Method for automatically identifying precipitated phase of single-crystal high-temperature alloy based on scanning electron microscopic image

Abstract

The invention relates to an automatic identification method for a precipitated phase of a single-crystal high-temperature alloy based on a scanning electron microscopic image, belongs to the technical field of image processing, and solves the problem that misjudgment or missed judgment is easy to occur during automatic identification and segmentation of the precipitated phase of the single-crystal high-temperature alloy in the prior art. The method comprises the following specific steps: carrying out graying treatment on an acquired scanning electron microscopic image of the single-crystal high-temperature alloy to obtain a grayscale image of precipitated phase grains; carrying out binarization processing on the grayscale image by using the first grayscale threshold and the second grayscale threshold to obtain a corresponding first binary image and a corresponding second binary image; all the grain areas on the two denoised binary images correspond to each other, and a single-layer binary image corresponding to the upper-layer grains is obtained based on the grains with the determined corresponding relation; and the crystal grains in the single-layer binary image are classified, and independent crystal grains, partially adhered crystal grains and / or completely adhered crystal grains are identified, so that efficient, stable and objective automatic identification of the precipitated phase is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and particularly to an automatic identification method for precipitation phases of single-crystal superalloys based on scanning electron microscopy images. Background Art

[0002] Due to their excellent high-temperature mechanical properties and oxidation and corrosion resistance, nickel-based single-crystal superalloys are widely used as the basic materials for key equipment such as aeroengine turbine blades and hot-end components of gas turbines. Their high-temperature mechanical properties (such as creep strength and fatigue life) are directly affected by the internal microstructure characteristics. The microstructure of nickel-based superalloys mainly consists of two phases, namely, the γ-Ni matrix phase and the γ'-Ni3Al precipitation phase (also known as the γ' precipitation phase or γ' phase). The γ' phase coherently precipitates on the matrix and is an important strengthening phase. During service, due to the combined action of temperature and stress, the γ' phase coarsens directionally along a certain direction to form a rafted structure, resulting in a significant reduction in the high-temperature mechanical properties of the alloy. The actual microscopic morphology of γ' phase precipitation and rafting is relatively complex, and multiple factors such as temperature field, plastic rheology, force field, ordering, and crystal defects may affect the morphology, distribution, and volume change of the γ' phase, thereby affecting the high-temperature mechanical properties of the alloy. Therefore, the quantitative identification and statistical analysis of the γ' phase are of great significance in material property evaluation, process optimization, and product quality control.

[0003] In traditional analysis methods, the observation and measurement of metal microstructures mainly rely on metallographic preparation and image acquisition techniques. After treatments such as corrosion, grinding, and polishing, observing and collecting images through a metallographic microscope or a scanning electron microscope (SEM) have become the main means for studying the microscopic characteristics of material grains, precipitation phases, and defect distributions. However, at present, most of the analysis of these images uses artificial quantitative metallographic methods, such as the grid intercept method and the micrometer eyepiece method. These methods are cumbersome, time-consuming, and highly dependent on the operator's experience, resulting in large errors in statistical results. Moreover, due to the limited field of view for a single observation, it is difficult to meet the engineering application requirements of large-sample and rapid detection.

[0004] In recent years, with the development of digital image processing technology and computational geometry analysis methods, automated analysis has gradually become a hot topic in the quantitative study of metallographic images. Existing automatic recognition methods mostly rely on traditional algorithms. However, in practical applications, when imaging metallographic samples, due to factors such as uneven illumination and noise interference in microscopic images, the boundaries of γ'-precipitated phases (which appear grain-like, and the precipitated phases will be referred to as grains hereafter) in the image are often blurred and discontinuous. Additionally, during the preparation of metallographic samples, the surface matrix needs to be etched away, and there are cases where the etching time is too long, resulting in excessive removal of the matrix, and further causing the γ'-phases at different depths to stack, presenting an overlapping phenomenon of the front and back layers (i.e., the upper layer and the lower layer). As a result, traditional algorithms are prone to misjudgment or missed judgment during automatic recognition and segmentation, affecting the accuracy of measurement results and the reliability of statistical data. Summary of the Invention

[0005] In view of the above analysis, the embodiments of the present invention aim to provide an automatic recognition method for precipitated phases of single-crystal superalloys based on scanning electron microscopy images, so as to solve the problem of easy misjudgment or missed judgment during the automatic recognition and segmentation of precipitated phases of single-crystal superalloys in the prior art.

[0006] The object of the present invention is mainly achieved through the following technical solutions:

[0007] The embodiments of the present invention provide an automatic recognition method for precipitated phases of single-crystal superalloys based on scanning electron microscopy images, including the following steps:

[0008] Perform grayscale processing on the collected scanning electron microscopy images of single-crystal superalloys to obtain grayscale images of the precipitated phase grains;

[0009] Perform binary processing on the grayscale image using a first grayscale threshold and a second grayscale threshold respectively to obtain corresponding first binary images and second binary images; wherein, the second grayscale threshold is greater than the first grayscale threshold;

[0010] Perform denoising processing on the first binary image and the second binary image respectively;

[0011] Correspond all the grain regions on the two denoised binary images, and obtain a single-layer binary image of the corresponding upper-layer grains based on the grains with determined corresponding relationships;

[0012] Classify the grains in the single-layer binary image to identify single grains, partially adhered grains, and / or completely adhered grains.

[0013] Further, obtaining the single-layer binary image of the corresponding upper-layer grains includes:

[0014] Identify completely upper-layer grains using the area of the regions of the corresponding relationship grains in the denoised first and second binary images;

[0015] Perform morphological closing and opening operations on the denoised second binary image respectively to obtain the corresponding output images \(I\) closed and \(I\) opened ;

[0016] Based on the output image \(I\) closed and the output image \(I\) opened , identify the upper-layer grains and lower-layer grains in the upper and lower overlapping layers;

[0017] Remove the lower-layer grains from the denoised second binary image, retain the complete upper-layer grains and the upper-layer grains in the upper and lower overlapping layers to obtain the single-layer binary image.

[0018] Furthermore, identifying the upper-layer grains and lower-layer grains in the upper and lower overlapping layers includes:

[0019] Based on the denoised second binary image with labeled grain numbers, obtain the output image \(I\) closed and the output image \(I\) opened ;

[0020] Judge each grain with the same number in the output image \(I\) closed and the output image \(I\) opened in sequence: If the ratio of the area of the grain in the output image \(I\) opened to the area of the grain in the output image \(I\) closed is greater than the preset lower-layer threshold, then the grain is the upper-layer grain in the upper and lower overlapping layers; otherwise, the grain is the lower-layer grain.

[0021] Furthermore, identifying the complete upper-layer grains includes:

[0022] Traverse each grain in the denoised second binary image. If the following conditions are met, then the grain belongs to the complete upper-layer grains:

[0023]

[0024] where \(Area(a)\) is the area of the grain in the denoised second binary image; \(Area(b)\) is the area of the corresponding grain in the denoised first binary image; \(E1\) is the preset first-level threshold.

[0025] Furthermore, identifying the individual grains includes:

[0026] Calculate the aspect ratio of each grain in the single-layer binary image in sequence. If the aspect ratio of the grain meets the preset first morphological threshold range, then identify the grain as an individual grain; otherwise, the grain is an adhered grain.

[0027] Further, identifying the partially adhered grains and completely adhered grains includes:

[0028] Based on the single-layer binary image, extract the identified adhered grains to obtain an adhered binary image;

[0029] Calculate the edges of each adhered grain in the adhered binary image to obtain the minimum convex hull corresponding to each adhered grain;

[0030] If any adhered grain and its corresponding minimum convex hull meet the morphological conditions, identify the adhered grain as partially adhered; otherwise, it is completely adhered.

[0031] Further, meeting the morphological conditions includes:

[0032] Respectively calculate the distance values from each pixel point in any adhered grain to the boundary pixel points of the adhered grain to obtain the maximum distance value h of the adhered grain area;

[0033] Respectively calculate the distance values from each pixel point of the minimum convex hull corresponding to the adhered grain to the boundary pixel points of the adhered grain to obtain the maximum distance value h' of the corresponding minimum convex hull;

[0034] If is less than a preset second morphological threshold, the morphological conditions are met.

[0035] Further, obtaining the first binary image and the second binary image includes:

[0036] Obtain the gray histogram of the gray image;

[0037] Based on the gray histogram, use the first gray threshold determined by the maximum between-class variance to segment the gray image to obtain the first binary image;

[0038] Use the peak with the largest gray value in the gray histogram as the second gray threshold to segment the gray image to obtain the second binary image.

[0039] Further, perform denoising processing on the first binary image and the second binary image respectively, including:

[0040] For any binary image, calculate the areas of all grain connected regions in the binary image respectively;

[0041] Traverse the area of each connected region in the binary image. If the area of the connected region is less than a preset connection threshold, remove the connected region;

[0042] Traverse each pixel point at the edge of the binary image. If the pixel at the edge is an incomplete boundary, remove the connected region.

[0043] Further, corresponding all grain regions on the two denoised binary images includes:

[0044] Determining the region coordinates of all grains on the two binary images respectively;

[0045] Traversing each grain region in the denoised second binary image, and determining the corresponding grains of each grain in the denoised first binary image based on the following method:

[0046] According to the coordinates of any pixel point in the grain region of the denoised second binary image, finding the corresponding pixel point coordinates in the denoised first binary image, and the grain where the corresponding pixel point is located has a corresponding relationship with the grain in the denoised second binary image.

[0047] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0048] 1. Through the statistics and analysis of a large amount of image data, the present invention proposes a method for automatically identifying and extracting the microscopic morphology parameters of precipitation-related keys from scanning electron microscopy images. By comparing and analyzing the two binary images obtained by strategic segmentation, the binary image of the upper precipitation phase is extracted, and the precipitation phase in the image is identified and analyzed. Finally, the automatic identification and statistical analysis of the precipitation phase of single crystal superalloys are realized, greatly improving the detection efficiency. It not only accurately identifies the morphological characteristics of the precipitation phase, but also quantitatively analyzes its volume fraction, size, distribution, shape parameters, etc.

[0049] 2. Selecting a reasonable segmentation threshold to obtain two binary images, combining the morphological characteristics of grains, comparing the area of grain (i.e., precipitation phase) regions in the two binary images, and screening out the completely upper-layer grains; performing morphological closing and opening operations on the obtained single-layer binary images respectively, and using the area of grains in the obtained corresponding images to identify the overlapping regions and lower-layer grains. By introducing advanced image processing and geometry technologies, the brightness and geometric characteristics presented in the SEM images due to sample preparation and imaging conditions are fully exploited, overcoming the overlapping interference caused by the lower precipitation phase to the upper precipitation phase, avoiding the subjective errors existing in the manual measurement process, and ensuring the accuracy of the data.

[0050] 3. For the shape type of the precipitation phase (i.e., identifying the grain adhesion situation) in the single-layer binary image with the upper precipitation phase extracted, combining the set characteristics of the grain shape, using different determination indexes to calculate the adhesion degree, realizing the division of single grains, edge-adhered grains and completely adhered grains, and then quantitatively statistically analyzing the grain size and shape. This not only shortens the analysis cycle, but also establishes a quantitative correlation between the microstructure and mechanical properties, providing data support for the quantitative research on the evolution of material microstructure.

[0051] 4. Remove the noise on the binary image by determining the connected regions, obtain an image with clearer grain boundaries, and at the same time remove the incomplete grains at the image boundary, reducing the interference of blurred edges and noise of the precipitates on the image, which helps to improve the accuracy of precipitate recognition.

[0052] In the present invention, the above technical solutions can also be combined with each other to achieve more preferred combination schemes. Other features and advantages of the present invention will be described in the subsequent specification, and some advantages can be made obvious from the specification, or understood by implementing the present invention. The objectives and other advantages of the present invention can be realized and obtained from the content specifically pointed out in the specification and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The drawings are only for the purpose of showing specific embodiments and are not considered to be a limitation of the present invention. Throughout the drawings, the same reference signs denote the same components.

[0054] Figure 1 It is a flowchart of the method for automatically identifying precipitates in a single-crystal superalloy based on a scanning electron microscope image according to an embodiment of the present invention;

[0055] Figure 2 It is a gray-scale schematic diagram of a single-crystal superalloy of an electron microscope image according to an embodiment of the present invention;

[0056] Figure 3 It is an example diagram of grains in a binary image obtained by dividing with different thresholds according to an embodiment of the present invention;

[0057] Figure 4 It is a schematic diagram of a binary image after preprocessing according to an embodiment of the present invention;

[0058] Figure 5a It is a schematic diagram of a single separated grain classified from the precipitates of a single-crystal superalloy according to an embodiment of the present invention;

[0059] Figure 5b It is a schematic diagram of some adhered grains classified from the precipitates of a single-crystal superalloy according to an embodiment of the present invention;

[0060] Figure 5c It is a schematic diagram of completely adhered grains classified from the precipitates of a single-crystal superalloy according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0061] The preferred embodiments of the present invention will be specifically described below with reference to the drawings, where the drawings form a part of this application and are used together with the embodiments of the present invention to explain the principles of the present invention, and are not used to limit the scope of the present invention.

[0062] Embodiment 1

[0063] A specific embodiment of the present invention discloses an automatic recognition method for precipitation phases of single-crystal superalloys based on scanning electron microscopy images, as follows Figure 1 shown, including the following steps:

[0064] Step S1: Perform grayscale processing on the collected scanning electron microscopy image of the single-crystal superalloy to obtain a grayscale image of the precipitation-phase grains;

[0065] Step S2: Perform binary processing on the grayscale image using a first grayscale threshold and a second grayscale threshold respectively to obtain corresponding first binary image and second binary image; wherein, the second grayscale threshold is greater than the first grayscale threshold;

[0066] Step S3: Perform denoising processing on the first binary image and the second binary image respectively;

[0067] Step S4: Correlate all the grain regions on the two denoised binary images, and obtain a single-layer binary image of the corresponding upper-layer grains based on the grains with the determined correspondence;

[0068] Step S5: Classify the grains in the single-layer binary image to identify single grains, partially adhered grains and / or completely adhered grains.

[0069] Through the above method, it is divided into two binary images by using different thresholds. By comparing the grain regions (i.e., precipitation phases) in the two denoised binary images, the upper-layer precipitation phases are identified, and the shapes of the precipitation phases in the single-layer binary image are classified to realize automatic and accurate identification of the precipitation phases and their states in the scanning electron microscopy images.

[0070] In the research of metallic alloys in materials science, for single-crystal superalloys, due to the uneven distribution of the precipitation-phase components in space under high-temperature service, the matrix phase in the corresponding scanning electron microscopy image is the dark background part, and the precipitation phases are shown as bright white and there will be upper and lower layers, and even the upper and lower layers overlap; at the same time, due to the depth difference of the precipitation phases, the brightness of the upper and lower layers is also different; among them, the upper-layer precipitation phases are closer to the surface, with a higher degree of corrosion, stronger reflected electron signals, and higher presented brightness, while the lower layer is located in the matrix gap, with significant signal attenuation and reduced brightness contrast (i.e., the upper layer is brighter and the lower layer is darker).

[0071] Exemplarily, the single-crystal alloy is nickel-based single crystal. Specifically, in step S1, when performing grayscale processing on the collected scanning electron microscopy image, each pixel value of the precipitation-phase tissue image of the nickel-based single-crystal material is converted into a single brightness value, removing the color information of the image and only retaining the brightness information to obtain a grayscale image, as Figure 2 shown.

[0072] Exemplarily, grayscale processing is achieved by the way of weighted average, as shown in the following formula:

[0073] Gray = 0.2989R + 0.5870G + 0.1440B,

[0074] wherein, R, G, and B respectively represent the red, green, and blue channel values of each pixel in the original image.

[0075] Specifically, in step S2, by selecting a reasonable gray-scale threshold to divide the gray-scale image, two binary images are obtained respectively for removing the influence of the lower-layer grains subsequently. Among them, since the selected second gray-scale threshold is greater than the first gray-scale threshold, the grain regions in the second binary image segmented by the second gray-scale threshold are all included in the corresponding grain regions in the first binary image. The specific steps are as follows:

[0076] S21. Obtain the gray-scale histogram of the gray-scale image, that is, count the number of pixels of each gray-scale value in the image;

[0077] It should be noted that during the binarization process, the gray-scale histogram is usually used to analyze the distribution of gray-scale image pixels. Since there are differences in the gray-scale values of the upper and lower layers in the collected scanning electron microscope images, the gray-scale histogram will present a form with two significant peaks. A peak with a low gray-scale value will appear on the left side of the bimodal histogram, corresponding to the background part of the image; a peak with a high gray-scale value will appear on the right side, corresponding to the foreground part in the image. Based on the distribution of gray-scale image pixels, a suitable threshold is selected for segmentation to obtain a binary image. Among them, the gray-scale threshold is between 0 and 255. If the gray-scale value of a pixel is greater than or equal to the gray-scale threshold, the pixel is classified as the foreground (usually assigned a value of 255, that is, white); if the gray-scale value of the pixel is less than the threshold, the pixel is classified as the background (usually assigned a value of 0, that is, black).

[0078] S22. Based on the gray-scale histogram, use the maximum between-class variance to determine the optimal threshold to obtain the first gray-scale threshold T1 for segmenting the gray-scale image to obtain the first binary image;

[0079] Exemplarily, each gray-scale value in the range of 0 - 255 is used as the threshold in turn, and the between-class variance between the foreground and background into which the image is divided under this threshold is calculated. The threshold determined by finding the maximum between-class variance between the foreground and background is used as the first gray-scale threshold for segmenting the gray-scale image.

[0080] S23. Use the peak with the largest gray-scale value in the gray-scale histogram as the second gray-scale threshold T2 to segment the gray-scale image to obtain the second binary image.

[0081] It should be noted that in the histogram, the gray value increases from left to right, and the higher the presented brightness. The core of using the maximum between-class variance to determine the threshold is to maximize the difference between the foreground and background of the image. For a histogram with bimodal characteristics, the threshold determined in this way will appear between the two peaks. Since the second gray threshold selects the right peak with the largest gray value in the bimodal histogram, the second gray threshold will be greater than the first gray threshold. Further, the grain regions (corresponding to the foreground part) in the second binary image are relatively smaller than the corresponding grain regions in the first binary image.

[0082] Through the above method, by using the preferred peak threshold combined with the maximum between-class variance threshold, the gray image is segmented to form a comparative analysis of the binary image. As Figure 3 shown, the first row (a) gives a comparison example of the upper-layer grains at the corresponding positions in the initial gray image, the first binary image, and the second binary image. The second row (b) gives a comparison example of the upper and lower overlapping grains at the corresponding positions in the above three images. The third row (c) gives a comparison example of the lower-layer grains at the corresponding positions in the above three images. It can be seen that this method can present obvious differences in grains of different layers, can effectively distinguish the overlapping regions, solves the problem that it is difficult for conventional algorithms to distinguish overlapping regions, and lays a foundation for subsequent grain recognition.

[0083] Specifically, in step S3, the two binary images are respectively denoised in the same way; among them, the specific steps for denoising any binary image include:

[0084] S31. Calculate the areas of all grain connected regions in the binary image respectively;

[0085] Specifically, mark the connected region H i in the binary image, where i = 1, 2,..., n represents the index of each connected region, and calculate the area A(H i ) corresponding to the connected region H i .

[0086] S32. Traverse the area of each connected region in the binary image. If the area of the connected region is less than the preset connected threshold, then remove the connected region;

[0087] Exemplarily, take the preset connected threshold A min as 10 pixels; for the area of any connected region, if A(H i ) is less than A min , then the connected region H i is regarded as noise and removed; by traversing the area of each connected region in the binary image, an image with clearer grain boundary details is obtained. Among them, through the analysis of the resolution of a large number of scanning electron micrographs, the magnification of the microscope during image sampling, and the structure of the grains themselves, the threshold is determined to be 10 pixel values.

[0088] S33. Traverse the pixel points at each edge of the binary image. If the pixel at the edge is an incomplete boundary, remove the connected region to remove the incomplete grains at the image boundary.

[0089] Exemplarily, for a binary image where the pixel values are only 0 or 1, and the pixel value 1 represents the grain part, traverse the pixel points at the image edge. When the pixel value is 1, delete its connected region. Finally, remove the grains at the image edge.

[0090] Through the above denoising process, remove the interference noise and incomplete grains at the boundary in the collected image to ensure the quantization accuracy of subsequent algorithm research.

[0091] Specifically, in step S4, based on the two denoised binary images, the specific steps to identify the grain regions (i.e., the upper-layer grains) suitable for subsequent quantization classification and obtain the single-layer binary image corresponding to the upper-layer grains are as follows:

[0092] S41. Perform position correspondence for all grain regions on the two denoised binary images to ensure the comparison of grains at the same position on the two images. It includes:

[0093] S411. Respectively determine the region coordinates of all grains on the two denoised binary images;

[0094] Exemplarily, since the spacing between many grains in the image is relatively narrow, the 4-neighborhood method is selected to determine the connected region as the target grain region, and then the region coordinates of the grains are obtained. Among them, the 4-neighborhood method means that for any pixel point (x, y) in the image, its 4 directly adjacent pixel points (up, down, left, and right) are defined as the 4-neighborhood of this pixel. The coordinates of these 4 neighborhood points can be expressed in turn as: (x - 1, y) (up); (x + 1, y) (down); (x, y - 1) (left); (x, y + 1) (right). If two pixel points are in the 4-neighborhood and have the same gray-scale attribute, they are considered connected, and at the same time, the coordinates of the target grain region are determined.

[0095] S412. Perform position correspondence between the coordinates of any grain in the second denoised binary image and the coordinates of each grain in the first denoised binary image;

[0096] Exemplarily, because the grain regions in the second binary image are relatively smaller than the corresponding grain regions in the first binary image. In other words, the positions of the grains in the two images in the unified coordinate system remain unchanged, and only the grain regions in the second binary image are all included in the corresponding grain regions in the first binary image. According to this characteristic, make the following judgment on any grain in the second denoised binary image to determine the corresponding grains of each grain in the first denoised binary image:

[0097] For any grain in the second binary image after denoising, take the coordinates of any pixel point within the grain region, and find the corresponding pixel point coordinates in the first binary image after denoising. The grain where the corresponding pixel point is located has a corresponding relationship with the grain in the second binary image after denoising.

[0098] S413. Traverse each grain region in the second binary image after denoising, and determine the corresponding grains of each grain in the first binary image after denoising.

[0099] S42. According to the analysis of the grains in the precipitated phase microstructure image of the nickel-based single crystal material collected, divide the grains into three types: single grains, upper and lower layer overlapping grains, and lower layer grains. Based on the grains with determined corresponding relationships, identify the upper and lower layer situations, including:

[0100] S421. Use the area of the corresponding grains in the first and second binary images after denoising to identify the completely upper layer grains;

[0101] Exemplarily, traverse each grain in the second binary image after denoising. If the following conditions are met, the grain belongs to the completely upper layer grains:

[0102]

[0103] Among them, Area(a) is the area of the grain in the second binary image after denoising; Area(b) is the area of the corresponding grain in the first binary image after denoising; E1 is a preset upper layer threshold.

[0104] It should be noted that as Figure 3 shown, for the completely upper layer grains with higher brightness, the difference in grain area between the two binary images obtained by segmentation with different thresholds is small. Combining geometric characteristics, statistical analysis is performed on the grain data in a large number of images, and the upper layer threshold is determined to be 90%, which can be adjusted according to the actual situation.

[0105] S422. Perform morphological closing operation and opening operation on the second binary image after denoising respectively to obtain the corresponding output images I closed and I opened ;

[0106] Specifically, the closing operation is completed by sequentially combining two steps of dilation and then erosion, which is expressed as:

[0107]

[0108] Among them, I closed is the output image after the morphological closing operation; I is the second binary image after denoising; B is the structural element; represents morphological dilation; represents morphological erosion.

[0109] In the above way, first perform a dilation operation on the image I to expand the bright regions, fill small holes, and connect the disconnected regions; then perform an erosion operation on the dilated image to perform erosion on the dilated image, thereby removing the noise caused by dilation and maintaining the contour integrity of the original object to obtain the output image I closed .

[0110] The opening operation is completed by the sequential combination of erosion and dilation, expressed as:

[0111]

[0112] First perform an erosion operation on the image I to remove small noises, make the object smaller, and disconnect the small connected regions; then perform a dilation operation on the eroded image, thereby restoring the main target regions left after erosion and not restoring the original noises to obtain the output image I opened .

[0113] S423. Based on the output image I closed and the output image I opened , identify the upper-layer grains and lower-layer grains in the upper and lower overlapping layers, remove the lower-layer grains on the denoised second binary image, retain the completely upper-layer grains and the upper-layer grains in the upper and lower overlapping layers, and obtain a single-layer binary image corresponding to the upper-layer grains, as Figure 4 shown.

[0114] Exemplarily, before performing the opening operation and closing operation on the denoised second binary image, pre-number the grain positions in its image, and sequentially judge the grains with the same number in the output image I closed and the output image I opened : If the area of the grain in the output image I opened and the area of the region in the output image I closed satisfy the following conditions, then the grain is the upper-layer grain in the upper and lower overlapping layers; otherwise, the grain is the lower-layer grain:

[0115]

[0116] where Area(a closed ) is the area of the region of grain a in the output image I closed ; Area(a opened ) is the area of the corresponding region of grain a in the output image I opened ; E2 is a preset second-level threshold, and this threshold is determined to be 95% according to statistical analysis, and can be adjusted according to the actual situation specifically.

[0117] Furthermore, after accurately extracting the upper precipitation phase region, calculating its area provides data support for the quantitative study of the material microstructure evolution.

[0118] Specifically, in step S5, there are three types of grain types in the scanning electron microscopy image of the single-crystal superalloy material, namely, separate grains without adhesion, grains with partial adhesion at the grain edges, and completely adhered grains with smooth adhesion edges. The steps for identifying based on the grain types in the single-layer binary image include:

[0119] S51. Identify separate grains and adhered grains;

[0120] Exemplarily, identify by the aspect ratio: sequentially calculate the aspect ratio of each grain in the single-layer binary image. If the aspect ratio of the grain satisfies the preset first morphology threshold range, then identify the grain as a separate grain; otherwise, the grain is an adhered grain. The identified separate grains are as Figure 5a shown.

[0121] Among them, the aspect ratio usually takes the ratio of the long axis to the short axis of the grain, expressed as: '

[0122] In materials science, the γ-precipitation phase separate grains are in a square state, and since the adhered grains usually consist of two or more separate grains, the first morphology threshold range is set to [1 - 1.5], that is, when 1 ≤ Aspect ≤ 1.5, the grain is determined to be a separate grain.

[0123] S52. Calculate and classify the adhesion degree of the adhered grains to identify the grains with partial adhesion at the edges and / or complete adhesion, specifically including:

[0124] S521. Based on the single-layer binary image, extract the identified adhered grains to obtain an adhered binary image;

[0125] S522. Calculate the edges of each adhered grain in the adhered binary image to obtain the minimum convex hull corresponding to each adhered grain;

[0126] Exemplarily, solve the edge pixels of the adhered grains by the divide-and-conquer method of the two-dimensional point set convex hull to obtain the convex hull. This algorithm uses a recursive method to find the "outer" points that form the convex hull, and finally obtains the smallest convex polygon that contains all points through continuous "removing" of the internal points, which is the convex figure of the adhered grain.

[0127] S523. If any adhered grain and its corresponding minimum convex hull satisfy the morphological condition, then identify the adhered grain as partially adhered, as Figure 5b shown; otherwise, it is completely adhered, as Figure 5c shown.

[0128] Specifically, perform distance transformation on the adhesive binary image to obtain an image containing distance information. For the region where the pixel values of the originally adhered grain part are 1 (i.e., inside the grain), after distance transformation, numerical values that can distinguish different distances to the background region are obtained.

[0129] Exemplarily, use the Euclidean distance to calculate the distance values of each pixel point in any adhered grain to the pixel points on the boundary of the adhered grain (i.e., the background region) respectively, and select the maximum distance value among them, denoted as h (since it is the distance value to the grain boundary, the distance value at the position of the grain geometric center is the largest).

[0130] Similarly, calculate the distance values of each pixel point of the corresponding minimum convex hull of the adhered grain to the pixel points on the boundary of the adhered grain respectively, and obtain the maximum distance value h' of the corresponding minimum convex hull (which is the depth of the grain adhesion).

[0131] h'

[0132] If h is less than the preset second morphology threshold, the morphological condition is satisfied, and the grain is determined to be partially adhered. Similarly, according to statistical analysis, the preset second morphology threshold is determined to be 90%, which can be adjusted according to the actual situation specifically.

[0133] Furthermore, after completing the determination of the grain shape type, the specific properties such as the size and shape of the precipitated phase can be further statistically analyzed correspondingly to ensure the accuracy of the data.

[0134] Compared with the prior art, an automatic recognition method for precipitated phases in single-crystal superalloys based on scanning electron microscopy images provided in this embodiment overcomes the overlapping interference of the lower precipitated phases relative to the upper precipitated phases in the scanning electron microscopy images by introducing advanced image processing and geometry technologies, extracts the two-phase diagram of the upper precipitated phases, and can automatically and quickly identify the state of the γ' phase in the images, and finally realizes the efficient and accurate extraction of the morphological features of the precipitated phases. On the one hand, the automatic analysis method of the method helps to reduce human intervention and improve the accuracy and consistency of the data; on the other hand, through the recognition of the degree of grain adhesion, the quantitative statistical analysis of the precipitated phases is realized, providing a scientific basis for optimizing the heat treatment process of single-crystal superalloys and improving their service performance.

[0135] Those skilled in the art can understand that all or part of the processes for implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium. Among them, the computer-readable storage medium is a magnetic disk, an optical disk, a read-only memory or a random access memory, etc.

[0136] As described above, it is only the preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. An automatic recognition method for precipitation phases of single crystal superalloys based on scanning electron microscopy images, characterized in that, It includes the following steps: Perform grayscale processing on the collected scanning electron microscopy image of the single crystal superalloy to obtain a grayscale image of the precipitated phase grains; Perform binarization processing on the grayscale image using a first grayscale threshold and a second grayscale threshold respectively to obtain corresponding first binary image and second binary image; wherein, the second grayscale threshold is greater than the first grayscale threshold; Perform denoising processing on the first binary image and the second binary image respectively; Correspond all grain regions on the two denoised binary images, and obtain a single-layer binary image of the corresponding upper-layer grains based on the grains with determined correspondence; Classify the grains in the single-layer binary image to identify individual grains, partially adhered grains, and / or completely adhered grains.

2. The automatic identification method of the precipitation phase of a single crystal superalloy based on a scanning electron microscope image according to claim 1, wherein Obtaining the single-layer binary image of the corresponding upper-layer grains includes: Use the area of the corresponding grains in the denoised first and second binary images to identify completely upper-layer grains; Perform morphological closing and opening operations on the denoised binary image respectively to obtain the corresponding output images I closed and I opened ; Based on the output image I closed and the output image I opened , identify the upper-layer grains and lower-layer grains in the upper and lower overlapping layers; Remove the lower-layer grains in the denoised second binary image, and retain the completely upper-layer grains and the upper-layer grains in the upper and lower overlapping layers to obtain the single-layer binary image.

3. The automatic identification method of precipitation phases of single crystal superalloy based on scanning electron microscope images according to claim 2, characterized in that, Identifying the upper-layer grains and lower-layer grains in the upper and lower overlapping layers includes: Based on the denoised second binary image with labeled grain numbers, an output image Output Image I is obtained closed and the output image Output Image I opened ; For each grain with the same number in the output image I closed and the output image I opened , make a judgment: If the ratio of the area of the grain in the output image I opened to the area of the grain in the output image I closed is greater than a preset lower layer threshold, then the grain is the upper layer grain in the upper and lower overlapping layers; otherwise, the grain is the lower layer grain.

4. The automatic recognition method for precipitation phases of single crystal superalloys based on scanning electron microscopy images according to claim 2, wherein Identifying the completely upper-layer grains includes: Traverse each grain in the denoised second binary image. If the following conditions are met, then this grain belongs to the completely upper-layer grains: Wherein, Area(a) is the area of the grain in the denoised second binary image; Area(b) is the area of the corresponding grain in the denoised first binary image; E1 is a preset first-level threshold.

5. The automatic identification method of the precipitation phase of the single crystal superalloy based on the scanning electron microscopy image according to claim 1, wherein, Identifying the individual grains includes: Calculate the aspect ratio of each grain in the single-layer binary image in turn. If the aspect ratio of the grain satisfies the preset first morphology threshold range, then identify this grain as an individual grain; otherwise, this grain is an adhered grain.

6. The automatic recognition method of the precipitation phase of the single crystal superalloy based on the scanning electron microscope image according to claim 5, wherein, Identifying the partially adhered grains and completely adhered grains includes: Based on the single-layer binary image, extract the identified adhered grains to obtain an adhered binary image; Calculate the edges of each adhered grain in the adhered binary image to obtain the minimum convex hull corresponding to each adhered grain; If any adhered grain and its corresponding minimum convex hull meet the morphological conditions, then identify this adhered grain as partially adhered; otherwise, it is completely adhered.

7. The automatic recognition method for precipitated phases of single crystal superalloy based on scanning electron microscopy images according to claim 6, wherein Meeting the morphological conditions includes: Calculate the distance values from each pixel point in any adhered grain to the boundary pixel points of this adhered grain respectively to obtain the maximum distance value h of this adhered grain region; Calculate the distance values from each pixel point of the minimum convex hull corresponding to this adhered grain to the boundary pixel points of this adhered grain respectively to obtain the maximum distance value h' of the corresponding minimum convex hull; h ' If h is less than a preset second morphology threshold, then the morphological conditions are met.

8. A method for automatically identifying the precipitated phases of a single-crystal superalloy based on scanning electron microscopy images according to any one of claims 1-7, characterized in that, Obtaining the first binary image and the second binary image includes: Obtain the grayscale histogram of the grayscale image; Based on the grayscale histogram, use the first grayscale threshold determined by the maximum between-class variance to segment the grayscale image to obtain the first binary image; Use the peak value with the largest grayscale value in the grayscale histogram as the second grayscale threshold to segment the grayscale image to obtain the second binary image.

9. The automatic identification method of the precipitation phase of a single crystal superalloy based on a scanning electron microscope image according to claim 1, wherein Performing denoising processing on the first binary image and the second binary image respectively includes: For any binary image, calculate the areas of all grain connected regions in the binary image respectively; Traverse the area of each connected region in the binary image. If the area of the connected region is less than a preset connection threshold, remove the connected region; Traverse each pixel point at the edge of the binary image. If the pixel at the edge is an incomplete boundary, remove the connected region.

10. A method for automatically identifying the precipitation phases of a single-crystal superalloy based on scanning electron microscopy images according to claim 1, characterized in that, Correspond all grain regions on the two denoised binary images, including: Determine the region coordinates of all grains on the two binary images respectively; Traverse each grain region in the second denoised binary image, and determine the corresponding grains of each grain in the first denoised binary image based on the following method: According to the coordinates of any pixel point in the grain region of the second denoised binary image, find the corresponding pixel point coordinates in the first denoised binary image. The grain where the corresponding pixel point is located has a corresponding relationship with the grain in the second denoised binary image.

Images