A method for detecting defects in concrete dams

Abstract

This invention discloses a method for detecting defects in concrete dams, belonging to the technical field of dam crack detection. The method includes: acquiring an image of the concrete dam; performing environmental adaptive compensation adjustment on the concrete dam image; preprocessing the concrete dam image; separating the concrete dam image into crack regions and background regions; extracting the contours of the crack regions; geometrically filling the contours of the crack regions to obtain key feature values ​​of the cracks; and evaluating the crack hazard based on the key feature values. This invention not only reduces the impact of environment, lighting, and weather on the captured images but also suppresses transmission noise, effectively improving image capture and transmission quality. Furthermore, the geometric filling of the crack region contours enables rapid crack hazard evaluation. The entire method is computationally simple, fast, has low computational cost, low resource consumption, and is adaptable to processing various types of dam image data, exhibiting strong generalization ability.

Description

Technical Field

[0001] This invention belongs to the technical field of dam crack detection, specifically relating to a method for detecting defects in concrete dams. Background Technology

[0002] Concrete dams are common hydraulic structures primarily used for water retention, storage, and flow regulation. Constructed of concrete, they possess high strength and durability. The safety of concrete dams is crucial to the lives and property of residents along their route, and dams play a vital role in power generation, irrigation, and flood control. A dam accident can cause not only direct loss of life and property but also a chain reaction of power outages, disrupted agricultural irrigation, and loss of flood control function, severely impacting local and regional economic and social development.

[0003] Concrete cracks can occur due to factors such as concrete shrinkage, environmental temperature differences, concrete material properties, and construction processes. Concrete cracks not only affect the integrity of the structure but also cause steel corrosion, reduce the structure's seismic resistance, and endanger the normal use and safe operation of the dam. In severe cases, they can lead to dam leakage and structural damage.

[0004] Existing traditional manual inspection and monitoring methods are inefficient, making it difficult to detect cracks in a timely manner. They are also susceptible to human factors, limiting the objectivity and accuracy of the detection results. Based on this, automated detection algorithms based on image segmentation have emerged. Although these algorithms can improve detection accuracy, they require specific network structure models, as well as complex algorithms and huge computing power. Their computing costs are enormous, resource consumption is high, and their generalization ability is still weak, making it impossible to comprehensively cover various types of dam image data. Summary of the Invention

[0005] The purpose of this invention is to address the aforementioned shortcomings of existing technologies by providing a method for detecting defects in concrete dams. This method solves the problems of existing image segmentation algorithms for detecting cracks in concrete dams, which suffer from complexity, high computational resource requirements, and weak generalizability.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] A method for detecting defects in concrete dams, comprising the following steps:

[0008] S1. Use drones to acquire images of the concrete dam;

[0009] S2. Perform environmental adaptive compensation adjustment on the concrete dam image to obtain the adjusted concrete dam image;

[0010] S3. Preprocess the adjusted concrete dam image;

[0011] S4. The preprocessed concrete dam image is separated into crack region and background region using a binarization method.

[0012] S5. The contour of the crack region is extracted using the Canny edge detection and connected component analysis methods.

[0013] S6. Perform geometric filling on the contour of the crack area to obtain the key feature values ​​of the crack;

[0014] S7. Conduct crack hazard assessment based on key characteristic values.

[0015] Furthermore, in S2, the concrete dam image undergoes adaptive environmental compensation adjustment. Specifically, based on illumination intensity, atmospheric light value, dynamic sharpening intensity, and transmission loss compensation, the original concrete dam image is adaptively corrected and compensated. This correction and compensation process is specifically expressed as follows:

[0016]

[0017] In the formula, This shows the adjusted image of the concrete dam. This represents the original image of the concrete dam. Indicates light intensity; This indicates the distance between the drone and the concrete dam; A represents the atmospheric light value. Indicates the dynamic sharpening intensity; This indicates compensation for transmission loss caused by the shooting equipment; Indicates the atmospheric transmittance coefficient; Indicates the brightness response index; This represents the color temperature deviation compensation factor.

[0018] Furthermore, dynamic sharpening intensity Represented as:

[0019]

[0020] In the formula, The base sharpening factor is set to 0.3; Indicates the incremental compensation for transmission loss; This indicates the transmission efficiency of the shooting equipment.

[0021] Furthermore, the transmission loss compensation E of the shooting equipment is expressed as:

[0022]

[0023] In the formula, This represents the transmission noise gain coefficient.

[0024] Furthermore, S3 specifically includes: performing Gaussian filtering on the adjusted concrete dam image:

[0025]

[0026] In the formula, H ij Let i be a convolution kernel of size i*j; i and j are the kernel size parameters; 2k+1 is the window width, k is a natural number; σ is the variance;

[0027] Downsampling is performed on the concrete dam image after Gaussian filtering:

[0028]

[0029] In the formula, The floating coordinates of the target point; , , , Let be the four adjacent integer coordinates of the floating coordinates in the image of the concrete dam, where The coordinate of the top-left corner among four adjacent integer coordinates; Floating coordinates Relative to the top left corner coordinates The offset; Image of a concrete dam in floating coordinates The pixel value at that location.

[0030] Furthermore, S4 specifically includes the following steps:

[0031] S41. Calculate the probability of each gray level in the preprocessed concrete dam image based on the histogram. ;

[0032] S42. Set an initial threshold t, and iterate through all gray levels. Pixels with gray values ​​less than or equal to t belong to the background area, and pixels with gray values ​​greater than t belong to the crack area.

[0033] For a threshold t, the probabilities of the crack region and the background region are expressed as:

[0034]

[0035]

[0036] In the formula, The probability of a crack region; The probability of the background region; Represents grayscale level The probability of M; M is the maximum value of the gray level;

[0037] S43. Calculate the average gray value of the crack area and the background area, and calculate the inter-class variance;

[0038] S44. Iterate through the threshold t and select the threshold that maximizes the inter-class variance. As a segmentation threshold;

[0039] S45. Use segmentation threshold. Binarize the image of the concrete dam, separating pixels with gray values ​​less than or equal to... The portion used as the background area is larger than Partially divided into crack regions:

[0040]

[0041] In the formula, This represents the pixel grayscale value.

[0042] Furthermore, in S43, the average grayscale values ​​of the crack region and the background region are expressed as follows:

[0043]

[0044]

[0045] In the formula, This represents the average gray value of the crack area; This represents the average grayscale value of the background area;

[0046] The variance between classes is expressed as:

[0047]

[0048] In the formula, The variance is between classes.

[0049] Furthermore, S6 specifically includes:

[0050] S6.1 Construct a two-dimensional rectangular coordinate system, mesh the two-dimensional rectangular coordinate system, and place the contour image of the crack region in the meshed two-dimensional rectangular coordinate system;

[0051] S6.2. Along the extension direction of the crack, several circles are sequentially filled into the outline of the crack area in an inscribed manner, and the maximum equivalent width and average equivalent width of the crack are calculated based on a two-dimensional rectangular coordinate system.

[0052] Among them, the maximum equivalent width W max The diameter of the largest inscribed circle;

[0053] Average equivalent width W avg This is the average diameter of all inscribed circles;

[0054] S6.3. The straight-line length between the two extreme points in the direction of crack extension is taken as the equivalent length of the crack.

[0055] Furthermore, S7 specifically includes: calculating the crack hazard index based on the maximum equivalent width, average equivalent width, and equivalent length of the crack.

[0056]

[0057] In the formula, Indicates the risk index of cracks; This indicates the equivalent length of the crack.

[0058] The method for detecting defects in concrete dams provided by this invention has the following beneficial effects:

[0059] 1. This invention comprehensively considers illumination intensity, atmospheric light value, dynamic sharpening intensity, and transmission loss compensation to perform environmental adaptive compensation adjustment on concrete dam images acquired by UAVs. This not only minimizes the impact of the environment on the captured images but also suppresses transmission noise, effectively improving image capture and transmission quality. Furthermore, by combining binarization processing and Canny edge detection, crack contours are obtained, and the critical feature values ​​of the crack contours are used to evaluate the current degree of crack danger. The entire algorithm process of this invention does not involve image segmentation using machine learning or deep learning, nor does it require a specific network structure model. The entire computation is simple, fast, has low computational cost, low resource consumption, and is adaptable to processing various types of dam image data, exhibiting strong generalization ability.

[0060] 2. This invention uses a circular filling of the entire crack to quickly and efficiently calculate the maximum and average equivalent width of the crack, and captures the coupling relationship between crack width and length, thereby defining a crack hazard index. This allows for a better understanding of the relationship between the absolute size of the current crack and its local expansion, and can also reflect the risk of abrupt width changes during crack extension, such as the instability of crack tip expansion. This has important guiding significance for crack morphological development monitoring and risk level assessment. Attached Figure Description

[0061] Figure 1 This is a flowchart of a concrete dam defect detection method according to an embodiment of the present invention.

[0062] Figure 2 This is a binarized image of the crack in an embodiment of the present invention.

[0063] Figure 3 This is a schematic diagram illustrating the geometric filling of the contour of a local crack region according to an embodiment of the present invention. Detailed Implementation

[0064] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0065] This embodiment provides a method for detecting defects in concrete dams. It can adaptively compensate and adjust the captured images according to the current shooting environment to obtain higher quality concrete dam images. Simultaneously, it couples the maximum equivalent width, average equivalent width, and equivalent length of cracks to quickly evaluate the current crack's hazard level and monitor its morphology. Figure 1 Specifically, it includes the following:

[0066] S1. Use drones to acquire images of the concrete dam;

[0067] Specifically, in this embodiment, the drone is equipped with at least corresponding shooting equipment and laser rangefinder to collect real-time images of the current concrete dam and the distance between the drone and the concrete dam, etc.

[0068] In addition, to facilitate the transmission of image data, a corresponding communication transmission module is also required to transmit the acquired concrete dam images to a remote host computer in real time for subsequent image processing.

[0069] S2. When the shooting equipment on the drone is collecting images in the external environment, it is often affected by multiple factors such as weather, environment and light, such as atmospheric light value, light intensity, haze concentration, rain, visibility, etc. Therefore, in order to improve the quality of the captured images, this embodiment performs environmental adaptive compensation adjustment on the concrete dam image to obtain the adjusted concrete dam image.

[0070] Specifically, based on illumination intensity, atmospheric light value, dynamic sharpening intensity, and transmission loss compensation, adaptive correction and compensation processing is performed on the original concrete dam image. This correction and compensation processing is specifically expressed as follows:

[0071]

[0072] In the formula, This shows the adjusted image of the concrete dam. This represents the original image of the concrete dam (specifically, the image pixel matrix). Indicates light intensity; This indicates the distance between the drone and the concrete dam; A represents the atmospheric light value. Indicates the dynamic sharpening intensity; This indicates compensation for transmission loss caused by the shooting equipment; Indicates the atmospheric transmittance coefficient; Indicates the brightness response index; This represents the color temperature deviation compensation factor.

[0073] Taking sandstorm weather as an example, when visibility is reduced during sandstorm weather, the brightness response index can be improved. Reduce color temperature deviation compensation factor Increase dynamic sharpening intensity The value of is adjusted through adaptive coupling of multiple environmental parameters to compensate for and adjust the light intensity, color, and sharpness of the image, thereby improving the quality of the captured image.

[0074] This embodiment introduces dynamic sharpening intensity. This enhances the edges and details of an image, making the image clearer, and is represented as follows:

[0075]

[0076] In the formula, The base sharpening factor is set to 0.3; This indicates the incremental value for transmission loss compensation, which can be set to 0.2~0.3; This indicates the transmission efficiency of the shooting equipment. When, it indicates lossless transmission. Transmission losses due to compression or packet loss, such as those caused by wireless interference or electromagnetic interference, are therefore... The value range is 0 to 1.

[0077] To compensate for transmission losses and suppress image noise amplification caused by decreased transmission rate, this embodiment introduces a transmission loss compensation E for the imaging device, which is expressed as:

[0078]

[0079] In the formula, This represents the transmission noise gain coefficient; in specific transmission loss scenarios, i.e., when the transmission rate decreases, it is expressed as... Reduce overall gain and suppress noise.

[0080] In this embodiment, when transmitting images to a remote host computer, under conditions of a 10% to 30% packet loss rate due to channel bandwidth fluctuations, this embodiment uses dynamic sharpening intensity. Under the coupled adjustment of transmission loss compensation E, the edge recovery capability is improved by 8%~12%, and the noise suppression is improved by 2dB~4dB.

[0081] S3. Preprocess the adjusted concrete dam image;

[0082] The preprocessing in this embodiment includes, firstly, removing images that do not meet the requirements in terms of clarity and shooting angle, and then selecting images that meet the quality requirements.

[0083] Secondly, the adjusted concrete dam image is subjected to Gaussian filtering and downsampling to eliminate high-frequency noise and reduce image resolution, so as to facilitate rapid image processing in the later stages.

[0084] Specifically, Gaussian filtering is expressed as follows:

[0085]

[0086] In the formula, H ij Let i be a convolution kernel of size i*j; i and j are the kernel size parameters; 2k+1 is the window width, k is a natural number; σ is the variance;

[0087] Downsampling is performed on the concrete dam image after Gaussian filtering:

[0088]

[0089] In the formula, The floating coordinates of the target point; , , , Let be the four adjacent integer coordinates of the floating coordinates in the image of the concrete dam, where The coordinate of the top-left corner among four adjacent integer coordinates; Floating coordinates Relative to the top left corner coordinates The offset; Image of a concrete dam in floating coordinates The pixel value at that location.

[0090] S4. The preprocessed concrete dam image is separated into crack areas and background areas using a binarization method, which includes the following steps:

[0091] S41. Calculate the probability of each gray level in the preprocessed concrete dam image based on the histogram;

[0092] Specifically, calculate each gray level of the concrete dam image. The number of pixels (from 0 to 255) is used to obtain a grayscale histogram. The probability of each grayscale level is then calculated based on the grayscale histogram. That is, pixel grayscale level The proportion of pixels in the total number of pixels.

[0093] S42. Set an initial threshold t, and iterate through all gray levels. Pixels with gray values ​​less than or equal to t belong to the background area, and pixels with gray values ​​greater than t belong to the crack area.

[0094] For a threshold t, the probabilities of the crack region and the background region are expressed as:

[0095]

[0096]

[0097] In the formula, The probability of a crack region; The probability of the background region; Represents grayscale level The probability of M; M is the maximum value of the gray level;

[0098] S43. Calculate the average gray value of the crack area and the background area, and calculate the inter-class variance;

[0099] The average gray values ​​of the crack area and the background area are represented as follows:

[0100]

[0101]

[0102] In the formula, This represents the average gray value of the crack area; This represents the average grayscale value of the background area;

[0103] The variance between classes is expressed as:

[0104]

[0105] In the formula, The variance is between classes.

[0106] S44. Iterate through the threshold t and select the threshold that maximizes the inter-class variance. As a segmentation threshold;

[0107] S45. Use segmentation threshold. Binarize the image of the concrete dam, separating pixels with gray values ​​less than or equal to... The portion used as the background area is larger than Partially divided into crack regions:

[0108]

[0109] In the formula, This represents the pixel grayscale value.

[0110] S5. The contour of the crack region is extracted using the Canny edge detection and connected component analysis methods.

[0111] After obtaining the binarized crack region image, Canny edge detection is performed on the crack region image to obtain the edge image of the crack region. Then, connected component analysis is used to obtain the closed contour image, thereby obtaining the contour of the crack region. Specifically, as follows... Figure 2 As shown, in Figure 2 In order to highlight the crack area, a binarization black-and-white conversion process can be performed, which results in a black crack area and a white background area.

[0112] The Canny edge detection and connected component analysis methods in this embodiment are conventional techniques in the field, so their detailed processes will not be described in detail.

[0113] S6. Geometrically fill the contour of the crack region to obtain the key feature values ​​of the crack. This includes the following steps:

[0114] S6.1 Construct a two-dimensional rectangular coordinate system, mesh the two-dimensional rectangular coordinate system, and place the contour image of the crack region in the meshed two-dimensional rectangular coordinate system;

[0115] S6.2 Along the direction of the crack extension, several circles are sequentially filled into the outline of the crack area in an inscribed manner. The diameter of each circle may be the same or different. Based on the two-dimensional rectangular coordinate system, the maximum width and average width of the crack are quickly calculated.

[0116] The inscribed circle in this embodiment is as follows: Figure 3 As shown, the requirement for the inscribed circle is that it must be internally tangent to both sides of the crack and externally tangent to the adjacent circles, so that the size of the inscribed circle can be uniquely determined.

[0117] In this embodiment, to quickly calculate the maximum and average width of the crack, the diameter corresponding to the largest circle among the inscribed circles is selected as the maximum equivalent width W of the crack. max The average diameter of all inscribed circles is selected as the average equivalent width W of the crack. avg Although the equivalent method in this embodiment will cause a certain decrease in accuracy, the purpose of this embodiment is to monitor the shape development of cracks, so the accuracy requirement is not particularly high. The equivalent calculation in this embodiment is also in line with the qualitative research on crack shape development and the physical law of crack shape development, and can achieve faster calculation.

[0118] S6.3. The straight-line length between the two extreme points in the direction of crack extension is taken as the equivalent length of the crack, such as... Figure 2The starting point of the crack is one extreme point, and the ending point of the crack at its lowest point is another extreme point. Using a two-dimensional rectangular coordinate system, the straight-line distance between the two extreme points can be quickly calculated, and then the equivalent length of the crack can be quickly calculated.

[0119] S7. Conduct crack hazard assessment based on key characteristic values;

[0120] The crack hazard index is calculated based on the maximum equivalent width, average equivalent width, and equivalent length of the crack:

[0121]

[0122] In the formula, Indicates the risk index of cracks; This indicates the equivalent length of the crack.

[0123] In the above crack risk assessment, from a local perspective, the formula... This item demonstrates the hazard posed by the maximum crack width. The maximum equivalent width is squared to further amplify the actual damage caused to the concrete dam by the maximum width; while for... This item measures the correlation between width difference and crack length. If the crack is long and its local width increases abruptly, this item increases significantly, reflecting the synergistic risk of "long cracks accompanied by local weak points," and embodying the shape development trend of abrupt width changes during crack extension. Figure 3 In the fourth inscribed circle (from top to bottom), the width of the crack suddenly increases due to an abrupt change at that location. Figure 3 Of all the inscribed circles, the fourth inscribed circle has the largest diameter.

[0124] Overall, if the crack danger index... The higher the value, the greater the risk factor of the crack, and the more worthy it is of supervision and priority repair.

[0125] Although specific embodiments of the invention have been described in detail with reference to the accompanying drawings, this should not be construed as limiting the scope of protection of this patent. Various modifications and variations that can be made by a person skilled in the art without inventive effort within the scope described in the claims still fall within the scope of protection of this patent.

Claims

1. A method for detecting defects in a concrete dam, characterized in that, The method comprises the following steps: S1, acquiring a concrete dam image by using a UAV; S2, performing environment self-adaptive compensation adjustment on the concrete dam image to obtain an adjusted concrete dam image; S3, preprocessing the adjusted concrete dam image; S4, separating the preprocessed concrete dam image into a crack region and a background region by using a binarization processing method; S5, extracting the contour of the crack region by using a canny edge detection and connected region analysis method; S6, performing geometric filling on the contour of the crack region to obtain key characteristic values of the crack; S7, performing crack risk assessment according to the key characteristic values; In the S2, the environment self-adaptive compensation adjustment on the concrete dam image is specifically: based on the light intensity, the atmospheric light value, the dynamic sharpening intensity and the transmission loss compensation, the original concrete dam image is subjected to adaptive correction and compensation processing, and the correction and compensation processing is specifically represented as: ； In the formula, represents the adjusted concrete dam image; represents the original concrete dam image; represents the light intensity; represents the distance of the unmanned aerial vehicle from the concrete dam; A is the atmospheric light value; represents the dynamic sharpening intensity; represents the transmission loss compensation of the shooting device; represents the atmospheric transmittance coefficient; represents the brightness response index; represents the color temperature deviation compensation factor; the dynamic sharpening strength is represented as: ； In the formula, is a base sharpening coefficient, taken as 0.3; represents a transmission loss compensation increment; represents a transmission efficiency of the shooting device; The transmission loss compensation E of the shooting device is represented as: ； In the formula, denotes the transmission noise gain factor.

2. The method for detecting defects in a concrete dam according to claim 1, characterized in that, The S3 specifically comprises: performing Gaussian filtering processing on the adjusted concrete dam image: ； In the formula, H ij is a convolution kernel with a size of i*j; i and j are size parameters of the convolution kernel; 2k+1 is a window width, k is a natural number; and σ is a variance. Performing down-sampling processing on the concrete dam image subjected to the Gaussian filtering processing: ； In the formula, is the floating coordinate of the target point; , , , are four adjacent integer coordinates of the floating coordinate in the concrete dam image, wherein, is the upper left corner coordinate in the four adjacent integer coordinates; are the offsets of the floating coordinate relative to the upper left corner coordinate ; is the pixel value of the concrete dam image at the floating coordinate .

3. The method for detecting defects in a concrete dam according to claim 2, characterized in that, The S4 specifically comprises the following steps: S41. Calculate the probability of each gray level of the pre-processed concrete dam image according to the histogram ; S42, setting an initial threshold t, and traversing all gray levels, pixels with a gray value less than or equal to t are attributed to the background region, and pixels with a gray value greater than t are attributed to the crack region; wherein, for the threshold t, the probability of the crack region and the background region is represented as: ； ； wherein is the crack region probability; is the background region probability; is the probability of a gray level M is the maximum value of the gray level. S43, calculating the average gray value of the crack region and the background region, and calculating the inter-class variance; S44, traverse the threshold t, select the threshold that maximizes the inter-class variance as a segmentation threshold; S45, using a segmentation threshold The concrete dam image is binarized, and the part with a pixel gray value less than or equal to is taken as a background region, and the part greater than is segmented into a crack region: ； In the formula, is the pixel gray value.

4. The method for detecting defects in a concrete dam according to claim 3, characterized in that, In the S43, the average gray value of the crack region and the background region is represented as: ； ； In the formula, represents the average gray value of the crack region; represents the average gray value of the background region; The inter-class variance is represented as: ； In the formula, is the interclass variance.

5. The method of claim 1, wherein, The S6 specifically comprises: S6.1, constructing a two-dimensional rectangular coordinate system, meshing the two-dimensional rectangular coordinate system, and placing the crack region contour image in the meshed two-dimensional rectangular coordinate system; S6.2, along the extension direction of the crack, a plurality of circles are sequentially filled in the crack region contour in the form of inscribed, and based on the two-dimensional rectangular coordinate system, the maximum equivalent width and the average equivalent width of the crack are calculated; wherein the maximum equivalent width W max is the diameter corresponding to the largest circle among all the inscribed circles; average equivalent width W avg is the average value of the diameter of all the inscribed circles; S6.3, taking the straight line length between the two extreme points farthest in the extension direction of the crack as the equivalent length of the crack.

6. The method for detecting defects in a concrete dam according to claim 5, characterized in that, The S7 specifically comprises: calculating the crack risk index according to the maximum equivalent width, the average equivalent width and the equivalent length of the crack: ； wherein represents a crack risk index; represents an equivalent length of a crack.

Images