A steel bar mesh construction quality detection method based on machine vision

Abstract

This invention relates to the field of construction inspection technology and discloses a machine vision-based method for inspecting the construction quality of steel mesh. The method includes: acquiring color images and depth maps of the steel mesh; locating adjacent intersection nodes and the mid-span sag points of the connecting lines; extracting depth values ​​and converting them into three-dimensional physical coordinates; setting the intersection nodes as the local coordinate origin; calculating the difference between the horizontal projected straight-line distance and the relative sag depth; using algebraic analytical methods to simultaneously solve for the coefficients of the local parabolic deflection shape; using the arc length integral analytical method based on these coefficients to obtain the true physical arc length of the steel reinforcement; and finally calculating the relative error ratio between the actual arc length and the standard spacing to output the judgment result. This invention effectively eliminates the projection shortening error caused by local steel reinforcement sag, the calculation process requires no iterative optimization, the data logic is strictly self-consistent and closed-loop, and high-precision true spacing detection is achieved.

Description

Technical Field

[0001] This invention relates to the field of building construction inspection technology, specifically a method for inspecting the construction quality of steel mesh based on machine vision. Background Technology

[0002] In the preparatory stage before concrete pouring in building construction, quality acceptance of the laid and tied steel mesh is a crucial step in ensuring structural safety. Traditional machine vision inspection technology is typically set up in a pre-defined application scenario directly above the steel mesh to be tested. It identifies steel mesh intersections and directly calculates the straight-line distance between nodes by acquiring two-dimensional color images or simple three-dimensional depth data to assess whether the mesh spacing meets the standards. However, in actual construction sites, steel mesh often experiences localized sagging and bending deformation due to its own weight, workers' foot traffic, or uneven placement of bottom pads. This uncommon sagging deformation leads to visual projection distortion when viewed from directly above, causing projection shortening errors in mesh spacing measurements. Existing two-dimensional planar vision algorithms or simple three-dimensional straight-line distance calculation schemes cannot accurately reflect the true physical spacing of the steel bars when sagging and bending occur, easily leading to errors in actual mesh spacing measurements. The steel mesh is sometimes misjudged as having excessively large spacing due to projection shortening. To overcome this error, some conventional techniques attempt to introduce complex three-dimensional spatial curve iterative fitting algorithms. However, these solutions heavily rely on algorithms that require iterative calculations or optimizations, resulting in high computational costs and the introduction of external weight parameters determined by experience, leading to a lack of strict self-consistency and closed-loop data at each step. Therefore, the industry currently lacks a detection scheme that can accurately reconstruct the local sag shape of steel bars without using iterative optimization. How to cleverly utilize machine vision equipment with depth sensors to accurately obtain the three-dimensional coordinates of intersection points and mid-span sag points, establish a local spatial coordinate system through dimensionality reduction, and then directly and accurately restore the true physical arc length of the steel bars in the sag state through algebraic analysis and the arc length integral rule, thereby effectively eliminating projection shortening errors and achieving accurate quality judgment with closed-loop data from the beginning and end, is a technical problem that has not yet been solved in current related technologies and urgently needs to be overcome. Summary of the Invention

[0003] This invention provides a machine vision-based method for inspecting the construction quality of steel mesh, which helps to solve the problems mentioned in the background art.

[0004] This invention provides the following technical solution: a machine vision-based method for inspecting the construction quality of reinforcing mesh, comprising:

[0005] A color image and depth map of the steel mesh to be tested are acquired. The intersection nodes of two adjacent steel meshes are located through image processing, and the depth value of the intersection node is extracted from the depth map.

[0006] By combining the camera's intrinsic parameter matrix, the two-dimensional pixel coordinates of the two adjacent intersection nodes and the corresponding extracted depth values ​​are converted into three-dimensional spatial physical coordinates.

[0007] Locate the center position of the line connecting two adjacent intersection nodes on the color image as the mid-span droop point, extract the depth value of the mid-span droop point in the depth map, and calculate its three-dimensional spatial physical coordinates.

[0008] Set one of the intersection nodes as the local coordinate origin, and calculate the physical projection straight-line distance of the other intersection node and the mid-span drooping point relative to the origin on the horizontal plane.

[0009] Using the local coordinate origin as the depth reference, calculate the relative sagging depth difference between the other intersection node and the mid-span sagging point relative to the origin;

[0010] Using the calculated straight-line distances of each of the physical projections and the corresponding differences in relative droop depths, the quadratic and linear coefficients of the local parabolic flexural shape are obtained simultaneously through algebraic analytical solutions.

[0011] Based on the coefficients of the obtained local parabolic flexural shape, the actual physical arc length of the steel bar is obtained by calculating the straight-line distance interval of the physical projection using the arc length integral analytical rule.

[0012] Calculate the relative error ratio between the actual physical arc length and the preset standard grid design spacing, and output the judgment result of whether the construction quality is qualified or not according to the preset tolerance threshold.

[0013] Optionally, acquiring a color image and depth map of the reinforcing mesh to be tested, locating the intersection nodes of two adjacent reinforcing meshes through image processing, and extracting the depth value of the intersection node from the depth map includes:

[0014] Set up a calibrated RGB-D camera with its lens mounted vertically downwards directly above the steel mesh to be measured; the RGB-D camera simultaneously acquires RGB color images containing the steel mesh and corresponding depth maps;

[0015] Obtain fixed parameters, including: camera focal length, principal point x and y coordinates, and rebar grid spacing threshold;

[0016] Preset error tolerance threshold;

[0017] Based on the two-dimensional pixel coordinates of the located intersection node on the color image, the data corresponding to that position is indexed and extracted from the depth map as the physical depth value of the intersection node in the camera coordinate system.

[0018] Optionally, the step of combining the camera's intrinsic parameter matrix to convert the acquired two-dimensional pixel coordinates of the two adjacent intersection nodes and the corresponding extracted depth values ​​into three-dimensional spatial physical coordinates includes:

[0019] The horizontal coordinate difference is obtained by subtracting the horizontal coordinate of the principal point from the horizontal coordinate of one of the intersection nodes in the color image. The horizontal coordinate difference is then multiplied by the physical depth value of the intersection node and divided by the focal length of the camera to obtain the horizontal physical coordinate of the intersection node in three-dimensional space. The vertical coordinate difference is obtained by subtracting the vertical coordinate of the principal point from the vertical coordinate of the intersection node in the color image. The vertical coordinate difference is then multiplied by the physical depth value of the intersection node and divided by the focal length of the camera to obtain the vertical physical coordinate of the intersection node in three-dimensional space.

[0020] Similarly, based on the two-dimensional pixel coordinates and physical depth value of another intersection node located on the same rebar and adjacent to it, the three-dimensional horizontal and vertical physical coordinates of that other intersection node are calculated.

[0021] Optionally, the step of locating the center position of the line connecting two adjacent intersection nodes on the color image as the mid-span sagging point, extracting the depth value of the mid-span sagging point in the depth map, and converting it into its three-dimensional spatial physical coordinates includes:

[0022] Calculate the average of the two-dimensional pixel horizontal coordinates of two adjacent intersection nodes in the color image, and use it as the pixel horizontal coordinate of the mid-span downward point;

[0023] Calculate the average of the two-dimensional pixel vertical coordinates of two adjacent intersection nodes in the color image, and use it as the pixel vertical coordinate of the mid-span downward point;

[0024] Based on the calculated pixel horizontal and vertical coordinates of the mid-span sagging point, the data at the corresponding position is extracted from the depth map and used as the physical depth value of the mid-span sagging point.

[0025] Subtract the horizontal coordinate of the principal point from the pixel horizontal coordinate of the mid-span drooping point, multiply by the physical depth value of the mid-span drooping point, and divide by the focal length to obtain the three-dimensional horizontal physical coordinate of the mid-span drooping point.

[0026] Subtract the principal point's ordinate from the pixel's vertical coordinate of the mid-span drooping point, multiply by the physical depth value of the mid-span drooping point, and divide by the focal length to obtain the three-dimensional spatial vertical physical coordinates of the mid-span drooping point.

[0027] Optionally, setting one of the intersection nodes as the local coordinate origin and calculating the physical projection straight-line distances of the other intersection node and the mid-span drooping point relative to the origin on the horizontal plane includes:

[0028] Calculate the difference between the lateral physical coordinates of the other intersection node and the lateral physical coordinates of the intersection node that serves as the local coordinate origin, and then square the difference.

[0029] Calculate the difference between the longitudinal physical coordinates of the other intersection node and the longitudinal physical coordinates of the intersection node that serves as the local coordinate origin, and then square the difference.

[0030] Add the two squares together and take the square root to obtain the physical projection straight-line distance of the other intersection node relative to the local coordinate origin on the horizontal plane.

[0031] Calculate the difference between the horizontal physical coordinates of the mid-span downward point and the horizontal physical coordinates of the intersection node (which serves as the local coordinate origin), and then square the difference. At the same time, calculate the difference between the vertical physical coordinates of the mid-span downward point and the vertical physical coordinates of the intersection node (which serves as the local coordinate origin), and then square the difference.

[0032] Add the two squares corresponding to the mid-span downward point and take the square root to obtain the physical projection straight-line distance of the mid-span downward point relative to the local coordinate origin on the horizontal plane.

[0033] Optionally, the step of using the local coordinate origin as a depth reference to calculate the relative sagging depth difference between the other intersection node and the mid-span sagging point relative to the origin includes:

[0034] Subtract the physical depth value of the intersection node that serves as the local coordinate origin from the physical depth value of the other intersection node to obtain the relative droop depth difference of the other intersection node relative to the local coordinate origin.

[0035] Subtracting the physical depth value of the intersection node, which serves as the local coordinate origin, from the physical depth value of the mid-span sagging point yields the relative sagging depth difference between the mid-span sagging point and the local coordinate origin.

[0036] Optionally, the step of using the calculated straight-line distances of each of the physical projections and the corresponding differences in relative sag depths to simultaneously solve the quadratic and linear coefficients of the local parabolic deflection shape using an algebraic analytical solution method includes:

[0037] Multiply the difference in relative sagging depth between the mid-sag points by the physical projection straight-line distance of the other intersection node to obtain the first product term;

[0038] Multiply the difference in relative droop depth of the other intersection node by the physical projected straight-line distance of the mid-span droop point to obtain the second product term;

[0039] Subtract the second product term from the first product term to obtain the numerator of the quadratic term; multiply the square of the physical projection straight-line distance of the mid-span downward point by the physical projection straight-line distance of the other intersection node to obtain the third product term;

[0040] Multiply the square of the physical projection straight-line distance of the other intersection node by the physical projection straight-line distance of the mid-span downward point to obtain the fourth product term;

[0041] Subtracting the fourth product term from the third product term yields the denominator of the quadratic term;

[0042] Dividing the numerator of the quadratic term by the denominator of the quadratic term yields the coefficients of the quadratic term in the local parabolic flexural shape.

[0043] Multiplying the coefficient of the quadratic term by the square of the physical projection straight-line distance of the other intersection node yields the fifth product term;

[0044] Subtract the fifth product term from the relative droop depth difference of the other intersection node, and then divide by the physical projection straight-line distance of the other intersection node to obtain the first-order coefficient of the local parabolic flexural shape.

[0045] Optionally, the step of calculating the true physical arc length of the reinforcing bar by using the arc length integral analytical rule to calculate the linear distance interval of the physical projection based on the coefficients of the obtained local parabolic deflection shape includes:

[0046] Multiply the coefficient of the quadratic term by two, then multiply by the physical projection straight-line distance of another intersection node, and finally add the coefficient of the linear term to obtain the substitution variable for the upper limit of integration.

[0047] Take the square root of the square of the substitution variable for the upper limit of integration plus one to get the first square term; calculate the natural logarithm by adding the first square term to the substitution variable for the upper limit of integration, and add the product of the substitution variable for the upper limit of integration and the first square term to get the original value of the upper limit of integration.

[0048] Take the square root of the square of the coefficient of the first term plus one to get the second radical term; calculate the natural logarithm by adding the coefficient of the first term to the second radical term, and then add the product of the coefficient of the first term and the second radical term to get the value of the original function with the lower limit of integration.

[0049] Subtracting the original function value of the lower limit of integration from the original function value of the upper limit of integration, and then dividing by four times the coefficient of the quadratic term, yields the true physical arc length of the reinforcing bars between two adjacent intersection nodes.

[0050] Optionally, the step of calculating the relative error ratio between the actual physical unfolded arc length and the preset standard grid design spacing, and outputting the judgment result of whether the construction quality is qualified or not according to the preset tolerance threshold, includes:

[0051] Calculate the absolute value of the difference between the actual physical arc length and the preset standard grid design spacing; divide the absolute value of the difference by the standard grid design spacing, and then multiply by 100% to obtain the percentage deviation of the actual spacing from the standard spacing;

[0052] Extract the preset error tolerance threshold; if the calculated percentage of deviation is less than or equal to the error tolerance threshold, the construction quality is deemed acceptable; if the calculated percentage of deviation is greater than the error tolerance threshold, the construction quality is deemed unacceptable.

[0053] The present invention has the following beneficial effects:

[0054] 1. By acquiring color images and depth maps of the reinforcing mesh to be tested, extracting the three-dimensional physical coordinates of intersection nodes and mid-span sagging points, and then constructing a local parabolic deflection model, this solution proposes a high-precision construction quality inspection algorithm using the arc length integral analytical method to calculate the true physical arc length of the reinforcing bars. This solution is designed because in the specific environment of a construction site, the tied reinforcing mesh often experiences localized non-planar sagging and deflection due to its own weight, uneven bottom pads, or foot traffic. In this specific environment, conventional machine vision's two-dimensional top-view detection or simple three-dimensional straight-line distance calculation will produce severe "projection shortening error," easily misjudging actually qualified reinforcing mesh as having excessive spacing. The core unique and beneficial effect of this technical solution lies in: cleverly reducing the dimensionality of the complex three-dimensional bending problem, eliminating reliance on external empirical parameters, and completely avoiding the extremely computationally intensive iterative optimization process, achieving complete data closure and self-consistency; it accurately "straightens" the bent and sagging reinforcing bars in digital space with extremely low computational cost, restoring their true physical layout length. This not only completely overcomes the measurement shortening error caused by visual projection distortion, but also significantly improves the accuracy and reliability of quality acceptance in complex construction environments, avoids misjudgment and omission, and effectively ensures the safety of building structures.

[0055] 2. By setting up a calibrated depth camera and simultaneously acquiring color images and corresponding depth maps of the steel mesh to be measured, image processing technology is used to accurately locate the intersection nodes of two adjacent steel meshes and extract the corresponding depth values ​​from the depth map. This step breaks through the limitations of traditional two-dimensional planar visual inspection technology from the source of data acquisition. Traditional pure two-dimensional images can often only capture planar projection information, which is prone to visual errors when faced with non-planar deformation of the steel mesh caused by gravity or external trampling. However, the method of extracting depth information gives each key intersection node a real spatial distance attribute. This method of integrating the clear texture features of color images with the accurate distance measurement capability of depth maps not only greatly improves the accuracy of key node identification and spatial positioning, but also provides an absolutely reliable initial distance reference for the subsequent accurate reconstruction of the true physical form of the steel bars in three-dimensional space, effectively avoiding the initial data distortion caused by the shortening of single visual projection.

[0056] 3. By combining the camera's intrinsic parameter matrix and converting the two-dimensional pixel coordinates of two adjacent intersection nodes and the corresponding extracted depth values ​​into three-dimensional physical coordinates, this processing successfully achieves a precise transition from virtual image pixel dimensions to real-world physical size dimensions. In actual calculations, the potential skew of the lens's optical center is corrected using the principal point coordinates, and the pixel difference is spatially mapped to the depth data using the camera's focal length parameters. This allows isolated image pixels to be restored to three-dimensional spatial coordinates with absolute physical scale. This rigorous coordinate system transformation mechanism not only completely eliminates the inherent perspective distortion during image acquisition, ensuring that data between nodes can be processed with high precision under the same real physical dimensions, but also constructs a stable and unified three-dimensional physical reference coordinate space for the entire detection scheme. This gives all subsequent morphological analysis and distance measurement extremely accurate real-world physical meaning, fundamentally guaranteeing the rigor and objectivity of the detection data.

[0057] 4. By accurately locating the center of the line connecting two adjacent intersections on a color image as the mid-span sag point, and further extracting the depth value of this mid-span sag point in the depth map to calculate its three-dimensional spatial physical coordinates, this operation keenly captures the physical essence of the steel mesh's mid-span deflection caused by its own gravity or external forces in actual construction scenarios. Traditional visual inspection often only focuses on the intersections at both ends and ignores the deformation of the middle section of the steel. This solution, by adding the spatial coordinate extraction of the lowest sag point in the mid-span, expands the original single straight line model consisting of only two ends into a three-point spatial structure that can truly reflect the bending characteristics. This expansion of data sampling points for specific sag situations can accurately locate the key position with the largest steel deformation amplitude without large-area dense point cloud scanning. It completely outlines the basic spatial contour of steel sag with extremely low data processing cost and computing power consumption, providing the most core key node data for the subsequent construction of an accurate local curve reconstruction model.

[0058] 5. By setting one of the intersection nodes as the local coordinate origin and calculating the physical projection straight-line distances of the other intersection node and the mid-span drooping point relative to the origin on the horizontal plane, this data processing method cleverly completes the dimensionality reduction transformation of the complex three-dimensional spatial problem into a two-dimensional local plane. Since the drooping and deflection shape of a single steel bar under the combined effects of gravity and other factors is usually concentrated in a vertical cross-section, directly fitting the curve in three-dimensional space is not only extremely cumbersome to calculate but also easily affected by multi-dimensional spatial noise. However, by calculating the physical projection straight-line distance on the horizontal plane, unnecessary redundant information in spatial dimensions is effectively eliminated, providing extremely accurate horizontal coordinate reference data for the subsequent construction of a two-dimensional planar parabolic model. This method of establishing a local coordinate system and performing accurate spatial dimensionality reduction greatly simplifies the subsequent mathematical model derivation and establishment process while ensuring the accuracy of steel bar shape restoration, and significantly improves the running efficiency and adaptability of the entire visual inspection algorithm in complex environments.

[0059] 6. By using the local coordinate origin as the absolute depth reference, and then calculating the relative sag depth difference between the other intersection node and the mid-span sag point relative to the origin, this step rigorously improves the construction of the relative coordinate system within the local two-dimensional vertical section. In actual inspection, directly using absolute depth coordinates is often affected by differences in camera installation height or large fluctuations in the overall elevation of the steel mesh. However, by calculating the depth difference between nodes, the absolute elevation data is successfully and smoothly transformed into relative morphological data that purely reflects the local deflection characteristics of the steel bars themselves. This relative sag depth difference directly constitutes the key input item of the ordinate in the subsequent local deflection curve model, enabling the algorithm to be completely immune to the systematic interference caused by the overall structural elevation translation. It focuses only on analyzing the actual bending deformation amplitude of the steel bars themselves due to local stress, further enhancing the self-consistency of the algorithm's internal data and ensuring the high stability of the morphological feature extraction model in various complex construction site environments.

[0060] 7. By utilizing the calculated straight-line distances of each physical projection and the corresponding differences in relative droop depth, the quadratic and linear coefficients of the local parabolic flexural shape are obtained simultaneously using algebraic analytical methods. This process completely eliminates the iterative optimization algorithm that consumes a lot of low-level computing power in traditional curve fitting techniques. This design cleverly utilizes the spatial relative coordinate properties of the parabola passing through the origin and two other known feature points to directly derive a unique and absolutely accurate analytical solution for the model parameters through rigorous mathematical algebraic equations. This analytical calculation method without iterative optimization not only completely avoids the introduction of external weight parameters that are difficult to interpret and heavily rely on manual experience for debugging, achieving a high degree of self-consistency and a tight closed loop in the data flow within the technical solution, but also, since the overall calculation process only includes basic algebraic operations, the entire detection scheme can achieve extremely fast real-time calculations on ordinary portable devices with very low hardware computing power consumption, thereby efficiently and accurately transforming discrete spatial coordinate points into a continuous mathematical expression of the local parabolic flexural shape.

[0061] 8. By calculating the linear distance interval of the physical projection based on the obtained coefficients of the local parabolic deflection morphology using the arc length integral analytical rule, the true physical arc length of the reinforcing bar is obtained. This is the core technical means to completely overcome the visual projection shortening error in the entire detection method. Faced with the non-planar sagging deformation of the reinforcing bar, the conventional linear distance measurement mechanism will inevitably lead to a serious reduction in the measured length. However, this step simplifies the high-order calculus solution process in advance in the background to a basic algebraic operation formula that includes the substitution of variables for the antiderivative. This operation process of directly substituting the upper and lower limits of the integral to obtain the analytical solution of the definite integral is equivalent to accurately and non-destructively physical straightening and restoring the bent and sagging reinforcing bar in the digital space. This process not only completely avoids the numerical approximation estimation method that the computer is forced to use when processing integral signs, effectively eliminating the cumulative calculation truncation error, but also restores the true physical layout length of the reinforcing bar mesh in the special state of sagging and deflection with unprecedented high precision, giving this detection scheme a strong technical advantage and reliable results.

[0062] 9. By calculating the relative error ratio between the actual physical arc length obtained from the reconstruction and the preset standard grid design spacing, and further outputting the final judgment result of whether the construction quality is qualified or not based on the preset tolerance threshold, this step finally completes the data closed-loop link from front-end machine vision perception to final engineering quality inspection decision; the extremely complex spatial curve distance is uniformly transformed into an intuitive and standardized percentage error index, which can seamlessly connect with the objective allowable deviation requirements in the current construction quality acceptance specifications; at the same time, setting a tolerance threshold for the final data comparison and judgment not only gives the detection algorithm a scientific tolerance for normal and reasonable construction tolerances, avoiding frequent quality false alarms due to extremely small reasonable construction disturbances, but also strictly and keenly controls the quality and safety red line of serious deviation from the drawing design specifications; this standardized result output mechanism makes the originally abstract and profound machine vision and spatial calculus algorithms finally successfully implemented into quantitative judgment instructions that can be directly understood and efficiently executed by front-line acceptance personnel on the construction site, comprehensively ensuring the efficiency and standardization of safety and quality inspection of building structures. Attached Figure Description

[0063] Figure 1 This is a schematic diagram of the basic process of the present invention. Detailed Implementation

[0064] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0065] Example 1, refer to Figure 1 A machine vision-based method for inspecting the construction quality of steel mesh includes:

[0066] A color image and depth map of the steel mesh to be tested are acquired. The intersection nodes of two adjacent steel meshes are located through image processing, and the depth value of the intersection node is extracted from the depth map.

[0067] By combining the camera's intrinsic parameter matrix, the two-dimensional pixel coordinates of the two adjacent intersection nodes and the corresponding extracted depth values ​​are converted into three-dimensional spatial physical coordinates.

[0068] Locate the center position of the line connecting two adjacent intersection nodes on the color image as the mid-span droop point, extract the depth value of the mid-span droop point in the depth map, and calculate its three-dimensional spatial physical coordinates.

[0069] Set one of the intersection nodes as the local coordinate origin, and calculate the physical projection straight-line distance of the other intersection node and the mid-span drooping point relative to the origin on the horizontal plane.

[0070] Using the local coordinate origin as the depth reference, calculate the relative sagging depth difference between the other intersection node and the mid-span sagging point relative to the origin;

[0071] Using the calculated straight-line distances of each of the physical projections and the corresponding differences in relative droop depths, the quadratic and linear coefficients of the local parabolic flexural shape are obtained simultaneously through algebraic analytical solutions.

[0072] Based on the coefficients of the obtained local parabolic flexural shape, the actual physical arc length of the steel bar is obtained by calculating the straight-line distance interval of the physical projection using the arc length integral analytical rule.

[0073] Calculate the relative error ratio between the actual physical arc length and the preset standard grid design spacing, and output the judgment result of whether the construction quality is qualified or not according to the preset tolerance threshold.

[0074] The process of acquiring a color image and depth map of the reinforcing mesh to be tested, locating the intersection nodes of two adjacent reinforcing meshes through image processing, and extracting the depth value of the intersection node from the depth map includes:

[0075] Set up a calibrated RGB-D camera with its lens mounted vertically downwards directly above the steel mesh to be measured; the RGB-D camera simultaneously acquires RGB color images containing the steel mesh and corresponding depth maps;

[0076] Obtain fixed parameters, including: camera focal length, principal point x and y coordinates, and rebar grid spacing threshold;

[0077] Preset error tolerance threshold;

[0078] Based on the two-dimensional pixel coordinates of the located intersection node on the color image, the data corresponding to that position is indexed and extracted from the depth map as the physical depth value of the intersection node in the camera coordinate system.

[0079] The process of combining the camera's intrinsic parameter matrix to convert the acquired two-dimensional pixel coordinates and corresponding extracted depth values ​​of two adjacent intersection nodes into three-dimensional spatial physical coordinates includes:

[0080] The horizontal coordinate difference is obtained by subtracting the horizontal coordinate of the principal point from the horizontal coordinate of one of the intersection nodes in the color image. The horizontal coordinate difference is then multiplied by the physical depth value of the intersection node and divided by the focal length of the camera to obtain the horizontal physical coordinate of the intersection node in three-dimensional space. The vertical coordinate difference is obtained by subtracting the vertical coordinate of the principal point from the vertical coordinate of the intersection node in the color image. The vertical coordinate difference is then multiplied by the physical depth value of the intersection node and divided by the focal length of the camera to obtain the vertical physical coordinate of the intersection node in three-dimensional space.

[0081] Similarly, based on the two-dimensional pixel coordinates and physical depth value of another intersection node located on the same rebar and adjacent to it, the three-dimensional horizontal and vertical physical coordinates of that other intersection node are calculated.

[0082] The step of locating the center position of the line connecting two adjacent intersection nodes on the color image as the mid-span droop point, extracting the depth value of the mid-span droop point in the depth map, and converting it into its three-dimensional spatial physical coordinates includes:

[0083] Calculate the average of the two-dimensional pixel horizontal coordinates of two adjacent intersection nodes in the color image, and use it as the pixel horizontal coordinate of the mid-span downward point;

[0084] Calculate the average of the two-dimensional pixel vertical coordinates of two adjacent intersection nodes in the color image, and use it as the pixel vertical coordinate of the mid-span downward point;

[0085] Based on the calculated pixel horizontal and vertical coordinates of the mid-span sagging point, the data at the corresponding position is extracted from the depth map and used as the physical depth value of the mid-span sagging point.

[0086] Subtract the horizontal coordinate of the principal point from the pixel horizontal coordinate of the mid-span drooping point, multiply by the physical depth value of the mid-span drooping point, and divide by the focal length to obtain the three-dimensional horizontal physical coordinate of the mid-span drooping point.

[0087] Subtract the principal point's ordinate from the pixel's vertical coordinate of the mid-span drooping point, multiply by the physical depth value of the mid-span drooping point, and divide by the focal length to obtain the three-dimensional spatial vertical physical coordinates of the mid-span drooping point.

[0088] The step of setting one of the intersection nodes as the local coordinate origin and calculating the physical projection straight-line distances of the other intersection node and the mid-span drooping point relative to the origin on the horizontal plane includes:

[0089] Calculate the difference between the lateral physical coordinates of the other intersection node and the lateral physical coordinates of the intersection node that serves as the local coordinate origin, and then square the difference.

[0090] Calculate the difference between the longitudinal physical coordinates of the other intersection node and the longitudinal physical coordinates of the intersection node that serves as the local coordinate origin, and then square the difference.

[0091] Add the two squares together and take the square root to obtain the physical projection straight-line distance of the other intersection node relative to the local coordinate origin on the horizontal plane.

[0092] Calculate the difference between the horizontal physical coordinates of the mid-span downward point and the horizontal physical coordinates of the intersection node (which serves as the local coordinate origin), and then square the difference. At the same time, calculate the difference between the vertical physical coordinates of the mid-span downward point and the vertical physical coordinates of the intersection node (which serves as the local coordinate origin), and then square the difference.

[0093] Add the two squares corresponding to the mid-span downward point and take the square root to obtain the physical projection straight-line distance of the mid-span downward point relative to the local coordinate origin on the horizontal plane.

[0094] The step of using the local coordinate origin as a depth reference to calculate the relative sagging depth difference between the other intersection node and the mid-span sagging point relative to the origin includes:

[0095] Subtract the physical depth value of the intersection node that serves as the local coordinate origin from the physical depth value of the other intersection node to obtain the relative droop depth difference of the other intersection node relative to the local coordinate origin.

[0096] Subtracting the physical depth value of the intersection node, which serves as the local coordinate origin, from the physical depth value of the mid-span sagging point yields the relative sagging depth difference between the mid-span sagging point and the local coordinate origin.

[0097] The calculation of the linear distances of each of the physical projections and the corresponding differences in relative sag depths is used to simultaneously solve the quadratic and linear coefficients of the local parabolic deflection shape using algebraic analytical methods, including:

[0098] Multiply the difference in relative sagging depth between the mid-sag points by the physical projection straight-line distance of the other intersection node to obtain the first product term;

[0099] Multiply the difference in relative droop depth of the other intersection node by the physical projected straight-line distance of the mid-span droop point to obtain the second product term;

[0100] Subtract the second product term from the first product term to obtain the numerator of the quadratic term; multiply the square of the physical projection straight-line distance of the mid-span downward point by the physical projection straight-line distance of the other intersection node to obtain the third product term;

[0101] Multiply the square of the physical projection straight-line distance of the other intersection node by the physical projection straight-line distance of the mid-span downward point to obtain the fourth product term;

[0102] Subtracting the fourth product term from the third product term yields the denominator of the quadratic term;

[0103] Dividing the numerator of the quadratic term by the denominator of the quadratic term yields the coefficients of the quadratic term in the local parabolic flexural shape.

[0104] Multiplying the coefficient of the quadratic term by the square of the physical projection straight-line distance of the other intersection node yields the fifth product term;

[0105] Subtract the fifth product term from the relative droop depth difference of the other intersection node, and then divide by the physical projection straight-line distance of the other intersection node to obtain the first-order coefficient of the local parabolic flexural shape.

[0106] The coefficients of the obtained local parabolic deflection shape are used to calculate the actual physical arc length of the reinforcing bar by applying the arc length integral analytical rule to the physical projection straight-line distance interval, including:

[0107] Multiply the coefficient of the quadratic term by two, then multiply by the physical projection straight-line distance of another intersection node, and finally add the coefficient of the linear term to obtain the substitution variable for the upper limit of integration.

[0108] Take the square root of the square of the substitution variable for the upper limit of integration plus one to get the first square term; calculate the natural logarithm by adding the first square term to the substitution variable for the upper limit of integration, and add the product of the substitution variable for the upper limit of integration and the first square term to get the original value of the upper limit of integration.

[0109] Take the square root of the square of the coefficient of the first term plus one to get the second radical term; calculate the natural logarithm by adding the coefficient of the first term to the second radical term, and then add the product of the coefficient of the first term and the second radical term to get the value of the original function with the lower limit of integration.

[0110] Subtracting the original function value of the lower limit of integration from the original function value of the upper limit of integration, and then dividing by four times the coefficient of the quadratic term, yields the true physical arc length of the reinforcing bars between two adjacent intersection nodes.

[0111] The calculation of the relative error ratio between the actual physical unfolded arc length and the preset standard grid design spacing, and the output of the judgment result on whether the construction quality is qualified or not based on the preset tolerance threshold, includes:

[0112] Calculate the absolute value of the difference between the actual physical arc length and the preset standard grid design spacing; divide the absolute value of the difference by the standard grid design spacing, and then multiply by 100% to obtain the percentage deviation of the actual spacing from the standard spacing;

[0113] Extract the preset error tolerance threshold; if the calculated percentage of deviation is less than or equal to the error tolerance threshold, the construction quality is deemed acceptable; if the calculated percentage of deviation is greater than the error tolerance threshold, the construction quality is deemed unacceptable.

[0114] Example 2. A machine vision-based method for inspecting the construction quality of reinforcing mesh, comprising:

[0115] The process of acquiring a color image and depth map of the reinforcing mesh to be tested, locating the intersection nodes of two adjacent reinforcing meshes through image processing, and extracting the depth value of the intersection node from the depth map includes:

[0116] A calibrated RGB-D camera is set up with its lens mounted vertically downwards directly above the steel mesh to be measured; the RGB-D camera simultaneously acquires RGB color images containing the steel mesh and corresponding depth maps.

[0117] Obtain fixed parameters, including: camera focal length. The principal point's horizontal and vertical coordinates represent the two-dimensional pixel coordinate projection position of the camera lens's true optical center onto the image sensor chip. Rebar mesh spacing threshold Preset error tolerance threshold The value is 5%;

[0118] Extract the first using the following formula The depth value corresponding to each node:

[0119] in, To obtain the target quantity, let represent the first... The physical depth values ​​of the intersecting nodes in the camera coordinate system; The amount of data directly measured represents the depth map matrix initially acquired by the RGB-D camera; For directly obtained known quantities, let represent the th The two-dimensional pixel coordinates of the intersection nodes on the image.

[0120] By setting up a calibrated depth camera and simultaneously acquiring color images and corresponding depth maps of the steel mesh to be measured, and then using image processing technology to accurately locate the intersection nodes of two adjacent steel meshes and extract the corresponding depth values ​​from the depth map, this step breaks through the limitations of traditional two-dimensional planar visual inspection technology from the source of data acquisition. Traditional pure two-dimensional images often can only capture planar projection information, which is prone to visual errors when faced with non-planar deformation of the steel mesh caused by gravity or external trampling. However, the method of extracting depth information gives each key intersection node a real spatial distance attribute. This method of integrating the clear texture features of color images with the accurate distance measurement capability of depth maps not only greatly improves the accuracy of key node identification and spatial positioning, but also provides an absolutely reliable initial distance reference for the subsequent accurate reconstruction of the true physical shape of the steel bars in three-dimensional space, effectively avoiding the initial data distortion caused by the shortening of single visual projection.

[0121] The process of combining the camera's intrinsic parameter matrix to convert the acquired two-dimensional pixel coordinates and corresponding extracted depth values ​​of two adjacent intersection nodes into three-dimensional spatial physical coordinates includes:

[0122] The number is calculated using the following formula. The horizontal coordinates of the intersection nodes in three-dimensional space and vertical coordinates :

[0123] in, and To obtain the target quantity, let represent the th The horizontal and vertical coordinates of each node in the 3D camera coordinate system; For the first The physical depth values ​​of the intersecting nodes in the camera coordinate system; For directly obtained known quantities, let represent the th The two-dimensional pixel horizontal coordinates of each intersection node on the image; For directly obtained known quantities, let represent the th The two-dimensional pixel vertical coordinates of each intersection node on the image; For directly obtained known quantities, the x-coordinate of the principal point is represented. is a directly obtained known quantity, representing the ordinate of the principal point; f is a directly obtained known quantity, representing the focal length.

[0124] Similarly, for nodes Another intersection node located on the same rebar and adjacent to each other Calculate its physical depth value and the three-dimensional physical horizontal coordinate with vertical coordinate .

[0125] By combining the camera's intrinsic parameter matrix and converting the two-dimensional pixel coordinates of two adjacent intersection nodes and the corresponding extracted depth values ​​into three-dimensional physical coordinates, this processing successfully achieves a precise transition from virtual image pixel dimensions to real-world physical size dimensions. In actual calculations, the potential skew of the lens's optical center is corrected using the principal point coordinates, and the pixel difference is spatially mapped to the depth data using the camera's focal length parameters. This allows isolated image pixels to be restored to three-dimensional spatial coordinates with absolute physical scale. This rigorous coordinate system transformation mechanism not only completely eliminates the inherent perspective distortion during image acquisition, ensuring that data between nodes can be processed with high precision under the same real physical dimensions, but also constructs a stable and unified three-dimensional physical reference coordinate space for the entire detection scheme. This gives all subsequent morphological analysis and distance measurement extremely accurate real-world physical meaning, fundamentally guaranteeing the rigor and objectivity of the detection data.

[0126] The step of locating the center position of the line connecting two adjacent intersection nodes on the color image as the mid-span droop point, extracting the depth value of the mid-span droop point in the depth map, and converting it into its three-dimensional spatial physical coordinates includes:

[0127] The midpoint depth value is calculated using the following formula:

[0128] in, To obtain the target quantity, let the node be represented. and nodes midpoint The depth value; For directly obtained known quantities, let represent the th The two-dimensional pixel horizontal coordinates of each intersection node on the image; For directly obtained known quantities, let represent the th The two-dimensional pixel vertical coordinates of each intersection node on the image; For directly obtained known quantities, let represent the th The two-dimensional pixel horizontal coordinates of each intersection node on the image; For directly obtained known quantities, let represent the th The two-dimensional pixel vertical coordinates of each intersection node on the image;

[0129] The three-dimensional horizontal and vertical coordinates are calculated using the following formulas:

[0130] in, To obtain the target quantity, let the node be represented. and nodes midpoint The three-dimensional horizontal physical coordinates; To obtain the target quantity, let the node be represented. and nodes midpoint The three-dimensional vertical physical coordinates; For directly obtained known quantities, let represent the th The two-dimensional pixel horizontal coordinates of each intersection node on the image; For directly obtained known quantities, let represent the th The two-dimensional pixel vertical coordinates of each intersection node on the image; For directly obtained known quantities, let represent the th The two-dimensional pixel horizontal coordinates of each intersection node on the image; For directly obtained known quantities, let represent the th The two-dimensional pixel vertical coordinates of each intersection node on the image; The depth value across the midpoint; For directly obtained known quantities, the x-coordinate of the principal point is represented. is a directly obtained known quantity, representing the ordinate of the principal point; f is a directly obtained known quantity, representing the focal length.

[0131] By precisely locating the center point of the line connecting two adjacent intersections on a color image as the mid-span sag point, and further extracting the depth value of this mid-span sag point in the depth map to calculate its three-dimensional spatial physical coordinates, this operation keenly captures the physical essence of the steel mesh's mid-span deflection caused by its own gravity or external forces in actual construction scenarios. Traditional visual inspection often only focuses on the intersections at both ends and ignores the deformation of the middle section of the steel. This solution, by adding the spatial coordinate extraction of the lowest sag point at the mid-span, expands the original single straight line model consisting of only two ends into a three-point spatial structure that can truly reflect the bending characteristics. This expansion of data sampling points for specific sag situations can accurately locate the key position with the largest steel deformation amplitude without large-area dense point cloud scanning. It completely outlines the basic spatial contour of steel sag with extremely low data processing cost and computing power consumption, providing the most core key node data for the subsequent construction of an accurate local curve reconstruction model.

[0132] The step of setting one of the intersection nodes as the local coordinate origin and calculating the physical projection straight-line distances of the other intersection node and the mid-span drooping point relative to the origin on the horizontal plane includes:

[0133] Nodes are calculated using the following formula. To the node Physical straight-line distance on the horizontal plane:

[0134] in, To obtain the target quantity, let the node be represented. To the node Physical straight-line distance on the horizontal plane; and They represent the first The horizontal and vertical coordinates of each node in the 3D camera coordinate system; and They represent the first The horizontal and vertical coordinates of each node in the 3D camera coordinate system;

[0135] Nodes are calculated using the following formula. and nodes midpoint To the node Horizontal straight-line distance:

[0136] in, To obtain the target quantity, let the node be represented. and nodes midpoint To the node The horizontal straight-line distance; and They represent the first The horizontal and vertical coordinates of each node in the 3D camera coordinate system; and Representing nodes respectively and nodes midpoint The three-dimensional horizontal physical coordinates and the three-dimensional vertical physical coordinates.

[0137] By setting one of the intersection nodes as the local coordinate origin and calculating the physical projection straight-line distances of the other intersection node and the mid-span drooping point relative to this origin on the horizontal plane, this data processing method cleverly completes the dimensionality reduction transformation of complex three-dimensional spatial problems into two-dimensional local plane problems. Since the drooping and deflection morphology of a single steel bar under the combined effects of gravity and other factors is usually concentrated in a vertical cross-section, directly fitting the curve in three-dimensional space is not only extremely cumbersome to calculate but also highly susceptible to interference from multi-dimensional spatial noise. However, by calculating the physical projection straight-line distance on the horizontal plane, unnecessary redundant spatial dimensional information is effectively eliminated, providing extremely accurate horizontal coordinate reference data for the subsequent construction of a two-dimensional planar parabolic model. This method of establishing a local coordinate system and performing precise spatial dimensionality reduction greatly simplifies the subsequent mathematical model derivation and establishment process while fully ensuring the accuracy of steel bar morphology restoration, and significantly improves the running efficiency and adaptability of the entire visual inspection algorithm in complex environments.

[0138] The step of using the local coordinate origin as a depth reference to calculate the relative sagging depth difference between the other intersection node and the mid-span sagging point relative to the origin includes:

[0139] Nodes are calculated using the following formula. relative nodes Depth difference:

[0140] in, To obtain the target quantity, let the node be represented. relative nodes The depth difference; For the first The physical depth values ​​of the intersecting nodes in the camera coordinate system; For the first The physical depth values ​​of the intersecting nodes in the camera coordinate system;

[0141] Nodes are calculated using the following formula. and nodes midpoint relative nodes Depth difference:

[0142] in, To obtain the target quantity, let the node be represented. and nodes midpoint relative nodes The depth difference; For the first The physical depth values ​​of the intersecting nodes in the camera coordinate system; For nodes and nodes midpoint The depth value.

[0143] By using the local coordinate origin as the absolute depth reference, and then calculating the relative sag depth difference between another intersection node and the mid-span sag point relative to the origin, this step rigorously improves the construction of the relative coordinate system within the local two-dimensional vertical section. In actual inspection, directly using absolute depth coordinates is often affected by differences in camera installation height or significant fluctuations in the overall elevation of the steel mesh. However, by calculating the depth difference between nodes, the absolute elevation data is successfully and smoothly transformed into relative morphological data that purely reflects the local deflection characteristics of the steel bars. This relative sag depth difference directly constitutes the key input item of the ordinate in the subsequent local deflection curve model, enabling the algorithm to be completely immune to systematic interference caused by the overall structural elevation translation. It focuses only on analyzing the actual bending deformation amplitude of the steel bars due to local stress, further enhancing the self-consistency of the algorithm's internal data and ensuring the high stability of the morphological feature extraction model in various complex construction site environments.

[0144] The calculation of the linear distances of each of the physical projections and the corresponding differences in relative sag depths is used to simultaneously solve the quadratic and linear coefficients of the local parabolic deflection shape using algebraic analytical methods, including:

[0145] The coefficients of the quadratic term of the local parabola are calculated using the following formula:

[0146] in, To obtain the target quantity, let represent the coefficient of the quadratic term of the local parabola; For nodes relative nodes The depth difference; For nodes and nodes midpoint relative nodes The depth difference; For nodes To the node Physical straight-line distance on the horizontal plane; For nodes and nodes midpoint To the node The horizontal straight-line distance;

[0147] The coefficients of the first term of the local parabola are calculated using the following formula:

[0148] in, To obtain the target quantity, let represent the coefficient of the first term of the local parabola; For nodes relative nodes The depth difference; For nodes To the node Physical straight-line distance on the horizontal plane; The coefficient of the quadratic term of the local parabola.

[0149] By utilizing the calculated straight-line distances of each physical projection and the corresponding relative droop depth differences, the quadratic and linear coefficients of the local parabolic flexural shape are obtained simultaneously using algebraic analytical methods. This process completely eliminates the iterative optimization algorithm that consumes a lot of low-level computing power in traditional curve fitting techniques. This design cleverly utilizes the spatial relative coordinate properties of the parabola passing through the origin and two other known feature points to directly derive a unique and absolutely accurate analytical solution for the model parameters through rigorous mathematical algebraic equations. This analytical calculation method without iterative optimization not only completely avoids the introduction of external weight parameters that are difficult to interpret and heavily rely on manual experience for debugging, achieving a high degree of self-consistency and a tight closed loop in the data flow within the technical solution, but also, because the overall calculation process only includes basic algebraic operations, the entire detection scheme can achieve extremely fast real-time calculations on ordinary portable devices with very low hardware computing power consumption, thereby efficiently and accurately transforming discrete spatial coordinate points into a continuous mathematical expression of the local parabolic flexural shape.

[0150] The coefficients of the obtained local parabolic deflection shape are used to calculate the actual physical arc length of the reinforcing bar by applying the arc length integral analytical rule to the physical projection straight-line distance interval, including:

[0151] The intermediate substitution variable at the upper limit of integration is calculated using the following formula:

[0152] in, The target quantity is represented by the intermediate substitution variable at the upper limit of integration; The coefficient of the first term of the local parabola; For nodes To the node Physical straight-line distance on the horizontal plane; The coefficients of the quadratic term of the local parabola;

[0153] The horizontal distance is calculated using the following formula. The upper limit of integration at point: the original function value.

[0154] in, The target quantity is represented by the horizontal distance. The upper limit of the integral at that point is the value of the original function. This is an intermediate substitution variable at the upper limit of integration;

[0155] The value of the antiderivative corresponding to the lower limit of integration at the origin is calculated using the following formula:

[0156] in, The target quantity is represented by the antiderivative value corresponding to the lower limit of integration at the origin; The coefficient of the first term of the local parabola;

[0157] Nodes are calculated using the following formula. and The actual physical mesh spacing between them:

[0158] in, To obtain the target quantity, let the node be represented. and The actual physical mesh spacing between them; For corresponding horizontal distance The upper limit of the integral at that point is the value of the original function. This is the value of the antiderivative corresponding to the lower limit of integration at the origin; The coefficient of the quadratic term of the local parabola.

[0159] By calculating the linear distance interval of the physical projection based on the obtained coefficients of the local parabolic flexural shape, the true physical arc length of the reinforcing bar is obtained by using the arc length integral analytical rule. This is the core technical means to completely overcome the visual projection shortening error in the entire detection method. Faced with the non-planar sagging deformation of the reinforcing bar, the conventional linear distance measurement mechanism will inevitably lead to a serious reduction in the measured length. However, this step simplifies the high-order calculus solution process in advance in the background to a basic algebraic operation formula that includes the substitution of variables for the antiderivative function. This operation process of directly substituting the upper and lower limits of the integral to obtain the analytical solution of the definite integral is equivalent to accurately and non-destructively physical straightening and restoring the bent and sagging reinforcing bar in the digital space. This process not only completely avoids the numerical approximation estimation method that the computer is forced to use when processing integral signs, effectively eliminating the cumulative calculation truncation error, but also restores the true physical layout length of the reinforcing bar mesh in the special state of sagging and flexing with unprecedented high precision, giving this detection scheme a strong technical advantage and reliable results.

[0160] The calculation of the relative error ratio between the actual physical unfolded arc length and the preset standard grid design spacing, and the output of the judgment result on whether the construction quality is qualified or not based on the preset tolerance threshold, includes:

[0161] The percentage deviation of the actual spacing from the standard spacing is calculated using the following formula:

[0162] in, The target quantity is represented by the percentage deviation of the actual spacing from the standard spacing. For nodes and The actual physical mesh spacing between them; For directly obtained known quantities, the standard design spacing is represented;

[0163] Obtain the error tolerance threshold ;

[0164] like If so, the construction quality is deemed acceptable;

[0165] like If so, the construction quality is deemed unqualified.

[0166] By calculating the relative error ratio between the actual physical arc length obtained from the reconstruction and the preset standard grid design spacing, and further outputting the final judgment result of whether the construction quality is qualified or not based on the preset tolerance threshold, this step finally completes the data closed-loop link from front-end machine vision perception to final engineering quality inspection decision. The extremely complex spatial curve distance is uniformly transformed into an intuitive and standardized percentage error index, which can seamlessly connect with the objective allowable deviation requirements in the current construction quality acceptance specifications. At the same time, setting a tolerance threshold for the final data comparison and judgment not only gives the detection algorithm a scientific tolerance for normal and reasonable construction tolerances, avoiding frequent quality false alarms due to extremely small reasonable construction disturbances, but also strictly and keenly controls the quality and safety red line of serious deviation from the drawing design specifications. This standardized result output mechanism makes the originally abstract and profound machine vision and spatial calculus algorithms successfully implemented into quantitative judgment instructions that can be directly understood and efficiently executed by front-line acceptance personnel on the construction site, comprehensively ensuring the efficiency and standardization of safety and quality inspection of building structures.

[0167] By acquiring color images and depth maps of the reinforcing mesh to be tested, extracting the three-dimensional physical coordinates of intersection nodes and mid-span sagging points, and then constructing a local parabolic deflection model, this solution proposes a high-precision construction quality inspection algorithm using the arc length integral analytical method to calculate the true physical arc length of the reinforcing bars. This solution is designed because in the specific environment of a construction site, the tied reinforcing mesh often experiences localized non-planar sagging and deflection due to its own weight, uneven bottom padding, or foot traffic. In this specific environment, conventional machine vision's two-dimensional top-view inspection or simple three-dimensional straight-line distance calculation will produce severe "projection shortening error," easily misjudging actually qualified reinforcing mesh as having excessive spacing. The core unique and beneficial effect of this technical solution lies in: cleverly reducing the dimensionality of the complex three-dimensional bending problem, eliminating reliance on external empirical parameters, and completely avoiding the computationally intensive iterative optimization process, achieving complete data closure and self-consistency; it accurately "straightens" the bent and sagging reinforcing bars in digital space with extremely low computational cost, restoring their true physical layout length. This not only completely overcomes the measurement shortening error caused by visual projection distortion, but also significantly improves the accuracy and reliability of quality acceptance in complex construction environments, avoids misjudgment and omission, and effectively ensures the safety of building structures.

[0168] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0169] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A machine vision-based method for inspecting the construction quality of reinforcing mesh, characterized in that, include: A color image and depth map of the steel mesh to be tested are acquired. The intersection nodes of two adjacent steel meshes are located through image processing, and the depth value of the intersection node is extracted from the depth map. By combining the camera's intrinsic parameter matrix, the two-dimensional pixel coordinates of the two adjacent intersection nodes and the corresponding extracted depth values ​​are converted into three-dimensional spatial physical coordinates. Locate the center position of the line connecting two adjacent intersection nodes on the color image as the mid-span droop point, extract the depth value of the mid-span droop point in the depth map, and calculate its three-dimensional spatial physical coordinates. Set one of the intersection nodes as the local coordinate origin, and calculate the physical projection straight-line distance of the other intersection node and the mid-span drooping point relative to the origin on the horizontal plane. Using the local coordinate origin as the depth reference, calculate the relative sagging depth difference between the other intersection node and the mid-span sagging point relative to the origin; Using the calculated straight-line distances of each of the physical projections and the corresponding differences in relative droop depths, the quadratic and linear coefficients of the local parabolic flexural shape are obtained simultaneously through algebraic analytical solutions. Based on the coefficients of the obtained local parabolic flexural shape, the actual physical arc length of the steel bar is obtained by calculating the straight-line distance interval of the physical projection using the arc length integral analytical rule. Calculate the relative error ratio between the actual physical arc length and the preset standard grid design spacing, and output the judgment result of whether the construction quality is qualified or not according to the preset tolerance threshold.

2. The method for inspecting the construction quality of reinforcing mesh based on machine vision according to claim 1, characterized in that, The process of acquiring a color image and depth map of the reinforcing mesh to be tested, locating the intersection nodes of two adjacent reinforcing meshes through image processing, and extracting the depth value of the intersection node from the depth map includes: Set up a calibrated RGB-D camera with its lens mounted vertically downwards directly above the steel mesh to be measured; the RGB-D camera simultaneously acquires RGB color images containing the steel mesh and corresponding depth maps; Obtain fixed parameters, including: camera focal length, principal point x and y coordinates, and rebar grid spacing threshold; Preset error tolerance threshold; Based on the two-dimensional pixel coordinates of the located intersection node on the color image, the data corresponding to that position is indexed and extracted from the depth map as the physical depth value of the intersection node in the camera coordinate system.

3. The method for inspecting the construction quality of reinforcing mesh based on machine vision according to claim 2, characterized in that, The process of combining the camera's intrinsic parameter matrix to convert the acquired two-dimensional pixel coordinates and corresponding extracted depth values ​​of two adjacent intersection nodes into three-dimensional spatial physical coordinates includes: The horizontal coordinate difference is obtained by subtracting the horizontal coordinate of the principal point from the horizontal coordinate of one of the intersection nodes in the color image. The horizontal coordinate difference is then multiplied by the physical depth value of the intersection node and divided by the focal length of the camera to obtain the horizontal physical coordinate of the intersection node in three-dimensional space. The vertical coordinate difference is obtained by subtracting the vertical coordinate of the principal point from the vertical coordinate of the intersection node in the color image. The vertical coordinate difference is then multiplied by the physical depth value of the intersection node and divided by the focal length of the camera to obtain the vertical physical coordinate of the intersection node in three-dimensional space. Similarly, based on the two-dimensional pixel coordinates and physical depth value of another intersection node located on the same rebar and adjacent to it, the three-dimensional horizontal and vertical physical coordinates of that other intersection node are calculated.

4. The method for inspecting the construction quality of reinforcing mesh based on machine vision according to claim 3, characterized in that, The step of locating the center position of the line connecting two adjacent intersection nodes on the color image as the mid-span droop point, extracting the depth value of the mid-span droop point in the depth map, and converting it into its three-dimensional spatial physical coordinates includes: Calculate the average of the two-dimensional pixel horizontal coordinates of two adjacent intersection nodes in the color image, and use it as the pixel horizontal coordinate of the mid-span downward point; Calculate the average of the two-dimensional pixel vertical coordinates of two adjacent intersection nodes in the color image, and use it as the pixel vertical coordinate of the mid-span downward point; Based on the calculated pixel horizontal and vertical coordinates of the mid-span sagging point, the data at the corresponding position is extracted from the depth map and used as the physical depth value of the mid-span sagging point. Subtract the horizontal coordinate of the principal point from the pixel horizontal coordinate of the mid-span drooping point, multiply by the physical depth value of the mid-span drooping point, and divide by the focal length to obtain the three-dimensional horizontal physical coordinate of the mid-span drooping point. Subtract the principal point's ordinate from the pixel's vertical coordinate of the mid-span drooping point, multiply by the physical depth value of the mid-span drooping point, and divide by the focal length to obtain the three-dimensional spatial vertical physical coordinates of the mid-span drooping point.

5. The method for inspecting the construction quality of reinforcing mesh based on machine vision according to claim 4, characterized in that, The step of setting one of the intersection nodes as the local coordinate origin and calculating the physical projection straight-line distances of the other intersection node and the mid-span drooping point relative to the origin on the horizontal plane includes: Calculate the difference between the lateral physical coordinates of the other intersection node and the lateral physical coordinates of the intersection node that serves as the local coordinate origin, and then square the difference. Calculate the difference between the longitudinal physical coordinates of the other intersection node and the longitudinal physical coordinates of the intersection node that serves as the local coordinate origin, and then square the difference. Add the two squares together and take the square root to obtain the physical projection straight-line distance of the other intersection node relative to the local coordinate origin on the horizontal plane. Calculate the difference between the horizontal physical coordinates of the mid-span downward point and the horizontal physical coordinates of the intersection node (which serves as the local coordinate origin), and then square the difference. At the same time, calculate the difference between the vertical physical coordinates of the mid-span downward point and the vertical physical coordinates of the intersection node (which serves as the local coordinate origin), and then square the difference. Add the two squares corresponding to the mid-span downward point and take the square root to obtain the physical projection straight-line distance of the mid-span downward point relative to the local coordinate origin on the horizontal plane.

6. The method for inspecting the construction quality of reinforcing mesh based on machine vision according to claim 5, characterized in that, The step of using the local coordinate origin as a depth reference to calculate the relative sagging depth difference between the other intersection node and the mid-span sagging point relative to the origin includes: Subtract the physical depth value of the intersection node that serves as the local coordinate origin from the physical depth value of the other intersection node to obtain the relative droop depth difference of the other intersection node relative to the local coordinate origin. Subtracting the physical depth value of the intersection node, which serves as the local coordinate origin, from the physical depth value of the mid-span sagging point yields the relative sagging depth difference between the mid-span sagging point and the local coordinate origin.

7. The method for inspecting the construction quality of reinforcing mesh based on machine vision according to claim 6, characterized in that, The calculation of the linear distances of each of the physical projections and the corresponding differences in relative sag depths is used to simultaneously solve the quadratic and linear coefficients of the local parabolic deflection shape using algebraic analytical methods, including: Multiply the difference in relative sagging depth between the mid-sag points by the physical projection straight-line distance of the other intersection node to obtain the first product term; Multiply the difference in relative droop depth of the other intersection node by the physical projected straight-line distance of the mid-span droop point to obtain the second product term; Subtract the second product term from the first product term to obtain the numerator of the quadratic term; multiply the square of the physical projection straight-line distance of the mid-span downward point by the physical projection straight-line distance of the other intersection node to obtain the third product term; Multiply the square of the physical projection straight-line distance of the other intersection node by the physical projection straight-line distance of the mid-span downward point to obtain the fourth product term; Subtracting the fourth product term from the third product term yields the denominator of the quadratic term; Dividing the numerator of the quadratic term by the denominator of the quadratic term yields the coefficients of the quadratic term in the local parabolic flexural shape. Multiplying the coefficient of the quadratic term by the square of the physical projection straight-line distance of the other intersection node yields the fifth product term; Subtract the fifth product term from the relative droop depth difference of the other intersection node, and then divide by the physical projection straight-line distance of the other intersection node to obtain the first-order coefficient of the local parabolic flexural shape.

8. The method for inspecting the construction quality of reinforcing mesh based on machine vision according to claim 7, characterized in that, The coefficients of the obtained local parabolic deflection shape are used to calculate the actual physical arc length of the reinforcing bar by applying the arc length integral analytical rule to the physical projection straight-line distance interval, including: Multiply the coefficient of the quadratic term by two, then multiply by the physical projection straight-line distance of another intersection node, and finally add the coefficient of the linear term to obtain the substitution variable for the upper limit of integration. Take the square root of the square of the substitution variable for the upper limit of integration plus one to get the first square term; calculate the natural logarithm by adding the first square term to the substitution variable for the upper limit of integration, and add the product of the substitution variable for the upper limit of integration and the first square term to get the original value of the upper limit of integration. Take the square root of the square of the coefficient of the first term plus one to get the second radical term; calculate the natural logarithm by adding the coefficient of the first term to the second radical term, and then add the product of the coefficient of the first term and the second radical term to get the value of the original function with the lower limit of integration. Subtracting the original function value of the lower limit of integration from the original function value of the upper limit of integration, and then dividing by four times the coefficient of the quadratic term, yields the true physical arc length of the reinforcing bars between two adjacent intersection nodes.

9. A method for inspecting the construction quality of reinforcing mesh based on machine vision according to claim 8, characterized in that, The calculation of the relative error ratio between the actual physical unfolded arc length and the preset standard grid design spacing, and the output of the judgment result on whether the construction quality is qualified or not based on the preset tolerance threshold, includes: Calculate the absolute value of the difference between the actual physical arc length and the preset standard grid design spacing; divide the absolute value of the difference by the standard grid design spacing, and then multiply by 100% to obtain the percentage deviation of the actual spacing from the standard spacing; Extract the preset error tolerance threshold; if the calculated percentage of deviation is less than or equal to the error tolerance threshold, the construction quality is deemed acceptable; if the calculated percentage of deviation is greater than the error tolerance threshold, the construction quality is deemed unacceptable.

Images