Tunnel concrete construction quality detection method based on digital image technology

Abstract

The present application relates to the field of tunnel engineering and concrete quality detection, in order to improve the accuracy of defect detection, provide a tunnel concrete construction quality detection method based on digital image technology, by integrating linear array camera with tunnel lining trolley, realize continuous, high-speed scanning of concrete surface, adapt to the dynamic scene of tunnel construction; Optimize the image processing algorithm, improve the identification accuracy of small cracks, overcome the influence of uneven tunnel lighting; At the same time, complete the detection of cracks, honeycomb surface and dislocation, and automatically calculate the crack length, width, honeycomb surface area, dislocation height and other parameters, realize the quantitative evaluation of quality; Establish real-time feedback mechanism, generate results in real time during detection process and trigger abnormal alarm, ensure that construction defects can be rectified in time, improve the detection efficiency and precision of tunnel concrete construction quality.

Description

Technical Field

[0001] This invention relates to the field of tunnel engineering and concrete quality inspection, specifically a method for inspecting the construction quality of tunnel concrete based on digital image technology. Background Technology

[0002] The construction quality of tunnel concrete structures (such as shotcrete initial support and secondary lining) directly affects the structural safety and service life of the tunnel. Surface defects (such as cracks) are a key indicator for quality assessment. Traditional inspection methods rely primarily on manual inspections. Inspectors must enter the tunnel to visually observe crack distribution, measure crack width with feeler gauges, measure misalignment height with rulers, and record the approximate extent of honeycomb and pitted surfaces. However, this method has significant limitations: the complex environment inside the tunnel, with high dust concentration and dim lighting, makes it easy to miss defects during manual observation; at the same time, manual inspection is extremely inefficient. In long tunnels, inspection results need to be processed afterward and cannot be fed back to the construction team in real time, which may lead to the continuous occurrence of similar defects.

[0003] In recent years, digital imaging technology has begun to be applied to concrete defect detection, but existing technologies still have shortcomings: due to environmental interference and algorithm limitations, the recognition rate of minute cracks of 0.2-0.5mm is low, and the measurement errors of misalignment height and pitted area are large. Summary of the Invention

[0004] To improve the accuracy of defect detection, this application provides a method for detecting the construction quality of tunnel concrete based on digital image technology.

[0005] The technical solution adopted by the present invention to solve the above problems is:

[0006] Methods for detecting the construction quality of tunnel concrete based on digital image technology include:

[0007] A detection system, including a light source and a line scan camera, is deployed on the tunnel lining trolley.

[0008] The detection system acquires images of the area to be detected, and stitches the acquired images together using a stitching algorithm to form a complete image of the area to be detected.

[0009] The stitched images are preprocessed, including noise removal, distortion correction, and contrast enhancement.

[0010] Defect detection is performed based on the preprocessed images, including cracks, honeycomb pitting, and misalignments; among which,

[0011] Crack detection steps include:

[0012] The gradient magnitude is calculated for the preprocessed image, and the high and low thresholds are determined based on the gradient magnitude. The crack contours are identified based on the high and low thresholds. Straight cracks are extracted based on Hough transform and the crack length and width are calculated.

[0013] The honeycomb surface detection steps include:

[0014] The Otsu algorithm is used to determine the optimal segmentation threshold T; regions with gray values ​​below T are considered honeycomb-like pitted surfaces; the pitting rate is calculated.

[0015] The steps for detecting misaligned platforms include:

[0016] Based on image registration, the relative displacement between two complete images of the region to be detected is calculated, and the displacement difference of each pixel is obtained. A displacement difference image is generated based on the displacement difference. The sliding window method is applied to the displacement difference image to calculate the displacement difference variance within each window. The region of abrupt change in variance is the misalignment boundary. The actual misalignment height is calculated based on Δy.

[0017] Furthermore, the gradient is calculated using the Sobel operator.

[0018] Furthermore, high threshold low threshold , This represents the maximum gradient magnitude.

[0019] Furthermore, for gradient magnitude G greater than The pixels are marked as defining the edges, between and Pixels that are connected to the defined edge are marked as candidate edges, and the final crack contour is composed of the defined edge and the candidate edges.

[0020] Furthermore, the specific steps for extracting straight cracks and calculating crack length and width based on Hough transform are as follows:

[0021] Map each edge point (x,y) in the crack profile to polar coordinates (ρ,θ);

[0022] Create an accumulator space to count the number of edge points corresponding to each (ρ,θ);

[0023] Find the peak value in the accumulator space. The (ρ,θ) corresponding to the peak value is the main direction and location of the crack.

[0024] The crack length is determined based on the maximum distance between consecutive edge points along the main direction and location of the crack.

[0025] A quadratic polynomial fit is performed on the two edges of the crack; the distance between the two edges is calculated based on θ to determine the crack width.

[0026] Furthermore, before using the Otsu algorithm to determine the optimal segmentation threshold T, the preprocessed image is divided into several sub-blocks, and the optimal segmentation threshold T is calculated separately for each sub-block.

[0027] Furthermore, when calculating the displacement difference, the average displacement difference of N consecutive frames of images is used as the final displacement difference.

[0028] Furthermore, it also includes related controls based on defect detection results.

[0029] Furthermore, relevant controls based on defect detection results specifically include: providing repair suggestions and / or triggering alarms.

[0030] Furthermore, triggering the alarm also includes suspending the movement of the tunnel lining trolley.

[0031] The advantages of this invention compared to existing technologies are as follows: By integrating a linear array camera with a tunnel lining trolley, continuous and high-speed scanning of the concrete surface is achieved, adapting to the dynamic scenarios of tunnel construction; the image processing algorithm is optimized to improve the accuracy of identifying micro-cracks and overcome the influence of uneven tunnel illumination; simultaneously, crack, honeycomb surface, and misalignment detection are completed, and parameters such as crack length and width, honeycomb surface area, and misalignment height are automatically calculated to achieve quantitative quality assessment; a real-time feedback mechanism is established to generate results and trigger abnormal alarms in real time during the detection process, ensuring that construction defects can be rectified in a timely manner, and improving the detection efficiency and accuracy of tunnel concrete construction quality. Attached Figure Description

[0032] Figure 1 This is a flowchart of a method for detecting the construction quality of tunnel concrete based on digital image technology.

[0033] Figure 2 Schematic diagram of the lining trolley inspection system installation;

[0034] Figure 3 This is a comparison chart of image preprocessing results;

[0035] Attached reference numerals: 1-Tunnel excavation outline, 2-Tunnel lining, 3-5G transmission module, 4-Odometer, 5-Line scan camera, 6-LED light source, 7-Dust particles, 8-Liner cracks, 9-Pockmarked surface. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0037] like Figure 1As shown, the tunnel concrete construction quality inspection method based on digital image technology includes:

[0038] A detection system, including a light source and a line scan camera, is deployed on the tunnel lining trolley.

[0039] A detection system, including light sources and line scan cameras, is installed on the top and sides of the trolley. High-resolution industrial-grade cameras (e.g., 4096 pixels) are selected to ensure an image pixel resolution of 0.1mm / pixel (i.e., 1 pixel corresponds to an actual size of 0.1mm). In this embodiment, a line scan camera is chosen because it is more suitable for continuous scanning of large-area, elongated objects, especially in dynamic scenes. The camera is installed perpendicular to the trolley's movement direction, and the scanning range covers the trolley's working surface (adjusted according to the tunnel cross-section). High-brightness LED line light sources are used on both sides, with the light source and camera optical axis at approximately a 30° angle to avoid glare caused by direct sunlight and to compensate for insufficient lighting in the dim tunnel environment. The system also integrates an odometer and a 5G transmission module. The odometer records the detection position (associated with the tunnel station number), and the 5G module transmits image data back to the terminal server in real time, supporting continuous detection while the trolley is moving. An overall installation diagram is shown below. Figure 2 As shown, 1 is the tunnel excavation outline, 2 is the tunnel lining, 3 is the 5G transmission module, 4 is the odometer, 5 is the line scan camera, and 6 is the LED light source.

[0040] The detection system acquires images of the area to be detected, and stitches the acquired images together using a stitching algorithm to form a complete image of the area to be detected.

[0041] The image acquisition process is synchronized with the trolley construction. When the trolley moves forward, the line scan camera is triggered by the odometer. One line image is acquired for every 0.01m movement. The acquisition parameters can be dynamically adjusted. For example, when the dust sensor detects that the concentration exceeds a certain level (such as 10mg / m³), the image can be ensured to meet the requirements by increasing the brightness of the light source and reducing the speed of the trolley movement.

[0042] The acquired linear array images are stitched together to generate a panoramic unfolded image. Feature point matching is used during stitching to eliminate image shifts caused by trolley vibration. The final panoramic image covers the entire surface of the detection area.

[0043] The stitched images are preprocessed, including noise removal, distortion correction, and contrast enhancement.

[0044] Image preprocessing is fundamental to defect detection, aiming to eliminate noise, correct distortion, and enhance defect features. Noise removal is performed first using a bilateral filtering algorithm, the formula of which is:

[0045] ,

[0046] in, For the original image in grayscale value at that location (Control spatial domain filtering range) (Control the grayscale filtering intensity) As a normalization coefficient, this algorithm can effectively remove high-frequency noise caused by dust particles while preserving the crack edges, thereby improving the peak signal-to-noise ratio of the filtered image.

[0047] Next, distortion correction is performed based on the camera intrinsic parameter matrix:

[0048] ,

[0049] in, ,

[0050] Image pixel coordinates Convert to distortion-free coordinates The correction formula is:

[0051] , ,

[0052] Where (X,Y,Z) are three-dimensional spatial coordinates, and lens distortion is eliminated through correction (radial distortion <0.1 pixels).

[0053] Finally, contrast enhancement is performed using the CLAHE algorithm to highlight cracks and honeycomb boundaries. The formula is as follows:

[0054] ,

[0055] Among them, the CLAHE operator is designed for tunnel scenes with extremely uneven lighting. It divides the image into local sub-blocks and equalizes them separately, solving the problem of "increased noise in dark areas and overexposure in bright areas" caused by traditional global equalization. It is adapted to the local contrast enhancement needs of tunnel lining surfaces (such as cracks and water seepage being more prominent in shadow areas). Setting clipLimit=3.0 (limiting the maximum number of grayscale pixels in a sub-block to 3% of the total pixels in the sub-block) avoids overexposure in areas with strong light in the tunnel. This method involves dividing the tunnel image into 8×8 local sub-blocks, with each sub-block being independently equalized and then interpolated and stitched together. This size ensures local adaptability (capturing local lighting changes on the lining surface, such as equipment shadows) while avoiding "block effects" (abrupt changes in sub-block boundaries) through interpolation. At typical tunnel resolutions (such as 1024×1024), the sub-block size is adapted to the local feature scale of cracks and water leakage, providing coherent and high-contrast preprocessing results for subsequent defect detection. This represents the image after filtering.

[0056] The processing effect diagram is as follows Figure 3As shown, 7 represents dust particles, 8 represents lining cracks, and 9 represents pitted surface.

[0057] Defect detection is performed based on the preprocessed images, including cracks, honeycomb pitting, and misalignments; among which,

[0058] Crack detection steps include:

[0059] The gradient magnitude is calculated on the preprocessed image to highlight the grayscale difference between the crack and the background. In this embodiment, the Sobel operator is used to calculate the gradient components in the x and y directions:

[0060] , ,

[0061] Where I is the input image, This represents the convolution operation; the Prewitt operator can also be used to calculate the gradient components, and there is no restriction here.

[0062] For each pixel, the gradient magnitude G can be calculated using the following formula: The gradient magnitude reflects the intensity of pixel grayscale changes, and the gradient direction... It indicates the direction of the edge, providing a directional basis for subsequent edge connections.

[0063] High and low thresholds are determined based on gradient magnitude, and crack contours are identified based on the high and low thresholds; in this embodiment, the high threshold... low threshold , This represents the maximum gradient magnitude. For gradient magnitude G greater than [value missing], [the value missing]. The pixels are marked as defining the edges, between and Pixels that are connected to the defined edge are marked as candidate edges, and the final crack contour is composed of the defined edge and the candidate edges.

[0064] The gradient of the background texture (such as pouring marks and exposed aggregate) on the surface of tunnel concrete is usually less than 20% of the gradient of the crack edge. Taking 0.2 times can effectively filter more than 80% of non-crack interference, while retaining the strong edges of the core crack area (continuous pixels with high gradient values), ensuring that the core features of the crack edge are not misjudged as noise in complex environments such as dust and uneven lighting.

[0065] Low threshold set to This ratio is designed based on the continuous linear characteristics of the crack. The gradient value at the crack ends or in the edge region (weak edge) where the gray-scale change is gradual may be lower than that of the crack. However, these weak edges exhibit connectivity with strong edges. By setting 50% of the high threshold as the low threshold, these weak edges can be included in the detection range, preventing cracks from being truncated and ensuring edge integrity. Simultaneously, 0.1 times... It can filter out isolated noise (whose gradient is mostly below this value), ensuring that only effective weak edges connected to strong edges are retained, especially suitable for the detection needs of 0.2mm fine cracks.

[0066] Extracting straight cracks based on Hough transform and calculating crack length and width; specifically including:

[0067] Based on each edge point (x, y) in the crack profile Mapped to polar coordinates (ρ, θ), where The distance from the origin to the line is denoted as . It is a straight-line angle;

[0068] Create an accumulator space to count the number of edge points corresponding to each (ρ,θ);

[0069] Find the peak value in the accumulator space. The (ρ,θ) corresponding to the peak value is the main direction and location of the crack.

[0070] The crack length is determined based on the maximum distance between consecutive edge points along the main direction and location of the crack.

[0071] A quadratic polynomial fit was performed on both sides of the crack. Left edge: Right edge: Calculate the distance between the two edges along a direction perpendicular to the crack direction (perpendicular to θ), and convert it to the actual width using a pixel scale (e.g., 0.1mm / pixel):

[0072] ,

[0073] Where N is the number of sampling points, ensuring that the width calculation error is < 0.05mm, meeting the recognition requirement of the 0.2mm threshold.

[0074] The honeycomb surface detection steps include:

[0075] The Otsu algorithm is used to determine the optimal segmentation threshold T by maximizing the inter-class variance. Determine the segmentation threshold T, and maximize the inter-class variance between the foreground (honeycomb) and background (normal concrete) using the following formula:

[0076] ,

[0077] in The proportion of the two types of pixels segmented by a threshold T. The average grayscale value is the sum of the two pixel classes. To address the uneven illumination in the tunnel, the image is divided into 16×16 sub-blocks, and a threshold is calculated separately for each sub-block to avoid missed detections caused by a global threshold.

[0078] After segmentation, regions with gray values ​​lower than T (honeycomb texture) are extracted. The total number of pixels in the honeycomb region is counted, and the total number of pixels in the detected region is compared with the total number of pixels in the honeycomb region to calculate the texture rate.

[0079] .

[0080] The steps for detecting misaligned platforms include:

[0081] Based on image registration, the relative displacement between two complete images of the region to be detected is calculated, and the displacement difference of each pixel is obtained. : Perform Fourier transform on the two images to obtain Calculate the cross power spectrum ,

[0082] Where conj is the conjugate operation; the peak position is obtained by performing an inverse Fourier transform on P. That is, the displacement difference.

[0083] A displacement difference image is generated based on the displacement difference. A sliding window method (e.g., 50×50 pixels) is applied to the displacement difference image to calculate the variance of the displacement difference within each window. Regions with abrupt changes in variance are the boundaries of the misalignment. The actual misalignment height h is calculated based on Δy. .

[0084] The detection method of this application can accurately identify excessive cracks with a width greater than 0.2 mm, excessive honeycomb pitting with a surface roughness greater than 5%, and excessive misalignment with a height greater than 5 mm.

[0085] After defects are detected, the system also includes relevant controls based on the defect detection results. For example, it provides corresponding repair suggestions based on preset repair recommendations, such as recommending epoxy resin injection repair for cracks. When serious defects are detected (such as crack width > 0.5 mm or surface roughness > 10%), it can also trigger an audible and visual alarm and send a signal to the lining trolley control system to suspend trolley movement until construction resumes after the defect is rectified.

Claims

1. A method for detecting the construction quality of tunnel concrete based on digital image technology, characterized in that, include: A detection system, including a light source and a line scan camera, is deployed on the tunnel lining trolley. The detection system acquires images of the area to be detected, and stitches the acquired images together using a stitching algorithm to form a complete image of the area to be detected. The stitched images are preprocessed, including noise removal, distortion correction, and contrast enhancement; among these, a bilateral filtering algorithm is used for noise removal, with the following parameters: , ,in, Control the spatial domain filtering range, The grayscale filtering intensity is controlled; distortion correction is performed based on the camera intrinsic parameter matrix; contrast enhancement is achieved using the CLAHE algorithm with clipLimit=3.

0. ; Defect detection is performed based on the preprocessed images, including cracks, honeycomb pitting, and misalignments; among which, Crack detection steps include: The gradient magnitude is calculated for the preprocessed image, and the high and low thresholds are determined based on the gradient magnitude. The crack outline is then identified based on the high and low thresholds. Straight cracks are extracted based on the Hough transform, and the crack length and width are calculated. The honeycomb surface detection steps include: The preprocessed image is divided into several sub-blocks, and the optimal segmentation threshold T is determined for each sub-block using the Otsu algorithm; regions with gray values ​​below T are considered honeycomb-like pitted surfaces; the pitting rate is calculated. The steps for detecting misaligned platforms include: Based on image registration, the relative displacement between two complete images of the region to be detected is calculated, and the displacement difference of each pixel is obtained. A displacement difference image is generated based on the displacement difference. The sliding window method is applied to the displacement difference image to calculate the displacement difference variance within each window. The region of abrupt change in variance is the misalignment boundary. The actual misalignment height is calculated based on Δy.

2. The method for detecting the construction quality of tunnel concrete based on digital image technology according to claim 1, characterized in that, The gradient is calculated using the Sobel operator.

3. The method for detecting the construction quality of tunnel concrete based on digital image technology according to claim 1, characterized in that, High threshold low threshold , This represents the maximum gradient magnitude.

4. The method for detecting the construction quality of tunnel concrete based on digital image technology according to claim 3, characterized in that, For gradient magnitude G greater than The pixels are marked as defining the edges, between and Pixels that are connected to the defined edge are marked as candidate edges, and the final crack contour is composed of the defined edge and the candidate edges.

5. The method for detecting the construction quality of tunnel concrete based on digital image technology according to claim 1, characterized in that, The specific steps for extracting straight cracks and calculating crack length and width based on Hough transform are as follows: Map each edge point (x,y) in the crack profile to polar coordinates (ρ,θ); Create an accumulator space to count the number of edge points corresponding to each (ρ,θ); Find the peak value in the accumulator space. The (ρ,θ) corresponding to the peak value is the main direction and location of the crack. The crack length is determined based on the maximum distance between consecutive edge points along the main direction and location of the crack. A quadratic polynomial fit is performed on the two edges of the crack; the distance between the two edges is calculated based on θ to determine the crack width.

6. The method for detecting the construction quality of tunnel concrete based on digital image technology according to claim 1, characterized in that, When calculating the displacement difference, the average displacement difference of N consecutive frames is used as the final displacement difference.

7. The method for detecting the construction quality of tunnel concrete based on digital image technology according to claim 1, characterized in that, It also includes related controls based on defect detection results.

8. The method for detecting the construction quality of tunnel concrete based on digital image technology according to claim 7, characterized in that, The relevant controls based on defect detection results specifically include: providing repair suggestions and / or triggering alarms.

9. The method for detecting the construction quality of tunnel concrete based on digital image technology according to claim 8, characterized in that, The alarm is triggered along with the suspension of the tunnel lining trolley movement.

Images