Image Recognition-Based Method for Evaluating the Construction Quality of Bridges with Fully Distributed Supports

Abstract

This invention discloses a method for evaluating the construction quality of bridge full-support scaffolding based on image recognition, comprising: S1. acquiring image information of the full-support scaffolding; S2. correcting the full-support scaffolding image to obtain a corrected image; S3. performing grayscale processing on the corrected image to obtain a grayscale image; S4. acquiring the deformation map of the grayscale image and determining whether the structural deformation of the full-support scaffolding meets the stiffness requirements based on the deformation map; S5. performing target matching on the corrected image to obtain the number of full-support scaffolding; and determining whether the number of full-support scaffolding equals the required number for construction. This invention can quickly determine whether the structural deformation of the full-support scaffolding meets the stiffness requirements and whether the number of scaffolding is correct, while ensuring safety.

Description

Technical Field

[0001] This invention relates to the field of bridge support construction, and specifically to a method for evaluating the construction quality of bridge full-coverage support based on image recognition. Background Technology

[0002] Bridge full-span scaffolding, also known as full-span scaffolding, consists of many scaffolding sections spliced ​​together according to certain specifications and heights. Its function is to provide temporary support for cast-in-place concrete beams, which are then dismantled once the concrete beams have reached the required strength for use.

[0003] The stability of a fully-supported scaffolding structure plays a decisive role in the construction of cast-in-place concrete beams. Insufficient structural load-bearing capacity of the lower scaffolding can lead to buckling instability, directly affecting the success or failure of the project. This can be caused by an under-design of the number of scaffolding members, or even by the sheer size of the fully-supported scaffolding itself, where the actual number of members during construction may be less than the designed number, resulting in discrepancies between the constructed and designed structures and insufficient load-bearing capacity. Furthermore, excessive deformation of some longer scaffolding members within the fully-supported scaffolding can alter the structure itself, causing the stiffness of the supporting structure to no longer meet the weight of the superstructure, leading to the full-supported structure breaking or collapsing.

[0004] To effectively monitor the safety of bridge scaffolding construction, it is essential to conduct efficient and reasonable evaluations of its construction quality. Currently, however, the evaluation of bridge scaffolding construction quality is often carried out by technical monitoring personnel through manual on-site inspections. This method is not only inefficient but also prone to safety accidents. Summary of the Invention

[0005] In view of this, the purpose of this invention is to overcome the defects in the prior art and provide a method for evaluating the construction quality of bridge full-support structure based on image recognition. This method can quickly determine whether the structural deformation of the full-support structure meets the stiffness requirements and whether the number of supports is correct, while ensuring safety.

[0006] The image recognition-based method for evaluating the construction quality of bridge full-support structure according to the present invention includes the following steps:

[0007] S1. Acquire image information of the full-coverage support structure;

[0008] S2. Correct the image of the full-coverage support structure to obtain the corrected image;

[0009] S3. Perform grayscale processing on the corrected image to obtain a grayscale image;

[0010] S4. Obtain the deformation map of the grayscale image, and determine whether the structural deformation of the fully covered support meets the stiffness requirements based on the deformation map;

[0011] S5. Perform target matching on the corrected image to obtain the number of full-coverage supports; and determine whether the number of full-coverage supports is equal to the number required for construction.

[0012] Furthermore, the image of the full-coverage support is corrected to obtain the corrected image, specifically including:

[0013] S21. Obtain the camera's internal parameters, external parameters, and distortion parameters;

[0014] S22. Perform coordinate transformation on the image of the full-coverage support, so that the world coordinate system is first transformed into the camera coordinate system, and then the camera coordinate system is transformed into the image coordinate system;

[0015] S23. Substitute the camera's internal parameters into the image coordinate system to perform distortion correction, thereby transforming the image coordinate system into the camera coordinate system;

[0016] S24. Then convert the camera coordinate system to the image coordinate system, and obtain the pixel coordinates of the full-coverage bracket image by interpolation in the image coordinate system.

[0017] Furthermore, the pixel coordinates of the full-coverage support image are determined according to the following formula:

[0018]

[0019] Where K is the camera intrinsic parameter matrix, M is the camera extrinsic parameter matrix, and P is the coordinate matrix in the world coordinate system; (X w ,Y w Z w ) represents the coordinates in the world coordinate system, R is the rotation matrix, t is the camera offset matrix, and f is the coordinates in the world coordinate system. x For f / dx, f y Let f / dy be the focal length of the camera during shooting, dx and dy be the number of pixels per millimeter in the x and y directions, respectively; s be the distortion parameter; u0 and v0 be the coordinates of the principal point, which is the origin in the image coordinate system; u and v be the pixel coordinates.

[0020] Furthermore, the corrected image is converted to grayscale to obtain a grayscale image, specifically including:

[0021] S31. Perform denoising processing on the corrected image to obtain the denoised image;

[0022] S32. Determine the gradient magnitude and orientation of the denoised image;

[0023] S33. Process all pixels in the denoised image according to the following steps:

[0024] S331. Within a given range of n×n, compare pixel P in the denoised image with its neighboring pixels along its gradient line. If the gradient magnitude of pixel P is greater than the gradient magnitude of the two neighboring pixels, then retain the gradient magnitude of pixel P. If it is smaller than the gradient magnitude of the two neighboring pixels, then set the gradient magnitude of pixel P to 0.

[0025] S332. If the gradient magnitude of pixel P is greater than the set maximum threshold, then pixel P takes the value 1; if the gradient magnitude of pixel P is less than the set minimum threshold, then pixel P takes the value 0.

[0026] S34. Use the image processed in step S33 as a grayscale image.

[0027] Furthermore, the deformation map of the grayscale image is obtained, and the structural deformation of the fully-supported structure is determined based on the deformation map to see if it meets the stiffness requirements. Specifically, this includes:

[0028] S41. Scale the grayscale image to its actual size to obtain a distorted version of the grayscale image;

[0029] S42. From the deformation map of the grayscale image, extract the actual deformation map of a single support in the full-coverage support, and calculate the actual bending deflection between the single supports in a single layer.

[0030] S43. Compare the actual bending deflection with a set deflection threshold. If the actual bending deflection is greater than the deflection threshold, the structure of the full-coverage support cannot meet the construction quality requirements; if the actual bending deflection is not greater than the deflection threshold, the structure of the full-coverage support meets the construction quality requirements.

[0031] Furthermore, target matching is performed on the corrected image to obtain the number of fully distributed supports, specifically including:

[0032] S51. Use the corrected image as the target image, and select a reference template in the target image;

[0033] S52. Move the reference template within the target image and compare the entire target image:

[0034] If the similarity coefficient of position (x,y) in the target image is greater than the set value, then position (x,y) is taken as the matching position, and the number of full-cloth brackets is recorded as 1.

[0035] S53. Count the number of full-coverage supports after processing in step S52 to obtain the number of full-coverage supports from a single viewpoint.

[0036] S54. Combining the dual-view matching diagram, and following steps S52-S53, the final number of full-coverage supports is obtained.

[0037] Furthermore, the similarity coefficient R(x,y) at position (x,y) in the target image is determined according to the following formula:

[0038]

[0039] Among them, the T(x′,y′) is the pixel value of the reference template at position (x′,y′). T(x″,y″) is the average pixel value of the reference template within a region of area w·h, T(x″,y″) is the pixel value of the reference module at position (x″,y″), (x″,y″) is the coordinate of the region of area w·h, w is the width of the region, and h is the height of the region.

[0040] The I(x+x′,y+y′) represents the pixel value of the target image at position (x+x′,y+y′). Let I(x+x″,y+y″) be the average pixel value of the target image within a region of area w·h, and let I(x+x″,y+y″) be the pixel value of the target image at position (x+x″,y+y″).

[0041] The beneficial effects of this invention are as follows: This invention discloses a method for evaluating the construction quality of bridge full-support scaffolding based on image recognition. By using image matching based on computer vision, it eliminates the need for contact with the full-support scaffolding, ensuring personnel safety and operability. It can quickly count the number of full-support scaffolding and monitor its deformation, thereby determining whether the structural deformation of the full-support scaffolding meets the stiffness requirements and whether the number of scaffolding is correct, thus achieving the purpose of evaluating construction quality. Attached Figure Description

[0042] The present invention will be further described below with reference to the accompanying drawings and embodiments:

[0043] Figure 1 This is a flowchart of the construction quality evaluation method of the present invention;

[0044] Figure 2 This is a schematic diagram of the image capture point arrangement according to the present invention;

[0045] Figure 3 This is a schematic diagram illustrating the conversion of the world coordinate system to the pixel coordinate system according to the present invention;

[0046] Figure 4 This is a flowchart of the camera calibration distortion correction process of the present invention. Detailed Implementation

[0047] The present invention will be further described below with reference to the accompanying drawings, as shown in the figures:

[0048] The image recognition-based method for evaluating the construction quality of bridge full-support structure according to the present invention includes the following steps:

[0049] S1. Acquire image information of the full-coverage support structure;

[0050] S2. Correct the image of the full-coverage support structure to obtain the corrected image;

[0051] S3. Perform grayscale processing on the corrected image to obtain a grayscale image;

[0052] S4. Obtain the deformation map of the grayscale image, and determine whether the structural deformation of the fully covered support meets the stiffness requirements based on the deformation map;

[0053] S5. Perform target matching on the corrected image to obtain the number of full-coverage supports; and determine whether the number of full-coverage supports is equal to the number required for construction.

[0054] By identifying the structural deformation of the fully-supported scaffolding, linear monitoring of its deformation is achieved, enabling the determination of whether the structural stiffness is within a safe range, thereby evaluating the construction quality. By obtaining the number of fully-supported scaffolds, it can be determined whether the number equals the required number for construction; if not, the number of fully-supported scaffolds does not meet the requirements, thus achieving the purpose of evaluating the construction quality.

[0055] In this embodiment, in step S1, components that are significantly different from other parts of the fully-fledged support are selected as control points. By setting control points, a clear image feature template can be provided when counting the number of disc-buckle supports in the subsequent process. This image feature template can then be used as a reference template. For example, the disc buckles of the disc-buckle support can be used as control points, as the disc buckles have a clear morphological difference from the vertical or horizontal bars. According to... Figure 2 As shown, use a camera or webcam to capture images of the full-coverage support structure, ensuring that you are looking directly at the support structure to avoid situations where the viewing angle does not match the original structure due to tilting.

[0056] In this embodiment, as Figure 4 As shown, in step S2, the image of the full-coverage support is corrected to obtain the corrected image, specifically including:

[0057] S21. Obtain the camera's internal parameters, external parameters, and distortion parameters; among which, the camera's internal parameters, external parameters, and distortion parameters are obtained by using Zhang Zhengyou's camera calibration method and a checkerboard calibration board; in the process of implementing machine vision, camera calibration plays a crucial role in the accuracy of tracking results, and the accuracy of its calibration results will directly affect the accuracy of the final count of the number of full-coverage supports.

[0058] S22. Perform coordinate transformation on the image of the full-coverage support, so that the world coordinate system is first transformed into the camera coordinate system, and then the camera coordinate system is transformed into the image coordinate system;

[0059] S23. Substitute the camera's internal parameters into the image coordinate system for distortion correction, transforming the image coordinate system into the camera coordinate system. This process involves eliminating radial and tangential distortions caused by variations in lens manufacturing precision and process variations. Radial distortion primarily results from light rays bending further away from the lens center than closer, causing the distortion to distribute along the lens's radius. Tangential distortion is mostly due to the camera lens not being parallel to the camera's imaging plane. Both radial and tangential distortions need to be eliminated during image processing.

[0060] S24. Then convert the camera coordinate system to the image coordinate system, and obtain the pixel coordinates of the full-coverage bracket image by interpolation in the image coordinate system.

[0061] Among them, such as Figure 3 As shown, this implements the world coordinate system (X) w ,Y w Z w Convert the coordinates to pixel coordinates (u, v), and determine the pixel coordinates of the full-coverage support image according to the following formula:

[0062]

[0063] Where K is the camera intrinsic parameter matrix (3×3 matrix), M is the camera extrinsic parameter matrix (3×4 matrix), and P is the coordinate matrix in the world coordinate system (4×1 matrix); (X w ,Y w Z w ) represents the coordinates in the world coordinate system, R is the rotation matrix (3×3 matrix), t is the camera offset matrix (3×1 matrix), and f x For f / dx, f y Here, f / dy represents the focal length of the camera during shooting, for example, f can be 35mm or 40mm. dx and dy represent the number of pixels per millimeter in the x and y directions, respectively. s is the distortion parameter. u0 and v0 are the coordinates of the principal point, which is the origin of the image coordinate system. The coordinates in the pixel coordinate system are obtained by translating the coordinate system. The value is half of the two dimensions of the image coordinate system, which is the coordinate intercepted by the lens optical axis and the sensor plane. The unit is pixels. u and v are the pixel coordinates.

[0064] In this embodiment, step S3 involves converting the corrected image to grayscale to obtain a grayscale image, specifically including:

[0065] S31. Denoise the corrected image to obtain the denoised image; specifically, Gaussian filters are commonly used for image smoothing. This method can effectively reduce image noise while preserving the details of the original image, giving the final image a frosted look.

[0066] The following formula is used for Gaussian filters, where σ is the standard deviation of the n×n values ​​around the point to be calculated.

[0067]

[0068] The above formula uses existing technology and will not be elaborated further here.

[0069] S32. Determine the gradient magnitude and orientation of the denoised image; specifically, use the Sobel operator to calculate the partial derivatives in the horizontal and vertical directions of the filtered image obtained in step S31, and obtain the gradient magnitude and orientation of the denoised image according to the following two formulas.

[0070] Formula for calculating gradient magnitude: Formula for finding directions:

[0071] Where G(x,y) is the gradient magnitude at (x,y), and Gx and Gy are the image grayscale values ​​of the horizontal and vertical edges obtained after convolving the image with convolution factors (3×3 or 5×5). The magnitude of the azimuth angle θ indicates that as θ increases, the gradient magnitude is directed more towards the y-direction, and vice versa.

[0072] S33. Process all pixels in the denoised image according to the following steps:

[0073] S331. Within a given range of n×n, compare pixel P in the denoised image with its neighboring pixels along its gradient line. If the gradient magnitude of pixel P is greater than the gradient magnitude of the two neighboring pixels, then retain the gradient magnitude of pixel P. If it is smaller than the gradient magnitude of the two neighboring pixels, then set the gradient magnitude of pixel P to 0.

[0074] S332. If the gradient magnitude of pixel P is greater than the set maximum threshold, then pixel P is set to 1; if the gradient magnitude of pixel P is less than the set minimum threshold, then pixel P is set to 0. That is, a pixel value of 1 indicates that it is an edge, and a pixel value of 0 indicates that it is not an edge. The maximum and minimum thresholds are set according to the actual working conditions, and will not be elaborated here.

[0075] S34. The image processed in step S33 is used as a grayscale image. The grayscale image includes a fine black-and-white outline image of the outer contour of a single scaffold member in a full-coverage support system.

[0076] In this embodiment, step S4 involves obtaining a deformation map of the grayscale image and determining whether the structural deformation of the fully-supported structure meets the stiffness requirements based on the deformation map. Specifically, this includes:

[0077] S41. Scale the grayscale image according to the actual size to obtain a deformed image of the grayscale image; wherein, the deformed image of the grayscale image includes the actual deformed image of a single support of the full-coverage support.

[0078] S42. From the deformation map of the grayscale image, extract the actual deformation map of a single support in the full-coverage support, and calculate the actual bending deflection between the single supports in a single layer.

[0079] S43. Compare the actual bending deflection with a set deflection threshold. If the actual bending deflection is greater than the deflection threshold, the structure of the full-support frame cannot meet the construction quality requirements; if the actual bending deflection is not greater than the deflection threshold, the structure of the full-support frame meets the construction quality requirements. The deflection threshold can be set according to the specifications in the "Technical Specification for Temporary Support Structures in Building Construction" (JGJ 300-2013), for example, setting the deflection threshold for bending members to the smaller of 1 / 150 of the span and 10 mm.

[0080] In this embodiment, in step S5, multi-target matching with a single template is used, that is, according to some specific algorithms, the same or similar images are identified in two or more images.

[0081] Target matching is performed on the corrected image to obtain the number of fully covered supports, specifically including:

[0082] S51. Use the corrected image as the target image, and select a reference template in the target image; wherein, the reference template can be made according to the control points set in step S1 (such as the disc buckle in the disc buckle bracket);

[0083] S52. Move the reference template within the target image and compare the entire target image:

[0084] If the similarity coefficient of position (x,y) in the target image is greater than the set value, then position (x,y) is taken as the matching position, and the number of full-coverage brackets is recorded as 1; wherein, the set value can be 0.95, of course, according to different working conditions, the set value can also be other suitable values;

[0085] S53. Count the number of full-coverage supports after processing in step S52 to obtain the number of full-coverage supports from a single viewpoint.

[0086] S54. Combining the dual-view matching diagram, and following steps S52-S53, the final number of fully-fledged supports is obtained. The fully-fledged supports include snap-on supports, disc-buckle supports, and cup-buckle supports; the dual-view matching diagram is any two viewpoint images from the three views, such as... Figure 2 The image shown is a viewpoint image captured from the camera's perspective.

[0087] The similarity coefficient at position (x, y) in the target image represents the degree of similarity between the template image corresponding to the reference template and the target image at position (x, y) during the target matching process. A similarity coefficient of 1 indicates a perfect match, while a similarity coefficient of -1 indicates the worst match.

[0088] The similarity coefficient R(x,y) at position (x,y) in the target image is determined using the following formula:

[0089]

[0090] Among them, the T(x′,y′) is the pixel value of the reference template at position (x′,y′). T(x″,y″) is the average pixel value of the reference template within a region of area w·h, T(x″,y″) is the pixel value of the reference module at position (x″,y″), (x″,y″) is the coordinate of the region of area w·h, w is the width of the region, and h is the height of the region.

[0091] The I(x+x′,y+y′) represents the pixel value of the target image at position (x+x′,y+y′). Let I(x+x″,y+y″) be the average pixel value of the target image within a region of area w·h, and let I(x+x″,y+y″) be the pixel value of the target image at position (x+x″,y+y″).

[0092] The construction quality evaluation method of the present invention can quickly evaluate the construction quality of a full-coverage scaffold, and the evaluation method is non-contact, which reduces the danger during quality evaluation, ensures the accuracy of construction quantity, and realizes the safety evaluation of the full-coverage scaffold structure.

[0093] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for evaluating the construction quality of bridge full-support scaffolding based on image recognition, characterized in that: Includes the following steps: S1. Acquire image information of the full-coverage support structure; S2. Correct the image of the full-coverage support structure to obtain the corrected image; S3. Perform grayscale processing on the corrected image to obtain a grayscale image; S4. Obtain the deformation map of the grayscale image, and determine whether the structural deformation of the fully covered support meets the stiffness requirements based on the deformation map; S5. Perform target matching on the corrected image to obtain the number of full-coverage supports; and determine whether the number of full-coverage supports is equal to the number required for construction. Target matching is performed on the corrected image to obtain the number of fully covered supports, specifically including: S51. Use the corrected image as the target image, and select a reference template in the target image; S52. Move the reference template within the target image and compare the entire target image: If the location in the target image If the similarity coefficient is greater than a set value, then the position will be... As the matching position, the number of full-coverage brackets is recorded as 1; S53. Count the number of full-coverage supports after processing in step S52 to obtain the number of full-coverage supports from a single viewpoint. S54. Combining the dual-view matching diagram, and following steps S52-S53, the final number of full-coverage supports is obtained; The location in the target image is determined using the following formula. similarity coefficient : ； Among them, the , For reference template in position pixel values, As a reference template, in an area of The average pixel value within the region, For reference module in position pixel values, The area is Coordinates in the region, For the area width, For area height; The , For the target image at position pixel values, For the target image in an area of The average pixel value within the region, For the target image at position The pixel value.

2. The method for evaluating the construction quality of bridge full-support structure based on image recognition according to claim 1, characterized in that: The image of the full-coverage support structure is corrected to obtain the corrected image, specifically including: S21. Obtain the camera's internal parameters, external parameters, and distortion parameters; S22. Perform coordinate transformation on the image of the full-coverage support, so that the world coordinate system is first transformed into the camera coordinate system, and then the camera coordinate system is transformed into the image coordinate system; S23. Substitute the camera's internal parameters into the image coordinate system to perform distortion correction, thereby transforming the image coordinate system into the camera coordinate system; S24. Then convert the camera coordinate system to the image coordinate system, and obtain the pixel coordinates of the full-coverage bracket image by interpolation in the image coordinate system.

3. The method for evaluating the construction quality of bridge full-support structure based on image recognition according to claim 2, characterized in that: The pixel coordinates of the full-coverage support image are determined using the following formula: ； in, For the camera's intrinsic parameter matrix, For the camera's extrinsic parameter matrix, The coordinate matrix in the world coordinate system; , , () represents the coordinates in the world coordinate system. Let be a rotation matrix. Let's say it's the camera's offset matrix. For f / dx, Let f / dy be the focal length of the camera when taking the picture, and dx and dy be the number of pixels per millimeter in the x and y directions, respectively. These are distortion parameters; , These are the coordinates of the principal image point, which is the origin of the image coordinate system. , These are pixel coordinates.

4. The method for evaluating the construction quality of bridge full-support structure based on image recognition according to claim 1, characterized in that: The corrected image is then converted to grayscale to obtain a grayscale image, specifically including: S31. Perform denoising processing on the corrected image to obtain the denoised image; S32. Determine the gradient magnitude and orientation of the denoised image; S33. Process all pixels in the denoised image according to the following steps: S331. Within a given range of n×n, compare pixel P in the denoised image with its neighboring pixels along its gradient line. If the gradient magnitude of pixel P is greater than the gradient magnitude of the two neighboring pixels, then retain the gradient magnitude of pixel P. If it is smaller than the gradient magnitude of the two neighboring pixels, then set the gradient magnitude of pixel P to 0. S332. If the gradient magnitude of pixel P is greater than the set maximum threshold, then pixel P takes the value 1; if the gradient magnitude of pixel P is less than the set minimum threshold, then pixel P takes the value 0. S34. Use the image processed in step S33 as a grayscale image.

5. The method for evaluating the construction quality of bridge full-support structure based on image recognition according to claim 1, characterized in that: Obtain the deformation map of the grayscale image, and determine whether the structural deformation of the fully-supported structure meets the stiffness requirements based on the deformation map. Specifically, this includes: S41. Scale the grayscale image to its actual size to obtain a distorted version of the grayscale image; S42. From the deformation map of the grayscale image, extract the actual deformation map of a single support in the full-coverage support, and calculate the actual bending deflection between the single supports in a single layer. S43. Compare the actual bending deflection with a set deflection threshold. If the actual bending deflection is greater than the deflection threshold, the structure of the full-coverage support cannot meet the construction quality requirements; if the actual bending deflection is not greater than the deflection threshold, the structure of the full-coverage support meets the construction quality requirements.

Images