CN115330802B - Method for extracting debonding defect of X-ray image of carbon fiber composite gas cylinder - Google Patents

Abstract

The invention relates to the technical field of engineering detection, in particular to a method for extracting debonding defects of an X-ray image of a carbon fiber composite gas cylinder, which comprises the following steps: acquiring an original gas cylinder X-ray image, and preprocessing the original gas cylinder X-ray image to obtain a first processed image; obtaining a gas cylinder debonding defect image characteristic based on the first processing image, and performing debonding edge detection based on the gas cylinder debonding defect image characteristic to obtain a second processing image; and performing defect extraction on the second processed image, performing contour extraction on the extracted defect features to obtain defect contour features, and marking the defect contour features on the original gas cylinder X-ray image to obtain a final defect marking image. The invention can extract the debonding defect in the X-ray image of the composite carbon fiber gas cylinder, and conveniently distinguish the number, shape and other characteristics of the defect; meanwhile, the outline of the defect can be marked on the X-ray image of the composite material carbon fiber gas cylinder, and the position of the defect can be observed more intuitively.

Description

A method for extracting debonding defects from X-ray images of carbon fiber composite cylinders

Technical field

The invention relates to the technical field of engineering inspection, and in particular to a method for extracting debonding defects in X-ray images of carbon fiber composite gas cylinders.

Background technique

A gas cylinder is a pressure vessel that stores gas. It is commonly used in the industrial field to store and transport gases such as hydrogen, oxygen or nitrogen. Its characteristics are that it is easy to transport and can be recycled. Since a huge pressure will be generated inside the filled gas cylinder, once an accident occurs, a violent explosion will often occur, causing immeasurable losses. Therefore, our country has extremely high requirements for the integrity, sealing and quality of gas cylinders. There are many methods for quality testing of gas cylinders, such as material chemical composition analysis, liner mechanical property testing, fatigue testing, etc. These testing projects have complex procedures and require a lot of time and money. In addition, these detection methods are destructive to a certain extent and are only suitable for sampling inspection of gas cylinders, and cannot guarantee the quality and safety of each gas cylinder. In recent years, with the advancement of non-destructive testing technology, non-destructive testing has gradually become the main method for quality inspection of gas cylinders.

Non-destructive testing is to use technologies such as rays, ultrasonic waves, infrared rays, electromagnetic waves, etc. to detect light, Changes in sound, heat, electricity, magnetism, etc. can be used to detect physical and chemical parameters of the object to be detected. Compared with the previously mentioned detection methods, the advantage of non-destructive testing is that physical damage to the gas cylinder is avoided to the greatest extent during the detection process, and the integrity of the organizational structure of the gas cylinder can be ensured, and the performance of the gas cylinder itself will not be affected. Affected, the gas cylinders that have been tested can continue to be used. Non-destructive testing also has the advantages of fast detection speed, low requirements on the testing environment, and the ability to conduct on-site testing at the production site.

X-ray non-destructive testing refers to a technology that uses X-rays of uniform intensity to transilluminate the object being inspected and image it to reveal internal defects in the object. Since there are defective areas inside the object that absorb rays with different degrees, the intensity of the rays passing through the object is unevenly distributed. The sensor on the back of the object receives these rays and generates images with different grayscale values. Generally speaking, the thinner the area, the lower the absorption of rays, and the higher the gray value of the image.

In the process of locating defects through X-ray non-destructive inspection images of carbon fiber cylinders, manual visual inspection and deep learning-based methods are often used to locate defects. The manual visual inspection method means that workers rely on their own experience to identify the type and location of defects with the naked eye, and manually mark them on the image. Methods based on deep learning use a large number of annotated images to use algorithms such as Faster-RCNN and YOLO to extract the features of defective images and automatically annotate the images.

Currently, the disadvantages of manual visual inspection are:

1. Relying on the experience of workers to identify defect areas in gas cylinder images, the accuracy and speed of defect identification are seriously insufficient.

2. Everyone may have different standards for measuring defects, and it is difficult to formulate unified standards for judgment.

The disadvantages of deep learning-based defect location methods are:

1. In the process of training the model, a large amount of data is required, and the probability of defects in normal gas cylinders is low, making it difficult to provide enough data for training.

2. The deep learning method takes a long time to train the model, requires professional software engineers to perform parameter adjustments and repeated experiments, and has higher maintenance costs in the future.

Contents of the invention

The purpose of the present invention is to provide a method for locating debonding defects in carbon fiber composite gas cylinder images, which can be used on images of different carbon fiber gas cylinders, and at the same time perform morphological processing on the images to better display the defect locations.

In order to achieve the above objects, the present invention provides the following solutions:

A method for extracting debonding defects from X-ray images of carbon fiber composite gas cylinders, including:

Obtain an original gas cylinder X-ray image, and preprocess the original gas cylinder X-ray image to obtain a first processed image;

Obtain debonding defect image features of the gas bottle based on the first processed image, perform debonding edge detection based on the debonded defect image features of the gas bottle, and obtain a second processed image;

Defect extraction is performed on the second processed image, and contour extraction is performed on the extracted defect features to obtain defect contour features. The defect contour features are marked on the original gas cylinder X-ray image to obtain the final defect annotation image. .

Preferably, preprocessing the original gas bottle X-ray image includes: performing grayscale processing on the original gas bottle X-ray image to obtain a grayscale image, and the grayscale image is the first processed image.

Preferably, obtaining the image characteristics of the gas bottle debonding defect based on the first processed image includes:

Perform denoising processing on the first processed image to obtain the grayscale matrix of the original gas cylinder X-ray image; perform gradient calculation processing on the denoised image to obtain the gradient matrix of the original gas cylinder X-ray image;

The grayscale matrix and the gradient matrix are subjected to debonding edge detection to obtain image features of the gas cylinder debonding defect.

Preferably, denoising the first processed image includes:

Initialize a Gaussian convolution template through a Gaussian function, convert the first processed image into a grayscale value matrix, and use a zero filling method to expand the grayscale image;

Based on the Gaussian convolution template, a weighted average calculation is performed on the zero-filled grayscale values in the form of a sliding window to obtain a denoised image.

Preferably, performing gradient calculation processing on the denoised image includes:

Select a gradient operator template, calculate the gradient value of each pixel in the gradient operator template, calculate the gradient direction through the gradient value, and obtain the gradient matrix of the original gas bottle X-ray image; wherein the gradient operator template It is the Sobel first-order gradient operator template.

Preferably, the method for calculating the gradient direction is:

in, is the gradient direction, is the gradient value in the y direction calculated using the Sobel first-order gradient operator, is the gradient value in the x direction calculated using the Sobel first-order gradient operator.

Preferably, obtaining the second processed image includes:

Set a grayscale threshold and a gradient threshold according to the grayscale matrix and the gradient matrix, combined with the grayscale information of the original gas cylinder X-ray image;

Traverse the original gas cylinder X-ray image based on the gradient threshold and the grayscale threshold to determine the edge of the liner layer of the gas bottle, where the edge of the liner layer includes a first edge and a second edge;

Calculate the average gray value of the middle area between the first edge and the second edge. In the horizontal direction, if the average gray value is greater than the gray value of the first edge and the second edge, then mark it as Strong edges, otherwise marked as weak edges;

Neighborhood traversal is performed on the strong edge, and a gradient threshold is set to connect the strong edge with a weak edge whose distance is one pixel, to obtain a complete edge image, which is the second processed image.

Preferably, performing defect extraction on the second processed image includes:

Perform binarization processing on the second processed image to obtain a binary image, and perform an erosion operation based on the binary image to obtain a refined edge image;

Perform a first closing operation based on the thinned edge image to obtain a first defective area image, perform a second closing operation on the first defective area image, and obtain a second defective area image;

Perform a difference operation on the first defective area image and the second defective area image to obtain a defective area image;

Wherein, the structural elements of the first closed operation are smaller than the structural elements of the second closed operation.

Preferably, performing contour extraction on the defective area image includes:

Set the minimum side length, minimum perimeter and minimum area of the defective area image, and after screening, remove the contours of the defective area that do not meet the conditions to obtain the determined contour of the defective area;

The determined defect area outline is drawn on the original gas cylinder X-ray image in the form of a rectangular frame and displayed to obtain the final defect annotation image.

The beneficial effects of the present invention are:

1. The present invention can extract the debonding defects in the X-ray image of the composite carbon fiber cylinder, and facilitate the identification of the number, shape and other characteristics of the defects; at the same time, it can mark the outline of the defect on the X-ray image of the composite carbon fiber cylinder, which is more convenient. Visually observe the location of defects;

2. The present invention uses a customized Gaussian filter to denoise grayscale images, achieving better denoising effects and eliminating the impact of noise generated during the X-ray image acquisition process on the identification and extraction of image defects. It solves the problem of inaccurate image edge positioning and extraction caused by noise interference when processing images;

3. The present invention uses a customized gradient operator method to perform gradient calculation processing on the denoised image, which can accurately extract the gradient features of the image, making the edge area easier to locate;

4. The present invention uses an edge detection method based on adaptive gradient-grayscale threshold to perform edge detection and binarization processing on the obtained denoised image, and determines the edges on both sides of the debonding area based on the gradient-grayscale information to solve the problem. In the existing technology, such as the method of intelligently identifying and analyzing image defects using Canny edge detection, there are too many false edge areas and it is impossible to accurately locate the edge of the defective area, which is prone to problems such as false detection and difficulty in locating the defective area;

5. The present invention uses a contour screening method based on side length and area to screen the extracted contours, which solves the problem in the existing technology, such as the OpenCV contour search method, that the area and side length of the contour area cannot be screened. This reduces false detections and achieves more accurate defect contour extraction.

Description of the drawings

In order to explain the embodiments of the present invention or the technical solutions in the prior art more clearly, the drawings needed to be used in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some of the drawings of the present invention. Embodiments, for those of ordinary skill in the art, other drawings can also be obtained based on these drawings without exerting creative efforts.

Figure 1 is a flow chart of a method for locating image debonding defects in carbon fiber composite gas cylinders according to an embodiment of the present invention;

Figure 2 is a denoised image according to the embodiment of the present invention;

Figure 3 is the first edge image obtained in the embodiment of the present invention;

Figure 4 is a second edge image obtained in the embodiment of the present invention;

Figure 5 is an image after morphological processing in an embodiment of the present invention;

Figure 6 is the defect annotation image finally obtained in the embodiment of the present invention.

Detailed ways

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 only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

In order to make the above objects, features and advantages of the present invention more obvious and understandable, the present invention will be described in further detail below with reference to the accompanying drawings and specific embodiments.

The invention combines artificial prior knowledge and adaptive algorithms to identify gas cylinder defects.

1. Based on prior knowledge, determine the location of the cylinder debonding defect. This defect will only exist in the area between the aluminum liner layer and the carbon fiber composite layer. According to the characteristics of the debonding area, the image grayscale features of the debonding defect area are extracted. For example, the aluminum liner layer is thicker and X-rays are difficult to penetrate, resulting in a low image grayscale value. Because the debonding area is not tightly coupled, there is a gap that can be penetrated by X-rays, resulting in a higher gray value in the image. The overall gray value is low on both sides and the middle part is high. Using features summarized from prior knowledge, the algorithm's accuracy in identifying debonding areas has been greatly improved.

2. Using the edge detection method of adaptive grayscale-gradient threshold, the threshold is automatically set according to the different characteristics of each picture. Even when the number of defect samples is insufficient, good detection results can still be obtained.

As shown in Figure 1, this embodiment provides a method for extracting debonding defects from X-ray images of carbon fiber composite gas cylinders, which includes the following processes:

1. Obtain the X-ray image to be processed;

2. Perform grayscale processing on the acquired X-ray image to obtain the processed grayscale image, which is the first processed image;

3. Obtain the image features of the gas bottle debonding defect based on the first processed image, perform debonding edge detection based on the image features of the gas bottle debonding defect, and obtain the second processed image;

Obtaining the image characteristics of the gas bottle debonding defect based on the first processed image includes:

Perform denoising processing on the first processed image to obtain the grayscale matrix of the original gas cylinder X-ray image; perform gradient calculation processing on the denoised image to obtain the gradient matrix of the original gas cylinder X-ray image;

The grayscale matrix and the gradient matrix are subjected to debonding edge detection to obtain image features of the gas cylinder debonding defect.

4. Use morphological methods to extract defects from images processed by edge detection;

5. Perform contour extraction on the extracted defect features, and mark the extracted defect features on the original X-ray image.

Denoising the resulting grayscale image includes the following processes:

(1) Use the Gaussian function to initialize the 3*3 Gaussian convolution template.

(2) Convert the entire image into a form represented by a grayscale value matrix, and use zero padding to expand the image.

(3) Use the generated Gaussian convolution template to perform a weighted average calculation on the gray value of the zero-filled image in the form of a sliding window to obtain the Gaussian filtered gray value matrix.

The weight calculation formula of Gaussian template is:

in, is the calculated coefficient of the corresponding position of the Gaussian template. is the standard deviation, is a natural constant. It is the X-axis and Y-axis coordinates of the relative position in a coordinate system with the center point as the coordinate origin and including the X-axis and Y-axis.

Further, denoising image gradient calculation processing includes the following processes:

(1) Select the gradient operator template and calculate the gradient value of each pixel.

(2) Determine the gradient direction.

The calculation formula for calculating the gradient value in the x direction using the Sobel first-order gradient operator is:

The calculation formula for calculating the gradient value in the y direction using the Sobel first-order gradient operator is:

in, is the grayscale value of the corresponding point in the image.

The total gradient calculation formula is:

in, is the gradient value in the x direction calculated using the Sobel first-order gradient operator, is the gradient value in the y direction calculated using the Sobel first-order gradient operator.

The gradient direction calculation formula is:

The obtained denoised image is shown in Figure 2;

Further, the image gradient matrix and grayscale matrix are used for debonding edge detection, including the following process:

(1) According to the characteristics of the X-ray image, combined with the grayscale information of the image, set the grayscale and gradient thresholds to traverse the image to determine the edge of the inner bladder layer of the cylinder.

(2) Near the first edge, look for the edge of the carbon fiber layer with a larger gradient value and a gradient direction opposite to the first edge.

(3) Calculate the average gray value of the middle area of the two edges. In the horizontal direction, ensure that the average gray value of the debonded area is greater than the two edges, otherwise the edge will be marked as a weak edge.

(4) Perform an 8-neighborhood traversal on the determined strong edge, set a gradient threshold to connect it with the surrounding weak edges, and obtain a complete edge image, which is the second processed image.

The obtained first edge and second edge images are shown in Figure 3 and Figure 4.

Further, the morphological method is used to extract defect features from the image processed by edge detection, including the following processes:

(1) Binarize the image with determined edges, set the edge gray value to 255, and set the gray value of other areas to 0;

(2) Use 3*3 structural elements to perform erosion operations on the binary image to eliminate isolated points and possible noise in the image to obtain a refined edge image.

(3) Use 5*5 size structural elements to perform a closing operation on the refined edge image, fill in the small defective parts while keeping the edge of the image unchanged, and obtain the initially processed defective area image M;

(4) Use 15*15 size structural elements to perform a closing operation on the defective area image M obtained in the previous step, and fill in the larger defective parts without causing major changes in the edge of the image. After further processing, The defective area image N;

(5) Perform a difference operation on the images M and N obtained in (4) and (3) to obtain an accurate defect area image Q;

The image after morphological processing is shown in Figure 5.

Further, the extracted defect area image Q is contour extracted, and the extracted defect area is marked on the original X-ray image, including the following processes:

(1) Extract the contour of the defective area image and save the contour position in the form of an array.

(2) Set the minimum side length, minimum perimeter, and minimum area of the defect area.

(3) According to the set threshold, traverse the defect area and filter out the contours of the defect area that do not meet the conditions.

The set thresholds include three constraints: minimum side length, minimum perimeter and minimum area. When delineating the defect area, a rectangular frame is used for fitting. For the rectangular frame fitted to the defect area, the minimum side length is first used as the filtering condition, and the minimum perimeter of these rectangular frames whose minimum side length is greater than the set threshold is then calculated. Filter, and finally perform minimum area screening. The remaining rectangular frame is the determined outline of the defect area.

(4) Draw the determined outline of the defect area on the original image in the form of a rectangular frame and display it.

The final defect annotation image is shown in Figure 6.

The method of the present invention can be used on images of different carbon fiber gas cylinders and has good detection effects. At the same time, the image is processed morphologically to better display the defect location.

The above-described embodiments are only descriptions of preferred modes of the present invention and do not limit the scope of the present invention. Without departing from the design spirit of the present invention, those of ordinary skill in the art can make various modifications to the technical solutions of the present invention. All deformations and improvements shall fall within the protection scope determined by the claims of the present invention.

Claims (3)

1. A method for extracting debonding defects from X-ray images of carbon fiber composite gas cylinders, which is characterized by including: Obtain an original gas cylinder X-ray image, and preprocess the original gas cylinder X-ray image to obtain a first processed image, wherein the original gas cylinder X-ray image is a carbon fiber composite gas cylinder X-ray image; Obtain debonding defect image features of the gas bottle based on the first processed image, perform debonding edge detection based on the debonded defect image features of the gas bottle, and obtain a second processed image; Defect extraction is performed on the second processed image, and contour extraction is performed on the extracted defect features to obtain defect contour features. The defect contour features are marked on the original gas cylinder X-ray image to obtain the final defect annotation image. ; Preprocessing the original gas cylinder X-ray image includes: performing grayscale processing on the original gas cylinder X-ray image to obtain a grayscale image, where the grayscale image is the first processed image; Obtaining the image characteristics of the gas bottle debonding defect based on the first processed image includes: Perform denoising processing on the first processed image to obtain the grayscale matrix of the original gas cylinder X-ray image; perform gradient calculation processing on the denoised image to obtain the gradient matrix of the original gas cylinder X-ray image; Perform debonding edge detection on the grayscale matrix and the gradient matrix to obtain the image characteristics of the gas cylinder debonding defect; Performing denoising on the first processed image includes: Initialize a Gaussian convolution template through a Gaussian function, convert the first processed image into a grayscale value matrix, and use a zero filling method to expand the grayscale image; Based on the Gaussian convolution template, perform a weighted average calculation on the zero-filled grayscale values in the form of a sliding window to obtain a denoised image; Obtaining the second processed image includes: Set a grayscale threshold and a gradient threshold according to the grayscale matrix and the gradient matrix, combined with the grayscale information of the original gas cylinder X-ray image; The original gas bottle X-ray image is traversed based on the gradient threshold and the grayscale threshold to determine the liner layer edge of the gas bottle, where the liner layer edge includes a first edge and a second edge, where the second edge The edge is the edge of the carbon fiber layer with a large gradient value and the gradient direction is opposite to the first edge; Calculate the average gray value of the middle area between the first edge and the second edge. In the horizontal direction, if the average gray value is greater than the gray value of the first edge and the second edge, then mark it as Strong edges, otherwise marked as weak edges; Perform neighborhood traversal on the strong edge, set a gradient threshold to connect the strong edge with a weak edge that is one pixel away, and obtain a complete edge image, which is the second processed image; Performing defect extraction on the second processed image includes: Perform binarization processing on the second processed image to obtain a binary image, and perform an erosion operation based on the binary image to obtain a refined edge image; Perform a first closing operation based on the thinned edge image to obtain a first defective area image, perform a second closing operation on the first defective area image, and obtain a second defective area image; Perform a difference operation on the first defective area image and the second defective area image to obtain a defective area image; Wherein, the structural elements of the first closed operation are smaller than the structural elements of the second closed operation; Gradient calculation processing of the denoised image includes: Select a gradient operator template, calculate the gradient value of each pixel in the gradient operator template, calculate the gradient direction through the gradient value, and obtain the gradient matrix of the original gas bottle X-ray image; wherein the gradient operator template It is the Sobel first-order gradient operator template; Wherein, the method for calculating the gradient value is: G _x (x,y)＝g(x+1,y-1)+2*g(x+1,y)+g(x+1,y+1)-g(x-1,y-1 )-2*g(x-1,y)-g(x-1,y+1) G _y (x,y)＝g(x-1,y+1)+2*g(x,y+1)+g(x+1,y+1)-g(x-1,y-1 )-2*g(x,y-1)-g(x+1,y-1) Among them, g(x,y) is the value of the grayscale matrix obtained after image denoising at coordinates (x,y), G _x (x, y) is the gradient value of the image in the x direction, G _y (x, y) is the gradient value of the image in the y direction; According to the degree of change of each pixel in the image in the x and y directions, calculate the value of the gradient matrix at coordinates (x, y): Among them, G _{(x, y)} is the value of the image gradient matrix at coordinates (x, y). 2. The method for extracting debonding defects from X-ray images of carbon fiber composite gas cylinders according to claim 1, characterized in that the method for calculating the gradient direction is: Among them, θ is the gradient direction, G _y (x, y) is the gradient value in the y direction calculated using the Sobel first-order gradient operator, and G _x (x, y) is the gradient value calculated using the Sobel first-order gradient operator. The gradient value in the x direction. 3. The method for extracting debonding defects from X-ray images of carbon fiber composite gas cylinders according to claim 1, wherein the contour extraction of the defect area image includes: Set the minimum side length, minimum perimeter and minimum area of the defective area image, and after screening, remove the contours of the defective area that do not meet the conditions to obtain the determined contour of the defective area; The determined defect area outline is drawn on the original gas cylinder X-ray image in the form of a rectangular frame and displayed to obtain the final defect annotation image.