Width identification calculation method based on pavement small cracks

Abstract

The invention discloses a width identification calculation method based on small cracks of a pavement. The method comprises the following steps: S1, acquiring a depth point cloud picture of the pavement; s2, obtaining a preliminary crack mask; s3, preprocessing the original image; s4, carrying out binaryzation on the preprocessed image; s5, intercepting a preliminary crack pattern; s6, dividing the preliminary crack graph into a plurality of small areas at equal ratio, and measuring and calculating the crack width in each small area to obtain the average width of the crack; and S7, obtaining the width of each small region through iterative corrosion, and calculating the average width and the maximum width of the crack. A preliminary crack pattern is intercepted from the obtained crack mask and the preprocessed image, then the preliminary crack pattern is divided into a plurality of small areas at equal ratio, and the width of each small area is obtained through iterative corrosion, so that key parameters such as the average width and the maximum width of the crack are calculated, automation, high efficiency and accuracy of pavement crack width calculation are achieved, and the pavement crack width calculation efficiency is improved. And technical support is provided for detection and maintenance of road engineering.

Description

Technical Field

[0001] The present invention relates to the technical field of pavement crack detection, and particularly to a method for identifying and calculating the width of small pavement cracks. Background Art

[0002] In the field of road engineering, the detection and evaluation of pavement cracks are important links to ensure road safety and service life. Accurate crack width information helps optimize road maintenance plans and resource allocation. The maintenance department can reasonably arrange maintenance personnel, equipment, and materials according to the distribution and change of crack widths, improving the efficiency and economy of maintenance work. At the same time, for newly built roads, the monitoring of crack widths can be used as an important means of construction quality control to timely detect potential problems and ensure road quality and safety. With the development of technology, methods based on deep learning have achieved certain results in identifying pavement cracks. However, the existing technologies have significant limitations. Although deep learning can currently identify pavement cracks, it is powerless in calculating crack widths. This is because when the deep learning model outputs crack results, it can often only provide the approximate position and shape of the cracks, unable to accurately define the edges of the cracks, and thus unable to accurately calculate the crack widths. Summary of the Invention

[0003] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a method for identifying and calculating the width of small pavement cracks.

[0004] The purpose of the present invention is achieved through the following technical solutions: A method for identifying and calculating the width of small pavement cracks, comprising the following steps:

[0005] S1: Obtain the depth point cloud map of the pavement through a depth camera acquisition device;

[0006] S2: Process the acquired image to obtain a preliminary crack mask;

[0007] S3: Preprocess the original image;

[0008] S4: Binarize the preprocessed image;

[0009] S5: Use the crack mask obtained in step S2 to intercept the preliminary crack pattern on the image preprocessed in step S3;

[0010] S6: Divide the preliminary crack pattern into several small regions in equal proportion, measure and calculate the crack width in each small region, and obtain the average crack width through average operation;

[0011] S7: Obtain the width of each small region through iterative erosion, and calculate the average width and maximum width of the crack.

[0012] Preferably, in step S2, the acquired image is processed by image segmentation technology in deep learning.

[0013] Preferably, in step S3, the preprocessing includes outlier removal, stripe noise removal, and normalization.

[0014] Preferably, for outlier removal, the depth value of each point is x i,j , for a depth map X with length and width of m and n respectively, its mean and standard deviation are:

[0015]

[0016] When |x i,j - mean| > 5 × std, then this point is considered an outlier and needs to be replaced;

[0017] For stripe noise removal, it includes the following steps:

[0018] S3.1: Fit a smooth surface according to the depth map;

[0019] S3.2: Represent the image after removing stripe noise according to the difference between the depth map and the fitted surface.

[0020] Preferably, for outlier replacement, the mean value of normal values in the neighborhood within a certain range of this outlier is selected. Let the length and width of the neighborhood be 2k + 1, and the outlier is denoted as x i,j , and the neighborhood is expressed as:

[0021]

[0022] Denote the non - outlier values in the neighborhood of the outlier x i,j as x d , the number is denoted as N, and the value of x i,j is replaced with the mean value of the non - outlier values in the neighborhood,

[0023]

[0024] Preferably, in step S3.1, denote a certain point in the depth map as x i,j , and the point at the same position in the smooth surface is x i,j which is the weighted sum of the neighborhood with length and width of 2k + 1, and the weights follow a normal distribution:

[0025]

[0026] Preferably, in step S5, denote the processed depth map as I p , the crack mask as I m , and the intercepted crack pattern I c is:

[0027] I c = I p ∩I m 。

[0028] Preferably, in step S7, the following steps are further included:

[0029] S7.1: On the binarized depth image, let a certain crack pixel point be P 0 , in the eight-neighborhood of P 0 , starting from the pixel directly above P 0 , mark the neighborhood of P 0 in a clockwise direction as P 1 -P 8 , the value of each pixel point is equal to the value of the corresponding point on the binarized image, and use N 1 to represent the maximum value of continuous non-background points in the eight-neighborhood of P 0 . When N 1 ≥5, it is determined as a non-boundary point;

[0030] S7.2: Mark the crack pixel points that do not satisfy N 1 ≥5 as boundary pixel points. At the end of one iteration, unify the values of the marked pixel points to 0, and count the number of iterations, which is denoted as T;

[0031] S7.3: Count the number of iterations T i for each effective block in step S6

[0032] W max = max(T 1 , T 2 , T 3 …T i …T n )×2d;

[0033]

[0034] The present invention has the following advantages: The present invention intercepts a preliminary crack pattern on the obtained crack mask and the preprocessed image, then equally divides the preliminary crack pattern into several small regions, and obtains the width of each small region through iterative erosion, so as to accurately calculate key parameters such as the average width and maximum width of the crack, realizing the automation, high efficiency and accuracy of the calculation of the pavement crack width, and providing strong technical support for the detection and maintenance of road engineering. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a schematic diagram of the width recognition calculation method process;

[0036] Figure 2 It is a schematic diagram of the neighborhood distribution of some non-boundary points. Specific embodiments

[0037] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Generally, the components of the embodiments of the present invention described and illustrated in the accompanying drawings here can be arranged and designed in various different configurations.

[0038] Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed present invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0039] It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0040] It should be noted that like reference numerals and letters denote like items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0041] In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship when the product of the invention is normally placed, or the orientation or positional relationship commonly understood by those skilled in the art. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation of the present invention. In addition, the terms "first", "second", etc. are only used for descriptive distinction and cannot be understood as indicating or implying relative importance.

[0042] In the description of the present invention, it should also be noted that unless otherwise clearly specified and defined, the terms "arrangement", "installation", "connection", and "coupling" should be understood in a broad sense. For example, it may be a fixed connection, a detachable connection, or an integral connection; it may be a mechanical connection or an electrical connection; it may be directly connected or indirectly connected through an intermediate medium, and it may be the communication inside two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0043] In this embodiment, as Figure 1 shown, a method for identifying and calculating the width of small cracks on the road surface includes the following steps:

[0044] S1: Obtain the depth point cloud map of the road surface through a depth camera acquisition device;

[0045] S2: Process the acquired image to obtain a preliminary crack mask; specifically, process the acquired image through image segmentation technology in deep learning. Here, the image segmentation technology can select existing semantic segmentation technology or instance segmentation technology. In this process, a deep learning model trained with a large amount of data can be used to improve the accuracy of mask generation.

[0046] S3: Preprocess the original image;

[0047] S4: Binarize the preprocessed image; specifically, convert the preprocessed image into an image with only two color values, namely black and white. By setting an appropriate threshold, the pixels in the image are divided into two categories, one representing the crack area and the other representing the non-crack area. The main function of binarization processing is to greatly simplify the subsequent extraction and analysis of crack features.

[0048] S5: Use the crack mask obtained in step S2 to intercept the preliminary crack pattern on the image preprocessed in step S3;

[0049] S6: Divide the preliminary crack pattern into several small regions in equal proportion, measure and calculate the crack width in each small region, and obtain the average crack width through average operation; preferably, divide the intercepted preliminary crack pattern into 16×16 small regions. This method of zoning calculation can more accurately reflect the distribution of crack widths.

[0050] S7: Obtain the width of each small area through iterative erosion, and calculate the average width and maximum width of the crack. On the obtained crack mask and the preprocessed image, intercept the preliminary crack pattern, and then divide the preliminary crack pattern into several small areas in equal proportion. Obtain the width of each small area through iterative erosion, so as to accurately calculate key parameters such as the average width and maximum width of the crack, realize the automation, high efficiency and accuracy of the calculation of the pavement crack width, and provide strong technical support for the detection and maintenance of road engineering.

[0051] Further, in step S3, the preprocessing includes outlier removal, stripe noise removal and normalization. Specifically, due to external elements such as light and environment or defects of the acquisition device itself, the acquired images often have problems such as uneven depth, outliers and high noise. Therefore, it is necessary to preprocess the original images. Still further, for outlier removal, the depth value of each point is x i,j , for the depth map X with length and width of m and n respectively, its mean value and standard deviation are:

[0052]

[0053] When |x i,j - mean| > 5×std, then this point is considered an outlier and needs to be replaced; specifically, to ensure that no new features are introduced at the outlier, the outlier cannot be simply replaced with the image mean value here, but is replaced by selecting the mean value of the normal values in a certain range near this point, that is, let the length and width of the neighborhood be 2k + 1, and the outlier is denoted as x i,j , to avoid the neighborhood exceeding the image boundary, the neighborhood is expressed as:

[0054]

[0055] Denote the non-outlier values in the neighborhood of the outlier x i,j as x d , the number is denoted as N, and the value of x i,j is replaced by the mean value of the non-outlier values in the neighborhood,

[0056]

[0057] Since the acquisition device collects data while moving forward, the fluctuation of the speed will also cause data anomalies, which is reflected as stripe noise in the depth point cloud map. Therefore, for stripe noise removal, it includes the following steps:

[0058] S3.1: Fit a smooth surface according to the depth map;

[0059] S3.2: Represent the image after removing stripe noise according to the difference between the depth map and the fitted surface. Specifically, in step S3.1, denote a certain point in the depth map as xi,j , points at the same position in the smooth surface is x i,j The weighted sum of the neighborhood with a length and width of 2k + 1, where the weights follow a normal distribution:

[0060]

[0061] Then subtract the smooth surface from the depth map. At this time, the obtained image has removed the wavy undulations. For the normalization process, an existing standardized normalization method is adopted to uniformly adjust and standardize the depth change characteristics of the image, so that the image data has consistency and comparability under different conditions.

[0062] In this embodiment, in step S5, the processed depth map is denoted as I p , the crack mask is I m , the intercepted crack pattern I c is:

[0063] I c = I p ∩I m .

[0064] In this embodiment, as Figure 2 shown, in step S7, the following steps are further included:

[0065] S7.1: On the binarized depth image, let a certain crack pixel point be P 0 , in the eight-neighborhood of P 0 , starting from the pixel directly above P 0 , mark the neighborhood of P 0 in a clockwise direction as P 1 -P 8 , the value of each pixel point is equal to the value of the corresponding point on the binarized image, and use N 1 to represent the maximum value of the continuous non-background points in the eight-neighborhood of P 0 . When N 1 ≥5, it is determined as a non-boundary point;

[0066] S7.2: Mark the crack pixel points that do not satisfy N 1 ≥5 as boundary pixel points. At the end of one iteration, unify the values of the marked pixel points to 0, and count the number of iterations. The number of iterations is denoted as T;

[0067] S7.3: Count the number of iterations T for each effective block in step S6 i , let the number of effective blocks be n, and the actual width represented by each pixel be d. Then the maximum width and average width of the crack are respectively:

[0068] W max = max(T1 , T 2 , T 3 … T i … T n ) × 2d;

[0069]

[0070] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for calculating the width of small cracks on a road surface, characterized in that: The following steps are involved: S1: Obtain the depth point cloud image of the road surface through the depth camera acquisition device; S2: Processing the collected image to obtain a preliminary crack mask; S3: preprocessing the original image; S4: binarize the preprocessed image; S5: using the crack mask obtained in step S2 to intercept a preliminary crack pattern on the image preprocessed in step S3; S6: Divide the preliminary crack graph into several small areas in equal proportion, measure and calculate the crack width in each small area, and obtain the average width of the crack through average calculation; S7: Obtain the width of each small area through iterative corrosion, and calculate the average width and maximum width of the crack.

2. The method for calculating the width of small cracks on a road surface according to claim 1 is characterized in that: In step S2, the collected image is processed by image segmentation technology in deep learning.

3. The method for calculating the width of small cracks on a road surface according to claim 2 is characterized in that: In step S3, the preprocessing includes outlier removal, stripe noise removal and normalization.

4. The method for calculating the width of small cracks on a road surface according to claim 3 is characterized in that: For outlier removal, the depth value of each point is x i,j , for a depth map X with a length and width of m and n respectively, its mean and standard deviation are: When |x i,j When -mean|>5×std, the point is considered an outlier and needs to be replaced; For stripe noise removal, the following steps are included: S3.1: Fit a smooth surface based on the depth map; S3.2: The image after stripe noise removal is represented according to the difference between the depth map and the fitted surface.

5. The method for calculating the width of small cracks on a road surface according to claim 4 is characterized in that: The outlier point replacement is to select the mean of the normal values ​​in the neighborhood within a certain range of the outlier point. Let the length and width of the neighborhood be 2k+1, and the outlier point be denoted by x i,j , the neighborhood is expressed as: The outlier point x i,j The non-outlier value in the neighborhood of d , the number is recorded as N, x i,j The value of is replaced by the mean of the non-outlier values ​​in the neighborhood.

6. The method for calculating the width of small cracks on a road surface according to claim 4 is characterized in that: In step S3.1, a certain point in the depth map is denoted as x. i,j , points at the same position on a smooth surface For x i,j The weighted sum of a neighborhood with a length and width of 2k+1, with weights following a normal distribution:

7. The method for calculating the width of small cracks on a road surface according to claim 1 is characterized in that: In step S5, the processed depth map is recorded as I p , the crack mask is I m , the captured crack pattern I c for: I c =I p ∩I m 。 8. The method for calculating the width of small cracks on a road surface according to claim 1, characterized in that: The step S7 further includes the following steps: S7.1: On the binarized depth image, let a crack pixel be P0. In the eight neighborhoods of P0, start from the pixel directly above P0 and mark the neighborhoods of P0 as P1-P8 in a clockwise direction. The value of each pixel is equal to the value of the corresponding point on the binarized image. N1 is used to represent the maximum value of continuous non-background points in the eight neighborhoods of P0. When N1 ≥ 5, it is considered a non-boundary point. S7.2: Mark the crack pixels that do not satisfy N1≥5 as boundary pixels. When one iteration is completed, the values ​​of the marked pixels are unified to 0, and the number of iterations is counted, which is recorded as T. S7.3: Count the number of iterations T for each valid block in step S6 i , assuming the number of valid blocks is n, and the actual width represented by each pixel is d, then the maximum width and average width of the crack are: W max =max(T1,T2,T3…T i …T n )×2d;

Images