An intelligent method and system for detecting cracks in an aero-engine blade
By calculating the neighborhood contrast and noise performance of each pixel in the image of an aero-engine blade and correcting the Sobel gradient value, the problem of crack detection accuracy under the influence of noise is solved, and higher detection accuracy is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- XIAN YINHAN KONGTIAN TECH CO LTD
- Filing Date
- 2026-05-08
- Publication Date
- 2026-07-14
AI Technical Summary
Existing edge detection algorithms based on the Canny operator are easily affected by noise in aero-engine blade images, resulting in low accuracy in crack detection.
By calculating the neighborhood contrast and noise performance of each pixel, the Sobel gradient value is corrected, increasing the difference between the gradient values of crack edges and noise-affected points, thereby improving detection accuracy.
It improves the accuracy of crack detection in aero-engine blades and reduces the impact of noise on the detection results.
Smart Images
Figure CN122156240B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of image processing, specifically relating to an intelligent detection method and system for cracks in aero-engine blades. Background Technology
[0002] As the "heart" of modern aircraft, aero-engines' blades endure long-term high temperatures, high pressures, and high-frequency vibrations, making them prone to cracking. These cracks directly impact flight safety and engine lifespan. Therefore, crack detection on aero-engine blades is crucial. Traditional detection methods, such as manual visual inspection, ultrasonic testing, and eddy current testing, suffer from low efficiency, large subjective errors, and difficulty in covering complex curved surfaces, especially exhibiting a high false negative rate in identifying micron-level cracks. Therefore, a computer vision-based method, such as an edge detection algorithm based on the Canny operator, is employed for intelligent crack detection on aero-engine blades.
[0003] Currently, when using edge detection algorithms based on the Canny operator to detect cracks in aero-engine blade images, the gradient values used are the Sobel gradient values of individual pixels. However, the calculation of the Sobel gradient value for each pixel in the image is easily affected by noise. This results in larger Sobel gradient values for pixels affected by noise and pixels at the crack edge, with potentially small differences between them. Consequently, the accuracy of crack detection using the Canny operator-based edge detection algorithm may be low. Summary of the Invention
[0004] To address the problem that the Sobel gradient obtained from each pixel in an image is greatly affected by noise, resulting in relatively low accuracy in detecting cracks in the image, this invention proposes an intelligent detection method and system for aero-engine blade cracks.
[0005] To achieve the above objectives, the present invention provides the following technical solution:
[0006] The air-engine blade to be inspected is photographed to obtain the image to be inspected; the Sobel gradient value of each pixel in the image to be inspected is obtained;
[0007] The neighborhood contrast of each pixel in the image to be detected is obtained based on the difference in grayscale values between each pixel and its surrounding pixels. The noise performance of each pixel in the image to be detected is obtained based on the difference between the neighborhood contrast of each pixel in the image to be detected and the mean of the neighborhood contrast of all pixels, as well as the distribution of the neighborhood contrast of all pixels in the image to be detected.
[0008] Based on the distribution of Sobel gradient values of pixels in the image to be detected, the suspected crack edge pixels and non-suspected crack edge pixels in the image to be detected are obtained; based on the Euclidean distance, gray value difference, and Sobel gradient value difference between each suspected crack edge pixel and its surrounding suspected crack edge pixels, the crack representation degree of each suspected crack edge pixel in the image to be detected is obtained; the crack representation degree of non-suspected crack edge pixels in the image to be detected is set to 0.
[0009] Based on the Sobel gradient value, crack representation degree, and noise representation degree of each pixel in the image to be detected, the corrected Sobel gradient value of each pixel in the image to be detected is obtained; based on the corrected Sobel gradient value of each pixel in the image to be detected, the edges in the image to be detected are obtained, and the detection of aero-engine blades is completed.
[0010] Furthermore, the specific steps for obtaining the neighborhood contrast of each pixel in the image to be detected based on the difference in grayscale values between each pixel and its surrounding pixels are as follows:
[0011] The first in the image to be detected The pixels within the 8-neighborhood of the i-th pixel in the image to be detected are denoted as the i-th pixel in the image to be detected. The neighboring pixels of a pixel;
[0012] Based on the first in the image to be detected The difference in grayscale value between the nth pixel and its neighboring pixels is used to obtain the nth pixel in the image to be detected. Neighborhood contrast of each pixel.
[0013] Furthermore, the step of... The difference in grayscale value between the nth pixel and its neighboring pixels is used to obtain the nth pixel in the image to be detected. The specific formula for calculating the neighborhood contrast of a pixel is as follows:
[0014] ;
[0015] In the formula, Indicates the first element in the image to be detected. Neighborhood contrast of individual pixels Indicates the first element in the image to be detected. The grayscale value of each pixel Indicates the first element in the image to be detected. The first pixel The grayscale value of each neighboring pixel Indicates the first element in the image to be detected. The number of neighboring pixels of a pixel.
[0016] Furthermore, based on the difference between the neighborhood contrast of each pixel in the image to be detected and the mean difference between the neighborhood contrast of all pixels, and the distribution of the neighborhood contrast of all pixels in the image to be detected, the specific calculation formula for the noise performance of each pixel in the image to be detected is as follows:
[0017] ;
[0018] In the formula, Indicates the first element in the image to be detected. Noise performance per pixel Indicates the first element in the image to be detected. Neighborhood contrast of individual pixels This represents the mean of the neighborhood contrast of all pixels in the image to be detected. This represents the standard deviation of the neighborhood contrast of all pixels in the image to be detected. This represents a function that takes the maximum value of two data points.
[0019] Furthermore, the specific steps for obtaining the suspected crack edge pixels and non-suspected crack edge pixels in the image to be detected based on the distribution of Sobel gradient values of pixels in the image to be detected are as follows:
[0020] The average Sobel gradient values of all pixels in the image to be detected are denoted as the gradient threshold. ;
[0021] If the first in the image to be detected The Sobel gradient value of each pixel is greater than If so, then record it as a suspected crack edge pixel;
[0022] If the first in the image to be detected The Sobel gradient value of each pixel is less than or equal to If it is not a suspected crack edge pixel, then it is recorded as a non-crack edge pixel.
[0023] Furthermore, the specific steps for obtaining the crack representation degree of each suspected crack edge pixel in the image to be detected based on the Euclidean distance, the difference in gray value, and the difference in Sobel gradient value between each suspected crack edge pixel and its surrounding suspected crack edge pixels are as follows:
[0024] Obtain the first element in the image to be detected. The Euclidean distance between a suspected crack edge pixel and other suspected crack edge pixels;
[0025] Preset the number of pixels near the suspected crack edge , will be compared with the first in the image to be detected The pixel with the smallest Euclidean distance at the suspected crack edge The first suspected crack edge pixel is denoted as the nth pixel in the image to be detected. The neighboring pixels of the suspected crack edge pixel;
[0026] The first in the image to be detected The mean of the Euclidean distances between the nth suspected crack edge pixel and all its neighboring suspected crack edge pixels is denoted as the nth pixel in the image to be detected. The density of the distribution of pixels suspected to be at the edge of a crack;
[0027] Based on the first in the image to be detected The following factors are considered: the Euclidean distance between the suspected crack edge pixel and all its neighboring suspected crack edge pixels; the difference in grayscale values; the difference in Sobel gradient values; and the distribution density of all suspected crack edge pixels in the image to be detected. These factors are used to obtain the information about the th suspected crack edge pixel in the image to be detected. The crack representation of a pixel at the edge of a suspected crack.
[0028] Furthermore, the step of... The following factors are considered: the Euclidean distance between the suspected crack edge pixel and all its neighboring suspected crack edge pixels; the difference in grayscale values; the difference in Sobel gradient values; and the distribution density of all suspected crack edge pixels in the image to be detected. These factors are used to obtain the information about the th suspected crack edge pixel in the image to be detected. The specific steps for calculating the crack representation of a suspected crack edge pixel are as follows:
[0029] Will Let be the th element in the image to be detected. The grayscale difference between neighboring pixels at the suspected crack edge;
[0030] Will Let be the th element in the image to be detected. Gradient difference between neighboring pixels suspected of being at the edge of a crack;
[0031] Obtain the first element in the image to be detected. The specific formula for calculating the crack representation of a suspected crack edge pixel is as follows:
[0032] ;
[0033] In the formula, Indicates the first element in the image to be detected. The crack representation of a pixel at the edge of a suspected crack. Indicates the first element in the image to be detected. The number of neighboring pixels that are suspected to be at the edge of a crack. Indicates the first element in the image to be detected. The grayscale value of a pixel suspected to be at the edge of a crack. Indicates the first element in the image to be detected. The first pixel suspected to be the edge of a crack The grayscale value of a pixel near the suspected crack edge. Indicates the first element in the image to be detected. Sobel gradient values of pixels suspected to be at the edge of a crack Indicates the first element in the image to be detected. The first pixel suspected to be the edge of a crack The Sobel gradient value of a pixel near the suspected crack edge. This represents the mean distribution density of all suspected crack edge pixels in the image to be detected. Indicates the first element in the image to be detected. The pixel point suspected to be the edge of a crack and its first The Euclidean distance between pixels near the suspected crack edge. It is an exponential function with the natural constant as its base. This represents the mean of the grayscale differences between neighboring pixels of all suspected crack edge points in the image to be detected. This represents the mean of the neighboring gradient differences of all suspected crack edge pixels in the image to be detected.
[0034] Furthermore, the specific calculation steps for obtaining the corrected Sobel gradient value of each pixel in the image to be detected based on the Sobel gradient value, crack representation degree, and noise representation degree of each pixel in the image to be detected are as follows:
[0035] ;
[0036] In the formula, Indicates the first element in the image to be detected. Corrected Sobel gradient values for each pixel Indicates the first element in the image to be detected. Sobel gradient values for each pixel. Indicates the first element in the image to be detected. Crack representation per pixel Indicates the first element in the image to be detected. Noise performance per pixel.
[0037] The intelligent detection method and system for aero-engine blade cracks provided by this invention has the following beneficial effects: When detecting aero-engine blades using an edge detection algorithm based on the Canny operator, this invention first utilizes the characteristic that the gray values of pixels affected by noise in the image differ from those of their surrounding normal pixels. Based on the difference in gray values between each pixel and its surrounding pixels, the noise performance of each pixel is obtained, laying the groundwork for reducing the impact of noise-affected pixels on the detection results. Since the gray values of pixels on the crack edge also differ from those of some surrounding pixels, the noise performance of some pixels on the crack edge may also be relatively high. Therefore, based on the characteristic that each crack edge pixel contains multiple crack edge pixels around it, and the difference between the gray value and gradient value of each crack edge pixel and its surrounding crack edge pixels is relatively small, the crack representation degree of each pixel is obtained. Then, based on the crack representation degree and noise representation degree of each pixel, the Sobel gradient of each pixel is corrected, which increases the difference between the gradient of the edge pixels on the crack in the image and the gradient of the pixels affected by noise. This solves the problem that the detection results may be greatly affected by noise when using the current edge detection algorithm based on the Canny operator to perform edge detection on the image of the aero-engine blade to complete the intelligent detection of cracks in aero-engine blades. Attached Figure Description
[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention 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 only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0039] Figure 1 This is a flowchart illustrating the steps of an intelligent detection method for cracks in aero-engine blades according to an embodiment of the present invention. Detailed Implementation
[0040] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an intelligent detection method and system for aero-engine blade cracks proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.
[0041] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.
[0042] Example 1:
[0043] This invention provides an intelligent detection method for cracks in aero-engine blades, specifically as follows: Figure 1 As shown, it includes:
[0044] Step S001: Take a picture of the aero-engine blade to be inspected to obtain the image to be inspected; obtain the Sobel gradient value of each pixel in the image to be inspected.
[0045] Specifically, an initial image of the aero-engine blade to be inspected is obtained by taking a picture of the blade using a camera. This initial image is then converted to grayscale, and subsequently filtered using a Gaussian filter to obtain the image to be inspected. The grayscale conversion and Gaussian filtering operations are both well-known existing techniques and will not be elaborated upon in this embodiment.
[0046] Furthermore, the Sobel gradient value of each pixel in the image to be detected is obtained. The method for obtaining the Sobel gradient value of each pixel in the image is a well-known existing technique and will not be elaborated upon in this embodiment.
[0047] At this point, the Sobel gradient value of each pixel in the image to be detected is obtained.
[0048] Step S002: Based on the difference in grayscale values between each pixel in the image to be detected and its surrounding pixels, obtain the neighborhood contrast of each pixel in the image to be detected; based on the difference between the neighborhood contrast of each pixel in the image to be detected and the mean of the neighborhood contrast of all pixels, and the distribution of the neighborhood contrast of all pixels in the image to be detected, obtain the noise performance of each pixel in the image to be detected.
[0049] It's important to note that current edge detection algorithms based on the Canny operator use the Sobel gradient value of each pixel in the image. However, this method is highly sensitive to noise. Furthermore, direct Gaussian filtering for image denoising cannot accurately remove all noise, leaving some interference in the image. This results in both the Sobel gradient values of noise-affected pixels and those at the crack edge being relatively large, with potentially small differences between them. Consequently, the accuracy of edge detection using the Canny operator-based algorithm is relatively low, leading to lower accuracy in crack detection. Therefore, when using the Canny operator-based edge detection algorithm to detect cracks in an image, the Sobel gradient value of each pixel is first corrected.
[0050] It should be further noted that, since the Sobel gradient value of each pixel in the image may be affected by noise, when correcting the Sobel gradient value of each pixel in the image, it is necessary to first calculate the noise performance of each pixel in the image before correcting the Sobel gradient value of the pixel. Therefore, the noise performance of each pixel in the image needs to be calculated.
[0051] It should be further explained that, under normal circumstances, when a pixel in an image is affected by noise, its grayscale value will differ from that of its surrounding pixels that are not affected by noise. Therefore, the noise performance of each pixel in the image is calculated based on the grayscale difference between each pixel and its surrounding pixels. That is, firstly, the neighborhood contrast of each pixel in the image is obtained based on its grayscale distribution, and then the noise performance of each pixel is obtained based on its neighborhood contrast.
[0052] It should be further noted that, under normal circumstances, the number of pixels affected by noise in an image is relatively small compared to pixels unaffected by noise. Furthermore, in this embodiment, when obtaining the Sobel gradient value of each pixel in the image, a Gaussian filter is first applied to the image, further reducing the number of noise-affected pixels. This means the mean of the neighborhood contrast of all pixels can be considered a normal value, and the standard deviation of the neighborhood contrast of all pixels can be seen as the difference between the neighborhood contrast of most noise-free pixels and the mean of the neighborhood contrast of all pixels. Therefore, the noise performance of each pixel in the image is obtained based on the neighborhood contrast of each pixel and the neighborhood contrast of all pixels.
[0053] Specifically, the first element in the image to be detected... The pixels within the 8-neighborhood of the i-th pixel in the image to be detected are denoted as the i-th pixel in the image to be detected. The neighboring pixels of a pixel.
[0054] Furthermore, obtain the first [element] in the image to be detected. The specific formula for calculating the neighborhood contrast of a pixel is as follows:
[0055] ;
[0056] In the formula, Indicates the first element in the image to be detected. Neighborhood contrast of individual pixels Indicates the first element in the image to be detected. The grayscale value of each pixel Indicates the first element in the image to be detected. The first pixel The grayscale value of each neighboring pixel Indicates the first element in the image to be detected. The number of neighboring pixels of a pixel.
[0057] It should be noted that, The larger the value, the more significant the first... The significant difference in grayscale between the first pixel and its surrounding pixels further illustrates that the first pixel... Individual pixels are more likely to be affected by noise.
[0058] Furthermore, obtain the first [element] in the image to be detected. The specific formula for calculating the noise performance of each pixel is as follows:
[0059] ;
[0060] In the formula, Indicates the first element in the image to be detected. Noise performance per pixel Indicates the first element in the image to be detected. Neighborhood contrast of individual pixels This represents the mean of the neighborhood contrast of all pixels in the image to be detected. This represents the standard deviation of the neighborhood contrast of all pixels in the image to be detected. This represents a function that takes the maximum value of two data points.
[0061] It should be noted that, The larger the value, the more significant the first... The greater the neighborhood contrast of a pixel is compared to the neighborhood contrast of a normal pixel, the more it illustrates that the... The greater the likelihood of a pixel being affected by noise, the more likely it is to be affected by noise; this is because it is calculated by comparing the neighborhood contrast of all pixels. The sum of squares of the differences is obtained by first averaging and then taking the square root. , making Neighborhood contrast that can represent normal pixels in an image The difference, therefore, The coming of the first Each pixel is compared to a normal pixel. Furthermore, because the number of pixels affected by noise in the image is relatively small, when... When, explain the first Each pixel must be a normal pixel; therefore, by... This makes the first When each pixel is a normal pixel, let . Greater than 1 indicates that the first The nth pixel may be a pixel affected by noise, and if its value is greater than 1, it indicates that the nth pixel in the image to be detected is a pixel affected by noise. The greater the probability that a pixel is affected by noise, the higher the probability in the denominator. It is used to prevent the denominator from being 0.
[0062] At this point, the noise performance of each pixel in the image to be detected is obtained.
[0063] Step S003: Based on the distribution of Sobel gradient values of pixels in the image to be detected, obtain the suspected crack edge pixels and non-suspected crack edge pixels in the image to be detected; based on the Euclidean distance, gray value difference, and Sobel gradient value difference between each suspected crack edge pixel and its surrounding suspected crack edge pixels in the image to be detected, obtain the crack representation degree of each suspected crack edge pixel in the image to be detected; set the crack representation degree of non-suspected crack edge pixels in the image to be detected to 0.
[0064] It should be noted that because the grayscale values of crack pixels at the crack edge in the image differ from those of surrounding non-crack pixels, the noise level of some crack pixels at the crack edge is relatively high. This makes it impossible to directly correct the Sobel gradient of each pixel based on its noise level in the image to be detected. Therefore, before correcting the Sobel gradient of each pixel, it is necessary to calculate the crack representation level of each pixel in the image to be detected.
[0065] It should be further noted that, because the gradient values of pixels located at crack edges in the image to be detected are relatively large, and the number of pixels at crack edges in the image is relatively small, the Sobel gradient values of pixels at crack edges in the image to be detected should be greater than the average Sobel gradient values of all pixels in the image to be detected. Therefore, when obtaining the crack representation degree of each pixel in the image to be detected, the suspected crack edge pixels in the image to be detected are first obtained based on the gradient of each pixel in the image to be detected, and then the crack representation degree of each suspected crack edge pixel in the image to be detected is obtained.
[0066] It should be further noted that because the gradient values of pixels affected by noise in the image are relatively large, these noise-affected pixels are also suspected crack edge pixels. Furthermore, since a crack is unlikely to be composed of a single isolated pixel under normal circumstances, each crack edge pixel should be surrounded by multiple crack edge pixels. Conversely, noise-affected pixels may be surrounded by multiple unaffected pixels. That is, pixels on a crack are not isolated, while noise-affected pixels are. Therefore, based on the distribution of suspected crack edge pixels around each suspected crack edge pixel in the image to be detected, the crack representation degree of each suspected crack edge pixel is calculated.
[0067] It should be further noted that the grayscale and gradient values of each edge pixel on the same crack should be relatively similar to those of its surrounding edge pixels. However, the grayscale and gradient values of pixels affected by noise are highly likely to differ from those of edge pixels on the crack. Furthermore, since the degree of noise influence may vary among different noise-affected pixels in the image, their grayscale and gradient values may also differ. Therefore, the crack representation degree of each suspected crack edge pixel is calculated based on the differences in grayscale and gradient values between each suspected crack edge pixel and its surrounding suspected crack edge pixels.
[0068] Specifically, the average Sobel gradient values of all pixels in the image to be detected are denoted as the gradient threshold. If the first element in the image to be detected... The Sobel gradient value of each pixel is greater than If the first pixel in the image to be detected is a suspected crack edge pixel, then it is recorded as such. The Sobel gradient value of each pixel is less than or equal to If it is not a suspected crack edge pixel, then it is recorded as a non-crack edge pixel.
[0069] Furthermore, obtain the first [element] in the image to be detected. The Euclidean distance between the first suspected crack edge pixel and other suspected crack edge pixels is calculated. This distance is then compared with the first pixel in the image to be detected. The pixel with the smallest Euclidean distance at the suspected crack edge The first suspected crack edge pixel is denoted as the nth pixel in the image to be detected. The number of neighboring suspected crack edge pixels is defined in this embodiment. This is because a pixel in two-dimensional space has a maximum of 8 pixels in its 8-neighborhood. This is used as an example; other values can be set in other implementations. Furthermore, when it is necessary to select the first pixel from the image to be detected... The Euclidean distances of the pixels suspected to be at the edge of a crack are the same. Select within the pixels suspected of being at the edge of a crack The nth pixel is used as the nth When a suspected crack edge pixel is adjacent to another suspected crack edge pixel, then the above The pixels suspected to be at the edge of a crack are all the first... The neighboring pixels of the suspected crack edge pixel. Among them, Furthermore, obtaining the Euclidean distance between two pixels in an image is a well-known existing technique, and will not be described in detail in this embodiment.
[0070] Furthermore, the first element in the image to be detected... The mean of the Euclidean distances between the nth suspected crack edge pixel and all its neighboring suspected crack edge pixels is denoted as the nth pixel in the image to be detected. The density of the distribution of pixels suspected to be at the edge of a crack.
[0071] Furthermore, Let be the th element in the image to be detected. The grayscale difference between neighboring pixels suspected of being at the edge of a crack; Let be the th element in the image to be detected. Gradient difference of neighboring pixels at the suspected crack edge.
[0072] Furthermore, obtain the first [element] in the image to be detected. The specific formula for calculating the crack representation of a suspected crack edge pixel is as follows:
[0073] ;
[0074] In the formula, Indicates the first element in the image to be detected. The crack representation of a pixel at the edge of a suspected crack. Indicates the first element in the image to be detected. The number of neighboring pixels that are suspected to be at the edge of a crack. Indicates the first element in the image to be detected. The grayscale value of a pixel suspected to be at the edge of a crack. Indicates the first element in the image to be detected. The first pixel suspected to be the edge of a crack The grayscale value of a pixel near the suspected crack edge. Indicates the first element in the image to be detected. Sobel gradient values of pixels suspected to be at the edge of a crack Indicates the first element in the image to be detected. The first pixel suspected to be the edge of a crack The Sobel gradient value of a pixel near the suspected crack edge. This represents the mean distribution density of all suspected crack edge pixels in the image to be detected. Indicates the first element in the image to be detected. The pixel point suspected to be the edge of a crack and its first The Euclidean distance between pixels near the suspected crack edge. This represents the mean of the grayscale differences between neighboring pixels of all suspected crack edge points in the image to be detected. This represents the mean of the gradient differences between neighboring pixels of all suspected crack edge pixels in the image to be detected. As an exponential function with the natural constant as its base, this embodiment uses it to represent an inverse proportional relationship.
[0075] It should be noted that, The smaller the value, the more... The pixel point suspected to be the edge of a crack and its first The smaller the difference in grayscale values between pixels adjacent to the suspected crack edge, the more it indicates that the... The weaker the isolation of a suspected crack edge pixel, that is, the weaker the isolation of the first pixel. The higher the probability that a pixel suspected of being a crack edge is actually a pixel on the crack edge, the better. To use functions to represent inverse proportional relationships, in order to avoid when When the value is large, The function output approaches 0, resulting in a significant difference in this term value between cracked pixels and noise pixels, with noise pixels showing a larger difference. The problem of very small output differences can be solved by dividing by... Adjust in this way, Divide by The same principle applies; The larger the value, the more significant the first... The pixel point suspected to be the edge of a crack and its first The smaller the difference in Sobel gradient values between pixels adjacent to the suspected crack edge, the more it indicates that the... The weaker the isolation of a suspected crack edge pixel, that is, the weaker the isolation of the first pixel. The greater the likelihood that a pixel suspected of being a crack edge is a pixel on the crack edge; The smaller the value, the more... The distance between the suspected crack edge pixel and other suspected crack edge pixels around it is relatively large, further illustrating that the first... The stronger the isolation of a suspected crack edge pixel, that is, the more isolated the first pixel... The less likely a pixel suspected to be on the edge of a crack is actually a pixel on the crack edge.
[0076] Furthermore, if the first element in the image to be detected... If the nth pixel is not a suspected crack edge pixel, then the nth pixel in the image to be detected... The crack representation of a single pixel is 0.
[0077] Thus, the crack representation of each pixel in the image to be detected is obtained.
[0078] Step S004: Based on the Sobel gradient value, crack representation degree, and noise representation degree of each pixel in the image to be detected, obtain the corrected Sobel gradient value of each pixel in the image to be detected; based on the corrected Sobel gradient value of each pixel in the image to be detected, obtain the edges in the image to be detected, and complete the detection of the aero-engine blade.
[0079] Specifically, the formula for calculating the corrected Sobel gradient value for each pixel in the image to be detected is as follows:
[0080] ;
[0081] In the formula, Indicates the first element in the image to be detected. Corrected Sobel gradient values for each pixel Indicates the first element in the image to be detected. Sobel gradient values for each pixel. Indicates the first element in the image to be detected. Crack representation per pixel Indicates the first element in the image to be detected. Noise performance per pixel.
[0082] It should be noted that when , When, explain the first These pixels are most likely normal pixels. No, not the first The gradient value of each pixel is corrected; when The larger the value, the higher the level of the detection value. The higher the probability that a pixel is a crack, the more likely it is to be. The value is relatively small, at which point... A major adjustment is made to better increase the gradient difference between crack pixels and noise pixels. The larger the value, the higher the level of the detection value. The greater the probability that a pixel is noise, the better. Adjustments were made to better increase the gradient difference between crack pixels and noise pixels.
[0083] Furthermore, based on the corrected Sobel gradient values of all pixels in the image to be detected, edge detection based on the Canny operator is used to perform edge detection on the image to obtain the edges in the image to be detected. Edge detection of the image is a well-known existing technique and will not be described in detail in this embodiment.
[0084] Furthermore, the edges contained in an image of a normal, crack-free aero-engine blade are acquired. Within the edges in the image to be detected, the edges contained in the image of the normal, crack-free aero-engine blade are removed. Then, the edges that were not removed are taken as crack edges, thereby obtaining the crack in the image to be detected. Among these methods, obtaining the crack in the aero-engine blade to be detected through the edges in the image to be detected is a known existing technique.
[0085] Another embodiment of the present invention provides an intelligent detection system for cracks in aero-engine blades. The system includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the above-described method steps S001 to S004.
Claims
1. A smart detection method for cracks in aero-engine blades, characterized in that, include: The air engine blades to be inspected are photographed to obtain images of the inspected components. Obtain the Sobel gradient value for each pixel in the image to be detected; The neighborhood contrast of each pixel in the image to be detected is obtained based on the difference in grayscale values between each pixel and its surrounding pixels. The noise performance of each pixel in the image to be detected is obtained based on the difference between the neighborhood contrast of each pixel in the image to be detected and the mean of the neighborhood contrast of all pixels, as well as the distribution of the neighborhood contrast of all pixels in the image to be detected. Based on the distribution of Sobel gradient values of pixels in the image to be detected, the suspected crack edge pixels and non-suspected crack edge pixels in the image to be detected are obtained; based on the Euclidean distance, gray value difference, and Sobel gradient value difference between each suspected crack edge pixel and its surrounding suspected crack edge pixels, the crack representation degree of each suspected crack edge pixel in the image to be detected is obtained; the crack representation degree of non-suspected crack edge pixels in the image to be detected is set to 0. Based on the Sobel gradient value, crack representation degree, and noise representation degree of each pixel in the image to be detected, the corrected Sobel gradient value of each pixel in the image to be detected is obtained; based on the corrected Sobel gradient value of each pixel in the image to be detected, the edges in the image to be detected are obtained, and the detection of aero-engine blades is completed. The specific calculation steps for obtaining the corrected Sobel gradient value of each pixel in the image to be detected based on the Sobel gradient value, crack representation degree, and noise representation degree of each pixel in the image to be detected are as follows: ； In the formula, Indicates the first element in the image to be detected. Corrected Sobel gradient values for each pixel Indicates the first element in the image to be detected. Sobel gradient values for each pixel. Indicates the first element in the image to be detected. Crack representation per pixel Indicates the first element in the image to be detected. Noise performance of each pixel; The edges contained in the image of a normal, crack-free aero-engine blade are obtained. Within the edges in the image to be detected, the edges contained in the image of the normal, crack-free aero-engine blade are removed. Then, the edges that were not removed are taken as crack edges, thus obtaining the crack in the image to be detected.
2. The intelligent detection method for cracks in aero-engine blades according to claim 1, characterized in that, The specific steps for obtaining the neighborhood contrast of each pixel in the image to be detected based on the difference in grayscale values between each pixel and its surrounding pixels are as follows: The first in the image to be detected The pixels within the 8-neighborhood of the i-th pixel in the image to be detected are denoted as the i-th pixel in the image to be detected. The neighboring pixels of a pixel; Based on the first in the image to be detected The difference in grayscale value between the nth pixel and its neighboring pixels is used to obtain the nth pixel in the image to be detected. Neighborhood contrast of each pixel.
3. The intelligent detection method for cracks in aero-engine blades according to claim 2, characterized in that, The first in the image to be detected The difference in grayscale value between the nth pixel and its neighboring pixels is used to obtain the nth pixel in the image to be detected. The specific formula for calculating the neighborhood contrast of a pixel is as follows: ； In the formula, Indicates the first element in the image to be detected. Neighborhood contrast of individual pixels Indicates the first element in the image to be detected. The grayscale value of each pixel Indicates the first element in the image to be detected. The first pixel grayscale values of neighboring pixels Indicates the first element in the image to be detected. The number of neighboring pixels of a pixel.
4. The intelligent detection method for cracks in aero-engine blades according to claim 1, characterized in that, The specific calculation formula for the noise performance of each pixel in the image to be detected, based on the difference between the neighborhood contrast of each pixel and the mean neighborhood contrast of all pixels in the image to be detected, and the distribution of the neighborhood contrast of all pixels in the image to be detected, is as follows: ； In the formula, Indicates the first element in the image to be detected. Noise performance per pixel Indicates the first element in the image to be detected. Neighborhood contrast of individual pixels This represents the mean of the neighborhood contrast of all pixels in the image to be detected. This represents the standard deviation of the neighborhood contrast of all pixels in the image to be detected. This represents a function that takes the maximum value of two data points.
5. The intelligent detection method for cracks in aero-engine blades according to claim 1, characterized in that, The specific steps for obtaining the suspected crack edge pixels and non-suspected crack edge pixels in the image to be detected based on the distribution of Sobel gradient values of pixels in the image to be detected are as follows: The average Sobel gradient values of all pixels in the image to be detected are denoted as the gradient threshold. ; If the first in the image to be detected The Sobel gradient value of each pixel is greater than If so, then record it as a suspected crack edge pixel; If the first in the image to be detected The Sobel gradient value of each pixel is less than or equal to If it is not a suspected crack edge pixel, then it is recorded as a non-crack edge pixel.
6. The intelligent detection method for cracks in aero-engine blades according to claim 1, characterized in that, The specific steps for obtaining the crack representation of each suspected crack edge pixel in the image to be detected based on the Euclidean distance, the difference in grayscale value, and the difference in Sobel gradient value between each suspected crack edge pixel and its surrounding suspected crack edge pixels are as follows: Obtain the first element in the image to be detected. The Euclidean distance between a suspected crack edge pixel and other suspected crack edge pixels; Preset the number of pixels near the suspected crack edge , will be compared with the first in the image to be detected The pixel with the smallest Euclidean distance at the suspected crack edge The first suspected crack edge pixel is denoted as the nth pixel in the image to be detected. The neighboring pixels of the suspected crack edge pixel; The first in the image to be detected The mean of the Euclidean distances between the nth suspected crack edge pixel and all its neighboring suspected crack edge pixels is denoted as the nth pixel in the image to be detected. The density of the distribution of pixels suspected to be at the edge of a crack; Based on the first in the image to be detected The following factors are considered: the Euclidean distance between the suspected crack edge pixel and all its neighboring suspected crack edge pixels; the difference in grayscale values; the difference in Sobel gradient values; and the distribution density of all suspected crack edge pixels in the image to be detected. These factors are used to obtain the information about the th suspected crack edge pixel in the image to be detected. The crack representation of a pixel at the edge of a suspected crack.
7. The intelligent detection method for cracks in aero-engine blades according to claim 6, characterized in that, The first in the image to be detected The following factors are considered: the Euclidean distance between the suspected crack edge pixel and all its neighboring suspected crack edge pixels; the difference in grayscale values; the difference in Sobel gradient values; and the distribution density of all suspected crack edge pixels in the image to be detected. These factors are used to obtain the information about the th suspected crack edge pixel in the image to be detected. The specific steps for calculating the crack representation of a suspected crack edge pixel are as follows: Will Let be the th element in the image to be detected. The grayscale difference between neighboring pixels at the suspected crack edge; Will Let be the th element in the image to be detected. Gradient difference between neighboring pixels suspected of being at the edge of a crack; Obtain the first element in the image to be detected. The specific formula for calculating the crack representation of a suspected crack edge pixel is as follows: ； In the formula, Indicates the first element in the image to be detected. The crack representation of a pixel at the edge of a suspected crack. Indicates the first element in the image to be detected. The number of neighboring pixels that are suspected to be at the edge of a crack. Indicates the first element in the image to be detected. The grayscale value of a pixel suspected to be at the edge of a crack. Indicates the first element in the image to be detected. The first pixel suspected to be the edge of a crack The grayscale value of a pixel near the suspected crack edge. Indicates the first element in the image to be detected. Sobel gradient values of pixels suspected to be at the edge of a crack Indicates the first element in the image to be detected. The first pixel suspected to be the edge of a crack The Sobel gradient value of a pixel near the suspected crack edge. This represents the mean distribution density of all suspected crack edge pixels in the image to be detected. Indicates the first element in the image to be detected. The pixel point suspected to be the edge of a crack and its first The Euclidean distance between pixels near the suspected crack edge. It is an exponential function with the natural constant as its base. This represents the mean of the grayscale differences between neighboring pixels of all suspected crack edge points in the image to be detected. This represents the mean of the neighboring gradient differences of all suspected crack edge pixels in the image to be detected.