System and method for rapidly representing grain boundary network by using series of transmission electron microscope dark field images
Through the series of transmission electron microscope dark field imaging technology, the drift of grain boundary networks is quickly collected and corrected, and the pixel similarity is calculated, which solves the problems of low resolution and radiation damage of grain boundary network characterization in the existing technology, and achieves efficient and accurate grain boundary network characterization.
Patent Information
- Application Number
- CN202510404130.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-01
- Publication Date
- 2025-07-04
AI Technical Summary
The existing grain boundary network characterization methods have low resolution, are sensitive to crystal defects, and are prone to radiation damage. The data acquisition time is long, making it difficult to accurately and quickly obtain nanocrystal grain boundary networks.
Using a series of transmission electron microscope dark field image technology, by collecting dark field images at different positions on multiple diffraction rings, correcting the drift and calculating the data similarity between pixels and adjacent pixels, integrating the similarity and visualizing the shape of the grain boundary network.
It realizes high spatial resolution grain boundary network characterization, small electron irradiation dose and short data acquisition time, which can accurately reflect grain morphology and grain boundary network, reducing irradiation damage.
Smart Images
Figure CN120253903A_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the field of microstructure characterization of polycrystalline materials, and specifically relates to a system and method for rapidly characterizing grain boundary networks using a series of transmission electron microscopy dark field images. Background Art
[0002] The progress of human civilization is inseparable from the development of materials. As the pillar of the material field, polycrystalline materials are widely used in aerospace, transportation, building construction, electronic information and other fields. In polycrystalline materials, grain boundaries have a significant impact on their properties. For example, at room temperature, the strength of grain boundaries is higher than that of the grain interior, while the corrosion resistance of grain boundaries is weaker than that of the grain interior. Therefore, the characterization of grain boundary networks helps people directly obtain the positions of grain boundaries and statistically analyze grain sizes. Thus, obtaining grain boundary networks from samples helps people further improve the properties of materials.
[0003] Currently, the methods for obtaining grain boundary networks mainly include metallography, Electron Backscatter Diffraction (EBSD), Precession Electron Diffraction (PED), and Transmission Kikuchi Diffraction (TKD). Metallography can quickly obtain the grain boundary network of a sample, but its resolution is low and it is only applicable to micron-sized grains. Secondly, in techniques that rely on orientation calibration algorithms to determine grain morphology, such as electron backscatter diffraction technology, precession electron diffraction technology, and transmission kikuchi diffraction technology, there are some problems that cannot be ignored, such as low spatial resolution and inability to characterize nanocrystals (EBSD), being sensitive to crystal defects (EBSD and TKD), high electron irradiation dose, and easy to cause irradiation damage to the sample (PED). Moreover, all of the above techniques that use orientation to determine grain boundary networks require a long time (about 30 minutes) to collect data, and it is difficult to correct sample drift during data collection, which may lead to distortion of the characterized grain morphology.
[0004] In summary, the existing methods for obtaining grain boundary networks of crystalline materials have poor effects. Summary of the Invention
[0005] Embodiments of this application provide a system and method for rapidly characterizing grain boundary networks using a series of transmission electron microscopy dark field images, which can effectively solve the above problems.
[0006] Embodiments of this application are implemented through the following technical solutions:
[0007] On the one hand, embodiments of this application provide a method for rapidly characterizing grain boundary networks using a series of transmission electron microscopy dark field images, including the following steps:
[0008] Collect a series of dark field images at different positions on multiple diffraction rings;
[0009] Correct the drift of the series of dark field images;
[0010] For any position in the dark field image, take the intensity change of the series of dark field images at the same pixel as the feature of this pixel, and calculate the data similarity of the features corresponding to this pixel and multiple adjacent pixels;
[0011] Integrate the similarities between any pixel and multiple adjacent pixels and visualize them to obtain the morphology of the grain boundary network.
[0012] In some embodiments, collecting a series of dark field images at different positions on multiple diffraction rings includes: using the transmission electron microscope conical scanning dark field imaging technique to collect a series of dark field images at different positions on multiple diffraction rings.
[0013] In some embodiments, correcting the drift of the series of dark field images includes:
[0014] Calculate the drift value of each dark field image and apply these drift values to each dark field image, that is, each dark field image is translationally corrected according to its corresponding drift value;
[0015] In some embodiments, after correcting the drift of the series of dark field images, it further includes: performing image processing on the series of dark field images, and the image processing includes at least one sub-step of image noise reduction, contrast enhancement, and binarization.
[0016] In some embodiments, for any position in the dark field image, taking the intensity change of the series of dark field images at the same pixel as the feature of this pixel and calculating the data similarity of the features corresponding to this pixel and multiple adjacent pixels includes:
[0017] Represent the pixel values at the same position of the registered series of dark field images as a first-order tensor, and calculate the data similarity of the first-order tensors corresponding to the pixels at any position of the series of dark field images and adjacent pixels:
[0018]
[0019] where D ijk and D xyk respectively represent the first-order tensors of the pixel positions (i,j) and adjacent pixels (x,y), and |D ijk | and |D xyk | are the norms of the first-order tensors D ijk and D xyk respectively.
[0020] In some embodiments, integrating the similarities between any pixel and multiple adjacent pixels and visualizing them to obtain the morphology of the grain boundary network includes:
[0021] Calculate the representative similarity to represent each pixel value, and visualize the finally obtained matrix to obtain the grain boundary network diagram. Among them, calculating the representative similarity includes:
[0022]
[0023] In the formula, R (i,j)|(x,y) represents the cosine similarity between the first-order tensors corresponding to the pixel (i,j) and the adjacent pixel (x,y). I and S respectively represent the three-dimensional tensor of the series of dark-field images and the three-dimensional tensor composed of multiple adjacent pixels of any pixel. N n is the number of adjacent pixels, and n = 1, 2, 3, 4,....
[0024] On the other hand, the embodiment of the present application provides a system for quickly characterizing the grain boundary network using a series of transmission electron microscopy dark-field images, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method for quickly characterizing the grain boundary network in any one of the above embodiments.
[0025] Compared with the prior art, the embodiment of the present application has the following advantages and beneficial effects:
[0026] (1) A series of conical scanning dark-field images contain two-dimensional projections of sub-grains and grains, and can accurately reflect the morphology of grains and the grain boundary network in the sample;
[0027] (2) The electron irradiation dose of this method is small, and it only takes a few minutes to collect a complete data set;
[0028] (3) The drift of the dark-field image is corrected. The spatial resolution of the grain boundary network diagram is directly related to the magnification of the transmission electron microscope (adjustable), and can range from several hundred picometers to several nanometers. Description of the Drawings
[0029] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the accompanying drawings in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.
[0030] Figure 1 It is a schematic flowchart of a method for quickly characterizing the grain boundary network using a series of transmission electron microscopy dark-field images provided by some preferred embodiments of the present application;
[0031] Figure 2Schematic diagram of calculating the first-order tensors corresponding to a pixel at any position and multiple adjacent pixels in some examples of the present application. Among them, (a) shows the adjacent pixels in different neighborhoods around any pixel (i,j), and their colors correspond to those in (b) and (c); (b) shows the third-order tensor S composed of the first-order tensors corresponding to multiple adjacent pixels of the pixel (i,j) at the same position in a series of dark-field images; (c) shows two first-order tensors D composed of the pixel values of the pixels at positions (i,j) and (x,y) in k dark-field images ijk and D xyk ;
[0032] Figure 3 Bright-field image and grain boundary network diagram of nanostructured TA7 titanium alloy with large plastic deformation provided in some examples of the present application. Among them, (a) is a bright-field image obtained by a conventional transmission electron microscope, and (b) is a grain boundary network diagram obtained by using the method of the present application Detailed implementation manners
[0033] To make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Apparently, the described embodiments are some but not all of the embodiments of the present application
[0034] The terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or modules is not limited to the listed steps or modules, but optionally further includes steps or modules not listed, or optionally further includes other steps or modules inherent to these processes, methods, products or devices
[0035] The present application is implemented through the following technical solutions
[0036] On the one hand, an embodiment of the present application provides a method for quickly characterizing a grain boundary network using a series of transmission electron microscope dark-field images, including the following steps
[0037] Collect a series of dark-field images at different positions on multiple diffraction rings
[0038] Correct the drift of the series of dark-field images
[0039] For any position in the dark-field image, use the intensity change of the series of dark-field images at the same pixel as the feature of the pixel, and calculate the data similarity of the feature corresponding to the pixel and multiple adjacent pixels
[0040] Integrate the similarity between any pixel and multiple adjacent pixels and visualize it to obtain the morphology of the grain boundary network
[0041] In the above embodiments, the conical scanning dark-field imaging technique of a transmission electron microscope can be used to collect a series of dark-field images at different positions on multiple diffraction rings. For the convenience of understanding, the following examples are further described: When using the conical scanning dark-field images, a series of dark-field images at different positions on multiple diffraction rings can be collected through a transmission electron microscope using an electron beam control system. One bright-field image can be collected every 10 dark-field images, and it only takes a few minutes to collect a complete set of data.
[0042] The drift of the series of dark-field images can be corrected using an image registration method. Specifically, when correcting the drift of the dark-field images and performing image processing, an image registration algorithm can be selected according to the characteristics of the sample to correct the drift of each dark-field image. Image processing algorithms can be selected based on the quality of the dark-field images and the grain morphology to perform operations such as image noise reduction, contrast enhancement, and binarization on the series of dark-field images.
[0043] In the step of taking the intensity change of the series of dark-field images at the same pixel as the feature of the pixel, the pixel values at the same position of the registered series of dark-field images can be represented as a first-order tensor D ijk , where i and j represent a certain pixel position of each image, and k represents the corresponding position in the reciprocal space of the electron beam during conical scanning. In the step of calculating the data similarity of the feature corresponding to the pixel and multiple adjacent pixels, the similarity between the first-order tensors corresponding to any pixel and multiple adjacent pixels can be calculated. Specifically, an evaluation function can be selected according to the characteristics of the first-order tensor to calculate the similarity between the first-order tensors corresponding to any position of the series of dark-field images and multiple adjacent pixels. In subsequent steps, a representative similarity can be obtained by combining the similarities between any pixel and multiple adjacent pixels, and the morphological structure of the grain boundary network can be obtained by visualizing the similarity values corresponding to the series of pixels.
[0044] In some of these embodiments, collecting a series of dark-field images at different positions on multiple diffraction rings includes: using the conical scanning dark-field imaging technique of a transmission electron microscope to collect a series of dark-field images at different positions on multiple diffraction rings.
[0045] In some of these embodiments, correcting the drift of the series of dark-field images includes: calculating the drift value of each dark-field image and applying these drift values to each dark-field image, that is, each dark-field image is translationally corrected according to its corresponding drift value;
[0046] In some of these embodiments, after correcting the drift of the series of dark-field images, it further includes performing image processing on the series of dark-field images, and the image processing includes at least one sub-step of image noise reduction, contrast enhancement, and binarization.
[0047] In the above embodiments, one of the steps of image noise reduction, contrast enhancement, and binarization can be adopted, or multiple of them can be adopted. The image processing process can be performed at any time after correcting the drift of the series of dark field images, and can also be performed once or multiple times. With this setting, the influence of image noise can be reduced.
[0048] In some of these embodiments, for any position in the dark field image, taking the intensity change of the series of dark field images at the same pixel as the feature of this pixel, calculating the data similarity of the feature corresponding to this pixel and multiple adjacent pixels includes:
[0049] Representing the pixel values at the same position of the registered series of dark field images as a first-order tensor, the following method can be used to calculate the data similarity between the pixel at any position of the series of dark field images and the corresponding first-order tensors of adjacent pixels:
[0050]
[0051] In the formula, D ijk and D xyk respectively represent the first-order tensors of the pixel positions (i, j) and the adjacent pixel (x, y), and |D ijk | and |D xyk | are the norms of the first-order tensors D ijk and D xyk respectively.
[0052] In some of these embodiments, synthesizing the similarity between any pixel and multiple adjacent pixels and visualizing it to obtain the morphology of the grain boundary network includes:
[0053] Calculating a representative similarity to represent each pixel value, and visualizing the finally obtained matrix to obtain the grain boundary network diagram. Among them, calculating the representative similarity includes:
[0054]
[0055] In the formula, R (i,j)|(x,y) represents the cosine similarity between the first-order tensors corresponding to the pixel (i, j) and the adjacent pixel (x, y), I and S respectively represent the three-dimensional tensor of the series of dark field images and the three-dimensional tensor composed of multiple adjacent pixels of any pixel, and N n is the number of adjacent pixels, n = 1, 2, 3, 4,....
[0056] Specific example:
[0057] Please refer to Figure 1 and Figure 2 , the embodiments of the present application provide a method for quickly characterizing the grain boundary network using a series of transmission electron microscope dark field images, mainly including the following steps:
[0058] S1. Use the electron beam control system in the transmission electron microscope to collect a series of conical scanning dark field images at different positions on multiple diffraction rings, and collect a bright field image every 10 dark field images. For example, for 10 diffraction rings of TA7 titanium alloy from (10-10) to (0004), the incident electron beam rotates around the main optical axis by about 2° or 1° every other position on the same diffraction ring, that is, 180 or 360 dark field images are collected on each diffraction ring. The rate of a typical CMOS camera installed in the transmission electron microscope to shoot 1K×1K images can reach 100 frames / second. Therefore, it only takes about 1 minute at the fastest to collect a complete set of data.
[0059] S2. Assume that the drift values of the 10 dark field images between two adjacent bright field images in the dataset are approximately linear. Use the interpolation method to calculate the drift value of each dark field image and apply these drift values to each dark field image; an appropriate image segmentation algorithm, such as the U-Net neural network, can be used to binarize the series of dark field images.
[0060] S3. Refer to Figure 2 (c), represent the pixel values at the same position of the registered series of dark field images as a first-order tensor D ijk , where i and j represent a certain pixel position of each image, and k represents the corresponding position in the reciprocal space when the electron beam is conically scanned.
[0061] S4. Use the cosine similarity function to calculate the data similarity between the first-order tensors corresponding to the pixels at any position in the series of dark field images and the adjacent pixels; the formula of the cosine similarity function is as follows:
[0062]
[0063] In the formula, D ijk and D xyk respectively represent the first-order tensors at the pixel positions (i,j) and the adjacent pixel (x,y), |D ijk | and |D xyk | are the norms of the first-order tensors D ijk and D xyk respectively;
[0064] S5. By synthesizing the similarity between each pixel and its adjacent pixel, one of the mathematical methods such as arithmetic mean, geometric mean, maximum / minimum value, etc. can be selected to calculate a representative similarity to represent each pixel value. Visualize the finally obtained matrix to get the grain boundary network diagram. The average formula adopted in this application is as follows:
[0065]
[0066] In the formula, R (i,j)|(x,y)It represents the cosine similarity between the first-order tensors corresponding to the pixel (i,j) and the adjacent pixel (x,y). I and S respectively represent the three-dimensional tensor of the series of dark-field images and the three-dimensional tensor composed of multiple adjacent pixels of any pixel. N n is the number of adjacent pixels. Refer to Figure 2 (a). Considering the adjacent pixels in order, n can take values of 1, 2, 3, 4, ….
[0067] Please refer to Figure 3 . Compared with the ordinary bright-field image that cannot clearly show the nanocrystal boundaries, the grain boundary network diagram obtained in this application clearly shows the positions of the grain boundaries, the morphologies of the grains, and the sub-grain boundaries inside the grains.
[0068] On the other hand, the embodiment of the present application provides a system for rapidly characterizing the grain boundary network using a series of transmission electron microscopy dark-field images, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method for rapidly characterizing the grain boundary network using a series of transmission electron microscopy dark-field images in any one of the above embodiments.
[0069] The embodiment of the present application also provides a computer medium, on which a computer program is stored. The computer program is loaded by the processor to execute the steps in the method for rapidly characterizing the grain boundary network using a series of transmission electron microscopy dark-field images in any one of the above embodiments.
[0070] As described above, it is only a preferred embodiment of the present application and does not impose any limitations on the progress of the present application. Any simple modification or equivalent change made to the above embodiments based on the technical essence of the present application falls within the protection scope of the present application.
Claims
1. A method for rapidly characterizing a grain boundary network using a series of transmission electron microscope dark field images, characterized in that It includes the following steps: Collect a series of dark field images at different positions on multiple diffraction rings; Correct the drift of the series of dark field images; For any position in the dark field image, take the intensity change of the series of dark field images at the same pixel as the feature of this pixel, and calculate the data similarity of the feature corresponding to this pixel and multiple adjacent pixels; Integrate the similarity between any pixel and multiple adjacent pixels and visualize it to obtain the morphology of the grain boundary network.
2. The method for rapidly characterizing a grain boundary network using a series of dark field images of a transmission electron microscope according to claim 1, wherein Collecting a series of dark field images at different positions on multiple diffraction rings includes: Use the transmission electron microscope conical scanning dark field imaging technique to collect a series of dark field images at different positions on multiple diffraction rings.
3. The method for rapidly characterizing a grain boundary network using a series of dark field images of a transmission electron microscope according to claim 1, characterized in that, Correcting the drift of the series of dark field images includes: Calculate the drift value of each dark field image and apply these drift values to each dark field image, that is, each dark field image is translationally corrected according to its corresponding drift value.
4. The method for rapidly characterizing a grain boundary network using a series of transmission electron microscope dark field images according to claim 1, wherein After correcting the drift of the series of dark field images, it further includes: Perform image processing on the series of dark field images, and the image processing includes at least one sub-step of image noise reduction, contrast enhancement, and binarization.
5. The method for rapidly characterizing a grain boundary network using a series of transmission electron microscope dark field images according to claim 1, wherein For any position in the dark field image, taking the intensity change of the series of dark field images at the same pixel as the feature of this pixel, and calculating the data similarity of the feature corresponding to this pixel and multiple adjacent pixels includes: Represent the pixel values at the same position of the registered series of dark field images as a first-order tensor, and calculate the data similarity of the first-order tensors corresponding to the pixels at any position of the series of dark field images and adjacent pixels; where D ijk and D xyk represent the first-order tensors at pixel positions (i, j) and adjacent pixel (x, y) respectively, |D ijk | and |D xyk | are the moduli of the first-order tensors D ijk and D xyk respectively.
6. The method for rapidly characterizing a grain boundary network using a series of dark field images of a transmission electron microscope according to claim 1, characterized in that Integrating the similarity between any pixel and multiple adjacent pixels and visualizing it to obtain the morphology of the grain boundary network includes: Calculate a representative similarity to represent each pixel value, and visualize the finally obtained matrix to obtain a grain boundary network diagram, where calculating the representative similarity includes: where R (i,h)|(x,y) represents the cosine similarity between the first-order tensors corresponding to the pixel (i, j) and the adjacent pixel (x, y), I and S respectively represent the three-dimensional tensor of the series of dark-field images and the three-dimensional tensor composed of multiple adjacent pixels of any pixel, N n is the number of adjacent pixels, and n = 1, 2, 3, 4,....
7. A system for rapidly characterizing a grain boundary network using a series of transmission electron microscopy dark field images, characterized in that, It includes a memory and a processor, the memory stores a computer program, and the processor executes the computer program to implement the method for rapidly characterizing the grain boundary network using a series of transmission electron microscope dark field images according to any one of claims 1-6.