A Smart Measurement Method for Cross-Sectional Reinforcing Mesh Dimensions

Abstract

The invention discloses an intelligent measurement method for the size of a cross-sectional steel mesh. The method is based on machine vision and artificial intelligence technology and is used in an automatic production line of a steel mesh to realize intelligent measurement of the size of a cross-sectional steel mesh. The method comprises the following steps: image acquisition, placing the steel mesh flat on a base with a white background, and using three horizontal industrial cameras mounted on a gantry to shoot downward at a vertical angle at a fixed interval, the gantry being arranged on tracks on both sides to slide, thereby ensuring that a large steel mesh can be photographed with a complete cross-sectional image by the camera, image preprocessing, edge detection, elimination of steel bar transverse ribs by an extreme value method, classification processing of transverse and longitudinal steel bar pixels, Hough linear transformation, and calculation of the overall size of the steel mesh. The intelligent overall measurement of the cross-sectional steel mesh is creatively realized, and the accuracy, speed, and degree of automation meet actual production needs.

Description

Technical Field

[0001] The invention relates to the technical field of intelligent measurement, and in particular to an intelligent measurement method for the size of a cross-sectional steel mesh. Background Art

[0002] In railway engineering construction, the widespread use of simply supported box girder structures has made steel mesh a key supporting skeleton material. These steel meshes are usually processed, arranged and tied manually, which is not only time-consuming and labor-intensive, but also highly dependent on the workers' operating skills. Therefore, the use of automated steel mesh welding equipment for production has become a necessary means to solve this problem.

[0003] However, during the production process, due to the influence of process steps such as straightening, shearing, welding and bending, the size of the steel mesh often deviates from the design requirements. These deviations may affect the performance of the simply supported box girder and may even cause safety hazards. For large steel meshes with a size of 30 meters by 15 meters, traditional manual inspection methods are time-consuming and cannot guarantee sufficient accuracy, while traditional small object size inspection methods are also inadequate in this case.

[0004] In addition, in a factory environment, the image quality and recognition accuracy are affected by the special requirements of lighting conditions and image acquisition. Although 3D point cloud modeling provides a possible solution, it is time-consuming and has limited accuracy. Therefore, it is particularly urgent to develop a fast and accurate size detection technology for large steel mesh sheets such as 30 meters by 15 meters to ensure the integrity and safety of the structure. Summary of the invention

[0005] The purpose of the present invention is to provide an intelligent measurement method for the size of a cross-sectional steel mesh to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A cross-sectional steel mesh size intelligent measurement method comprises the following steps:

[0008] Step 1: Image acquisition

[0009] Place the steel mesh flat on a base with a white background, and use three horizontal industrial cameras mounted on a gantry to shoot downward at a fixed interval from a vertical perspective. The gantry is placed on the tracks on both sides to slide, so as to ensure that the large steel mesh can be captured by the camera with a complete cross-sectional image;

[0010] Step 2: Image preprocessing

[0011] The steel mesh image composed of RGB colors is assigned different weights to each color channel using the weighted average method. The final gray value is Gray, Gray = 0.299 × R + 0.587 × G + 0.114 × B, by multiplying the gray values ​​of the three components by the corresponding weights and summing them up.

[0012] Multi-scale bilateral filtering is used to decompose the steel mesh image into different scale levels, and then bilateral filtering is applied independently at each level to adapt to the local features of the image at different scales, thereby maintaining the important structural information of the image while denoising;

[0013] Step 3: Edge Detection

[0014] The optimized Canny algorithm is used to detect the edge of the steel mesh image. The traditional Canny edge detection algorithm includes Gaussian filtering denoising, 2-direction Sobel operator gradient calculation, where the 2 directions refer to the horizontal and vertical directions, non-maximum suppression, double threshold detection, and edge tracking. We change the 2-direction Sobel operator gradient calculation to 8-direction Sobel operator gradient calculation, and use an innovative method to combine the gradient vector voting method with the 8-direction Sobel operator to further improve the accuracy of the image gradient. The threshold is calculated using the maximum inter-class variance Otsu algorithm instead of the double threshold detection.

[0015] Step 4: Remove the steel bar transverse ribs using the extreme value method

[0016] In the previous step, we determined the pixel points at the edge of the steel mesh through edge detection, and then used a quadratic polynomial to fit the local area of ​​the edge pixel points, and calculated the curvature K of the polynomial. The positive and negative values ​​of the curvature are used to distinguish between convex ribs and concave points. Positive curvature indicates convex points, and negative curvature indicates concave points. After the ribs are identified, the values ​​of these points in the image are set to the background color to remove them. The background color is black or white. When the starting threshold is set to 0.07, more potential transverse ribs can be captured;

[0017] Step 5: Steel Mesh Image Stitching

[0018] Calibrate three fixed-position industrial cameras to determine their intrinsic and extrinsic parameters;

[0019] Calculate the expected overlap area of ​​the images based on the field of view of each industrial camera and the fixed spacing between the cameras, and use the camera calibration data to more accurately determine the pixel range of the overlap area;

[0020] Horizontal direction: The field of view width W is calculated using the horizontal field of view angle HFOV of each camera and the distance D to the shooting plane. The formula is: To solve the problems raised in the above background technology.

[0021] To achieve the above object, the present invention provides the following technical solutions:

[0022] A cross-sectional steel mesh size intelligent measurement method comprises the following steps:

[0023] Step 1: Image acquisition

[0024] Place the steel mesh flat on a base with a white background, and use three horizontal industrial cameras mounted on a gantry to shoot downward at a fixed interval from a vertical perspective. The gantry is placed on the tracks on both sides to slide, so as to ensure that the large steel mesh can be captured by the camera with a complete cross-sectional image;

[0025] Step 2: Image preprocessing

[0026] The steel mesh image composed of RGB colors is assigned different weights to each color channel using the weighted average method. The final gray value is Gray, Gray = 0.299 × R + 0.587 × G + 0.114 × B, by multiplying the gray values ​​of the three components by the corresponding weights and summing them up.

[0027] Multi-scale bilateral filtering is used to decompose the steel mesh image into different scale levels, and then bilateral filtering is applied independently at each level to adapt to the local features of the image at different scales, thereby maintaining the important structural information of the image while denoising;

[0028] Step 3: Edge Detection

[0029] The optimized Canny algorithm is used to detect the edge of the steel mesh image. The traditional Canny edge detection algorithm includes Gaussian filtering denoising, 2-direction Sobel operator gradient calculation, where the 2 directions refer to the horizontal and vertical directions, non-maximum suppression, double threshold detection, and edge tracking. We change the 2-direction Sobel operator gradient calculation to 8-direction Sobel operator gradient calculation, and use an innovative method to combine the gradient vector voting method with the 8-direction Sobel operator to further improve the accuracy of the image gradient. The threshold is calculated using the maximum inter-class variance Otsu algorithm instead of the double threshold detection.

[0030] Step 4: Remove the steel bar transverse ribs using the extreme value method

[0031] In the previous step, we determined the pixel points at the edge of the steel mesh through edge detection, and then used a quadratic polynomial to fit the local area of ​​the edge pixel points, and calculated the curvature K of the polynomial. The positive and negative values ​​of the curvature are used to distinguish between convex ribs and concave points. Positive curvature indicates convex points, and negative curvature indicates concave points. After the ribs are identified, the values ​​of these points in the image are set to the background color to remove them. The background color is black or white. When the starting threshold is set to 0.07, more potential transverse ribs can be captured;

[0032] Step 5: Steel Mesh Image Stitching

[0033] Calibrate three fixed-position industrial cameras to determine their intrinsic and extrinsic parameters;

[0034] Calculate the expected overlap area of ​​the images based on the field of view of each industrial camera and the fixed spacing between the cameras, and use the camera calibration data to more accurately determine the pixel range of the overlap area;

[0035] Horizontal direction: The field of view width W is calculated using the horizontal field of view angle HFOV of each camera and the distance D to the shooting plane. The formula is:

[0036] For two adjacent cameras, the field of view overlap width O is determined by the formula O = WS, where S is the fixed spacing between the cameras. Then, the number of pixels per unit length P is calculated based on the camera resolution and the field of view overlap width, and the formula is used:

[0037] O pixels =O×P

[0038] O pixels Represents the number of pixels after the conversion overlap width;

[0039] Vertical direction: Use encoder to synchronize the position of the moving steel mesh and camera shooting to determine the overlapping area. The encoder is installed on the track of the steel mesh to accurately track the moving distance and ensure that its output can be read and recorded by the camera system or control system;

[0040] Step 6: Pixel classification of horizontal and vertical steel bars

[0041] Based on the spatial distribution characteristics that the same horizontal rib has similar vertical coordinates and the same vertical rib has similar horizontal coordinates, the obtained pixel points are analyzed and classified, points are selected in the array for comparison and judgment one by one, and the elements that meet the conditions are selected to form a new classification array and are removed from the original array at the same time to shorten the next retrieval time. The number of elements in the newly obtained classification array is statistically judged, and it is judged as the pixel edge information of a horizontal rib only when the number is greater than a certain threshold. By using this method, the pixel edge information of each horizontal rib can be obtained in turn;

[0042] Step 7: Hough Linear Transform

[0043] Through the above steps, the pixel edge information of each horizontal and vertical steel bar has been obtained. Hough transform straight line detection is performed on each steel bar, that is, each set of pixel coordinates, so that the spatial position information of each steel bar can be obtained more intuitively. Due to the structural characteristics of the steel mesh, the straight line inclination angle of the horizontal bar is about 0°, and the inclination angle of the total longitudinal bar is about 90°. Some slope constraints can be appropriately added to the Hough transform to make the detection more accurate and improve the detection speed;

[0044] Step 8. Calculate the overall size of the steel mesh

[0045] The edge straight lines of the steel mesh are converted into mathematical straight line equations through the improved Hough transform. Each straight line will be stored in the computer memory in the form of parameters. The straight line equations are converted into a set of linear equations, and the coordinates of the intersection points of each straight line equation are calculated. The pixel spacing between the horizontal bars of the steel mesh is obtained by the difference in the horizontal coordinates of the intersection points, and the pixel spacing between the longitudinal bars of the steel mesh is obtained by the difference in the vertical coordinates of the intersection points. Then the real size conversion method is used, and the real distance is obtained according to the actual distance = pixel spacing × conversion factor, where the conversion factor = length / measured length or pixel.

[0046] As a further solution of the present invention: in step three, in order to further improve the accuracy of image gradients, the algorithm introduces an innovative gradient vector voting method. In this method, the gradient vector of each pixel is not only calculated by the Sobel operator, but also votes on the surrounding pixels according to its direction and size, thereby enhancing the continuity and clarity of the edge. The improved algorithm uses the maximum inter-class variance Otsu algorithm to automatically determine the optimal threshold in the dual threshold detection step. The Otsu algorithm selects the threshold by maximizing the inter-class variance between foreground and background pixels, which can more accurately distinguish between edge and non-edge areas, and is more intelligent and efficient than the traditional dual threshold method.

[0047] As a further solution of the present invention: in step five, the internal parameters include: focal length, principal point, distortion coefficient, and the external parameters include position and posture relative to a global reference frame.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] In view of the complexity of large-scale steel mesh measurement and the inability to meet the measurement accuracy, the present invention adopts a cross-sectional steel mesh measurement method through multi-camera shooting, avoiding the problems of insufficient field of view and insufficient resolution of a single camera, and ensuring the measurement requirements; in view of the inability of traditional canny edge detection to comprehensively analyze the edge features of steel mesh images, the present invention uses an 8-direction Sobel operator combined with a voting method and Otus threshold calculation to better extract the steel mesh contour;

[0050] In view of the problem that there are few methods for measuring the size of cross-sectional steel mesh, the present invention creatively realizes the overall size detection of steel mesh through improved Canny edge detection, extreme value method to eliminate steel cross ribs, image splicing, size conversion and other steps, and the speed and accuracy meet actual needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 The figure is an overall flow chart of an intelligent measurement method for cross-sectional steel mesh size.

[0052] Figure 2 The invention discloses a steel mesh image acquisition environment for an intelligent measurement method of cross-sectional steel mesh size.

[0053] Figure 3 The present invention is a steel mesh image of a specific embodiment of a cross-sectional steel mesh size intelligent measurement method.

[0054] Figure 4 The present invention is a pre-processed image of a steel mesh of a specific embodiment of a method for intelligently measuring the size of a cross-sectional steel mesh.

[0055] Figure 5 This is an edge detection image of a specific embodiment of a method for intelligently measuring the size of a cross-sectional steel mesh.

[0056] Figure 6 It is an inspection result image of a specific embodiment of an intelligent measurement method for cross-sectional steel mesh size. DETAILED DESCRIPTION

[0057] The present invention is further explained below in conjunction with the embodiments, i.e. the accompanying drawings, but this is not intended to limit the scope of protection of the present application.

[0058] In addition, in order to better illustrate the present application, numerous specific details are provided in the specific embodiments below. It should be understood by those skilled in the art that the present application can also be implemented without certain specific details. In some examples, methods, means, and elements well known to those skilled in the art are not described in detail in order to highlight the subject matter of the present application.

[0059] Step 1: Image acquisition

[0060] As the steel mesh is 30 meters long and 15 meters wide with large horizontal and vertical spans, the camera cannot obtain a complete and clear image of the entire steel mesh. Therefore, multiple mobile cameras are used to collect cross-sectional images of the steel bars in different parts of the steel mesh.

[0061] Fixed section measurement methods were selected in the longitudinal and transverse directions. Four sections were taken in the longitudinal direction, two at the ends, and two section measurement points were set at a distance of about 10m in the middle; three section measurement points were set in the transverse direction, one on each side, and a measurement interface was set 7.5m away from the side.

[0062] As in the longitudinal position, a mobile rail vehicle is set up, and three cameras are set up on the crossbeam. Cameras 1 and 3 cover the boundary 2 side areas of a measurement interface of the box beam, and camera 2 measures the middle area of ​​the interface to form image information of a measurement interface. The rail vehicle moves along the longitudinal direction of the steel mesh. During the movement, four different positions are selected to collect cross-sectional images of four different areas at the longitudinal ends and the middle position of about 10 cm of the steel mesh.

[0063] Step 2: Image preprocessing

[0064] The steel mesh image composed of RGB colors is assigned different weights to each color channel using the weighted average method, where R, G, and B represent the brightness values ​​of the Red channel, Green channel, and Blue channel, respectively. The final gray value is Gray, Gray = 0.299 × R + 0.587 × G + 0.114 × B, by multiplying the gray values ​​of the three components by the corresponding weights and summing them up.

[0065] Multi-scale bilateral filtering is used to decompose the steel mesh image into different scale levels, and the original image is downsampled to different resolutions to produce a series of images with varying scales. Each scale image captures the details and features of the original image at different levels. Lower scale (i.e., lower resolution) images reveal broader structural features, while higher scale (i.e., higher resolution) images retain more subtle details. Bilateral filtering is then applied independently at each level to adapt to the local features of the image at different scales, thereby smoothing out the noise in the image while maintaining important structural information of the image.

[0066] Step 3: Edge Detection

[0067] The optimized canny algorithm is used to detect the edge of the steel mesh image. The traditional canny edge detection algorithm includes Gaussian filtering denoising, 2-direction Sobel operator for gradient, where the 2 directions refer to the horizontal and vertical directions, non-maximum suppression, double threshold detection, and edge tracking. In this optimized algorithm, the traditional 2-direction Sobel operator (only considering the horizontal and vertical directions) is replaced by an 8-direction Sobel operator. This means that when the algorithm calculates the gradient, it not only considers the horizontal and vertical directions, but also the diagonal and other intermediate directions. This method can more comprehensively capture the edge information in the image, especially those edges in non-main directions.

[0068] The traditional Sobel operator convolution module in the two directions is:

[0069]

[0070]

[0071] in and is the convolution of the roll factor and the grayscale image f, θ is the gradient direction, is the gradient magnitude.

[0072] In the horizontal and vertical directions of the traditional Sobel operator, 3x3 convolution templates in other directions (including 45°, 135°, 180°, 225°, 270° and 315°) are added to obtain the gradient map. The 8-direction Sobel operator 3x3 convolution template is:

[0073] Based on the original steel mesh gradient map, the gradient vector voting method is used. This method is based on the fact that the pixels of the steel mesh image in a specific gradient direction are not isolated points on the edge, but part of the adjacent pixels on the edge. Therefore, we "vote" the surrounding pixels of the steel mesh image in the gradient direction to strengthen their confidence and weight that they may belong to the edge.

[0074] First, the voting matrix is ​​initialized and all its elements are set to zero. Then, according to the gradient direction of each pixel, the gradient magnitude is voted for the corresponding adjacent pixels. Finally, the voting matrix is ​​used to strengthen the adjacent pixels in the gradient direction in the image to improve the accuracy and robustness of edge detection.

[0075] Step 4: Remove the steel bar transverse ribs using the extreme value method

[0076] In the previous step, we determined the pixel points at the edge of the steel mesh through edge detection, and then used a quadratic polynomial to fit the local area of ​​the edge pixel points, calculated the curvature K of the polynomial, and used the positive and negative values ​​of the curvature to distinguish between convex points (ribs) and concave points. Positive curvature indicates convex points, and negative curvature indicates concave points. After the ribs are identified, the values ​​of these points in the image are set to the background color (usually black or white) to remove them. Setting the starting threshold to 0.07 can capture more potential transverse ribs.

[0077] Step 5: Steel Mesh Image Stitching

[0078] Three fixed-position industrial cameras are calibrated to determine their intrinsic parameters (such as focal length, principal point, distortion coefficient) and extrinsic parameters.

[0079] Calculate overlapping area

[0080] The expected overlap of the images is calculated based on the field of view of each camera and the fixed spacing between the cameras. The camera calibration data is used to more accurately determine the pixel range of the overlap area.

[0081] Horizontal direction: The field of view width W is calculated using the horizontal field of view (HFOV) of each camera and the distance D to the shooting plane. For two adjacent cameras, the field of view overlap width O is determined by the formula O=WS, where S is the fixed spacing between cameras. Then, the number of pixels per unit length P is calculated based on the camera resolution and the field of view width, and the overlap width is converted to pixels using the formula.

[0082] The horizontal field of view (HFOV) of each camera is 90 degrees.

[0083] The distance D from the camera to the shooting plane is 10 meters.

[0084] The fixed distance S between cameras is 15 meters.

[0085] The camera resolution is 4000 pixels wide.

[0086] First, we calculate the actual width W of a single field of view:

[0087] Then, we calculate the overlap width O of the two camera fields of view: O = WS O = 20-15 O = 5 meters Now, convert the overlap width O to pixels O pixels :

[0088] We need to determine how many pixels there are per meter. If the camera's field of view is 20 meters wide and the width is 4000 pixels, then there are 200 pixels per meter.

[0089] O pixels =O×PO pixels =O×PO pixels =1000 pixels

[0090] This results in an overlap of approximately 1000 pixels between each camera image.

[0091] Vertical direction: Use encoders to synchronize the position of the moving steel mesh and camera shooting to determine the overlap area. The encoder is installed on the track of the steel mesh to accurately track the moving distance and ensure that its output can be read and recorded by the camera system or control system.

[0092] Step 6: Pixel classification of horizontal and vertical steel bars

[0093] Based on the spatial distribution characteristics that the same horizontal rib has similar vertical coordinates and the same vertical rib has similar horizontal coordinates, the obtained pixel points are analyzed and classified. Points are selected in the array for comparison and judgment one by one, and the elements that meet the conditions are selected to form a new classification array and are removed from the original array at the same time to shorten the next retrieval time. The number of elements in the newly obtained classification array is statistically judged, and it is judged as the pixel edge information of a horizontal rib only when the number is greater than a certain threshold. This method can be used to obtain the pixel edge information of each horizontal rib in turn.

[0094] Step 7: Hough Linear Transform

[0095] Through the above work, the pixel edge information of each horizontal and vertical steel bar has been obtained. Hough transform straight line detection is performed on each steel bar, that is, each set of pixel coordinates, to obtain the spatial position information of each steel bar more intuitively. Due to the structural characteristics of the steel mesh, the straight line inclination angle of the horizontal bar is about 0°, and the inclination angle of the total longitudinal bar is about 90°. Some slope constraints can be appropriately added to the Hough transform to make the detection more accurate and improve the detection speed.

[0096] Step 8. Calculate the overall size of the steel mesh

[0097] The edge straight line of the steel mesh is converted into a mathematical straight line equation through the improved Hough transform. Each straight line is stored in the computer memory in the form of parameters. The straight line equation is converted into a linear equation system, and the coordinates of the intersection of each straight line equation are calculated. The pixel spacing between the horizontal bars of the steel mesh is obtained by the difference of the horizontal coordinates of the intersection, and the pixel spacing between the vertical bars of the steel mesh is obtained by the difference of the vertical coordinates of the intersection. Then the real size conversion method is used, and the real distance is obtained according to the actual distance = pixel spacing × conversion factor, where the conversion factor = length / measured length (pixel), where the conversion factor is 0.1 mm / pixel.

[0098] Some results of the actual distance:

[0099]

[0100] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the present invention can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, the embodiments should be considered exemplary and non-restrictive in all respects, and the scope of the present invention is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims be included in the present invention.

[0101] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A cross-sectional steel mesh size intelligent measurement method, characterized in that: The following steps are involved: Step 1: Image acquisition Place the steel mesh flat on a base with a white background, and use three horizontal industrial cameras mounted on a gantry to shoot downward at a fixed interval from a vertical perspective. The gantry is placed on the tracks on both sides to slide, so as to ensure that the large steel mesh can be captured by the camera with a complete cross-sectional image; Step 2: Image preprocessing The steel mesh image composed of RGB colors is assigned different weights to each color channel using the weighted average method. The final gray value is Gray, Gray = 0.299 × R + 0.587 × G + 0.114 × B, by multiplying the gray values ​​of the three components by the corresponding weights and summing them up. Multi-scale bilateral filtering is used to decompose the steel mesh image into different scale levels, and then bilateral filtering is applied independently at each level to adapt to the local features of the image at different scales, thereby maintaining the important structural information of the image while denoising; Step 3: Edge Detection The optimized Canny algorithm is used to detect the edge of the steel mesh image. The traditional Canny edge detection algorithm includes Gaussian filtering denoising, 2-direction Sobel operator gradient calculation, where the 2 directions refer to the horizontal and vertical directions, non-maximum suppression, double threshold detection, and edge tracking. We change the 2-direction Sobel operator gradient calculation to 8-direction Sobel operator gradient calculation, and use an innovative method to combine the gradient vector voting method with the 8-direction Sobel operator to further improve the accuracy of the image gradient. The threshold is calculated using the maximum inter-class variance Otsu algorithm instead of the double threshold detection. Step 4: Remove the steel bar transverse ribs using the extreme value method In the previous step, we determined the pixel points at the edge of the steel mesh through edge detection, and then used a quadratic polynomial to fit the local area of ​​the edge pixel points, and calculated the curvature K of the polynomial. The positive and negative values ​​of the curvature are used to distinguish between convex ribs and concave points. Positive curvature indicates convex points, and negative curvature indicates concave points. After the ribs are identified, the values ​​of these points in the image are set to the background color to remove them. The background color is black or white. When the starting threshold is set to 0.07, more potential transverse ribs can be captured; Step 5: Steel Mesh Image Stitching Calibrate three fixed-position industrial cameras to determine their intrinsic and extrinsic parameters; Calculate the expected overlap area of ​​the images based on the field of view of each industrial camera and the fixed spacing between the cameras, and use the camera calibration data to more accurately determine the pixel range of the overlap area; Horizontal direction: The field of view width W is calculated using the horizontal field of view angle HFOV of each camera and the distance D to the shooting plane. The formula is: For two adjacent cameras, the field of view overlap width O is determined by the formula O = WS, where S is the fixed spacing between the cameras. Then, the number of pixels per unit length P is calculated based on the camera resolution and the field of view overlap width, and the formula is used: O pixels =O×P;O pixels Represents the number of pixels after the conversion overlap width; Vertical direction: Use encoder to synchronize the position of the moving steel mesh and camera shooting to determine the overlapping area. The encoder is installed on the track of the steel mesh to accurately track the moving distance and ensure that its output can be read and recorded by the camera system or control system; Step 6: Pixel classification of horizontal and vertical steel bars Based on the spatial distribution characteristics that the same horizontal rib has similar vertical coordinates and the same vertical rib has similar horizontal coordinates, the obtained pixel points are analyzed and classified, points are selected in the array for comparison and judgment one by one, and the elements that meet the conditions are selected to form a new classification array and are removed from the original array at the same time to shorten the next retrieval time. The number of elements in the newly obtained classification array is statistically judged, and it is judged as the pixel edge information of a horizontal rib only when the number is greater than a certain threshold. By using this method, the pixel edge information of each horizontal rib can be obtained in turn; Step 7: Hough Linear Transform Through the above steps, the pixel edge information of each horizontal and vertical steel bar has been obtained. Hough transform straight line detection is performed on each steel bar, that is, each set of pixel coordinates, so that the spatial position information of each steel bar can be obtained more intuitively. Due to the structural characteristics of the steel mesh, the straight line inclination angle of the horizontal bar is about 0°, and the inclination angle of the total longitudinal bar is about 90°. Some slope constraints can be appropriately added to the Hough transform to make the detection more accurate and improve the detection speed; Step 8. Calculate the overall size of the steel mesh The edge straight lines of the steel mesh are converted into mathematical straight line equations through the improved Hough transform. Each straight line will be stored in the computer memory in the form of parameters. The straight line equations are converted into a set of linear equations, and the coordinates of the intersection points of each straight line equation are calculated. The pixel spacing between the horizontal bars of the steel mesh is obtained by the difference in the horizontal coordinates of the intersection points, and the pixel spacing between the longitudinal bars of the steel mesh is obtained by the difference in the vertical coordinates of the intersection points. Then the real size conversion method is used, and the real distance is obtained according to the actual distance = pixel spacing × conversion factor, where the conversion factor = length / measured length or pixel.

2. The cross-sectional steel mesh size intelligent measurement method according to claim 1 is characterized in that: In step three, in order to further improve the accuracy of image gradients, this algorithm introduces an innovative gradient vector voting method. In this method, the gradient vector of each pixel is not only calculated by the Sobel operator, but also votes for the surrounding pixels according to its direction and size, thereby enhancing the continuity and clarity of the edge. The improved algorithm uses the maximum inter-class variance Otsu algorithm to automatically determine the optimal threshold in the dual threshold detection step. The Otsu algorithm selects the threshold by maximizing the inter-class variance between foreground and background pixels, which can more accurately distinguish between edge and non-edge areas, and is more intelligent and efficient than the traditional dual threshold method.

3. The intelligent measurement method for cross-sectional steel mesh size according to claim 1 is characterized in that: In step five, the internal parameters include focal length, principal point, distortion coefficient, and the external parameters are position and attitude relative to a global reference frame.

Images