capture the gradient direction differences between adjacent pixels to effectively detect porosity defects. The calculation of directional difference typically involves calculating the angle between the gradient directions of adjacent pixels within a region.

[0047] Based on the above logic, the directional differences of the expanded neighborhood window satisfy the following relation:

[0048] ;

[0049] in, It is the first The directional differences of each extended neighborhood window It is the first The total number of pixels within an extended neighborhood window For explanation It is the first Expanded neighborhood window pixels, , , They are the first Pixels within an extended neighborhood window , , The gradient direction angle, It is the absolute value symbol.

[0050] In this formula, It is the first Pixels within an extended neighborhood window The sum of the absolute differences in the gradient direction angles of the adjacent right and top pixels is used to focus on the gradient changes in the horizontal and vertical directions. It is a normalization factor, due to the sum of the absolute differences of a single set of angles. The maximum value is Combined with normalization factor Can Limited to Interval. Directional differences. The larger the value, the more drastic the local directional abrupt change within the extended neighborhood window, and the more likely this region is to be a rough surface area or a porosity defect area; directional differences The smaller the value, the smoother the directional transition, and the more likely the area is to be a background area.

[0051] Calculating only the gradient difference in the horizontal and vertical directions reduces computational complexity and aligns with the characteristic that the texture of precast components is mostly distributed along the vertical or horizontal direction. This allows for efficient capture of gradient abrupt changes at the boundaries of local texture breaks or pore defect areas in pitted regions.

[0052] Furthermore, the directional differences of the extended neighborhood window can also be calculated based on the gradient orientation histogram of the extended neighborhood window. This calculation method is existing technology and will not be elaborated here.

[0053] Thus, the target neighborhood window and the gradient consistency and orientation differences of the target neighborhood window have been obtained.

[0054] Step S300: Using the center pixel of the target neighborhood window as the center of symmetry, preset several axes of symmetry, calculate the geometric symmetry of the pixel distribution within the target neighborhood window through the axes of symmetry, and record it as the symmetry score of the target neighborhood window.

[0055] It should be noted that symmetry is another key feature distinguishing porosity defect areas, rough surfaces, and background areas, with significant differences in their geometric symmetry characteristics. Porosity defect areas, as depressions on the surface of precast components, are mostly circular or elliptical in shape, with obvious geometric symmetry in their edge contours and internal grayscale distribution. Rough surfaces consist of dense and chaotic tiny pits, with an irregular overall shape and lacking stable symmetry. The texture of the background area often extends along a fixed direction, but it mostly exhibits a unidirectional, continuous, zigzag shape, lacking a symmetrical distribution around a central point. This difference in symmetry characteristics complements the gradient consistency and directional differences mentioned earlier, further improving the accuracy of distinguishing the three types of areas.

[0056] Specifically, based on the symmetrical characteristic of the porosity defect region, the symmetry score of the target neighborhood window satisfies the following relationship:

[0057] ;

[0058] in, It is the first Symmetrical scores for each target neighborhood window It is the first The average number of pixels across all axes of symmetry of the target neighborhood window It is the first The total number of pixels on the axis of symmetry of each target neighborhood window. It is the first The total number of symmetry axes of each target neighborhood window. It is the first The first on the axis of symmetry The coordinates of each pixel It is the first The first on the axis of symmetry The coordinates of the symmetrical point of each pixel It is the absolute value symbol. It is the grayscale value of a pixel.

[0059] This formula calculates the first... The grayscale differences between all pixels along the symmetry axis within a target neighborhood window and their symmetrical points are averaged, and then subtracted from this average by 1 to obtain the symmetry score. The higher the symmetry of the pixels within the target neighborhood window, the smaller the sum of the grayscale differences. The closer the value is to 1, the higher the probability that the target neighborhood window is a pore defect region; when the pixel distribution within the target neighborhood window is disordered and the sum of the grayscale differences is larger, the higher the probability. The closer the value is to 0, the higher the probability that the target neighborhood window is a textured area; the background area, due to the unidirectional extension of the texture, has a symmetry between the two, therefore... The values ​​are mostly centered.

[0060] It should be noted that the symmetry axes should be set based on the center pixel of the target neighborhood window. The number can be adjusted according to the geometric characteristics of the target neighborhood window: if the target neighborhood window is a centrally symmetrical shape such as a circle, it is recommended to set 4 symmetry axes, specifically including: horizontal, vertical, main diagonal, and secondary diagonal. The number of symmetry axes can also be set according to needs. If the target neighborhood window is a square or other geometric shape, the number of symmetry axes can be determined based on its own symmetry characteristics.

[0061] At this point, the symmetric score of the target neighborhood window has been obtained.

[0062] Step S400: Combine the gradient consistency, orientation difference, symmetry score, and grayscale standard deviation of the target neighborhood window to calculate the probability that the target neighborhood window belongs to the porosity defect region. When the probability is greater than a preset threshold, mark the target neighborhood window as a porosity defect region.

[0063] It should be noted that the features of the porosity defect region, the pitted region, and the background region have all been quantified. The probability that the target neighborhood window belongs to the porosity defect region is calculated by combining the above indicators. In addition, the grayscale standard deviation is also taken into account. This probability satisfies the following relationship:

[0064] ;

[0065] in, It is the first The probability that a target neighborhood window belongs to a porosity defect region. It is the first Normalized values ​​of directional differences among target neighborhood windows It is the first Gradient consistency of each target neighborhood window It is the first The normalized value of the gray standard deviation of each target neighborhood window. It is the first Symmetrical scores of each target neighborhood window , , They are , , The weight.

[0066] This formula amplifies the characteristic differences in porosity defect regions through the coordinated design of the numerator, denominator, and exponent. (Numerator) This integrates the directional difference information of the target neighborhood window. Due to the abrupt gradient change at the edge of the pore defect region, the normalized value of its directional difference is... It will be located between the background area and the textured area, through weighting. Its contribution to the molecule can be adjusted to avoid excessive suppression or amplification of the effect of this feature. Denominator Combining information from gradient consistency and grayscale standard deviation, the texture of the pore defect area is weak. Small, and with uniform grayscale distribution The smaller the sum of the two, the larger the overall fractional value. Meanwhile, the background area has strong texture. Large, pitted areas have drastic fluctuations in grayscale. The larger the value of the target neighborhood window, the larger the denominator value, and the smaller the fractional value. For the exponential term... The high symmetry of the porosity defect region was utilized to improve the symmetry score of the porosity defect region. Significantly higher than the background and pitted areas, through weighting By amplifying the power of the fractional value, the probability difference between the porosity defect area and the pitted area and background area is further widened.

[0067] In conclusion, The larger the value, the more closely the characteristics of the target neighborhood window match the porosity defect region, meaning that the probability of the target neighborhood window being a porosity defect region is higher.

[0068] It should be added that, It is the first The ratio of the directional difference of a target neighborhood window to the maximum directional difference of all target neighborhood windows in the image. It is the first The ratio of the standard deviation of the gray level of each target neighborhood window to the maximum standard deviation of the gray level of all target neighborhood windows in the image. Weight , , It was calculated using the Analytic Hierarchy Process (AHP), which is an existing technology and will not be discussed in detail here.

[0069] After calculating the probability of porosity defect regions in all target neighborhood windows of the grayscale image as described above, a porosity defect threshold needs to be set: when When the value exceeds the porosity defect threshold, the corresponding target neighborhood window is marked as a porosity defect region.

[0070] Thus, the porosity defect area in the surface image of the precast component was obtained.

[0071] Step S500: Evaluate whether the precast component is qualified based on the ratio of the total area of ​​the porosity defect area to the total area of ​​the surface image of the precast component.

[0072] It should be noted that area can intuitively reflect the actual distribution range of porosity defect areas, while area ratio can comprehensively reflect the dual impact of the quantity and size of porosity defect areas, which is more in line with the assessment needs of surface integrity in engineering. This invention considers quantifying the proportion of porosity defect areas on the surface of precast components by calculating the ratio of the total area of ​​all areas marked as porosity defects to the total area of ​​the surface image of the precast component.

[0073] Specifically, the construction quality is assessed by the ratio of the total area of ​​the porosity defect area to the total area of ​​the precast component surface image. A threshold value needs to be set for this ratio; an empirical value is [value missing]. It can also be set according to needs; exceeding this ratio threshold can be considered as the prefabricated component being of substandard quality.

[0074] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A construction quality assessment method based on the distribution analysis of pore areas on the surface of precast components, characterized in that, Including the following steps: Acquire surface images of precast components and preprocess them to obtain grayscale images of the precast components; For the grayscale image of the precast component, the initial neighborhood window of each pixel is expanded by a preset step size. The gradient consistency and orientation difference of the expanded neighborhood window during the expansion process are calculated. When the expansion termination condition is met, the expanded neighborhood window at the termination is the target neighborhood window. Using the center pixel of the target neighborhood window as the center of symmetry, several symmetry axes are preset. The geometric symmetry of the pixel distribution within the target neighborhood window is calculated through these symmetry axes to obtain the symmetry score of the target neighborhood window, including: Construct several axes of symmetry with the center pixel of the target neighborhood window as the center of symmetry; Calculate the grayscale difference between a pixel on several axes of symmetry and its symmetrical point, and record it as the symmetry score of the target neighborhood window; Taking any pixel as the target pixel, calculate the sum of the absolute differences in gradient direction angles between the target pixel and its immediate right and top pixels, denoted as the gradient direction angle difference between the target pixel and its neighboring pixels. The directional difference of the extended neighborhood window is obtained by accumulating the gradient direction angle differences between all pixels within the extended neighborhood window and their neighboring pixels, and then normalizing the result, satisfying the following relationship: ； in, It is the first The directional differences of each extended neighborhood window It is the first The total number of pixels within an extended neighborhood window For explanation It is the first Expanded neighborhood window pixels, , , They are the first Pixels within an extended neighborhood window , , The gradient direction angle, It is the absolute value symbol; By integrating gradient consistency, orientation difference, symmetry score, and grayscale standard deviation of the target neighborhood window, and taking into account the texture characteristics, orientation abrupt changes, and geometric symmetry of the target neighborhood window, the probability that the target neighborhood window belongs to a porosity defect region is calculated. The expression is: ； in, It is the first The probability that a target neighborhood window belongs to a porosity defect region. It is the first Normalized values ​​of directional differences among target neighborhood windows It is the first Gradient consistency of each target neighborhood window It is the first The normalized value of the gray standard deviation of each target neighborhood window. It is the first Symmetrical scores for each target neighborhood window , , They are , , The weights; When the probability is greater than a preset threshold, the target neighborhood window is marked as a porosity defect region. The qualification of a precast component is assessed by the ratio of the total area of ​​the porosity defect area to the total area of ​​the surface image of the precast component.

2. The construction quality assessment method based on the distribution analysis of pore areas on the surface of precast components according to claim 1, characterized in that, The gradient consistency of the extended neighborhood window is the ratio of the magnitude of the vector sum of the gradient vectors of all pixels within the extended neighborhood window to the sum of the magnitudes.

3. The construction quality assessment method based on the distribution analysis of pore areas on the surface of precast components according to claim 1, characterized in that, The calculation of gradient consistency of the extended neighborhood window includes: Construct the structure tensor of the extended neighborhood window; Calculate the two eigenvalues ​​of the structure tensor; The consistency index of the two eigenvalues ​​of the structure tensor is calculated and denoted as the gradient consistency of the extended neighborhood window.

4. The construction quality assessment method based on the analysis of the porosity distribution on the surface of precast components according to claim 1, characterized in that, The , , It was calculated using the Analytic Hierarchy Process (AHP).

5. The construction quality assessment method based on the distribution analysis of pore areas on the surface of precast components according to claim 1, characterized in that, The extended termination condition includes: When the absolute difference in gradient consistency and the absolute difference in orientation difference between two adjacent extended neighborhood windows are both less than their respective preset thresholds.

6. The construction quality assessment method based on the analysis of the porosity distribution on the surface of precast components according to claim 1, characterized in that, The directional difference of the extended neighborhood window is calculated based on the gradient orientation histogram of the extended neighborhood window.

Images