Defect identification method and system for molybdenum-rhenium alloy pipe fitting based on high-definition microscopic image

Abstract

The invention discloses a molybdenum-rhenium alloy pipe fitting defect identification method and system based on a high-definition microscopic image. The method comprises the following operation steps that high-definition microscopic equipment is used for collecting the high-definition image of a molybdenum-rhenium alloy pipe fitting; preprocessing the high-definition image of the molybdenum-rhenium alloy pipe fitting to obtain a molybdenum-rhenium image to be detected; performing execution processing of a region growing algorithm on the molybdenum-rhenium image to be detected to obtain a defect region to be detected; performing defect identification on the to-be-detected defect area to obtain type defects; and judging the defect problem of the molybdenum-rhenium alloy pipe fitting according to the type defect.

Description

Technical Field

[0001] The present invention relates to anomaly detection, and particularly to a method and system for defect identification of molybdenum-rhenium alloy pipe fittings based on high-definition microscopic images. Background Art

[0002] As an important high-temperature alloy material, molybdenum-rhenium alloy has been highly regarded due to its excellent high-temperature performance, good corrosion resistance and mechanical strength. As one of its main forms, molybdenum-rhenium alloy pipe fittings are often affected by factors such as temperature changes, mechanical stress and chemical reactions during manufacturing and use, resulting in various types of defects, mainly including pores and cracks. Pore defects are usually caused by incomplete escape of gas during the melting process, resulting in voids inside or on the surface of the metal. Cracks may be caused by overstretching, fatigue or uneven cooling, etc., resulting in cracking on the surface or inside of the metal. These defects not only affect the mechanical properties of molybdenum-rhenium alloy pipe fittings, but also may affect their application reliability in extreme environments.

[0003] Traditional defect detection methods, such as manual visual inspection or methods based on ordinary image processing, often have a high risk of missed detection and misjudgment due to problems such as insufficient image resolution and difficulty in identifying tiny defects, and it is difficult to meet the requirements of high precision and high reliability.

[0004] Currently, as an important defect detection means, image segmentation algorithms have been widely used in various industrial inspections. Especially the region growing algorithm, by starting from a specific seed point and gradually expanding the detection area, can effectively extract the defect area in the image. However, the surface defects of molybdenum-rhenium alloy pipe fittings have complex morphologies and may be similar to the surrounding background. How to improve the segmentation accuracy of the region growing algorithm and further identify the types of defects based on these segmented areas is the key issue in current research. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for defect identification of molybdenum-rhenium alloy pipe fittings based on high-definition microscopic images, which solve the above-mentioned technical problems pointed out in the prior art.

[0006] The present invention provides a method for defect identification of molybdenum-rhenium alloy pipe fittings based on high-definition microscopic images, including the following operating steps:

[0007] Use a high-definition microscopic device to collect high-definition images of molybdenum-rhenium alloy pipe fittings; preprocess the high-definition images of molybdenum-rhenium alloy pipe fittings to obtain the molybdenum-rhenium images to be detected;

[0008] Perform processing of the region growing algorithm on the molybdenum-rhenium images to be detected to obtain the defect regions to be detected; identify the defects in the defect regions to be detected to obtain the type of defects;

[0009] Judge whether there are defect problems in the molybdenum-rhenium alloy pipe fitting through the described type of defect.

[0010] Preferably, as an implementable solution; the described type of defect includes: pore type defect and crack type defect.

[0011] Preferably, as an implementable solution; perform processing of the region growing algorithm on the molybdenum-rhenium image to be detected to obtain the defect region to be detected. The specific operation steps are as follows:

[0012] Use the Otsu algorithm to calculate the gray value of each pixel point in the molybdenum-rhenium image to be detected, statistically analyze the pixel frequency through the gray value of the pixel point, and perform normalization processing on the pixel frequency to obtain the pixel probability;

[0013] Construct a gray histogram of the molybdenum-rhenium image to be detected through the pixel frequency;

[0014] Calculate the between-class variance of the separation degree between the foreground and the background of the molybdenum-rhenium image to be detected through the gray histogram. The calculation formula of the between-class variance of the separation degree is:

[0015]

[0016] In the formula, t is the threshold, expressed as the gray value from 0 to 255;

[0017] ω 0 (t) represents the sum of pixel probabilities below the threshold t;

[0018] ω 1 (t) represents the sum of pixel probabilities above the threshold t;

[0019] μ 0 (t) represents the average gray value below the threshold t;

[0020] μ 1 (t) represents the average gray value above the threshold t;

[0021] Traverse all gray values of the threshold t from 0 to 255, and calculate the between-class variance corresponding to the separation degree Obtain the standard threshold T, and the calculation formula is:

[0022]

[0023] Set the low threshold T according to the standard threshold T -1 and the high threshold T +1 ;

[0024] According to the low threshold T -1 、standard threshold T and high threshold T +1Construct a multi-level threshold sequence;

[0025] Calculate the blurriness index of each pixel point in the molybdenum-rhenium image to be detected, and construct a fuzzy entropy matrix; extract the feature map of the pixel points through the fuzzy entropy matrix; perform weighted scoring on each pixel point in the pixel point features to obtain the confidence score of each pixel point;

[0026] Screen strong pixel points and weak pixel points for each pixel point of the confidence score by constructing a multi-level threshold sequence to obtain seed points;

[0027] The seed points are divided into: strong seed points and weak seed points;

[0028] Use the region growing algorithm to grow the weak pixel points in the neighborhood of the weak seed points; calculate the Euclidean distance between the weak seed points and the weak pixel points in the neighborhood;

[0029] Preset an Euclidean distance threshold h; judge whether the Euclidean distance between the weak seed points and the weak pixel points in the neighborhood is less than the preset Euclidean distance threshold h;

[0030] If not, the weak seed point does not meet the growth condition and stops growing;

[0031] If so, the weak seed point grows the weak pixel points in the neighborhood until the weak seed point does not meet the growth condition and stops growing, and finally obtains the defect region to be detected as the weak defect region to be detected;

[0032] Grow the strong pixel points in the neighborhood of the strong seed points; calculate the Euclidean distance between the strong seed points and the strong pixel points in the neighborhood;

[0033] Judge whether the Euclidean distance between the strong seed points and the strong pixel points in the neighborhood is less than the preset Euclidean distance threshold h;

[0034] If not, the strong seed point does not meet the growth condition and stops growing;

[0035] If so, the strong seed point grows the weak pixel points in the neighborhood until the strong seed point does not meet the growth condition and stops growing, and finally obtains the defect region to be detected as the strong defect region to be detected.

[0036] Preferably, as an implementable solution; identify the defects in the defect region to be detected to obtain the type of defect, and the specific operation steps are as follows:

[0037] Cluster the pixel points with the same pixel point gray value in the weak defect region to be detected to obtain a plurality of first pixel point clustering clusters; use the first pixel point clustering clusters as the suspected pore contours;

[0038] Perform morphological preprocessing on the weak defect area to be detected to enhance the boundary contour of the suspected pore contour in the weak defect area to be detected;

[0039] Calculate the area and perimeter of the suspected pore contour for the boundary contour of the suspected pore contour;

[0040] Analyze the pore circularity of the suspected pore contour based on the area and perimeter of the suspected pore contour;

[0041] Preset a pore circularity threshold o, and determine whether the pore circularity of the suspected pore contour is greater than the pore circularity threshold o;

[0042] If not, the suspected pore contour presents a regular shape, and it is determined that there is no pore type defect in the weak defect area to be detected;

[0043] If so, it is determined that the suspected pore contour presents an irregular shape, and it is determined that there is a pore type defect in the weak defect area to be detected;

[0044] Select the pixel point with the maximum gray value in the strong defect area to be detected as the starting pixel point, and use the edge detection algorithm to judge whether the gray value of the starting pixel point is less than or equal to the neighborhood pixel points;

[0045] If not, the growth of the starting pixel point stops, and there is no crack type defect in the strong defect area to be detected;

[0046] If so, the starting pixel point continues to grow until it grows to form the crack contour of the strong defect area to be detected, and it is determined that there is a crack type defect in the strong defect area to be detected;

[0047] If the growth of the starting pixel point stops during the growth process, it is determined that there is no crack type defect in the strong defect area to be detected.

[0048] Preferably, as an implementable solution; calculate the blurriness index of each pixel point in the molybdenum-rhenium image to be detected, and construct a fuzzy entropy matrix; extract the feature map of the pixel points through the fuzzy entropy matrix, and the specific operation steps are as follows:

[0049] Set a multiplication ratio coefficient, and decompose the molybdenum-rhenium image to be detected according to the multiplication ratio coefficient by using a Gaussian pyramid to obtain multiple scale molybdenum-rhenium images;

[0050] Perform window smoothing with Gaussian kernels of different sizes on the multiple scale molybdenum-rhenium images, and continue to perform downsampling using bilinear interpolation to obtain the downsampled multiple scale molybdenum-rhenium images;

[0051] Set a pixel neighborhood window, and calculate the local mean and local standard deviation of the neighborhood pixel points for each pixel point in the multiple scale molybdenum-rhenium images according to the pixel neighborhood window;

[0052] Calculate the difference degree between the pixel grayscale value of each pixel and the local mean value; calculate the normalized ambiguity index of the difference degree to generate the ambiguity mapping matrix of each scale molybdenum-rhenium image;

[0053] Obtain the fuzzy entropy feature map through the ambiguity index of each pixel of the fuzzy entropy matrix;

[0054] Further calculate the variance of the neighborhood pixels through the local mean value of the neighborhood pixels obtained by the pixel neighborhood window to obtain the grayscale dispersion degree of the neighborhood pixels; traverse the variance of the neighborhood pixels of each pixel to obtain the grayscale dispersion degree of the neighborhood pixels of each pixel, and obtain the grayscale feature map of each pixel according to the grayscale dispersion degree of the neighborhood pixels of each pixel, and obtain the grayscale feature map;

[0055] Determine the maximum pixel grayscale value and the minimum pixel grayscale value of the pixel neighborhood within the pixel neighborhood window; calculate the difference contrast according to the maximum pixel grayscale value and the minimum pixel grayscale value; traverse and calculate the difference contrast within the pixel neighborhood window of each pixel to obtain the contrast feature map.

[0056] Preferably, as an implementable solution; perform weighted scoring on each pixel in the pixel features to obtain the confidence score of each pixel; screen each pixel of the confidence score by constructing a multi-level threshold sequence to obtain seed points, and the specific operation steps are as follows:

[0057] Perform weighted evaluation on the pixel using the fuzzy entropy feature map, the grayscale feature map and the contrast feature map of the pixels in the molybdenum-rhenium image to be detected, and calculate the confidence score of the pixel;

[0058] Traverse the confidence score of each pixel to determine whether it is greater than the low threshold T in the multi-level threshold sequence -1 ;

[0059] If not, then this pixel is a weak pixel and serves as a weak seed point;

[0060] If so, then continue to determine whether the confidence score of the remaining pixels is greater than the high threshold T +1 ;

[0061] If not, then this pixel is a normal pixel in the molybdenum-rhenium image to be detected;

[0062] If so, then take the remaining pixels as strong pixels and serve as strong seed points.

[0063] Preferably, as an implementable solution; a weighted evaluation is performed on the pixel points of the molybdenum-rhenium image to be detected by using the fuzzy entropy feature map, the grayscale feature map, and the contrast feature map of the pixel points, and the confidence score of the pixel points is calculated. The specific operation steps are as follows:

[0064] The fuzzy entropy feature map, the grayscale feature map, and the contrast feature map are normalized, and the fuzzy entropy feature map, the grayscale feature map, and the contrast feature map are merged by using a linear weighting method to obtain a fused feature;

[0065] The confidence score of the pixel points is obtained through weighted evaluation by using the fused feature.

[0066] Preferably, as an implementable solution; it is determined whether there is a defect problem in the molybdenum-rhenium alloy pipe fitting by using the type of defect. The specific operation steps are as follows:

[0067] When there is any type of defect such as a pore type defect or a crack type defect in the molybdenum-rhenium image to be detected, it is determined that there is a defect problem in the molybdenum-rhenium image to be detected.

[0068] Correspondingly, the present invention provides a defect recognition system for molybdenum-rhenium alloy pipe fittings based on high-definition microscopic images, including: a collection module; an identification module; a conclusion module;

[0069] The collection module is used to collect high-definition images of molybdenum-rhenium alloy pipe fittings by using high-definition microscopic equipment; preprocess the high-definition images of molybdenum-rhenium alloy pipe fittings to obtain a molybdenum-rhenium image to be detected;

[0070] The identification module is used to perform execution processing on the molybdenum-rhenium image to be detected by using a region growing algorithm to obtain a defect region to be detected; identify the defect in the defect region to be detected to obtain a type of defect;

[0071] The conclusion module is used to determine whether there is a defect problem in the molybdenum-rhenium alloy pipe fitting by using the type of defect.

[0072] Compared with the prior art, the embodiments of the present invention have at least the following technical advantages:

[0073] Analyze the above-mentioned defect recognition method for molybdenum-rhenium alloy pipe fittings based on high-definition microscopic images provided by the present invention. In specific applications, it obtains high-resolution images of molybdenum-rhenium alloy pipe fittings by using high-definition microscopic equipment, ensuring that tiny defect features (such as pores, cracks, etc.) can be clearly captured, effectively avoiding missed detections or misjudgments caused by blurred images, and improving the basic quality of detection. At the same time, by applying an improved region growing algorithm, the defect regions in molybdenum-rhenium alloy pipe fittings can be accurately extracted, especially defects with complete morphology and strong connectivity (such as pores and cracks). By setting appropriate thresholds and growth conditions, the algorithm shows higher sensitivity and accuracy in the recognition of fine cracks, effectively improving the comprehensiveness and meticulousness of defect detection. By identifying pore type defects and crack type defects, the detection means for abnormal defects in molybdenum-rhenium alloy pipe fittings is enhanced. Whether there is any type of defect, it can accurately determine that there is a defect problem in the pipe fitting, thereby improving the overall recognition quality and the reliability of judgment.

[0074] In summary, the defect recognition method for molybdenum-rhenium alloy pipe fittings based on high-definition microscopic images provided by the present invention realizes high-precision and high-efficiency defect detection through high-quality image acquisition, optimized preprocessing steps, accurate region growing algorithm, and comprehensive judgment of multiple types of defects, significantly improving the quality control level of molybdenum-rhenium alloy pipe fittings. Brief Description of the Drawings

[0075] Figure 1 It is the overall flowchart of a defect recognition method for molybdenum-rhenium alloy pipe fittings based on high-definition microscopic images in Embodiment 1 of the present invention;

[0076] Figure 2 It is a schematic diagram for calculating the roundness of pores in a defect recognition method for molybdenum-rhenium alloy pipe fittings based on high-definition microscopic images in Embodiment 1 of the present invention;

[0077] Figure 3 It is a fish-scale pattern schematic diagram of irregular pores in a defect recognition method for molybdenum-rhenium alloy pipe fittings based on high-definition microscopic images in Embodiment 1 of the present invention;

[0078] Figure 4 It is a schematic diagram of a weld bead of regular pores in a defect recognition method for molybdenum-rhenium alloy pipe fittings based on high-definition microscopic images in Embodiment 1 of the present invention;

[0079] Figure 5 It is the schematic diagram of the principle of a defect recognition system for molybdenum-rhenium alloy pipe fittings based on high-definition microscopic images in Embodiment 2 of the present invention. Detailed Description of the Invention

[0080] The technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0081] The present invention will be further described in detail below through specific embodiments in conjunction with the accompanying drawings.

[0082] Embodiment 1

[0083] See Figure 1 , the present invention provides a method for defect identification of molybdenum-rhenium alloy pipe fittings based on high-definition microscopic images, including the following operation steps:

[0084] S1: Use a high-definition microscopic device to collect high-definition images of molybdenum-rhenium alloy pipe fittings; preprocess the high-definition images of molybdenum-rhenium alloy pipe fittings to obtain the molybdenum-rhenium images to be detected;

[0085] It should be noted that using a high-definition microscopic device can obtain high-resolution images of the surface of molybdenum-rhenium alloy pipe fittings, ensuring that minute defect features such as pores and cracks can be captured; this is the basis for defect detection, capable of providing sufficiently clear image data to avoid missing or misjudging defects due to blurred images;

[0086] By preprocessing the high-definition images collected of molybdenum-rhenium alloy pipe fittings (such as denoising, enhancing contrast, grayscale conversion, etc.), the noise in the images can be eliminated, enhancing the identifiability of defect features; this is very important for subsequent region growing algorithms and defect identification, capable of improving the accuracy of detection;

[0087] S2: Perform processing using the region growing algorithm on the molybdenum-rhenium images to be detected to obtain the defect regions to be detected; identify the defects in the defect regions to be detected to obtain the type of defects;

[0088] The type of defects includes: pore type defects and crack type defects;

[0089] It should be noted that the region growing algorithm is a common image segmentation method, which starts from a seed point and gradually adds adjacent pixel points to the growing region. For the defect detection of molybdenum-rhenium alloy pipe fittings, the improved region growing algorithm can effectively extract possible defect regions from the images. The technical solution of this application can, by setting appropriate thresholds and growth conditions, identify morphologically complete and connected defect regions (such as pores or cracks), and can also accurately identify fine cracks; and separate these regions from the background, making sufficient preparations for subsequent defect identification types;

[0090] S3: Determine whether there are defect problems in the molybdenum-rhenium alloy pipe fittings based on the type of defect.

[0091] It should be noted that when there is any type of defect such as a pore type defect or a crack type defect in the molybdenum-rhenium image to be detected, it is determined that there are defect problems in the molybdenum-rhenium image to be detected;

[0092] Regardless of whether the molybdenum-rhenium image to be detected has a pore type defect or a crack type defect, it can be determined that there are abnormal defects in the molybdenum-rhenium image to be detected, which increases the means for judging the abnormal defects of molybdenum-rhenium alloy parts and also improves the recognition quality of molybdenum-rhenium alloy parts;

[0093] The following introduces the usage process of the improved region growing algorithm as follows:

[0094] Specifically, in step S2, perform the execution process of the region growing algorithm on the molybdenum-rhenium image to be detected to obtain the defect region to be detected; identify the defects in the defect region to be detected to obtain the type of defect. The specific operation steps are as follows:

[0095] S21: Use the Otsu algorithm to calculate the gray value of each pixel point in the molybdenum-rhenium image to be detected, statistically analyze the pixel frequency through the gray value of the pixel point, and perform normalization processing on the pixel frequency to obtain the pixel probability;

[0096] Construct a gray histogram of the molybdenum-rhenium image to be detected through the pixel frequency;

[0097] It should be noted that first, there are different gray regions in the molybdenum-rhenium image to be detected. When detecting weak boundaries or strong boundaries, a single threshold may not be able to effectively distinguish these regions, resulting in the loss of important information and affecting the subsequent processing effect; therefore, before performing the operation of the region growing algorithm, it is necessary to use the Otsu algorithm to identify the gray levels of different scenarios of the molybdenum-rhenium image to be detected, construct multi-level thresholds, and solve the problem that a single threshold for different gray regions, weak boundaries or strong boundaries cannot effectively identify the regions; this avoids the problem of errors in the defect region to be detected obtained subsequently.

[0098] Therefore, first traverse each pixel in the molybdenum-rhenium image to be detected; count the number of times the gray value of each pixel appears in the molybdenum-rhenium image to be detected to obtain the pixel frequency; then calculate the ratio of the pixel frequency of the gray value of a certain pixel in the molybdenum-rhenium image to be detected to all pixels in the molybdenum-rhenium image to be detected to obtain the pixel probability (for example, assuming that a molybdenum-rhenium image to be detected contains 1000 pixels, and the gray value of 200 pixels is 128, dividing 200 pixels by the total number of pixels 1000 in the molybdenum-rhenium image to be detected gives the probability 0.2 of the pixel with a gray value of 128); thus, construct a gray histogram. The gray histogram is a chart that describes the number of pixels at each gray level (usually from 0 to 255) in the molybdenum-rhenium image to be detected. The purpose of constructing the gray histogram is to provide basic data for subsequent between-class variance calculation and help understand the distribution of different gray levels in the molybdenum-rhenium image to be detected;

[0099] S22: Calculate the between-class variance of the separation degree between the foreground and the background of the molybdenum-rhenium image to be detected through the gray histogram. The calculation formula for the between-class variance of the separation degree is:

[0100]

[0101] In the formula, t is the threshold, representing the gray level from 0 to 255;

[0102] ω 0 (t) represents the sum of pixel probabilities below the threshold t (i.e., the foreground pixel probability);

[0103] ω 1 (t) represents the sum of pixel probabilities above the threshold t (i.e., the background pixel probability);

[0104] μ 0 (t) represents the average gray value below the threshold t (i.e., the average gray value of the foreground);

[0105] μ 1 (t) represents the average gray value above the threshold t (i.e., the average gray value of the background);

[0106] Traverse all gray levels of the threshold t from 0 to 255, and calculate the between-class variance corresponding to the separation degree Obtain the standard threshold T, and the calculation formula is:

[0107]

[0108] It should be noted that in the above formula, the grayscale histogram can reflect the difference between the foreground and background of the molybdenum-rhenium image to be detected through the grayscale level of the pixel points (i.e., the grayscale from 0 to 255). Therefore, the between-class variance is used to calculate the separation degree between the foreground and background, so as to select the threshold between the two as the judgment criterion for subsequent selection of pixel point seeds, and at the same time, it can also reduce the selection deviation caused by the mutual influence between the foreground and background of the molybdenum-rhenium image to be detected;

[0109] The Otsu algorithm is an automatic threshold selection method that determines the optimal threshold by maximizing the between-class variance. It can evaluate the discrimination degree between the foreground and background under different thresholds, and then find the optimal threshold. The threshold that maximizes the between-class variance is selected as T,

[0110] First, determine the grayscale threshold t, select the segmentation point between the foreground and background in the grayscale range of 0 - 255, and calculate the foreground pixel probability (that is, the foreground pixel probability refers to the sum of the probabilities of all pixels with grayscale values less than or equal to t: ) and the background pixel probability (that is, the background pixel probability refers to the sum of the probabilities of all pixels with grayscale values greater than t: In the formula, ω(i) represents the probability of a pixel point with grayscale value i, and N is the total number of pixel points in the molybdenum-rhenium image to be detected) to understand the distribution of different grayscale levels in the molybdenum-rhenium image to be detected; Calculate the average grayscale value of the foreground and the average grayscale value of the background respectively using the foreground pixel probability and the background pixel probability (average grayscale value of the foreground: Average grayscale value of the background: ); Then continue to calculate the between-class variance, which measures the difference degree between the foreground and background, and select the optimal threshold (i.e., the standard threshold T);

[0111] S23: Set the low threshold T -1 and the high threshold T +1 ;

[0112] According to the low threshold T -1 , the standard threshold T, and the high threshold T +1 to construct a multi-level threshold sequence;

[0113] It should be noted that the low threshold T -1 is used to detect features with lower contrast or weak boundaries. This threshold allows more pixel points to be assigned as the foreground, which helps to capture the finer structures and weak edges in the molybdenum-rhenium image to be detected; The high threshold T +1It is mainly used to ensure the detection of strong boundaries, suitable for the segmentation of high-contrast regions, which can filter out noise and ensure accurate segmentation in high-contrast regions; the low threshold allows for capturing weak edge information, while the high threshold ensures the retention of strong edges. This flexible segmentation strategy enhances the comprehensiveness of edge detection; the constructed multi-level threshold design enables the algorithm to better adapt to various features existing in the image, especially in the image of complex molybdenum-rhenium alloy parts, and can effectively distinguish regions with different gray levels; it is found that when performing the segmentation and subsequent processing of the defective area to be detected, the accurate selection of seed points depends on a good threshold setting. Therefore, it can be said that the multi-level threshold can ensure that the selected seed points are representative in the feature space, providing a technical basis for the subsequent more precise fine segmentation of the defective morphology and the surrounding background;

[0114] The embodiments of the present invention believe that constructing a multi-level threshold can more accurately identify those subtle differences manifested under different lighting conditions or surface characteristics, which is crucial for the subsequent region growing algorithm because it provides a more accurate basis for seed point selection and helps to achieve a more refined extraction of the defective area;

[0115] S24: Calculate the blurriness index of each pixel point in the molybdenum-rhenium image to be detected, and construct a fuzzy entropy matrix; extract the feature map of the pixel points through the fuzzy entropy matrix (that is, extract the gray level feature map, fuzzy entropy feature map, and contrast feature map); perform weighted scoring on each pixel point in the feature map of the pixel points to obtain the confidence score of each pixel point;

[0116] Screen strong pixel points and weak pixel points for each pixel point of the confidence score through constructing a multi-level threshold sequence to obtain seed points;

[0117] The seed points are divided into: strong seed points and weak seed points;

[0118] It should be noted that by calculating the gray deviation between the gray value of each pixel point and the standard threshold T, and constructing a membership function through the gray deviation of the gray value of each pixel point, the fuzziness of each pixel point can be reflected, so as to calculate the fuzzy entropy value of each pixel point using the membership function, which can indicate the uncertainty of each pixel point and can be clearer and more accurate when screening pixel points to become seed points;

[0119] Furthermore, construct a fuzzy entropy matrix corresponding to the fuzzy entropy value of each pixel point; then extract the feature map of the pixel points through the fuzzy entropy matrix: gray level feature map, fuzzy entropy feature map (that is, it can evaluate the information richness of each pixel point; measure the complexity and uncertainty of a certain local area in the molybdenum-rhenium image to be detected, and help to identify the details and texture information of the molybdenum-rhenium image to be detected) and contrast feature map; and perform normalized weighted calculation on the feature map of each pixel point to calculate the confidence scores of all pixel points;

[0120] Calculate the low threshold T in the multi-level threshold sequence of all pixel points -1 the standard threshold T, and the high threshold T +1 and make judgments to screen out suitable pixel points as seed points; if the confidence score of a pixel point is greater than the high threshold T +1 , then this pixel point is used as a strong pixel point; if the confidence score is less than the low threshold T -1 , then this pixel point is used as a weak pixel point. The strong pixel points and weak pixel points are used as strong seed points and weak seed points respectively, so as to obtain a seed point set, which can adapt to different gray distributions and ensure the representativeness of the seed points (that is, for example, through the screening of the high threshold, noise and misselection can be effectively suppressed, and the selected seed points are guaranteed to have a high degree of credibility); at the same time, the selection mechanism of the seed points ensures the uniform distribution of the points, which is beneficial to subsequent region growing;

[0121] Use the confidence score to screen out strong pixel points and weak pixel points for the low threshold T -1 and the high threshold T +1 in the multi-level threshold sequence of pixel points, and use the strong pixel points and weak pixel points as strong seed points and weak seed points respectively, which can also identify crack defects and pore defects in the defects respectively;

[0122] S25: Use the region growing algorithm to grow the weak pixel points in the neighborhood of the weak seed points; calculate the Euclidean distance between the weak seed points and the weak pixel points in the neighborhood

[0123] Preset the Euclidean distance threshold h; judge whether the Euclidean distance between the weak seed points and the weak pixel points in the neighborhood is less than the preset Euclidean distance threshold h

[0124] If not, then this weak seed point does not meet the growth condition and stops growing

[0125] If so, the weak seed point grows the weak pixel points in the neighborhood until this weak seed point does not meet the growth condition and stops growing, and finally obtains the defect region to be detected as the weak defect region to be detected

[0126] Grow the strong pixel points in the neighborhood of the strong seed points; calculate the Euclidean distance between the strong seed points and the strong pixel points in the neighborhood

[0127] Judge whether the Euclidean distance between the strong seed points and the strong pixel points in the neighborhood is less than the preset Euclidean distance threshold h

[0128] If not, then this strong seed point does not meet the growth condition and stops growing

[0129] If so, the strong seed points grow on the neighboring weak pixels until the strong seed points no longer meet the growth conditions, at which point the growth stops, and finally the defect area to be detected is obtained as the strong defect area to be detected.

[0130] It should be noted that the stronger strong pixels in the seed points are more clearly shown in the molybdenum-rhenium image to be detected, and even magnify relatively small defects, making them easier to identify. For example, crack defects. In molybdenum-rhenium alloy parts, some crack defects are obvious. Due to corrosion and other reasons, the cracks are larger, so they are more obvious. However, some crack defects are relatively small and not easily found because the production of molybdenum-rhenium alloy parts does not meet the standards and is relatively fragile. Therefore, strong pixels are needed to highlight such relatively small crack defects.

[0131] Weak pixels are weakly shown in the molybdenum-rhenium image to be detected and do not have such strong gray-scale brightness. However, in defect recognition, the pore defects of molybdenum-rhenium alloy parts are relatively prominent and obvious on the surface. Therefore, the embodiments of the present invention believe that using weak pixels to express relatively prominent and easily recognizable pore defects can better reflect the defect recognition.

[0132] S26: Cluster the pixel points with the same pixel gray value in the weak defect area to be detected to obtain multiple first pixel point clustering clusters; use the first pixel point clustering clusters as suspected pore contours.

[0133] Perform morphological preprocessing on the weak defect area to be detected to enhance the boundary contours of the suspected pore contours in the weak defect area to be detected.

[0134] Calculate the area and perimeter of the suspected pore contour for the boundary contour of the suspected pore contour.

[0135] Analyze the pore circularity of the suspected pore contour based on the area and perimeter of the suspected pore contour.

[0136] Preset a pore circularity threshold o, and determine whether the pore circularity of the suspected pore contour is greater than the pore circularity threshold o.

[0137] If not, the suspected pore contour presents a regular shape, and it is determined that there are no pore type defects in the weak defect area to be detected.

[0138] If so, it is determined that the suspected pore contour presents an irregular shape, and it is determined that there are pore type defects in the weak defect area to be detected.

[0139] It should be noted that the pixels in the weak defect area to be detected are clustered with the same pixel gray value, and the first pixel clusters obtained are used as suspected pores that may exist indirectly, so that the characteristics of whether there are pores can be quickly detected; and the suspected pore contours in the weak defect area to be detected are further morphologically preprocessed to highlight the pore contours and enhance some fuzzy pore boundaries, so that it is more convenient to calculate the subsequent suspected pore contour area and perimeter (that is, the suspected pore contour area and perimeter are calculated by calculating the distance between the central pixel of the first pixel cluster cluster and the edge pixel, so as to obtain the radius of the first pixel cluster cluster, and the suspected pore contour area and perimeter can be obtained by the radius), obtain the basic size characteristics of the suspected pore contour, and analyze the pore circularity (that is, the pore circularity is calculated by the suspected pore contour area and perimeter to obtain the pore circularity, and the calculation formula is: pore circularity = 4π×area / (perimeter) 2 ; For a perfect circle, such as Figure 2 As shown, the value is equal to 1; the more irregular the shape, the closer the value is to 0; the calculation is simple and suitable for quick evaluation) to evaluate the regularity of the pores. The more regular the pores, the more serious the defects (i.e. the welding of molybdenum-rhenium alloy parts is often fish-shaped, such as Figure 3 As shown, if there are pores, they will appear in a circular shape, so the more regular the pore shape, the more serious the condition. Figure 4 shown);

[0140] S27: selecting a pixel with the maximum grayscale value of pixels in the strong defect area to be detected as a starting pixel, and using an edge detection algorithm to determine whether the grayscale value of the starting pixel is less than or equal to that of the neighboring pixel;

[0141] If not, the starting pixel point stops growing, and the strong defect area to be detected has no crack type defect;

[0142] If yes, the starting pixel point continues to grow until it grows to form a crack outline of a strong defect area to be detected, and it is determined that a crack type defect exists in the strong defect area to be detected;

[0143] If the starting pixel stops growing during its growth, it is determined that the strong defect area to be detected has no crack type defect (that is, when the starting pixel grows halfway, the grayscale value of the pixel in the middle is greater than the grayscale value of the last grown pixel, then the growth is stopped at this time, and the crack outline of the strong defect area to be detected cannot be formed, so it is determined that the strong defect area to be detected has no crack type defect);

[0144] It should be noted that the edge detection algorithm is used to form a contour for the strong defect area to be detected. If a complete strong defect area to be detected is formed, it is equivalent to a coincidence verification of the crack defect in the defect area of the strong defect area to be detected. When the two coincide (i.e., the defect area of the strong defect area to be detected is wrapped, indicating double verification), it means that there is a crack-type defect in the strong defect area to be detected; if the crack contour only grows to half, it means that the two do not coincide, and there is no crack-type defect in the strong defect area to be detected;

[0145] The research in the embodiments of this application finds that the traditional region growing algorithm depends on the selection of the initial seed point. If the seed point is selected improperly, it may lead to misjudgment or missed detection. Especially in the complex molybdenum-rhenium alloy pipe fitting images, the gray-scale difference between the defect area and the background may be very small, making the selection of the seed point more difficult. Therefore, the embodiments of this application propose a processing operation to improve the accuracy of seed point selection and the ability to distinguish detailed feature defects on this basis. See the subsequent steps S241 - S244 for details.

[0146] Specifically, in step S24, calculate the blurriness index of each pixel point in the molybdenum-rhenium image to be detected, and construct a fuzzy entropy matrix; extract the feature map of the pixel points through the fuzzy entropy matrix; perform weighted scoring on each pixel point in the pixel point features to obtain the confidence score of each pixel point; screen each pixel point of the confidence score through constructing a multi-level threshold sequence to obtain the seed point. The specific operation steps are as follows:

[0147] S241: Set the multiplication ratio coefficient, and decompose the molybdenum-rhenium image to be detected according to the multiplication ratio coefficient by using the Gaussian pyramid to obtain multiple scale molybdenum-rhenium images;

[0148] Perform window smoothing on the multiple scale molybdenum-rhenium images with Gaussian kernels of different sizes, and continue to perform downsampling by using bilinear interpolation to obtain the downsampled multiple scale molybdenum-rhenium images;

[0149] Set the pixel point neighborhood window, and calculate the local mean and local standard deviation of the neighborhood pixel points for each pixel point in the multiple scale molybdenum-rhenium images (i.e., the downsampled multiple scale molybdenum-rhenium images) according to the pixel point neighborhood window;

[0150] Calculate the difference degree between the pixel point gray value of each pixel point and the local mean; calculate the blurriness index of the normalized difference degree to generate the blurriness mapping matrix (i.e., the fuzzy entropy matrix) of each scale molybdenum-rhenium image;

[0151] It should be noted that a multiplication ratio coefficient is set (i.e., 4 scale levels [1, 2, 4, 8] are constructed to cover multiple receptive fields to ensure capturing image features at different scales), and the molybdenum-rhenium image to be detected is decomposed according to the multiplication ratio coefficient using a Gaussian pyramid to obtain molybdenum-rhenium images at multiple scales, improving multi-scale analysis and enhancing feature description; different-sized Gaussian kernels are used for smoothing at each scale to avoid over-smoothing and information loss; downsampling is performed on each scale level to generate image representations with different resolutions (i.e., the molybdenum-rhenium images at multiple scales after downsampling); a pixel neighborhood window is set (i.e., the window can be set to 3×3, and the feature information of the local area is calculated for the pixel points within this window), the local mean μ and standard deviation σ of the pixel points within the 3×3 window are calculated, and then the difference degree is calculated by dividing the absolute value of the difference between the gray value of the pixel point and the local mean by the local standard deviation. The purpose of dividing by the standard deviation is to normalize the value of the difference degree, and the difference degree is converted into a fuzziness metric (i.e., a fuzziness index or fuzziness value), where the fuzziness represents the membership degree of the pixel point belonging to a certain region; a fuzziness mapping matrix is generated for each scale, and the size of the matrix changes with the scale; assuming a 100×100 image: the matrix sizes at different scales are: scale 1: 100×100, scale 2: 50×50, scale 4: 25×25, scale 8: 13×13, and the meaning of the value at each position is: 0.9 represents a highly uncertain region, 0.5 represents a medium-uncertain region, and 0.1 represents a relatively certain region;

[0152] S242: Obtain a fuzzy entropy feature map through the fuzziness index of each pixel point of the fuzzy entropy matrix;

[0153] The variance of the neighborhood pixel points is further calculated through the local mean of the neighborhood pixel points obtained through the pixel neighborhood window to obtain the gray dispersion degree of the neighborhood pixel points; the variance of the neighborhood pixel points of each pixel point is traversed to obtain the gray dispersion degree of the neighborhood pixel points of each pixel point, and a gray feature map of each pixel point is obtained according to the gray dispersion degree of the neighborhood pixel points of each pixel point, obtaining a gray feature map;

[0154] Determine the maximum pixel gray value and the minimum pixel gray value of the pixel neighborhood within the pixel neighborhood window; calculate the difference contrast according to the maximum pixel gray value and the minimum pixel gray value; traverse and calculate the difference contrast within the pixel neighborhood window of each pixel point to obtain a contrast feature map;

[0155] It should be noted that the fuzziness index of the extracted pixel points is obtained by using the fuzzy entropy matrix. First, since the fuzziness index of each pixel point has been calculated, the fuzziness index of the pixel point is directly used as the extracted fuzzy entropy feature map. The fuzzy entropy feature map can evaluate the information richness of each pixel point. Pixel points with high fuzzy entropy values are usually located in complex regions, can provide more information, and help improve the representativeness of seed points. Thus, a fuzzy entropy feature map is constructed to measure the complexity and uncertainty of a local region in the molybdenum-rhenium image to be detected, and to help identify the details and texture information of the molybdenum-rhenium image to be detected.

[0156] The local gray feature map of the molybdenum-rhenium image to be detected is identified through the pixel neighborhood window, and a gray feature map is constructed to provide information on the brightness distribution in the image, which is used to represent the basic visual features of the molybdenum-rhenium image to be detected.

[0157] By calculating the gray value contrast between pixel points, the overall visual effect of the molybdenum-rhenium image to be detected can be reflected, making the molybdenum-rhenium image to be detected more distinct, with more prominent details, and making it more attractive in visual perception. The contrast feature map measures the intensity of brightness changes in a local region and reflects the structural and edge information in the image.

[0158] S243: Weighted evaluation of the pixel points in the molybdenum-rhenium image to be detected is performed on the fuzzy entropy feature map, gray feature map, and contrast feature map of the pixel points, and the confidence score of the pixel points is calculated.

[0159] It should be noted that through the features extracted by pixel points, the fuzzy entropy feature map improves detail recognition, the gray feature map provides basic detection capabilities, and the contrast feature enhances edge detection. Therefore, by normalizing the fuzzy entropy feature map, gray feature map, and contrast feature map, the three can be weighted in one dimension. The pixel points are evaluated through the recognition capabilities of the three, so as to determine the position or recognition information (i.e., defect information) of each pixel point in the molybdenum-rhenium image to be detected, and better perform the fusion calculation of the pixel point confidence score.

[0160] S244: Traverse the confidence scores of each pixel point to determine whether they are greater than the low threshold T in the multi-level threshold sequence -1 ;

[0161] If not, then this pixel point is a weak pixel point and serves as a weak seed point.

[0162] If so, then continue to determine whether the confidence scores of the remaining pixel points are greater than the high threshold T +1 ;

[0163] If not, then this pixel point is a normal pixel point in the molybdenum-rhenium image to be detected.

[0164] If so, the remaining pixel points are regarded as strong pixel points and used as strong seed points.

[0165] It should be noted that through the screening layer by layer using the thresholds in the multi-level threshold sequence, first, the confidence scores of pixel points are judged by the low threshold T -1 to screen out the pixel points with relatively low confidence scores as weak pixel points. The weak pixel points are used as weak seed points to generate the defect area to be detected for pore defects, reflecting the identification of pore defects; and then it is judged whether the remaining pixel points are greater than the high threshold T +1 , and the strong pixel points are screened out as strong seed points to identify the defect area to be detected for crack defects, reflecting the identification of crack defects; while the pixel points with confidence scores between the low threshold T -1 and the high threshold T +1 are regarded as normal pixel points, and there may be no defect problem.

[0166] Further research finds that, for example, the fuzzy entropy feature map may be more suitable for identifying subtle changes in complex regions, while the contrast feature map is more suitable for capturing edge information. How to reasonably combine these features so that they can play their respective advantages in the final confidence score is a problem to be solved. There may be various complex textures and micro-defects on the surface of molybdenum-rhenium alloy pipe fittings. Therefore, based on step S243, the embodiment of the present invention proposes the information processing process of steps S2431 - S242 after expansion:

[0167] Specifically, in step S243, the fuzzy entropy feature map, grayscale feature map, and contrast feature map of the pixel points in the molybdenum-rhenium image to be detected are used to perform weighted evaluation on the pixel points, and the confidence score of the pixel points is calculated. The specific operation steps are as follows:

[0168] S2431: Normalize the fuzzy entropy feature map, grayscale feature map, and contrast feature map, and use the linear weighting method to combine the fuzzy entropy feature map, grayscale feature map, and contrast feature map to obtain a fused feature.

[0169] It should be noted that the fuzzy entropy feature map, grayscale feature map, and contrast feature map of each pixel point are extracted. Since the value ranges of different features may be different, we need to normalize each feature to make their value ranges consistent for subsequent weighted calculation.

[0170] The range of grayscale values is usually [0, 255]. To unify the range of the grayscale feature map, the grayscale feature map is normalized to [0, 1]: The value of fuzzy entropy is usually within a certain range. Assuming its value range is [0, max(H)], it can be normalized by the following formula: Unify the range of the fuzzy entropy feature map values to [0, 1]; the range of local contrast may be relatively large, and it can be normalized by a similar method: Also unify the range of the contrast feature map to [0, 1]; once the eigenvalue normalization is completed, these features can be combined using linear weighting to obtain the fused feature;

[0171] S2432: Obtain the confidence score of the pixel point through weighted evaluation using the fused feature;

[0172] It should be noted that when weighting the fused feature, assume that the fuzzy entropy feature map, grayscale feature map, and contrast feature map are assigned weights wH, wG, and wC, satisfying the condition: wH + wG + wC = 1, and calculate the confidence score for each pixel point in the molybdenum-rhenium image to be detected;

[0173] To improve the rationality of feature fusion and overcome the possible differences in the importance of different features for defect detection. In the above step S2431, by introducing the linear weighting method, different weights can be assigned to each feature according to actual needs. For example, crack defects usually appear as slender structures with high contrast, so a higher weight can be assigned to the contrast feature map; for pore defects, the fuzzy entropy feature map may be more important, so the weight of the fuzzy entropy feature map can be appropriately increased. This flexible weighting method enables the algorithm to dynamically adjust the importance of features according to different types of defects, thereby improving the rationality of feature fusion. Through normalization, the value ranges of all features are unified to [0, 1], ensuring that each feature has the same weight contribution in subsequent weighted calculations. This can prevent a certain feature from dominating the entire evaluation process due to its too large value range and ensure that each feature can fairly participate in the calculation of the final confidence score;

[0174] At the same time, through the method of multi-feature fusion, combining multiple features such as fuzzy entropy, grayscale, and local contrast, the characteristics of each pixel point can be more comprehensively described. The introduction of normalization processing and weighted evaluation enables the algorithm to maintain stable performance under different conditions and enhances its adaptability to complex images. At the same time, reasonable weight assignment also improves the robustness of the algorithm and reduces the influence of noise and other interference factors.

[0175] Generally speaking, in the technical solution of the embodiment of the present invention, in the process of using the Otsu algorithm to calculate the grayscale histogram and select the optimal threshold T in step S21, it is to better separate the foreground (i.e., possible defects) and the background (i.e., the normal surface of the pipe fitting), so as to avoid the loss of defect information caused by the inability of a single threshold to effectively distinguish regions with different grayscale levels. By constructing a multi-level threshold, the subtle differences manifested under different lighting conditions or surface characteristics can be more accurately identified, which is crucial for the subsequent region growing algorithm, because it provides a more accurate basis for seed point selection and helps to achieve more refined defect region extraction.

[0176] In addition, by introducing the concepts of fuzzy entropy and confidence score, the understanding of pixel point features is further enhanced, allowing the system to perform weighted scoring according to the uncertainty of each pixel point, so as to screen out suitable strong seed points and weak seed points. This method not only improves the recognition ability of different types of defects such as cracks and pores, but also ensures high detection accuracy even in the case of low contrast or more noise. Finally, through the circularity analysis of the suspected pore contour and the use of edge detection algorithm to confirm the crack contour, the effective classification of the two main types of defects is realized, increasing the reliability and accuracy of the quality assessment of molybdenum-rhenium alloy pipe fittings.

[0177] Embodiment 2

[0178] As Figure 5 shown, the present invention also provides a defect recognition system for molybdenum-rhenium alloy pipe fittings based on high-definition microscopic images, including: a collection module 10; an identification module 20; a conclusion module 30;

[0179] The collection module 10 is used to collect high-definition images of molybdenum-rhenium alloy pipe fittings by using high-definition microscopic equipment; preprocess the high-definition images of molybdenum-rhenium alloy pipe fittings to obtain the molybdenum-rhenium images to be detected;

[0180] The identification module 20 is used to perform execution processing of the region growing algorithm on the molybdenum-rhenium images to be detected to obtain the defect regions to be detected; identify the defects in the defect regions to be detected to obtain the type of defects;

[0181] The conclusion module 30 is used to judge whether there are defect problems in the molybdenum-rhenium alloy pipe fittings through the type of defects.

[0182] The defect recognition method for molybdenum-rhenium alloy pipe fittings based on high-definition microscopic images provided by the present invention realizes high-precision and high-efficiency defect detection through high-quality image acquisition, optimized preprocessing steps, accurate region growing algorithm and comprehensive judgment of multiple types of defects, significantly improving the quality control level of molybdenum-rhenium alloy pipe fittings.

[0183] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; those of ordinary skill in the art can modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A defect identification method for molybdenum-rhenium alloy pipes based on high-definition microscopic images, characterized in that: The steps are as follows: Using high-definition microscope equipment to collect high-definition images of molybdenum-rhenium alloy pipe fittings; pre-processing the high-definition images of molybdenum-rhenium alloy pipe fittings to obtain images of molybdenum-rhenium to be detected; Performing a regional growing algorithm on the molybdenum-rhenium image to be detected to obtain a defect region to be detected; identifying defects in the defect region to be detected to obtain a type of defect; It can be judged that the molybdenum-rhenium alloy pipe fitting has defects based on the type of defects.

2. According to claim 1, a defect identification method for molybdenum-rhenium alloy pipe fittings based on high-definition microscopic images is characterized in that: The types of defects include: pore type defects and crack type defects.

3. According to claim 2, a defect identification method for molybdenum-rhenium alloy pipe fittings based on high-definition microscopic images is characterized in that: The molybdenum-rhenium image to be detected is processed by a region growing algorithm to obtain a defect region to be detected. The specific operation steps are as follows: The gray value of each pixel in the molybdenum-rhenium image to be detected is calculated using the Otsu algorithm, the pixel frequency is counted according to the gray value of the pixel, and the pixel frequency is normalized to obtain the pixel probability; The grayscale histogram of the molybdenum-rhenium image to be detected is constructed by pixel frequency; The inter-class variance is calculated by using the grayscale histogram to separate the foreground and background of the molybdenum-rhenium image to be detected. The calculation formula of the inter-class variance of the separation degree is: Where t is the threshold, expressed as a grayscale from 0 to 255; ω0(t) is expressed as the sum of pixel probabilities below threshold t; ω1(t) is expressed as the sum of pixel probabilities above threshold t; μ0(t) represents the average gray value below the threshold t; μ1(t) represents the average gray value above the threshold t; Traverse all gray levels of threshold t from 0 to 255 and calculate the inter-class variance corresponding to the degree of separation The standard threshold T is obtained, and the calculation formula is: Set the low threshold T according to the standard threshold T -1 With high threshold T +1 ; According to the lower threshold T -1 , standard threshold T and high threshold T +1 Construct a multi-level threshold sequence; Calculate the fuzziness index of each pixel in the molybdenum-rhenium image to be detected, and construct a fuzzy entropy matrix; extract the feature map of the pixel through the fuzzy entropy matrix; Performing a weighted score on each pixel in the pixel feature to obtain a confidence score for each pixel; By constructing a multi-level threshold sequence, strong pixels and weak pixels are screened for each pixel of the confidence score to obtain the seed point; The seed points are divided into: strong seed points and weak seed points; Using a region growing algorithm to grow the weak seed point to the neighboring weak pixel point; calculating the Euclidean distance between the weak seed point and the neighboring weak pixel point; Preset the Euclidean distance threshold h; determine whether the Euclidean distance between the weak seed point and the neighboring weak pixel point is less than the preset Euclidean distance threshold h; If not, the weak seed point does not meet the growth conditions and growth stops; If yes, the weak seed point grows the neighboring weak pixel points until the weak seed point does not meet the growth conditions, and the growth is stopped. Finally, the defect area to be detected is obtained as the weak defect area to be detected; Grow the strong pixel points in the neighborhood of the strong seed point; Calculate the Euclidean distance between the strong seed point and the strong pixel point in the neighborhood; Determine whether the Euclidean distance between the strong seed point and the strong pixel point in the neighborhood is less than a preset Euclidean distance threshold h; If not, the strong seed point does not meet the growth conditions and stops growing; If so, the strong seed point grows the neighborhood weak pixel points until the strong seed point does not meet the growth conditions, and the growth is stopped. Finally, the defect area to be detected is obtained as the strong defect area to be detected.

4. According to claim 3, a defect identification method for molybdenum-rhenium alloy pipes based on high-definition microscopic images is characterized in that: Defects are identified in the defect area to be detected to obtain the type of defects. The specific operation steps are as follows: Clustering the pixels with the same grayscale value in the weak defect area to be detected to obtain a plurality of first pixel clusters; and using the first pixel clusters as suspected pore contours; Performing morphological preprocessing on the weak defect region to be detected, and enhancing the boundary contour of the suspected pore contour in the weak defect region to be detected; Calculating the area and perimeter of the suspected pore contour based on the boundary contour of the suspected pore contour; Analyzing the pore circularity of the suspected pore contour according to the area and perimeter of the suspected pore contour; A pore circularity threshold o is preset to determine whether the pore circularity of the suspected pore contour is greater than the pore circularity threshold o; If not, the suspected pore contour presents a regular shape, and it is determined that there is no pore type defect in the weak defect area to be detected; If yes, it is determined that the suspected pore contour is irregular in shape, and it is determined that there is a pore type defect in the weak defect area to be detected; Select a pixel with the maximum grayscale value as the starting pixel in the strong defect area to be detected, and use an edge detection algorithm to determine whether the grayscale value of the starting pixel is less than or equal to that of the neighboring pixel; If not, the starting pixel point stops growing, and the strong defect area to be detected has no crack type defect; If yes, the starting pixel point continues to grow until it grows to form a crack outline of a strong defect area to be detected, and it is determined that a crack type defect exists in the strong defect area to be detected; If the starting pixel stops growing during the growth, it is determined that the strong defect area to be detected has no crack type defect.

5. The defect identification method of molybdenum-rhenium alloy pipe fittings based on high-definition microscopic images according to claim 4 is characterized in that: Calculate the fuzziness index of each pixel in the molybdenum-rhenium image to be detected, and construct a fuzzy entropy matrix; extract the feature map of the pixel through the fuzzy entropy matrix. The specific operation steps are as follows: Setting a multiplication scale coefficient, using a Gaussian pyramid to decompose the molybdenum-rhenium image to be detected according to the multiplication scale coefficient, to obtain molybdenum-rhenium images of multiple scales; Performing window smoothing of Gaussian kernels of different sizes on the multiple scale molybdenum-rhenium images, and continuing to perform downsampling using bilinear interpolation to obtain multiple scale molybdenum-rhenium images after downsampling; Set a pixel neighborhood window, and calculate the local mean and local standard deviation of the neighboring pixels for each pixel in the molybdenum-rhenium images of multiple scales according to the pixel neighborhood window; Calculate the difference between the grayscale value of each pixel and the local mean; calculate the fuzziness index of the normalized difference to generate a fuzziness mapping matrix for each scale molybdenum-rhenium image; Obtaining a fuzzy entropy feature map through the fuzziness index of each pixel point of the fuzzy entropy matrix; The local mean of the neighborhood pixels obtained by the pixel neighborhood window is further used to calculate the variance of the neighborhood pixels to obtain the grayscale dispersion of the neighborhood pixels; the variance of the neighborhood pixels of each pixel is traversed to obtain the grayscale dispersion of the neighborhood pixels of each pixel, and the grayscale feature map of each pixel is obtained according to the grayscale dispersion of the neighborhood pixels of each pixel to obtain the grayscale feature map; Determine the maximum pixel grayscale value and the minimum pixel grayscale value of the pixel neighborhood within the pixel neighborhood window; The difference contrast is calculated based on the maximum pixel grayscale value and the minimum pixel grayscale value; the difference contrast within the pixel neighborhood window of each pixel is traversed and calculated to obtain a contrast feature map.

6. The defect identification method of molybdenum-rhenium alloy pipe fittings based on high-definition microscopic images according to claim 5 is characterized in that: Perform a weighted score on each pixel in the pixel feature to obtain a confidence score for each pixel; filter each pixel with a confidence score by constructing a multi-level threshold sequence to obtain a seed point. The specific operation steps are as follows: Performing a weighted evaluation on the pixel points using the fuzzy entropy feature map, the grayscale feature map, and the contrast feature map of the pixel points in the molybdenum-rhenium image to be detected, and calculating the confidence score of the pixel points; Traverse the confidence score of each pixel to determine whether it is greater than the low threshold T in the multi-level threshold sequence -1 ; If not, the pixel is a weak pixel and serves as a weak seed point; If so, the remaining pixels are further judged whether the confidence score is greater than the high threshold T +1 ; If not, the pixel is a normal pixel in the molybdenum-rhenium image to be detected; If so, the remaining pixels are taken as strong pixels and strong seed points.

7. The defect identification method of molybdenum-rhenium alloy pipe fittings based on high-definition microscopic images according to claim 6 is characterized in that: The fuzzy entropy feature map, the grayscale feature map, and the contrast feature map of the pixel points in the molybdenum-rhenium image to be detected are weightedly evaluated to obtain the confidence score of the pixel points. The specific operation steps are as follows: Normalizing the fuzzy entropy feature map, the grayscale feature map, and the contrast feature map, and merging the fuzzy entropy feature map, the grayscale feature map, and the contrast feature map using a linear weighting method to obtain a fusion feature; The confidence score of the pixel point is obtained by performing weighted evaluation through the fusion feature.

8. The defect identification method of molybdenum-rhenium alloy pipe fittings based on high-definition microscopic images according to claim 7 is characterized in that: The molybdenum-rhenium alloy pipe fitting is judged to have defects through the type of defects, and the specific operation steps are as follows: When the molybdenum-rhenium image to be detected has any type of defect, either a pore type defect or a crack type defect, it is determined that the molybdenum-rhenium image to be detected has a defect problem.

9. A defect recognition system for molybdenum-rhenium alloy pipes based on high-definition microscopic images, comprising: Acquisition module; Identify the module; Conclusion module; The acquisition module is used to acquire high-definition images of molybdenum-rhenium alloy pipes using high-definition microscope equipment; Preprocess the high-definition image of the molybdenum-rhenium alloy pipe fitting to obtain the molybdenum-rhenium image to be detected; The recognition module is used to perform a regional growing algorithm on the molybdenum-rhenium image to be detected to obtain a defect region to be detected; identify defects in the defect region to be detected to obtain a type of defect; The conclusion module is used to determine whether the molybdenum-rhenium alloy pipe has a defect problem based on the type of defect.

Images