A method of material growth analysis based on image features

Abstract

The application discloses a material growth analysis method based on image features, comprising: obtaining the growth condition of a nanometer thin film material under a microscope and selecting a late frame for pixel labeling; using a tree model to classify and aggregate pixels of a key frame according to the labeling, to obtain simplified position and shape information of a correction point, a domain, and a liquid growth center; obtaining a video jitter correction matrix by analyzing the center point coordinates of the correction point and applying the matrix to the domain edge and the liquid growth core; matching the coordinates of each liquid growth center to obtain the growth trajectories of different paths, filtering part of the abnormal and error value interference, and drawing the liquid growth center trajectory and the domain mask diagram. The method can capture the growth condition of the liquid growth process of the nanometer thin film material in real time, and provide traversal for subsequent analysis.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of nanometer thin film material growth result analysis, and particularly relates to a material growth analysis method based on image features. BACKGROUND

[0002] Nanometer thin film materials have wide application prospects in the fields of electronics, optics, catalysis, energy storage, and sensors. In particular, in electronics, nanometer thin films are widely used in the manufacture of high-performance semiconductor devices, memories, and sensors, driving the development of information technology and intelligent hardware. Therefore, how to efficiently and controllably prepare high-quality nanometer thin film materials is a key research topic in materials science.

[0003] The preparation method of liquid nuclei as growth centers has attracted attention as a special thin film preparation technology. In this process, materials first form small liquid nuclei in a liquid state, and then these liquid nuclei transform into crystal domains on a solid substrate, and finally grow into nanometer thin films. Compared with traditional gas-phase or solid-phase growth methods, liquid nucleus growth has significant advantages in growth speed. The formation of liquid nuclei makes the growth process more efficient, while improving the uniformity and stability of the thin film.

[0004] The introduction of machine learning and image processing significantly improves the monitoring and optimization capabilities of the thin film growth process. Through the method based on image feature extraction, the dynamic changes in the liquid nucleus growth process can be captured in real time, key data can be extracted, and the growth of the thin film can be analyzed. The processing and analysis of these image data, combined with machine learning algorithms, can predict the growth trend of the thin film, optimize the growth parameters, and thus improve the preparation efficiency and the quality of the thin film. SUMMARY

[0005] The problem to be solved by the present application is to provide a material growth analysis method based on image features, which is used to capture the dynamic changes in the liquid growth process of nanometer thin film materials in real time, extract key data, and analyze the growth of the thin film.

[0006] The application adopts the following technical scheme: a material growth analysis method based on image features, comprising the following steps:

[0007] S1, obtaining the liquid growth of nanometer thin film materials under a microscope, and selecting a late frame for pixel labeling;

[0008] S2, using a tree model to classify and aggregate pixels of key frames according to pixel labeling, constructing a decision tree model, and obtaining simplified position and shape information of inert impurities, crystal domains, and liquid growth centers;

[0009] S3, taking the geometric center point of the inert impurity as a correction point and the geometric center point of the liquid growth center as a liquid growth nucleus, obtaining a video jitter correction matrix by analyzing the coordinates of the correction point, and applying the video jitter correction matrix to the domain wall edge and the liquid growth nucleus to obtain corrected domain wall edge information and liquid growth nucleus coordinates;

[0010] S4, matching each liquid growth nucleus coordinate to obtain growth trajectories of different paths;

[0011] S5, filtering each path obtained in S4 to alleviate abnormal and error value interference, and drawing a liquid growth center trajectory and a domain mask according to the domain wall edge and the liquid growth nucleus in S3.

[0012] Preferably, in step S1, one frame before growth and one frame during growth are selected as key frames, and the inert impurities, the domain, the liquid growth center, and the background used for correction are pixel-labeled to obtain a labeled data set Piexl_Matrix (Piexl-Matrix = [R1, R2, R3, R4]).

[0013] Preferably, in step S2, a tree model is used to extract features of R1, R2, R3, and R4 from Piexl_Matrix and construct a classifier. The tree model is based on pixel position and color channel features in the training set to divide and calculate the Gini score:

[0014]

[0015] wherein R i represents the source of the data for calculation (inert impurities R1, domain R2, liquid growth center R3, and background R4), Gini represents the Gini score calculation formula, X j represents the feature used for division, and t represents the division threshold in the decision tree model.

[0016] The selection of the leaf node is recursively updated to find the optimal division feature and division point. Specifically, by minimizing the division Gini score of the training set, the leaf node is constructed, and finally a decision tree is formed:

[0017]

[0018] wherein, t * represents the best division feature and the best division point, Gini split represents the Gini score of each feature X j corresponding to the division threshold.

[0019] Finally, the pixel scale of the inert impurities, the domain, the liquid growth nucleus, and the background is divided.

[0020] Further, using Canny operator to extract the inert impurity edge Edge_r ([f, N0, 2] on the division result, only one impurity point is selected for correction, f is the number of selected key frames, Ni is the number of edge fitting points), domain edge Edge_s ([f, n1, N1, 2], n1 is the number of domains, and -1 is used to complete the matrix to ensure full elements in the case of domain disappearance), liquid growth core Edge_l ([f, n2, N2, 2], n2 is the number of liquid growth cores, and -1 is used to complete the matrix to ensure full elements in the case of growth core disappearance).

[0021] Preferably, in step S3, according to the inert impurity edge Edge-r zero dimension, f edge information is extracted, and since the area of the inert impurity is small, the average value of i (i = 0, 1, …, N0-1) contour coordinates is taken as the impurity center center_i, and all center_i are stored in the center point matrix Center_Matrix according to the sequence order, that is:

[0022] Center_Matrix = [center_i | i ∈ {0, 1, …, N0-1}]

[0023] Similarly, considering the small size of the liquid growth core, the geometric center of the liquid growth center is used as the liquid growth core, and for each of the n2 growth centers, the weighted average value of the edge points (the number is N2) is taken as the coordinate of the liquid growth core, and finally the liquid growth core matrix Center_liquid is obtained, which is in the shape of [f, n2, 2].

[0024] Further, by comparing the changes of the correction impurity center point matrix Center_Matrix value in the time scale, the picture jitter correction matrix Stable_Matrix of the selected key frame can be obtained, which is specifically:

[0025] Stabled_Matrix = [0, Center_Matrix[i+1] - Center_Matrix[i] | i ∈ {0, 1, …, N0-2}]

[0026] Then, Stabled_Matrix is applied to Edge_s and Edge_l to obtain the corrected domain edge information and the corrected liquid growth core matrix Center_liquid′:

[0027] Edge s ′ = Edge_s - Stabled_Matrix

[0028] Center_liquid' = Center_liquid - Stabled_Matrix

[0029] Preferably, in step S4, the corrected liquid growth nucleus matrix Center_liquid' is investigated, and a certain interval is selected in time from the original video to frame extraction to convert the video into multiple images as key frames. Let f represent the number of selected key frames, and n2 represent the number of liquid growth nuclei in each frame.

[0030] There are practical operations: 1. The interval of the key frame is not large (not more than 2s); 2. Multiple growth nuclei are due to the division of a growth center master nucleus, and the role of the liquid growth nucleus is mainly to promote the acceleration of growth around the nucleus, and the liquid nucleus is always in motion until it disappears.

[0031] For the case of f>1, find the nearest neighbor matching point in the previous frame for all points in n2 whose value is not (-1, -1), allow multiple points to match to the same point in the previous frame (for example, Center_liquid'[i+1,j], Center_liquid'[i+1,k] match to Center_liquid'[i,m] at the same time, i, i+1, j, k are all within the index range), and set a distance threshold at the same time.

[0032] If the Euclidean distance between the points of the current frame and the points matched with them in the previous frame is greater than the threshold, terminate the matching, and this path will no longer be matched subsequently. In order to ensure the consistency of the length of the remaining paths, select the missing points to be (-1, -1).

[0033] If the longest path contains L points, and there is no point to be matched or exceeds the distance threshold subsequently, terminate all matching and store all matching results in Paths, which is in the shape of [M, L, 2], where M is the number of paths selected, and L is the number of liquid growth centers contained in the longest path.

[0034] Preferably, in step 5, for the path matrix Paths, each element in the zero dimension represents a path, which is represented as: (path_0, path_1, … path_M-1).

[0035] When the points in the path are not (-1, -1), a Gaussian kernel is used to filter the path:

[0036]

[0037] Where x is the horizontal or vertical coordinate, k represents the size of the Gaussian kernel, σ is the smoothing weight, which is used to control the smoothing strength, i represents the current position number, and j selects the number of points around the current position i for smoothing.

[0038] Further, according to the non-1 coordinate in the crystal domain edge Edge_s, a white mask is drawn, and the background is drawn as black; according to the Paths' obtained after filtering and each path in the Paths', a path graph is drawn on the mask.

[0039] The technical scheme of the present application further provides an electronic device comprising:

[0040] one or more processors;

[0041] a storage device having one or more programs stored thereon;

[0042] When the one or more programs are executed by the one or more processors, the one or more processors implement the image feature-based material growth analysis method described above.

[0043] The technical scheme of the present application further provides a computer readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the image feature-based material growth analysis method described above.

[0044] Compared with the prior art, the technical scheme of the present application has the following technical effects:

[0045] 1. As a novel growth form in the field of material growth, the analysis method of the present application can effectively extract the information contained in this novel growth method, which is convenient for researchers to analyze the growth process and growth mechanism to promote the progress of this growth method.

[0046] 2. The analysis method of the present application adopts a machine learning method to improve the efficiency of information collection, which provides the possibility for high-throughput and large-batch data processing. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 The flowchart of the image feature-based material growth analysis method of the present application;

[0048] Figure 2 The schematic diagram of the comparison between the mask result after fitting of the pixel level and the original graph in the tree model of the present application;

[0049] Figure 3 The crystal domain mask and the liquid growth path graph drawn by the embodiment of the present application. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical scheme and advantages of the present application clearer, the technical scheme of the application will be further described in detail below with reference to the drawings. The described embodiments are only a part of the embodiments involved in the present application. All non-innovative embodiments of other researchers in the field on the basis of the embodiments belong to the protection scope of the present application. Meanwhile, the step numbers in the embodiments are only set for the convenience of description and explanation, and the order between the steps is not limited in any way. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0051] The present application focuses on the liquid growth form of nanometer thin film material, integrates the technical assistance of machine learning and computer vision, realizes efficient growth information extraction and presentation for subsequent analysis.

[0052] In an embodiment of the present application, the liquid nucleation growth information of nanometer thin film material is extracted based on image features. Taking MoS2 as an example, the nanometer thin film material liquid growth path capture method, as shown in Figure 1 , includes the following steps:

[0053] S1, acquire the growth of MoS2 under a microscope and select a late frame for pixel labeling;

[0054] S2, use a tree model to classify and aggregate pixels of key frames according to the labeling, and obtain simplified position and shape information of correction points, domains, and liquid growth centers;

[0055] S3, obtain a video jitter correction matrix by analyzing the center point coordinates of the correction points and apply it to the domain edges and liquid growth nuclei;

[0056] S4, match the coordinates of each liquid growth center to obtain the growth trajectories of different paths, and then filter to relieve part of the abnormal and error value interference;

[0057] S5, draw the liquid growth center trajectory and the domain mask map according to the domain edges in S3 and the liquid growth nuclei in S4.

[0058] Specifically, in step S1, the key frame interval is set to 5 frames, a total of 18 key frames, and the 3rd and 17th frames are selected as key frames for pixel labeling of inert impurities, domains, and liquid growth nuclei, and saved as a labeled data set.

[0059] Specifically, in step S2, a decision tree model is constructed with Gini score as the division parameter, and the maximum depth is set to 4 to ensure generalization. The decision tree model is trained on the labeled data set obtained in S1.

[0060] The model is saved and loaded into other data to be analyzed to obtain the pixel classification map of inert impurities, domains, and liquid growth centers.

[0061] The pixel is fitted into a closed figure using the Canny operator to obtain the inert impurity edge information Edge_r([18, 151, 2], the domain edge information Edge_s([18, 1, 104, 2]), and the liquid growth core Edge_l([18, 2, 189, 2]). The schematic diagram after fitting of the growth core in the embodiment is as shown in FIG. 4. Figure 2 .

[0062] Specifically, in step S3, 18 edge information is extracted according to the zero dimension of the inert impurity edge Edge_r. Since the area of the inert impurity is small, the average value of the i(i=0, 1,…, 17) contour coordinates is taken to obtain the center coordinates, which are regarded as the impurity center center_i, and all center_i are stored in the center point matrix Center_Matrix in sequence.

[0063] Considering that the size of the liquid growth core is small, the geometric center of the liquid growth center is used as the liquid growth core. For each of the two growth cores, the weighted average value of the edge points (the maximum value is 189) is taken as the growth center of the liquid growth core, and finally the liquid core growth core matrix Center_liquid is obtained, which is shaped as [18, 2, 2]. After Center_Matrix correction, Center_liquid’ is obtained, and the shape remains unchanged.

[0064] Specifically, in step S4, the key frame interval is 0.25s, the number of paths in Center_liquid’ is 2, the distance threshold is set to 250 pixels, the point matching is performed, the path is split at two points, and is stored as Paths=[path0, path1], wherein the length of path0 is 8 points, and the length of path1 is 18 points, that is, the liquid growth core of the first path disappears at the 8th key frame, and the second path maintains until the last key frame.

[0065] Specifically, in step S5, for path0 and path1, a Gaussian kernel with k=2 and σ=1 is selected for filtering, and the domain mask path is drawn, as shown in FIG. 5. Figure 3 .

[0066] It can be seen that the method of the embodiment can effectively extract the information contained in the growth mode of liquid growth, which is convenient for researchers to study the growth process and growth mechanism to promote the progress of this growth method.

[0067] In the embodiments of the present application, an electronic device is also provided, which comprises one or more processors, a storage device having one or more programs stored thereon, and when the one or more programs are executed by the one or more processors, the one or more processors implement the image feature-based material growth analysis method described in any of the above embodiments.

[0068] In the embodiments of the present application, a computer readable storage medium having a computer program stored thereon is also provided, and when the program is executed by a processor, the steps of the image feature-based material growth analysis method in any of the above embodiments are implemented.

[0069] The above merely describes the preferred embodiments of the present application, and it should be noted that those of ordinary skill in the art can make several improvements and refinements without departing from the principles of the present application, and these improvements and refinements should also be considered within the protection scope of the present application.

Claims

1. A method of material growth analysis based on image features, characterized by, It comprises the following steps: S1, acquire the liquid growth of the nanometer film material under the microscope, and select the late frame for pixel labeling; S2, use the tree model to classify and aggregate the key frames according to the pixel labeling, construct a decision tree model, and obtain the simplified position and shape information of the inert impurities, crystal domains, and liquid growth centers; S3, take the geometric center point of the inert impurities as the correction point and the geometric center point of the liquid growth center as the liquid growth nucleus, analyze the coordinates of the correction point to obtain a video jitter correction matrix, and apply it to the crystal domain edge and the liquid growth nucleus to obtain the corrected crystal domain edge information and the liquid growth nucleus coordinates; S4, match the liquid growth nucleus coordinates to obtain the growth trajectories of different paths; S5, filter each path obtained in S4 to alleviate abnormal and error value interference, and draw the liquid growth center trajectory and the crystal domain mask according to the crystal domain edge and the liquid growth nucleus in S3.

2. The image feature based material growth analysis method of claim 1, wherein, The pixel labeling in step S1 is as follows: Select a pre-growth frame and an intermediate growth frame as key frames, label the inert impurities R1, the crystal domains R2, the liquid growth centers R3, and the background R4, and save them as a labeled data set Piexl-Matrix=[R1, R2, R3, R4].

3. The image feature based material growth analysis method of claim 2, wherein, The method for constructing the decision tree model in step S2 is as follows: Use the tree model to extract the features of R1, R2, R3, and R4 from Piexl_Matrix and construct a classifier, divide based on the pixel positions and color channel features in the training set, calculate the Gini score as the recursive index of the decision tree, find the optimal division feature and division point, recursively update the leaf nodes, and form a decision tree model.

4. The image feature based material growth analysis method of claim 3, wherein: Step S2 divides the pixel scale of the inert impurities, crystal domains, liquid growth centers, and background based on the trained decision tree model, uses the Canny operator to fit the pixels into closed graphics, and extracts the inert impurity edge Edge_r, the crystal domain edge Edge_s, and the liquid growth nucleus Edge_l.

5. The image feature based material growth analysis method of claim 4, wherein, The method in step S3 is as follows: S3.1, extract the edge information according to the zeroth dimension of the inert impurity edge Edge-r, obtain the center coordinates based on the average value of the contour coordinates, obtain the impurity center, and store all the impurity centers in the impurity center point matrix Center_Matrix in sequence; S3.2, use the geometric center of the liquid growth center as the liquid growth nucleus, take the edge points for weighted average for each liquid growth center to obtain the coordinates of the liquid growth nucleus, and construct the liquid growth nucleus matrix Center_liquid; S3.3, compare the changes of the correction impurity center point matrix Center_Matrix value in the time scale to obtain the picture jitter correction matrix Stable_Matrix of the key frame; S3.4, apply Stabled_Matrix to Edge_s and Edge_l to obtain the corrected crystal domain edge information Edge_s' and the corrected liquid growth nucleus matrix Center_liquid'.

6. The image feature based material growth analysis method of claim 5, wherein, Before matching the coordinates of each liquid growth center, a certain interval is selected from the original video based on the corrected liquid growth center matrix Center_liquid' to convert the video into multiple images as key frames according to time, the number of key frames is f, the time interval is not more than 2s, and the number of liquid growth centers in each frame is n2; The plurality of liquid growth centers are divided from a liquid growth center main nucleus, the position of the liquid growth center main nucleus is relatively unchanged, and the liquid growth center is used to promote the growth acceleration around the nucleus and is always in motion until disappearing.

7. The image feature based material growth analysis method of claim 6, wherein: In step S4, the coordinates of each liquid growth center are matched, and the method is as follows: For all points in n2 whose values are not (-1, -1), find the nearest neighbor matching point in the previous frame, allow multiple liquid growth centers at the same time to match to the same point in the previous frame, and set a distance threshold to limit the matching range; If the Euclidean distance between the points in the current frame and the points in the previous frame matched therewith is greater than the threshold, the matching is terminated, the missing points are filled as (-1, -1) to ensure the consistency of the path length, and the missing points are filled as (-1, -1) to ensure the consistency of the path length; If the longest path contains L points, since there is no point to be matched or the distance threshold is exceeded, all matching is terminated, and all matching results are stored in the path matrix Paths, which has a shape of [M, L, 2], wherein M is the number of selected paths, and L is the number of liquid growth centers contained in the longest path.

8. The image feature based material growth analysis method of claim 7, wherein, In step S5, each path is filtered, and the method is as follows: For the path matrix Paths, each element in the zero dimension represents a path, and when the points in the path are not (-1, -1), the path is subjected to Gaussian kernel filtering to reduce the interference of abnormal values.

9. An electronic device, comprising: One or more processors; Storage device having one or more programs stored thereon; When the one or more programs are executed by the one or more processors, the one or more processors implement the image feature-based material growth analysis method of any one of claims 1 to 8. A computer program is stored thereon, which is executed by a processor to implement the steps in the image feature-based material growth analysis method of any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, ​

Images