A method and device for extracting roughness of ternary positive electrode materials by multi-feature fusion

Abstract

The present invention discloses a method and device for extracting the roughness of a ternary cathode material SEM image with multi-feature fusion. The method comprises the following steps: obtaining SEM images of the ternary cathode material at different sintering temperatures, preprocessing the acquired SEM images; extracting the fractal features of the preprocessed SEM images using a double blanket method, and expressing them using fractal dimensions; analyzing and extracting the texture features of the preprocessed SEM images using a grayscale co-occurrence matrix; and fusing the extracted fractal dimensions with the texture features to obtain the roughness of the SEM image of the ternary cathode material with multi-feature fusion. The present invention proposes a method of combining fractal dimensions with information entropy to characterize the roughness of the ternary cathode material, thereby realizing the extraction of the texture features of the SEM images. The feature quantities extracted by various methods complement each other, making the roughness extraction of the SEM images more comprehensive and effective.

Description

Technical Field

[0001] The invention relates to the technical field of image processing, and in particular discloses a method and a device for extracting roughness from a SEM image of a ternary positive electrode material by fusion of multiple features. Background Art

[0002] In order to meet the growing needs, research on high-performance battery cathode materials has become an important task in recent years. The production capacity and market share of ternary cathode materials, especially nickel-rich ternary cathode materials, are constantly increasing. They are the most promising lithium-ion battery cathode materials in the next few years and have received great attention and extensive research.

[0003] The performance of ternary cathode materials is closely related to their sintering process. Any change in a factor during the sintering process may cause the sintered material to fail to meet the requirements. Therefore, the extraction and detection of various parameters during the sintering process of ternary cathode materials is also of utmost importance. Among them, the change in the surface roughness of ternary cathode materials is an important factor affecting the quality of the finished product.

[0004] There are four main methods commonly used to describe texture features: statistical method, model method, structural method and spectral method. However, the traditional image texture feature extraction method cannot fully characterize the changes in the surface roughness of the positive electrode material. Therefore, it is necessary to combine multiple texture feature extraction methods to more accurately characterize the roughness of the positive electrode material.

[0005] Therefore, the inability of existing image texture feature extraction methods to comprehensively characterize the changes in the surface roughness of positive electrode materials is a technical problem that needs to be solved urgently. Summary of the invention

[0006] The present invention provides a multi-feature fused SEM image roughness extraction method and device for ternary cathode material, aiming to solve the technical problem that the existing image texture feature extraction method cannot comprehensively characterize the change of the surface roughness of the cathode material.

[0007] One aspect of the present invention relates to a multi-feature fusion ternary cathode material roughness extraction method, comprising the following steps:

[0008] Acquire SEM images of the ternary cathode material at different sintering temperatures, and pre-process the acquired SEM images;

[0009] The fractal features of the pre-processed SEM images were extracted using the double blanket method and expressed using fractal dimensions;

[0010] Gray-level co-occurrence matrix is ​​used to analyze and extract the texture features of the preprocessed SEM images;

[0011] The extracted fractal dimension is fused with the texture features to obtain the roughness of the SEM image of the ternary cathode material with multi-feature fusion.

[0012] Further, the SEM images of the ternary cathode material at different sintering temperatures are obtained, and the steps of preprocessing the obtained SEM images include:

[0013] The labels on the SEM images of the ternary cathode materials at different sintering temperatures were cropped;

[0014] Extract the uneven brightness background in the SEM image after the label is cut out, and then use the original SEM image before cutting out to subtract the uneven background to obtain the SEM image after the uneven brightness is eliminated;

[0015] The SEM image after eliminating uneven brightness is filtered to remove noise on the surface of the SEM image.

[0016] Further, the steps of extracting fractal features of the pre-processed SEM image using the double blanket method and expressing them using fractal dimensions include:

[0017] Convert the input SEM image into a grayscale surface image, assuming f(x,y) is a surface in three-dimensional space, (i,j) represents the pixel position, and the height of a point on the surface is the grayscale value of the pixel corresponding to the position in the image;

[0018] Assume that a certain point in the surface is taken as the center, and the pixels that are greater than r from the center point are covered with a plane with a thickness of 2r. The surface area of ​​the surface can be obtained by dividing the volume between the upper and lower surfaces of the covering plane by 2r. This plane is called a carpet.

[0019] Assuming that the image f(x, y) is at scale r, the upper and lower surfaces are represented as Ur(x, y) and Br(x, y) respectively, then the upper and lower surfaces can be represented as:

[0020]

[0021] in, For the upper surface, is the lower surface, f(x,y) is the surface in three-dimensional space, d[(x,y),(m,n)] is the distance from point (x,y) to point (m,n), The upper surface of the space formed by the carpet covering, To cover the lower surface of the space formed by the carpet;

[0022] Calculate the volume of the space formed by the carpet and the area of ​​the surface respectively:

[0023]

[0024] in, The volume of the space created by the carpet covering, is the area of ​​the surface, The upper surface of the space formed by the carpet covering, To cover the lower surface of the space formed by the carpet;

[0025] The fractal dimension of the preprocessed SEM image is:

[0026]

[0027] in, is the area of ​​the surface, is the carpet dimension, is a constant, is the interval value.

[0028] Furthermore, the steps of analyzing and extracting the texture features of the preprocessed SEM image using a gray level co-occurrence matrix include:

[0029] The grayscale symbiosis matrix of the SEM image of the ternary cathode material is constructed using grayscale quantization levels of different sizes. The grayscale symbiosis matrix can be expressed as:

[0030]

[0031] Where P(i,j) is a point pair in the gray-level co-occurrence matrix, f(x,y) represents a two-dimensional image, S is a set of pixel pairs with a specific spatial relationship in the image area, its size is M*N, and the number of gray levels is G;

[0032] The texture feature parameter entropy value Ent of the SEM image of the ternary cathode material is extracted from the constructed grayscale co-occurrence matrix. The entropy value Ent represents the amount of information in the image. The entropy value Ent is calculated as follows:

[0033]

[0034] Where G represents the gray level of the image, and P(i,j) is the point pair in the gray level co-occurrence matrix;

[0035] Select appropriate step size d and generation direction θ to construct gray-level co-occurrence matrix P;

[0036] The entropy values ​​in the i-direction of the gray-level co-occurrence matrix are calculated to obtain the entropy values ​​in i different directions, and the average is calculated to obtain the entropy value of the SEM image calculated by the gray-level co-occurrence matrix;

[0037] By averaging the fractal dimension and entropy value, a new comprehensive indicator fractal-entropy value is obtained to characterize the surface roughness of ternary positive electrode materials.

[0038] Furthermore, the steps of averaging and fusing the fractal dimension and the entropy value to obtain a comprehensive new index fractal-entropy value for characterizing the surface roughness of the ternary cathode material include:

[0039] The calculated fractal dimension and entropy value are first normalized and then averaged together. The formula is as follows:

[0040]

[0041] Among them, D is the fractal dimension, Ent is the entropy value, R is the fractal entropy value, D_NORM is the normalized fractal dimension, and Ent_NORM is the normalized entropy value.

[0042] Another aspect of the present invention relates to a multi-feature fusion ternary cathode material roughness extraction device, comprising:

[0043] An acquisition module, used to acquire SEM images of the ternary cathode material at different sintering temperatures and to pre-process the acquired SEM images;

[0044] A first extraction module is used to extract fractal features of the pre-processed SEM image using a double blanket method and express them using fractal dimensions;

[0045] The second extraction module is used to analyze and extract the texture features of the pre-processed SEM image using a gray level co-occurrence matrix;

[0046] The fusion module is used to fuse the extracted fractal dimension with the texture feature to obtain the roughness of the SEM image of the ternary cathode material with multi-feature fusion.

[0047] Furthermore, the acquisition module includes:

[0048] A cutting unit, used for cutting labels on SEM images of ternary cathode materials at different sintering temperatures;

[0049] The first extraction unit is used to extract the uneven brightness background in the SEM image after the label is cut out, and then use the original SEM image before cutting out to subtract the uneven background to obtain the SEM image after the uneven brightness is eliminated;

[0050] The filtering processing unit is used to filter the SEM image after eliminating the uneven brightness, so as to remove the noise points on the surface of the SEM image.

[0051] Furthermore, the first extraction module includes:

[0052] Define a unit for converting an input SEM image into a grayscale surface image. Let f(x,y) be a surface in three-dimensional space, (i,j) represent pixel positions, and the height of a point on the surface be the grayscale value of the pixel corresponding to the position in the image.

[0053] The first assumption unit is used to assume that a certain point in the surface is taken as the center, and for the pixel points that are greater than r from the center point, a plane with a thickness of 2r is used to cover them. Then the surface area of ​​the surface can be obtained by dividing the volume between the upper and lower surfaces of the covering plane by 2r. This plane is called a carpet.

[0054] The second assumption unit is used to assume that the image f(x, y) is at scale r, and the upper and lower surfaces are represented as Ur(x, y) and Br(x, y) respectively. Then the upper and lower surfaces can be expressed as:

[0055]

[0056] in, For the upper surface, is the lower surface, f(x,y) is the surface in three-dimensional space, d[(x,y),(m,n)] is the distance from point (x,y) to point (m,n), The upper surface of the space formed by the carpet covering, To cover the lower surface of the space formed by the carpet;

[0057] The first calculation unit is used to calculate the volume of the space formed by the carpet and the area of ​​the curved surface, respectively:

[0058]

[0059] Among them, The volume of the space created by the carpet covering, is the area of ​​the surface, The upper surface of the space formed by the carpet covering, To cover the lower surface of the space formed by the carpet;

[0060] The fractal dimension of the preprocessed SEM image is:

[0061]

[0062] in, is the area of ​​the surface, is the carpet dimension, is a constant, is the interval value.

[0063] Furthermore, the second extraction module includes:

[0064] The first construction unit is used to construct a grayscale co-occurrence matrix of the SEM image of the ternary cathode material using grayscale quantization levels of different sizes. The grayscale co-occurrence matrix can be expressed as:

[0065]

[0066] Where P(i,j) is a point pair in the gray-level co-occurrence matrix, f(x,y) represents a two-dimensional image, S is a set of pixel pairs with a specific spatial relationship in the image area, its size is M*N, and the number of gray levels is G;

[0067] The second extraction unit is used to extract the texture feature parameter entropy value Ent of the SEM image of the ternary positive electrode material from the constructed grayscale co-occurrence matrix. The entropy value Ent represents the amount of information of the image. The entropy value Ent is calculated as follows:

[0068]

[0069] Where G represents the gray level of the image, and P(i,j) is the point pair in the gray level co-occurrence matrix;

[0070] The second construction unit is used to select a suitable step size d and a generation direction θ to construct a gray level co-occurrence matrix P;

[0071] The second calculation unit is used to calculate the entropy value of the gray-level co-occurrence matrix in i directions, obtain the entropy values ​​in i different directions, and take the average value thereof to obtain the entropy value of the SEM image calculated by the gray-level co-occurrence matrix;

[0072] The fusion unit is used to average the fractal dimension and the entropy value to obtain a comprehensive new indicator fractal-entropy value for characterizing the surface roughness of the ternary positive electrode material.

[0073] Furthermore, the fusion unit comprises:

[0074] The processing subunit is used to normalize the calculated fractal dimension and entropy value first, and then average and merge the two. The formula is as follows:

[0075]

[0076] Among them, D is the fractal dimension, Ent is the entropy value, R is the fractal entropy value, D_NORM is the normalized fractal dimension, and Ent_NORM is the normalized entropy value.

[0077] The beneficial effects achieved by the present invention are:

[0078] The present invention provides a method and device for extracting the roughness of a ternary cathode material SEM image with multi-feature fusion. The method and device are used to obtain SEM images of the ternary cathode material at different sintering temperatures, and the obtained SEM images are preprocessed; the fractal features of the preprocessed SEM images are extracted using a double blanket method, and are represented by fractal dimensions; the texture features of the preprocessed SEM images are analyzed and extracted using a grayscale co-occurrence matrix; the extracted fractal dimensions are fused with the texture features to obtain the roughness of the ternary cathode material SEM image with multi-feature fusion. The method and device for extracting the roughness of a ternary cathode material SEM image with multi-feature fusion provided by the present invention propose a method of combining fractal dimensions with information entropy to characterize the roughness of the ternary cathode material, thereby realizing the extraction of texture features of the SEM image. The feature quantities extracted by various methods complement each other, making the roughness extraction of the SEM image more comprehensive and effective. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 A schematic flow chart of an embodiment of a method for extracting roughness from SEM images of ternary cathode materials by multi-feature fusion provided by the present invention;

[0080] Figure 2 The SEM image of the ternary positive electrode material taken by a scanning electron microscope in the multi-feature fusion ternary positive electrode material SEM image roughness extraction method provided by the present invention;

[0081] Figure 3 The SEM image that has been preprocessed in the multi-feature fusion ternary cathode material SEM image roughness extraction method provided by the present invention;

[0082] Figure 4 The relationship between the fractal dimension and temperature calculated by the double blanket method in the multi-feature fusion ternary cathode material SEM image roughness extraction method provided by the present invention;

[0083] Figure 5 The relationship between entropy and temperature calculated by gray-level co-occurrence matrix in the multi-feature fusion ternary cathode material SEM image roughness extraction method provided by the present invention;

[0084] Figure 6 The relationship between the fractal entropy value and the temperature in the multi-feature fusion ternary cathode material SEM image roughness extraction method provided by the present invention;

[0085] Figure 7 for Figure 1 A schematic diagram of a detailed process of obtaining SEM images of ternary cathode materials at different sintering temperatures and preprocessing the obtained SEM images;

[0086] Figure 8A functional block diagram of an embodiment of a multi-feature fusion ternary cathode material SEM image roughness extraction device provided by the present invention;

[0087] Fig. 9 for Figure 8 The functional module diagram of the acquisition module 1 embodiment shown in FIG.

[0088] Fig.10 for Figure 8 A functional module diagram of an embodiment of a first extraction module shown in FIG.

[0089] Fig.11 for Figure 8 A functional module diagram of an embodiment of a second extraction module is shown in FIG.

[0090] Description of Figure Numbers:

[0091] 10. Acquisition module; 20. First extraction module; 30. Second extraction module; 40. Fusion module; 11. Cropping unit; 12. First extraction unit; 13. Filtering unit; 21. Definition unit; 22. First assumption unit; 23. Second assumption unit; 24. First calculation unit; 31. First construction unit; 32. Second extraction unit; 33. Second construction unit; 34. Second calculation unit; 35. Fusion unit. DETAILED DESCRIPTION

[0092] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0093] like Figures 1 to 6 As shown, the first embodiment of the present invention proposes a method for extracting roughness of a ternary positive electrode material by fusion of multiple features, comprising the following steps:

[0094] Step S100, obtaining SEM images of the ternary positive electrode material at different sintering temperatures, and preprocessing the obtained SEM images.

[0095] SEM images at different sintering temperatures were obtained by scanning electron microscopy, and the SEM images obtained during the sintering process of the ternary positive electrode material were preprocessed.

[0096] Step S200: extracting fractal features of the pre-processed SEM image using a double blanket method and expressing them using fractal dimensions.

[0097] The fractal dimension of the SEM image of the ternary positive electrode material after pretreatment is calculated. Fractals are morphological features that fill space in a non-integer form. Fractal dimension, as a quantitative characterization and basic parameter of fractals, is an important principle of fractal theory. The size of the fractal dimension reflects the complexity, irregularity, fineness and degree of filling of the contour space. In materials science, fractal dimensions are also often used to characterize many parameters of materials. In this embodiment, the blanket dimension is obtained by the double blanket method.

[0098] Step S300: Analyze and extract texture features of the preprocessed SEM image using a gray level co-occurrence matrix.

[0099] The grayscale co-occurrence matrix of the pre-processed SEM image of the ternary cathode material is extracted to obtain the texture features of the pre-processed SEM image.

[0100] Step S400: Fusing the extracted fractal dimension with the texture feature to obtain the roughness of the multi-feature fused SEM image of the ternary cathode material.

[0101] The fractal dimension and entropy value are averaged and fused to obtain a new comprehensive indicator fractal-entropy value that characterizes the surface roughness of the ternary cathode material, and the roughness of the SEM image of the ternary cathode material after multi-feature fusion is obtained.

[0102] The present embodiment provides a method for extracting the roughness of the SEM image of a ternary positive electrode material with multi-feature fusion. The SEM images of the ternary positive electrode material at different sintering temperatures are obtained, and the obtained SEM images are preprocessed; the fractal features of the preprocessed SEM images are extracted using the double blanket method, and represented by the fractal dimension; the texture features of the preprocessed SEM images are analyzed and extracted using the grayscale co-occurrence matrix; the extracted fractal dimension is fused with the texture features to obtain the roughness of the SEM image of the ternary positive electrode material with multi-feature fusion. The method for extracting the roughness of the SEM image of a ternary positive electrode material with multi-feature fusion provided in the present embodiment proposes to characterize the roughness of the ternary positive electrode material by combining the fractal dimension and information entropy, and realizes the extraction of the texture features of the SEM image. The feature quantities extracted by various methods complement each other, making the roughness extraction of the SEM image more comprehensive and effective.

[0103] Further, see Figure 7 , Figure 7 for Figure 1 In the detailed flow chart of step S100 shown in FIG. 1 , in this embodiment, step S100 includes:

[0104] Step S110, cutting out labels on SEM images of the ternary positive electrode material at different sintering temperatures.

[0105] The SEM image label bar is cropped to avoid unnecessary label information affecting image feature extraction.

[0106] Step S120 , extracting the uneven brightness background in the SEM image after the label is cropped, and then subtracting the uneven background from the original SEM image before cropping to obtain a SEM image after eliminating the uneven brightness.

[0107] Since the scanning electron microscope may cause uneven brightness of the image due to its imaging principle during the shooting process, resulting in inaccurate image feature extraction, the uneven brightness background of the SEM image is extracted, and then the uneven background is subtracted from the original film to obtain the SEM image after eliminating the uneven brightness.

[0108] Step S130 , filtering the SEM image after eliminating uneven brightness to remove noise points on the surface of the SEM image.

[0109] Since the image noise is interfered by different random signals during the image acquisition and transmission process, it may cause image blur, pockmarks and other phenomena, which will affect the received image information and is not conducive to the subsequent quantitative statistics of the image and the extraction of image features. Therefore, it is necessary to filter the SEM image to remove the noise on the image surface.

[0110] The method for extracting roughness from SEM images of ternary positive electrode materials with multi-feature fusion provided in this embodiment is to crop the labels on the SEM images of ternary positive electrode materials at different sintering temperatures; extract the uneven brightness background in the SEM image after cropping the labels, and then use the original SEM image before cropping to subtract the uneven background to obtain the SEM image after eliminating the uneven brightness; filter the SEM image after eliminating the uneven brightness to remove the noise on the surface of the SEM image. The method for extracting roughness from SEM images of ternary positive electrode materials with multi-feature fusion provided in this embodiment proposes to characterize the roughness of ternary positive electrode materials by combining fractal dimension and information entropy, and realizes the extraction of texture features of SEM images. The feature quantities extracted by multiple methods complement each other, making the roughness extraction of SEM images more comprehensive and effective.

[0111] Preferably, see Figures 1 to 7 The multi-feature fusion ternary cathode material SEM image roughness extraction method provided in this embodiment, step 200 includes:

[0112] Step S210, converting the input SEM image into a grayscale surface image, assuming that f(x, y) is a surface in three-dimensional space, (i, j) represents a pixel position, and the height of a point on the surface is the grayscale value of the pixel point corresponding to the position of the point in the image.

[0113] Step S220, assuming that a certain point in the surface is taken as the center, for the pixel points that are greater than r from the center point, a plane with a thickness of 2r is used to cover them, then the surface area of ​​the surface can be obtained by dividing the volume between the upper and lower surfaces of the covering plane by 2r, and this plane is called a carpet.

[0114] Step S230: Assuming that the image f(x, y) is at scale r, the upper and lower surfaces are represented as Ur(x, y) and Br(x, y) respectively, then the upper and lower surfaces can be represented as:

[0115]

[0116] In formula (1), For the upper surface, is the lower surface, f(x,y) is the surface in three-dimensional space, d[(x,y),(m,n)] is the distance from point (x,y) to point (m,n), The upper surface of the space formed by the carpet covering, The lower surface of the space formed is covered by a carpet.

[0117] Step S240: Calculate the volume of the space formed by the carpet and the area of ​​the curved surface:

[0118]

[0119] In formula (2), The volume of the space created by the carpet covering, is the area of ​​the surface, The upper surface of the space formed by the carpet covering, The lower surface of the space formed is covered by a carpet.

[0120] The fractal dimension of the preprocessed SEM image is:

[0121]

[0122] In formula (3), is the area of ​​the surface, is the carpet dimension, is a constant, is the interval value.

[0123] Through calculation, multiple groups of point pairs (Sr, r) are obtained, and the slope of the straight line is fitted. The slope obtained is the blanket dimension D. The calculation results are shown in Table 1 below:

[0124]

[0125] Table 1

[0126] The multi-feature fusion ternary cathode material SEM image roughness extraction method provided in this embodiment converts the input SEM image into a grayscale curved surface image; assuming that a certain point in the curved surface is taken as the center, for the pixel points with a distance greater than r from the center point, a plane with a thickness of 2r is used to cover, then the surface area of ​​the curved surface can be obtained by dividing the volume between the upper and lower surfaces of the covered plane by 2r, and the plane is called a carpet; assuming that the image f(x, y) is at a scale of r, the upper and lower surfaces are represented as Ur(x, y) and Br(x, y) respectively; calculate the volume of the space formed by the carpet coverage and the area of ​​the curved surface. The multi-feature fusion ternary cathode material SEM image roughness extraction method provided in this embodiment proposes to use a combination of fractal dimension and information entropy to characterize the roughness of the ternary cathode material, realize the extraction of SEM image texture features, and the feature quantities extracted by multiple methods complement each other, making the SEM image roughness extraction more comprehensive and effective.

[0127] See also Figures 1 to 7 The multi-feature fusion ternary cathode material SEM image roughness extraction method provided in this embodiment, step S300 includes:

[0128] Step S310: construct a grayscale co-occurrence matrix of the SEM image of the ternary cathode material using grayscale quantization levels of different sizes. The grayscale co-occurrence matrix can be expressed as:

[0129]

[0130] In formula (4), P(i, j) is a pair of points in the gray-level co-occurrence matrix, f(x, y) represents a two-dimensional image, and S is a set of pixel pairs with a specific spatial relationship in the image region, with a size of M*N and a gray level of G.

[0131] The gray level co-occurrence matrix is ​​used to describe the joint distribution of gray levels of two pixels with a spatial position relationship, which reflects the gray level spatial correlation characteristics of the image. The gray level co-occurrence matrix can reflect the information of the image gray level direction, adjacent intervals and change amplitude.

[0132] Step S320, extracting the texture feature parameter entropy value Ent of the SEM image of the ternary positive electrode material from the constructed gray level co-occurrence matrix, where the entropy value Ent represents the amount of information in the image, and the entropy value Ent is calculated as follows:

[0133]

[0134] In formula (5), G represents the grayscale level of the image, and P(i,j) is the point pair in the grayscale co-occurrence matrix.

[0135] For images with slowly changing textures, the values ​​on the diagonal of the gray-level co-occurrence matrix are larger; for images with rapidly changing textures, the values ​​on the diagonal of the gray-level co-occurrence matrix are smaller, and the values ​​on both sides of the diagonal are larger. Due to the large amount of data in the gray-level co-occurrence matrix, it is generally not directly used as a feature to distinguish textures, but some statistics constructed based on it are used as texture classification features.

[0136] When all elements in the gray-level co-occurrence matrix have the greatest randomness and the elements in the co-occurrence matrix are dispersed, the entropy value is large. The entropy value represents the degree of non-uniformity or complexity of the texture in the image. The larger the entropy value, the more non-uniform the image texture and the more complex the image.

[0137] Step S330: Select a suitable step size d and generation direction θ to construct a gray level co-occurrence matrix P.

[0138] For a grayscale image, when calculating the grayscale co-occurrence matrix, different parameters can be selected, such as step size d and generation direction θ. If different step sizes and generation directions are selected, the calculation results of the texture feature parameters obtained will be different. Select appropriate step size d and generation direction θ to construct the grayscale co-occurrence matrix P. In this embodiment, four directions of 0°, 45°, 90°, and 135° are selected, and the step size is 2.

[0139] Step S340, calculating the entropy values ​​in i directions of the gray-level co-occurrence matrix to obtain the entropy values ​​in i different directions, taking the average thereof, and obtaining the entropy value of the SEM image calculated by the gray-level co-occurrence matrix.

[0140] By calculating the entropy values ​​in the four directions of the gray-level co-occurrence matrix, the entropy values ​​in four different directions are obtained, and the average is taken to obtain the entropy value of the SEM image calculated by the gray-level co-occurrence matrix. The results are shown in Table 2 below:

[0141]

[0142] Table 2

[0143] Step S350, average and fuse the fractal dimension and the entropy value to obtain a comprehensive new indicator fractal-entropy value that characterizes the surface roughness of the ternary positive electrode material.

[0144] The calculated fractal dimension and entropy value are first normalized and then averaged together. The formula is as follows:

[0145]

[0146] In formula (6), D is the fractal dimension, Ent is the entropy value, R is the fractal entropy value, D_NORM is the normalized fractal dimension, and Ent_NORM is the normalized entropy value. The calculation results are shown in Table 3 below:

[0147]

[0148] Table 3

[0149] After combining the model method and statistical method in the texture feature extraction method, the calculated fractal entropy value results more comprehensively describe the changes in the surface roughness of the ternary positive electrode material, improve the accuracy of roughness evaluation, and are more in line with the actual roughness change law.

[0150] The roughness extraction method of the ternary positive electrode material SEM image with multi-feature fusion provided in this embodiment is to construct the grayscale co-occurrence matrix of the ternary positive electrode material SEM image by using grayscale quantization levels of different sizes; extract the texture feature parameter entropy value Ent of the ternary positive electrode material SEM image from the constructed grayscale co-occurrence matrix; select the appropriate step size d and generation direction θ to construct the grayscale co-occurrence matrix P; calculate the entropy values ​​in the i directions of the grayscale co-occurrence matrix to obtain the entropy values ​​in i different directions, and take the average to obtain the entropy value of the SEM image calculated by the grayscale co-occurrence matrix; average the fractal dimension and the entropy value to obtain a comprehensive new indicator fractal-entropy value for characterizing the surface roughness of the ternary positive electrode material. The roughness extraction method of the ternary positive electrode material SEM image with multi-feature fusion provided in this embodiment proposes to characterize the roughness of the ternary positive electrode material by combining the fractal dimension and information entropy, realizes the extraction of the texture features of the SEM image, and the feature quantities extracted by multiple methods complement each other, making the roughness extraction of the SEM image more comprehensive and effective.

[0151] like Figure 8 As shown, Figure 8 A functional block diagram of an embodiment of a roughness extraction device for ternary positive electrode material SEM image with multi-feature fusion provided by the present invention. In this embodiment, the roughness extraction device for ternary positive electrode material with multi-feature fusion includes an acquisition module 10, a first extraction module 20, a second extraction module 30 and a fusion module 40, wherein the acquisition module 10 is used to acquire SEM images of the ternary positive electrode material at different sintering temperatures and pre-process the acquired SEM images; the first extraction module 20 is used to extract fractal features of the pre-processed SEM image using a double blanket method and express them with fractal dimensions; the second extraction module 30 is used to analyze and extract texture features of the pre-processed SEM image using a gray level co-occurrence matrix; the fusion module 40 is used to fuse the extracted fractal dimension with the texture features to obtain the roughness of the SEM image of the ternary positive electrode material with multi-feature fusion.

[0152] The acquisition module 10 acquires SEM images at different sintering temperatures through a scanning electron microscope, and pre-processes the acquired SEM images during the sintering process of the ternary positive electrode material.

[0153] The first extraction module 20 calculates the fractal dimension of the SEM image of the ternary positive electrode material after pretreatment. Fractals are morphological features that fill space in a non-integer form. Fractal dimension, as a quantitative characterization and basic parameter of fractals, is an important principle of fractal theory. The size of the fractal dimension reflects the complexity, irregularity, refinement and degree of filling of the contour space. In materials science, fractal dimensions are also often used to characterize many parameters of materials. In this embodiment, the double blanket method is used to obtain the blanket dimension.

[0154] The second extraction module 30 extracts the grayscale co-occurrence matrix of the pre-processed SEM image of the ternary cathode material to obtain the texture features of the pre-processed SEM image.

[0155] The fusion module 40 averages and fuses the fractal dimension and the entropy value to obtain a comprehensive new indicator fractal-entropy value for characterizing the surface roughness of the ternary positive electrode material, and obtains the roughness of the SEM image of the ternary positive electrode material after multi-feature fusion.

[0156] The multi-feature fused ternary cathode material SEM image roughness extraction device provided in this embodiment obtains SEM images of the ternary cathode material at different sintering temperatures, pre-processes the obtained SEM images; uses the double blanket method to extract the fractal features of the pre-processed SEM images, and represents them with fractal dimensions; uses the grayscale co-occurrence matrix to analyze and extract the texture features of the pre-processed SEM images; and fuses the extracted fractal dimensions with the texture features to obtain the roughness of the multi-feature fused ternary cathode material SEM image. The multi-feature fused ternary cathode material SEM image roughness extraction device provided in this embodiment proposes to use a combination of fractal dimensions and information entropy to characterize the roughness of the ternary cathode material, realizes the extraction of SEM image texture features, and the feature quantities extracted by multiple methods complement each other, making the SEM image roughness extraction more comprehensive and effective.

[0157] Further, see Fig. 9 , Fig. 9 for Figure 8, in which the acquisition module 10 includes a cropping unit 11, a first extraction unit 12 and a filtering unit 13, wherein the cropping unit 11 is used to crop the label on the SEM image of the ternary positive electrode material at different sintering temperatures; the first extraction unit 12 is used to extract the uneven brightness background in the SEM image after the label is cropped, and then use the original SEM image before cropping to subtract the uneven background to obtain the SEM image after the uneven brightness is eliminated; the filtering unit 13 is used to filter the SEM image after the uneven brightness is eliminated to remove the noise on the surface of the SEM image.

[0158] The cropping unit 11 crops the SEM image label column to prevent unnecessary label information from affecting image feature extraction.

[0159] Since the scanning electron microscope may cause uneven brightness of the image due to its imaging principle during the shooting process, resulting in inaccurate image feature extraction, the first extraction unit 12 extracts the uneven brightness background of the SEM image, and then uses the original film to subtract the uneven background to obtain the SEM image after eliminating the uneven brightness.

[0160] Since the image noise is interfered by different random signals during the image acquisition and transmission process, it may cause image blur, pockmarks and other phenomena, which will affect the received image information and be unfavorable for subsequent quantitative statistics of the image and extraction of image features. Therefore, the filtering processing unit 13 is required to filter the SEM image to remove the noise on the image surface.

[0161] The multi-feature fused ternary cathode material SEM image roughness extraction device provided in this embodiment cuts out the labels on the SEM images of the ternary cathode material at different sintering temperatures; extracts the uneven brightness background in the SEM image after cutting out the labels, and then uses the original SEM image before cutting out to subtract the uneven background to obtain the SEM image after eliminating the uneven brightness; and filters the SEM image after eliminating the uneven brightness to remove the noise on the surface of the SEM image. The multi-feature fused ternary cathode material SEM image roughness extraction device provided in this embodiment proposes to use a combination of fractal dimension and information entropy to characterize the roughness of the ternary cathode material, and realizes the extraction of SEM image texture features. The feature quantities extracted by multiple methods complement each other, making the SEM image roughness extraction more comprehensive and effective.

[0162] Further, see Fig.10 , Fig.10 For Figure 8The functional module diagram of an embodiment of the first extraction module 20 is shown in the figure. In this embodiment, the first extraction module 20 includes a definition unit 21, a first assumption unit 22, a second assumption unit 23 and a first calculation unit 24, wherein the definition unit 21 is used to convert the input SEM image into a grayscale surface image, assuming that f(x, y) is a surface in three-dimensional space, (i, j) represents a pixel position, and the height of a point on the surface is the grayscale value of the pixel point corresponding to the position of the point in the image.

[0163] The first assumption unit 22 is used to assume that a certain point in the surface is taken as the center, and the pixel points with a distance greater than r from the center point are covered with a plane with a thickness of 2r. The surface area of ​​the surface can be obtained by dividing the volume between the upper and lower surfaces of the covering plane by 2r, and the plane is called a carpet.

[0164] The second assumption unit 23 is used to assume that the image f(x, y) is at a scale r, and the upper and lower surfaces are represented as Ur(x, y) and Br(x, y) respectively, then the upper and lower surfaces can be represented as:

[0165]

[0166] In formula (7), For the upper surface, is the lower surface, f(x,y) is the surface in three-dimensional space, d[(x,y),(m,n)] is the distance from point (x,y) to point (m,n), The upper surface of the space formed by the carpet covering, The lower surface of the space formed is covered by a carpet.

[0167] The first calculation unit 24 is used to calculate the volume of the space formed by the carpet covering and the area of ​​the curved surface respectively:

[0168]

[0169] In formula (8), The volume of the space created by the carpet covering, is the area of ​​the surface, The upper surface of the space formed by the carpet covering, The lower surface of the space formed is covered by a carpet.

[0170] The fractal dimension of the preprocessed SEM image is:

[0171]

[0172] In formula (9), is the area of ​​the surface, is the carpet dimension, is a constant, is the interval value.

[0173] Through calculation, multiple groups of point pairs (Sr, r) are obtained, and the slope of the straight line is fitted. The slope obtained is the blanket dimension D. The calculation results are shown in Table 4 below:

[0174]

[0175] Table 4

[0176] The multi-feature fusion ternary cathode material SEM image roughness extraction device provided in this embodiment converts the input SEM image into a grayscale curved surface image; assuming that a certain point in the curved surface is taken as the center, for the pixel points with a distance greater than r from the center point, a plane with a thickness of 2r is used to cover, then the surface area of ​​the curved surface can be obtained by dividing the volume between the upper and lower surfaces of the covered plane by 2r, and the plane is called a carpet; assuming that the image f(x, y) is at a scale of r, the upper and lower surfaces are represented as Ur(x, y) and Br(x, y) respectively; calculate the volume of the space formed by the carpet covering and the area of ​​the curved surface. The multi-feature fusion ternary cathode material SEM image roughness extraction device provided in this embodiment proposes to use a combination of fractal dimension and information entropy to characterize the roughness of the ternary cathode material, realize the extraction of SEM image texture features, and the feature quantities extracted by multiple methods complement each other, making the SEM image roughness extraction more comprehensive and effective.

[0177] Further, see Fig.11 , Figure 1 for Figure 8 , in this embodiment, the second extraction module 30 includes a first construction unit 31, a second extraction unit 32, a second construction unit 33, a second calculation unit 34 and a fusion unit 35, wherein the first construction unit 31 is used to construct a grayscale co-occurrence matrix of the SEM image of the ternary positive electrode material using grayscale quantization levels of different sizes, and the grayscale co-occurrence matrix can be expressed as:

[0178]

[0179] In formula (10), P(i,j) is a pair of points in the gray-level co-occurrence matrix, f(x,y) represents a two-dimensional image, and S is a set of pixel pairs with a specific spatial relationship in the image region, with a size of M*N and a gray level of G.

[0180] The gray level co-occurrence matrix is ​​used to describe the joint distribution of gray levels of two pixels with a spatial position relationship, which reflects the gray level spatial correlation characteristics of the image. The gray level co-occurrence matrix can reflect the information of the image gray level direction, adjacent intervals and change amplitude.

[0181] The second extraction unit 32 is used to extract the texture feature parameter entropy value Ent of the SEM image of the ternary positive electrode material from the constructed gray level co-occurrence matrix. The entropy value Ent represents the information amount of the image. The entropy value Ent is calculated as follows:

[0182]

[0183] In formula (11), G represents the grayscale level of the image, and P(i,j) is the point pair in the grayscale co-occurrence matrix.

[0184] For images with slowly changing textures, the values ​​on the diagonal of the gray-level co-occurrence matrix are larger; for images with rapidly changing textures, the values ​​on the diagonal of the gray-level co-occurrence matrix are smaller, and the values ​​on both sides of the diagonal are larger. Due to the large amount of data in the gray-level co-occurrence matrix, it is generally not directly used as a feature to distinguish textures, but some statistics constructed based on it are used as texture classification features.

[0185] When all elements in the gray-level co-occurrence matrix have the greatest randomness and the elements in the co-occurrence matrix are dispersed, the entropy value is large. The entropy value represents the degree of non-uniformity or complexity of the texture in the image. The larger the entropy value, the more non-uniform the image texture and the more complex the image.

[0186] The second construction unit 33 is used to select a suitable step size d and a generation direction θ to construct a gray level co-occurrence matrix P.

[0187] For a grayscale image, when calculating the grayscale co-occurrence matrix, different parameters can be selected, such as step size d and generation direction θ. If different step sizes and generation directions are selected, the calculation results of the texture feature parameters obtained will be different. Select appropriate step size d and generation direction θ to construct the grayscale co-occurrence matrix P. In this embodiment, four directions of 0°, 45°, 90°, and 135° are selected, and the step size is 2.

[0188] The second calculation unit 34 is used to calculate the entropy values ​​of the gray-level co-occurrence matrix in i directions, obtain the entropy values ​​in i different directions, and take the average thereof to obtain the entropy value of the SEM image calculated by the gray-level co-occurrence matrix.

[0189] By calculating the entropy values ​​in the four directions of the gray-level co-occurrence matrix, the entropy values ​​in four different directions are obtained, and the average is taken to obtain the entropy value of the SEM image calculated by the gray-level co-occurrence matrix. The results are shown in Table 5 below:

[0190]

[0191] Table 5

[0192] The fusion unit 35 is used to averagely fuse the fractal dimension and the entropy value to obtain a comprehensive new indicator fractal-entropy value that characterizes the surface roughness of the ternary positive electrode material. The fusion unit 35 includes a processing subunit, which is used to first normalize the calculated fractal dimension and entropy value, and then averagely fuse the two. The formula is as follows:

[0193]

[0194] In formula (12), D is the fractal dimension, Ent is the entropy value, R is the fractal entropy value, D_NORM is the normalized fractal dimension, and Ent_NORM is the normalized entropy value. The calculation results are shown in Table 6 below:

[0195]

[0196] Table 6

[0197] After combining the model method and statistical method in the texture feature extraction method, the calculated fractal entropy value results more comprehensively describe the changes in the surface roughness of the ternary positive electrode material, improve the accuracy of roughness evaluation, and are more in line with the actual roughness change law.

[0198] The roughness extraction device of the ternary positive electrode material SEM image with multi-feature fusion provided in this embodiment constructs the grayscale co-occurrence matrix of the ternary positive electrode material SEM image by using grayscale quantization levels of different sizes; extracts the texture feature parameter entropy value Ent of the ternary positive electrode material SEM image from the constructed grayscale co-occurrence matrix; selects a suitable step size d and generation direction θ to construct the grayscale co-occurrence matrix P; calculates the entropy values ​​in the i directions of the grayscale co-occurrence matrix to obtain the entropy values ​​in i different directions, takes the average value, and obtains the entropy value of the SEM image calculated by the grayscale co-occurrence matrix; averages and fuses the fractal dimension and the entropy value to obtain a comprehensive new indicator fractal-entropy value for characterizing the surface roughness of the ternary positive electrode material. The roughness extraction device of the ternary positive electrode material SEM image with multi-feature fusion provided in this embodiment proposes to characterize the roughness of the ternary positive electrode material by combining the fractal dimension and information entropy, realizes the extraction of the texture features of the SEM image, and the feature quantities extracted by multiple methods complement each other, making the roughness extraction of the SEM image more comprehensive and effective.

[0199] Although preferred embodiments of the present invention have been described, additional changes and modifications may be made to these embodiments by those skilled in the art once the basic inventive concepts are known. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention. Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A multi-feature fusion ternary cathode material roughness extraction method, characterized in that: The following steps are involved: Acquire SEM images of the ternary cathode material at different sintering temperatures, and pre-process the acquired SEM images; The fractal features of the pre-processed SEM images were extracted using the double blanket method and expressed using fractal dimensions; Gray-level co-occurrence matrix is ​​used to analyze and extract the texture features of the preprocessed SEM images; The extracted fractal dimension is fused with the texture feature to obtain the roughness of the SEM image of the ternary cathode material with multiple features fused; The step of extracting the fractal features of the pre-treated SEM image using the double blanket method and expressing them using fractal dimensions comprises: Convert the input SEM image into a grayscale surface image, assuming f(x,y) is a surface in three-dimensional space, (i,j) represents the pixel position, and the height of a point on the surface is the grayscale value of the pixel corresponding to the position in the image; Assume that a certain point in the surface is taken as the center, and the pixels that are greater than r from the center point are covered with a plane with a thickness of 2r. The surface area of ​​the surface can be obtained by dividing the volume between the upper and lower surfaces of the covering plane by 2r. This plane is called a carpet. Assuming that the image f(x, y) is at scale r, the upper and lower surfaces are represented as Ur(x, y) and Br(x, y) respectively, then the upper and lower surfaces can be represented as: in, For the upper surface, is the lower surface, f(x,y) is the surface in three-dimensional space, d[(x,y),(m,n)] is the distance from point (x,y) to point (m,n), The upper surface of the space formed by the carpet covering, To cover the lower surface of the space formed by the carpet; Calculate the volume of the space formed by the carpet and the area of ​​the surface respectively: in, The volume of the space created by the carpet covering, is the area of ​​the surface, The upper surface of the space formed by the carpet covering, To cover the lower surface of the space formed by the carpet; The fractal dimension of the preprocessed SEM image is: in, is the area of ​​the surface, is the carpet dimension, is a constant, is the interval value; The step of analyzing and extracting the texture features of the pre-processed SEM image using the gray level co-occurrence matrix comprises: The grayscale symbiosis matrix of the SEM image of the ternary cathode material is constructed using grayscale quantization levels of different sizes. The grayscale symbiosis matrix can be expressed as: Where P(i,j) is a point pair in the gray-level co-occurrence matrix, f(x,y) represents a two-dimensional image, S is a set of pixel pairs with a specific spatial relationship in the image area, its size is M*N, and the number of gray levels is G; The texture feature parameter entropy value Ent of the SEM image of the ternary cathode material is extracted from the constructed grayscale co-occurrence matrix. The entropy value Ent represents the amount of information in the image. The entropy value Ent is calculated as follows: Where G represents the gray level of the image, and P(i,j) is the point pair in the gray level co-occurrence matrix; Select appropriate step size d and generation direction θ to construct gray-level co-occurrence matrix P; The entropy values ​​in the i-direction of the gray-level co-occurrence matrix are calculated to obtain the entropy values ​​in i different directions, and the average is calculated to obtain the entropy value of the SEM image calculated by the gray-level co-occurrence matrix; By averaging the fractal dimension and entropy value, a new comprehensive indicator fractal-entropy value is obtained to characterize the surface roughness of ternary positive electrode materials.

2. The multi-feature fusion ternary cathode material roughness extraction method according to claim 1, characterized in that: The step of obtaining SEM images of the ternary cathode material at different sintering temperatures and preprocessing the obtained SEM images comprises: The labels on the SEM images of the ternary cathode materials at different sintering temperatures were cropped; Extract the uneven brightness background in the SEM image after the label is cut out, and then use the original SEM image before cutting out to subtract the uneven background to obtain the SEM image after the uneven brightness is eliminated; The SEM image after eliminating uneven brightness is filtered to remove noise on the surface of the SEM image.

3. The multi-feature fusion roughness extraction method for ternary cathode materials according to claim 1, characterized in that: The step of averaging and fusing the fractal dimension and the entropy value to obtain a comprehensive new index fractal-entropy value for characterizing the surface roughness of the ternary positive electrode material comprises: The calculated fractal dimension and entropy value are first normalized and then averaged together. The formula is as follows: Among them, D is the fractal dimension, Ent is the entropy value, R is the fractal entropy value, D_NORM is the normalized fractal dimension, and Ent_NORM is the normalized entropy value.

4. A multi-feature fusion ternary cathode material roughness extraction device, characterized in that: include: An acquisition module (10) is used to acquire SEM images of the ternary positive electrode material at different sintering temperatures and to pre-process the acquired SEM images; A first extraction module (20) is used to extract fractal features of the pre-processed SEM image using a double blanket method and express them using fractal dimensions; A second extraction module (30) is used to analyze and extract texture features of the pre-processed SEM image using a gray level co-occurrence matrix; A fusion module (40) is used to fuse the extracted fractal dimension with the texture feature to obtain the roughness of the multi-feature fused ternary positive electrode material SEM image; The first extraction module (20) comprises: A unit (21) is defined, which is used to convert an input SEM image into a grayscale surface image, where f(x, y) is a surface in three-dimensional space, (i, j) represents a pixel position, and the height of a point on the surface is the grayscale value of the pixel point corresponding to the position of the point in the image; The first assumption unit (22) is used to assume that a certain point in the surface is taken as the center, and for the pixel points that are greater than r from the center point, a plane with a thickness of 2r is used to cover them. Then, the surface area of ​​the surface can be obtained by dividing the volume between the upper and lower surfaces of the covering plane by 2r. This plane is called a carpet. The second assumption unit (23) is used to assume that the image f(x, y) is at a scale r, and the upper and lower surfaces are represented as Ur(x, y) and Br(x, y) respectively. Then the upper and lower surfaces can be represented as: in, For the upper surface, is the lower surface, f(x,y) is the surface in three-dimensional space, d[(x,y),(m,n)] is the distance from point (x,y) to point (m,n), The upper surface of the space formed by the carpet covering, To cover the lower surface of the space formed by the carpet; The first calculation unit (24) is used to calculate the volume of the space formed by the carpet covering and the area of ​​the curved surface respectively: in, The volume of the space created by the carpet covering, is the area of ​​the surface, The upper surface of the space formed by the carpet covering, To cover the lower surface of the space formed by the carpet; The fractal dimension of the preprocessed SEM image is: in, is the area of ​​the surface, is the carpet dimension, is a constant, is the interval value; The second extraction module (30) comprises: The first construction unit (31) is used to construct a grayscale co-occurrence matrix of the SEM image of the ternary positive electrode material using grayscale quantization levels of different sizes. The grayscale co-occurrence matrix can be expressed as: Where P(i,j) is a point pair in the gray-level co-occurrence matrix, f(x,y) represents a two-dimensional image, S is a set of pixel pairs with a specific spatial relationship in the image area, its size is M*N, and the number of gray levels is G; The second extraction unit (32) is used to extract the texture feature parameter entropy value Ent of the SEM image of the ternary positive electrode material from the constructed grayscale co-occurrence matrix. The entropy value Ent represents the amount of information in the image. The entropy value Ent is calculated as follows: Where G represents the gray level of the image, and P(i,j) is the point pair in the gray level co-occurrence matrix; The second construction unit (33) is used to select a suitable step size d and a generation direction θ to construct a gray level co-occurrence matrix P; A second calculation unit (34) is used to calculate the entropy value of the gray level co-occurrence matrix in i directions, obtain the entropy values ​​in i different directions, and calculate the average value to obtain the entropy value of the SEM image calculated by the gray level co-occurrence matrix; The fusion unit (35) is used to average and fuse the fractal dimension and the entropy value to obtain a comprehensive new indicator fractal-entropy value for characterizing the surface roughness of the ternary positive electrode material.

5. The multi-feature fusion ternary cathode material roughness extraction device according to claim 4, characterized in that: The acquisition module (10) comprises: A cutting unit (11) is used to cut labels on SEM images of ternary positive electrode materials at different sintering temperatures; A first extraction unit (12) is used to extract the uneven brightness background in the SEM image after the label is cut out, and then use the original SEM image before cutting out to subtract the uneven background to obtain the SEM image after the uneven brightness is eliminated; The filtering processing unit (13) is used to perform filtering processing on the SEM image after eliminating the uneven brightness, so as to remove noise points on the surface of the SEM image.

6. The multi-feature fusion ternary cathode material roughness extraction device according to claim 4, characterized in that: The fusion unit (35) comprises: The processing subunit is used to normalize the calculated fractal dimension and entropy value first, and then average and merge the two. The formula is as follows: Among them, D is the fractal dimension, Ent is the entropy value, R is the fractal entropy value, D_NORM is the normalized fractal dimension, and Ent_NORM is the normalized entropy value.

Images