Box culvert inner wall concrete quality detection method based on image recognition technology

Abstract

The invention discloses a box culvert inner wall concrete quality detection method based on an image recognition technology, and the method comprises the steps: collecting the three-dimensional image data of the inner wall of a box culvert, segmenting the three-dimensional image data, and obtaining a plurality of segmented sub-image regions; converting the sub-image areas into an HSV color space, and screening the sub-image areas conforming to concrete inner wall characteristics as effective image areas; features of the effective image area are obtained, feature vectors of the effective image area are obtained, and the features comprise at least one of edge features, corner features, morphological features and texture features; inputting the feature vector into a target classification model for classification, and outputting a disease classification result of the effective image region; and according to a disease classification result, evaluating a disease part in the effective image area, and according to a preset judgment standard, judging the quality grade of the box culvert inner wall concrete. The method disclosed by the invention can be used for efficiently and accurately detecting the quality of the concrete on the inner wall of the box culvert in the later period of construction and maintenance.

Description

Technical Field

[0001] This application belongs to the field of image recognition technology, and more specifically, relates to a method for detecting the quality of concrete on the inner wall of a box culvert based on image recognition technology. Background Technology

[0002] As a crucial infrastructure component, the quality of the inner concrete walls of box culverts directly impacts the structural safety and durability. However, over time, various defects often appear in the inner concrete walls of box culverts. These defects not only restrict the strength development of the concrete but can also lead to structural failure. Common concrete defects include cracks, honeycombing, pitting, voids, and exposed reinforcement. These problems are usually closely related to a variety of factors, including improper design, material quality defects, insufficient construction techniques, and external environmental factors.

[0003] To ensure the construction quality of the inner wall concrete of box culverts, current domestic quality inspection mainly focuses on predicting potential defects from aspects such as raw materials and structural design, including concrete mix proportions, reinforcement selection and placement. However, there is a lack of refined management methods for the inner wall concrete of box culverts in the later stages of construction and curing, making it impossible to clearly define its quality grade or effectively identify and resolve potential quality problems. Existing technologies typically use manual visual inspection to assess the quality of the inner wall of box culverts. This method is greatly affected by subjective factors, has low inspection efficiency, and is difficult to accurately identify and quantify subtle defects on the concrete surface. Although some non-contact inspection technologies, such as ultrasonic testing and infrared thermal imaging, can be used to detect concrete defects, their application to the inner wall inspection of box culverts often suffers from limited detection range, complex operation, and high cost, making it difficult to achieve large-scale, high-efficiency inspection.

[0004] Therefore, there is an urgent need for an efficient, reliable, and economical method for testing the quality of the inner wall concrete of box culverts to ensure structural safety, improve construction quality, and provide a basis for developing reasonable maintenance and repair plans, thereby extending the service life of box culverts and reducing later maintenance and replacement costs. Summary of the Invention

[0005] The objective of this application is to provide a method for quality inspection of the inner wall concrete of box culverts based on image recognition technology, so as to achieve efficient and accurate quality inspection of the inner wall concrete of box culverts in the later stage of construction and curing, thereby ensuring structural safety and improving construction quality.

[0006] To achieve the above-mentioned technical effects, this application provides a method for quality inspection of the inner wall concrete of a box culvert based on image recognition technology, comprising: acquiring three-dimensional image data of the inner wall of the box culvert; segmenting the three-dimensional image data to obtain multiple sub-image regions after segmentation; converting the sub-image regions into HSV color space and selecting the sub-image regions that conform to the characteristics of the inner wall of the concrete as valid image regions; obtaining the features of the valid image regions to obtain feature vectors of the valid image regions, wherein the features include at least one of edge features, corner features, morphological features, and texture features; inputting the feature vectors into a target classification model for classification and outputting the defect classification results of the valid image regions; evaluating the defective parts in the valid image regions according to the defect classification results, and determining the quality grade of the inner wall concrete of the box culvert according to a preset judgment standard.

[0007] The solution provided in this application acquires three-dimensional image data of the inner wall of a box culvert, extracts edge, corner, morphological, and texture features of the images, combines this with HSV color space filtering, and then uses a target classification model to identify concrete defects, ultimately assessing the concrete quality grade. This solution overcomes the shortcomings of existing technologies, such as the strong subjectivity, low efficiency, and difficulty in quantifying minute defects associated with manual visual inspection. It achieves automated and refined inspection of the concrete quality of the inner wall of box culverts, improving inspection efficiency and accuracy, reducing inspection costs, and is suitable for large-scale, high-efficiency inspection of the concrete quality of the inner wall of box culverts.

[0008] As an improvement to the technical solution of this application, before inputting the feature vector of the sub-image region into the target classification model for classification, the method further includes: acquiring a dataset of concrete disease images of multiple categories; dividing the sample images in the concrete disease image dataset into a training set and a test set; labeling the sample images in the training set with disease category labels; inputting the training set and the test set into the image classification model for training, validation, and testing; optimizing the loss function and prediction accuracy by adjusting the learning rate and fitting training of the image classification model; determining the accuracy, recall, and F1 score based on the confusion matrix of the image classification model; and obtaining the target classification model when the performance of the image classification model reaches the preset conditions.

[0009] As an improvement to the technical solution of this application, the above-mentioned acquisition of the features of the effective image region and the resulting feature vector of the effective image region includes: extracting the texture features of the effective image region based on the gray-level co-occurrence matrix; and determining the feature vector of the effective image region based on the texture features.

[0010] Further, the step of extracting texture features of the effective image region based on the gray-level co-occurrence matrix includes: calculating the contrast of pixels in the effective image region at various angles according to the following formula, where the contrast is used to represent the abrupt changes in texture gray levels in the effective image region: Where i = f(x,y), j = f(x+dx,y+dy), and f(x,y) is the gray value of pixel (x,y).

[0011] Further, the step of extracting texture features of the effective image region based on the gray-level co-occurrence matrix includes: calculating the angular second moment of the pixels in the effective image region according to the following formula, wherein the angular second moment is used to represent the texture uniformity of the effective image region: Where i = f(x,y), j = f(x+dx,y+dy), and f(x,y) is the gray value of pixel (x,y).

[0012] Further, the extraction of texture features of the effective image region based on the gray-level co-occurrence matrix includes: calculating the entropy value of the pixels in the effective image region according to the following formula, wherein the entropy value is used to represent the texture complexity of the effective image region: Where i = f(x,y), j = f(x+dx,y+dy), and f(x,y) is the gray value of pixel (x,y).

[0013] Further, the step of extracting texture features of the effective image region based on the gray-level co-occurrence matrix includes: calculating the inverse difference moment of pixels in the effective image region according to the following formula, wherein the inverse difference moment is used to represent the local texture consistency of the effective image region: Where i = f(x,y), j = f(x+dx,y+dy), and f(x,y) is the gray value of pixel (x,y).

[0014] As an improvement to the technical solution of this application, the defect classification results include, but are not limited to, cracks, honeycombing, pitting, holes, and exposed reinforcement; the evaluation of defect locations in the effective image area based on the defect classification results, and the determination of the quality grade of the inner wall concrete of the box culvert according to preset judgment criteria, includes: when the crack width is less than 0.2 mm, or the honeycomb size is less than 25 mm and the depth is less than 10 mm, or the pitted area is less than 100 cm². 2If the depth is less than 1mm, or the diameter of the hole is less than 25mm, or there is no exposed reinforcement, the quality grade of the inner wall concrete of the box culvert is deemed qualified; if the crack width is between 0.2mm and 0.3mm, or the honeycomb size is between 25mm and 500mm and the depth is between 1mm and 3mm, or the area of ​​the pitted area is greater than 100cm², the quality grade is deemed qualified. 2 Up to 500cm 2 If the crack width is greater than 0.3mm, the concrete honeycomb size is greater than 500mm, or the pitted area is greater than 500cm², the quality grade of the inner wall concrete of the box culvert needs further monitoring. 2 If the diameter of the concrete holes is greater than 500mm, or if exposed reinforcement is found, the quality grade of the concrete on the inner wall of the box culvert is deemed unqualified.

[0015] As an improvement to the technical solution of this application, the features of the concrete inner wall include: a preset color range of the concrete inner wall, wherein the H component of the preset color range of the concrete inner wall is in the range of [10,50] or [155,170] in the HSV color space, the S component is in the range of [0.1,0.4], and the V component is in the range of [0.3,0.5].

[0016] The beneficial effects of this application are as follows: The solution provided in this application acquires three-dimensional image data of the inner wall of a box culvert, extracts edge, corner, morphological, and texture features of the images, combines this with HSV color space filtering, and then uses a target classification model to identify concrete defects, ultimately assessing the concrete quality grade. This solution overcomes the shortcomings of existing technologies, such as the strong subjectivity, low efficiency, and difficulty in quantifying minute defects associated with manual visual inspection. It achieves automated and refined inspection of the concrete quality of the inner wall of box culverts, improving inspection efficiency and accuracy, reducing inspection costs, and is suitable for large-scale, high-efficiency inspection of the concrete quality of the inner wall of box culverts. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating a method for detecting the quality of concrete on the inner wall of a box culvert based on image recognition technology, as described in an embodiment of this application. Detailed Implementation

[0019] The embodiments of the technical solution of this application will be described in detail below. The following embodiments are only used to more clearly illustrate the technical solution of this application, and are therefore merely examples and should not be used to limit the scope of protection of this application. Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0020] The following is in conjunction with the appendix Figure 1 This application provides a detailed description of the technical solutions provided in each embodiment.

[0021] Please see Figure 1 The following is a flowchart illustrating a method for detecting the quality of concrete inner wall of a box culvert based on image recognition technology, as provided in one embodiment of this application. The method may include the following steps: S1: Acquire three-dimensional image data of the inner wall of the box culvert, and segment the three-dimensional image data to obtain multiple sub-image regions after segmentation; Specifically, the 3D image data of the inner wall of the box culvert is obtained by scanning the inner wall of the box culvert multiple times from different angles and positions to ensure coverage of the entire inner wall surface, and then fusing the data from the multiple scans using a stitching technique. Optionally, the 3D image data can be any of point cloud data, stereo image pairs, or depth images. The 3D image data is divided into multiple sub-image regions, each representing a local area of ​​the inner wall of the box culvert. In this way, the quality of each local area can be detected more precisely. Sub-image regions can be obtained using conventional image segmentation methods in the art. For example, several identical sub-image regions can be divided according to a fixed grid size, or sub-image regions with large differences can be effectively segmented based on image segmentation algorithms. For example, an image segmentation algorithm based on region growing is used to segment the 3D image data of the inner wall of the box culvert. First, the normal vector and curvature of each point are calculated based on the point cloud data. Points with similar normal vectors and curvatures are grouped into the same region, thereby effectively segmenting local areas on the inner wall surface of the box culvert that may have cracks, voids, protrusions, etc., into independent sub-image regions, avoiding the influence of noise and background interference in the overall image.

[0022] Optionally, after acquiring the three-dimensional image data of the inner wall of the box culvert, the method further includes cleaning the three-dimensional image data. Specifically, this includes at least one of the following cleaning methods: (1) Detect the sharpness of the image using the Laplacian algorithm and filter out blurry images; (2) By performing feature point matching on multiple sets of the three-dimensional image data, or by calculating the hash value between the images, duplicate images are identified and filtered out; (3) Use target detection algorithms, such as YOLO or Faster R-CNN based on deep learning, to identify and exclude images that are not related to concrete defects; (4) Based on the quality indicators such as brightness, contrast and noise level of the image, use the quality assessment model to screen out images with poor quality.

[0023] Further, optionally, the retained 3D image data can be preprocessed after image cleaning, for example, by using statistical filtering algorithms to remove noise, or by using the Laplacian algorithm to smooth the image, thereby enhancing the edge and texture features of the image.

[0024] S2: Convert the sub-image region to the HSV color space, and select the sub-image region that matches the characteristics of the concrete inner wall as the valid image region; Specifically, to filter out sub-image regions that do not contain the concrete inner wall, this step performs feature filtering after converting the sub-image regions to the HSV color space. Compared to the RGB color space, the color of concrete is relatively uniform, while the HSV color space can better separate color information in the image, including hue, saturation, and brightness, to obtain the color features of the concrete inner wall in the sub-image region. In one implementation, the concrete inner wall features include: a preset concrete inner wall color range, wherein the preset concrete inner wall color range has an H component in the range of [10, 50] or [155, 170], an S component in the range of [0.1, 0.4], and a V component in the range of [0.3, 0.5] in the HSV color space.

[0025] S3: Obtain the features of the effective image region to obtain the feature vector of the effective image region; Specifically, the features include at least one of edge features, corner features, morphological features, and texture features. For edge features of the effective image region, the Canny edge detection algorithm can be used for extraction. For corner features of the effective image region, the Harris corner detection algorithm or the Shi-Tomasi corner detection algorithm can be used for identification; these corners typically correspond to variations or defects in the concrete surface structure. For morphological features of the effective image region, contour extraction algorithms can be used to extract the contour of the concrete surface from the edge detection results and calculate its shape features, such as contour area, perimeter, and shape index. For color features of the effective image region, the color distribution of the concrete surface is analyzed by calculating color histograms and color moments. For texture features of the effective image region, the gray-level co-occurrence matrix can be used to reflect the roughness and smoothness of the concrete surface, identifying defects such as holes and pitting.

[0026] Through the above steps, the feature vector of the final effective image region will contain at least one type of information, such as edge features, corner features, morphological features, color features, and texture features.

[0027] In one implementation, for the extraction of texture features from the effective image region, step S3 further includes the following sub-steps: S31: Extract texture features of the effective image region based on the gray-level co-occurrence matrix; S32: Determine the feature vector of the effective image region based on the texture features.

[0028] Further, the step of extracting texture features of the effective image region based on the gray-level co-occurrence matrix includes: calculating the contrast of pixels in the effective image region at various angles according to the following formula, where the contrast is used to represent the abrupt changes in texture gray levels in the effective image region: Where i = f(x,y), j = f(x+dx,y+dy), and f(x,y) is the gray value of pixel (x,y).

[0029] Further, the step of extracting texture features of the effective image region based on the gray-level co-occurrence matrix includes: calculating the angular second moment of the pixels in the effective image region according to the following formula, wherein the angular second moment is used to represent the texture uniformity of the effective image region: Where i = f(x,y), j = f(x+dx,y+dy), and f(x,y) is the gray value of pixel (x,y).

[0030] Further, the extraction of texture features of the effective image region based on the gray-level co-occurrence matrix includes: calculating the entropy value of the pixels in the effective image region according to the following formula, wherein the entropy value is used to represent the texture complexity of the effective image region: Where i = f(x,y), j = f(x+dx,y+dy), and f(x,y) is the gray value of pixel (x,y).

[0031] Further, the step of extracting texture features of the effective image region based on the gray-level co-occurrence matrix includes: calculating the inverse difference moment of pixels in the effective image region according to the following formula, wherein the inverse difference moment is used to represent the local texture consistency of the effective image region: Where i = f(x,y), j = f(x+dx,y+dy), and f(x,y) is the gray value of pixel (x,y).

[0032] The histogram values ​​of contrast, second moment of angle, entropy, and inverse difference moment obtained above are combined into a feature vector. The dimension of the feature vector depends on the selected angle, gray level range, and dimension of each statistic. For example, to obtain the texture features of the three-dimensional image data of the inner wall of a box culvert in four directions (horizontal, vertical, and two diagonals), with 256 gray levels and the mean of each statistic, a feature vector with a dimension of 4 × 256 × 4 = 4096 can be obtained.

[0033] S4: Input the feature vector into the target classification model for classification, and output the disease classification result of the effective image region; Specifically, the disease classification results include, but are not limited to, cracks, honeycombing, pitting, holes, and exposed reinforcement.

[0034] To achieve more accurate classification by the target classification model, one implementation includes training the image classification model before step S4 to obtain the target classification model. The specific training steps include the following steps: a. Obtain image datasets of concrete defects in multiple categories, and divide the sample images in the concrete defect image datasets into training sets and test sets; Specifically, the sample images in the training set are labeled with disease category tags. These disease category tags are created by professionals who categorize and label the images, marking different disease types such as cracks, erosion, and peeling. Bounding boxes are drawn in the sample images to accurately indicate the location and size of the diseased areas, and a tagging review mechanism is established to ensure the accuracy and consistency of the tags and avoid labeling errors.

[0035] Optionally, since the images come from different shooting conditions and devices, their sizes and resolutions vary. Before labeling the sample images in the concrete disease image dataset, the sample images can be normalized in size using the following steps: detect the size and resolution of each image; use image interpolation algorithms, such as bilinear interpolation or Lanczos resampling, to scale the images to a uniform size; for images that cannot be scaled to a uniform size, use image cropping methods to ensure that the key disease areas are located in the center of the cropped image.

[0036] b. Input the training set and the test set into the image classification model for training, validation, and testing; c. Optimize the loss function and prediction accuracy by adjusting the learning rate and fitting training of the image classification model; d. Based on the confusion matrix of the image classification model, determine the precision, recall, and F1 score, and obtain the target classification model when the performance of the image classification model meets the preset conditions.

[0037] Specifically, a confusion matrix technique was used to evaluate the classification accuracy of the model and identify defect types including cracks, honeycombing, pitting, voids, and exposed reinforcement. The classification results are shown in Table 1 below.

[0038] Table 1. Classification results of image classification model for cracks, honeycombing, pitting, holes, and exposed reinforcement. Each row in the table represents an actual disease category, and each column represents a disease category predicted by the model. The number in each cell indicates the correspondence between the actual and predicted categories. The numbers on the main diagonal represent the number of correctly classified instances (true cases). The numbers off-diagonal represent the number of misclassified instances. The model's accuracy, recall, and F1 score are calculated based on the correctly predicted positive category (TP), the incorrectly predicted negative category (FP), the correctly predicted negative category (TN), and the incorrectly predicted positive category (FN). Accuracy is... The recall rate was The F1 value is .

[0039] Based on the confusion matrix data in Table 1 above, the recall, precision, and F1 score for each category can be calculated, as shown in Table 2 below: Table 2. Recall, precision, and F1 scores for cracks, honeycombing, pitting, voids, and exposed reinforcement. Category / Result TP FP TN FN accuracy Recall rate F1 score crack 45 8 184 8 0.942 0.849 0.895 honeycomb 48 9 178 2 0.922 0.96 0.937 Pockmarked 46 6 179 4 0.957 0.92 0.928 Hole 45 5 180 5 0.957 0.9 0.914 Exposed reinforcement 45 6 179 5 0.933 0.9 0.895 As can be seen from the data in Table 2, the target classification model achieves a classification accuracy of over 90% for each disease category, demonstrating good classification accuracy.

[0040] S5: Based on the disease classification results, evaluate the diseased parts in the effective image area, and determine the quality grade of the inner wall concrete of the box culvert according to the preset judgment criteria.

[0041] In one implementation, if the crack width is less than 0.2 mm, or the honeycomb size is less than 25 mm and the depth is less than 10 mm, or the pitted area is less than 100 cm2 and the depth is less than 1 mm, or the hole diameter is less than 25 mm, or there is no exposed reinforcement, then the quality grade of the concrete on the inner wall of the box culvert is determined to be qualified. If the crack width is between 0.2mm and 0.3mm, or the honeycomb size is between 25mm and 500mm and the depth is between 1mm and 3mm, or the pitted area is between 100cm2 and 500cm2 and the depth is greater than 3mm, or the hole diameter is between 25mm and 500mm, then the quality grade of the concrete on the inner wall of the box culvert needs to be further monitored. If the crack width is greater than 0.3mm, or the concrete honeycomb size is greater than 500mm, or the pitted area is greater than 500cm2, or the concrete hole diameter is greater than 500mm, or there is exposed reinforcement, the quality grade of the inner wall concrete of the box culvert is determined to be unqualified.

[0042] In summary, this application's embodiments acquire three-dimensional image data of the inner wall of a box culvert, extract edge, corner, morphological, and texture features of the images, combine this with HSV color space filtering, and then use a target classification model to identify concrete defects, ultimately assessing the concrete quality grade. This solution overcomes the shortcomings of existing technologies, such as the strong subjectivity, low efficiency, and difficulty in quantifying subtle defects associated with manual visual inspection. It achieves automated and refined inspection of the concrete quality of the inner wall of box culverts, improving inspection efficiency and accuracy while reducing inspection costs. It is suitable for large-scale, high-efficiency inspection of the concrete quality of the inner wall of box culverts and has good prospects for widespread application.

[0043] 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 them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for inspecting the quality of concrete inner wall of a box culvert based on image recognition technology, characterized in that, include: Three-dimensional image data of the inner wall of the box culvert is acquired, and the three-dimensional image data is segmented to obtain multiple sub-image regions. The sub-image regions are converted to the HSV color space, and the sub-image regions that conform to the characteristics of the inner wall of concrete are selected as valid image regions. The features of the effective image region are obtained to obtain the feature vector of the effective image region. The features include at least one of edge features, corner features, morphological features and texture features. The feature vector is input into the target classification model for classification, and the disease classification result of the effective image region is output. Based on the disease classification results, the diseased parts in the effective image area are evaluated, and the quality grade of the inner wall concrete of the box culvert is determined according to the preset judgment criteria.

2. The method as described in claim 1, characterized in that, Before inputting the feature vector of the sub-image region into the target classification model for classification, the method further includes: A dataset of concrete defects images of multiple categories is obtained, and the sample images in the concrete defects image dataset are divided into a training set and a test set; the sample images in the training set are labeled with defects category labels; The training set and the test set are input into the image classification model for training, validation, and testing; The loss function and prediction accuracy are optimized by adjusting the learning rate and fitting training of the image classification model. Precision, recall, and F1 score are determined based on the confusion matrix of the image classification model, and the target classification model is obtained when the performance of the image classification model meets the preset conditions.

3. The method as described in claim 1, characterized in that, The step of obtaining the features of the effective image region to obtain the feature vector of the effective image region includes: Texture features of the effective image region are extracted based on the gray-level co-occurrence matrix; Based on the texture features, the feature vector of the effective image region is determined.

4. The method as described in claim 3, characterized in that, The extraction of texture features of the effective image region based on the gray-level co-occurrence matrix includes: The contrast of pixels in the effective image region at various angles is calculated using the following formula, whereby the contrast is used to represent the abrupt changes in texture grayscale levels in the effective image region: Where i = f(x,y), j = f(x+dx,y+dy), and f(x,y) is the gray value of pixel (x,y).

5. The method as described in claim 3, characterized in that, The extraction of texture features of the effective image region based on the gray-level co-occurrence matrix includes: The second angular moment of each pixel in the effective image region is calculated using the following formula, whereby the second angular moment represents the texture uniformity of the effective image region: Where i = f(x,y), j = f(x+dx,y+dy), and f(x,y) is the gray value of pixel (x,y).

6. The method as described in claim 3, characterized in that, The extraction of texture features of the effective image region based on the gray-level co-occurrence matrix includes: The entropy value of the pixels in the effective image region is calculated according to the following formula, where the entropy value represents the texture complexity of the effective image region: Where i = f(x,y), j = f(x+dx,y+dy), and f(x,y) is the gray value of pixel (x,y).

7. The method as described in claim 3, characterized in that, The extraction of texture features of the effective image region based on the gray-level co-occurrence matrix includes: The inverse difference moment of the pixels in the effective image region is calculated according to the following formula, whereby the inverse difference moment is used to represent the local texture consistency of the effective image region: Where i = f(x,y), j = f(x+dx,y+dy), and f(x,y) is the gray value of pixel (x,y).

8. The method as described in claim 1, characterized in that, The defect classification results include, but are not limited to, cracks, honeycombing, pitting, holes, and exposed reinforcement; the evaluation of defect locations in the effective image area based on the defect classification results, and the determination of the quality grade of the inner wall concrete of the box culvert according to preset judgment criteria, includes: If the crack width is less than 0.2mm, or the honeycomb size is less than 25mm and the depth is less than 10mm, or the pitted area is less than 100cm2 and the depth is less than 1mm, or the hole diameter is less than 25mm, or there is no exposed reinforcement, then the quality grade of the concrete on the inner wall of the box culvert is deemed to be qualified. If the crack width is between 0.2mm and 0.3mm, or the honeycomb size is between 25mm and 500mm and the depth is between 1mm and 3mm, or the pitted area is between 100cm2 and 500cm2 and the depth is greater than 3mm, or the hole diameter is between 25mm and 500mm, then the quality grade of the concrete on the inner wall of the box culvert needs to be further monitored. If the crack width is greater than 0.3mm, or the concrete honeycomb size is greater than 500mm, or the pitted area is greater than 500cm2, or the concrete hole diameter is greater than 500mm, or there is exposed reinforcement, the quality grade of the inner wall concrete of the box culvert is determined to be unqualified.

9. The method as described in claim 1, characterized in that, The features of the concrete inner wall include: a preset color range for the concrete inner wall, wherein the H component of the preset color range for the concrete inner wall is in the range of [10,50] or [155,170] in the HSV color space, the S component is in the range of [0.1,0.4], and the V component is in the range of [0.3,0.5].

Images