A method for extracting conductive carbon black lattice structure

Abstract

The present invention discloses a method for extracting the lattice structure of conductive carbon black, and relates to the field of image processing technology. The present invention extracts the lattice structure of conductive carbon black and collects relevant data, and adopts a high-resolution transmission electron microscope as the original sample for analysis. Compared with the traditional method, the image processing method of the present invention has higher flexibility and automation level, and the operation process is simple. First, by writing a Python script and using a sliding window method, the original image can be divided into multiple small image blocks, which increases the flexibility of processing and adapts to conductive carbon black lattice structures of different sizes and shapes. Secondly, through Fourier-inverse Fourier transform and grayscale processing, the contrast between the lattice structure and the image background can be enhanced, making the lattice structure more clearly visible, which is conducive to subsequent extraction and analysis. In addition, by using threshold adjustment and coloring processing, the lattice structure stripes can be secondary enhanced.

Description

Technical Field

[0001] The invention relates to the technical field of image processing, and in particular to a method for extracting a conductive carbon black lattice structure. Background Art

[0002] Conductive carbon black is a special type of carbon black with special conductive properties. It is widely used in electronic devices, mining pipes, cable shielding materials, aerospace industry, petrochemical industry, power storage and other fields. Compared with other highly conductive fillers such as graphene and carbon nanotubes, conductive carbon black is particularly favored in the energy field due to its excellent cost performance. It is used as a conductive agent in power lithium-ion batteries, lead-acid batteries and supercapacitors.

[0003] At present, the evaluation of conductive carbon black mainly adopts a variety of characterization methods, including high-resolution transmission electron microscopy (HRTEM), conventional Raman spectroscopy, X-ray diffraction, infrared spectroscopy and DBP oil absorption value and other basic property characterization methods. Among them, HRTEM technology is widely used to study the microstructural characteristics of carbon materials, especially the observation and study of fine structures such as aromatic hydrocarbon layers. As an important tool, HRTEM technology provides important information about the morphology of conductive carbon black, but the information acquisition in other aspects is relatively limited and is usually limited to qualitative analysis.

[0004] An important topic in the study of carbon-based materials is to extract the lattice fringe parameters of coal aromatic hydrocarbons, and many processing methods have been developed for this purpose. Generally speaking, these methods can be divided into two types. The first method is a simple image processing based on high-resolution transmission electron microscope images. Generally, the fringe parameters are manually drawn and extracted directly using mapping software such as ArcGIS or Image J. The disadvantage of this method is that it is time-consuming and requires manual operation, and the error of manual extraction may be large.

[0005] The second method is mainly based on programs developed in computer languages. Although the results can be obtained quickly, when the program with determined parameters processes different images, due to the brightness difference of the images, a large amount of information may be lost in the filtering and threshold noise reduction process, and some noise may remain, resulting in inaccurate extracted parameter information. Therefore, manual intervention is still required to avoid unnecessary problems with subsequent results.

[0006] In summary, there are currently two main methods for extracting the lattice structure parameters of coal aromatic hydrocarbons. The first is a manual processing method based on high-resolution transmission electron microscope images. Although it is time-consuming and has a high degree of error, it can flexibly process different images. The second is an automated processing method developed based on computer language. Although it is fast, the processing results of different images may not be accurate enough and still require manual intervention.

[0007] Therefore, corresponding solutions need to be proposed. Summary of the invention

[0008] The purpose of the present invention is to provide a method for extracting the lattice structure of conductive carbon black.

[0009] To achieve the above object, the present invention provides the following technical solution: a method for extracting the lattice structure of conductive carbon black, comprising at least the following steps:

[0010] Step 1: Prepare a dispersion liquid with conductive carbon black and ethanol in a certain ratio, perform a high-resolution transmission electron microscope test after sample preparation, obtain an electron microscope original image and perform a precursor treatment to obtain a pre-processed image;

[0011] Step 2: Extract features from the preprocessed image obtained in step 1 through a trained convolutional neural network to enhance the contrast between the lattice structure stripes and the background;

[0012] Step 3: The image extracted from the features in step 2 is combined using different weights to enhance the contrast between the main lattice part and the image background to obtain a grayscale processed image;

[0013] Step 4: Using Adobe Photoshop to adjust the threshold of the grayscale image obtained in step 3, and color the stripes to enhance the stripes for a second time;

[0014] Step 5: Binarize the image after enhancing the stripes in step 4 and invert the pixel values;

[0015] Step 6: In step 5, the eight-neighborhood labeling algorithm is used to extract the main frame of the image after the pixel value is inverted, and the corrosion algorithm is further used to remove the main edge pixels, and finally the white main line is displayed on the black background;

[0016] Step 7: Use the linear fitting algorithm to calculate the length and angle of the main lines of the lattice structure on the image obtained in step 6, and collect and calculate the statistical data;

[0017] Step 8: Repeat the steps to piece the cropped images together to complete the extraction process of the original tif image.

[0018] Preferably, the precursor treatment in step 1 includes at least the following steps:

[0019] First, compile a Python script to crop the original image;

[0020] Using the sliding window method, a window is slid on the image with a specified step size, the image is cropped into 512*512 pixel image blocks, and the image blocks within the window are extracted.

[0021] Preferably, the step 1 at least includes the following steps:

[0022] Prepare a Python script to process the original image. The original image is generally a tif format file with a pixel value of 4096*4096. In basic processing, the original image needs to be cropped to generate a 512*512 pixel image and converted to jpg format.

[0023] Preferably, the step 2 at least includes the following steps:

[0024] Convolutional neural networks are trained using HRTEM images with clear lattice structures. Convolutional neural networks (CNNs) are a deep learning model or a multi-layer perceptron similar to an artificial neural network, which is often used to analyze visual images. They include the following parts: input layer: input image and other information; convolution layer: used to extract the underlying features of the image; pooling layer: to prevent overfitting and reduce the data dimension; fully connected layer: summarizes the underlying features and information of the image obtained by the convolution layer and the pooling layer; output layer: obtains the result with the highest probability based on the information of the fully connected layer.

[0025] Determine the convolution kernel parameters through supervised learning methods and further process the image;

[0026] Since the input image has three channels, convolution calculation will be performed on a per-channel basis, which will restore the original shape and texture of the image and enhance the contrast between the lattice structure and the background.

[0027] Preferably, the step three at least includes the following steps:

[0028] For the three obtained three-channel images, appropriate weights are selected to combine them to obtain a result image with enhanced contrast between the main lattice part and the background;

[0029] The color distribution histograms of the three images are exported before combination, and the final weights are determined according to the overall light and dark ratio of the obtained distribution histograms.

[0030] Preferably, the step 4 at least includes the following steps:

[0031] Use Adobe Photoshop software for further processing;

[0032] Using Photoshop's threshold adjustment tool, according to the image features and needs, by adjusting the threshold value, the stripes of the lattice structure are separated to form a more obvious contrast with the image background, and the color processing function of Photoshop is further used to enhance the stripes for a second time;

[0033] By choosing the right color and adjusting the color parameters, you can give the striped part a more vivid color.

[0034] Preferably, the step five at least includes the following steps:

[0035] Use the Threshold Adjustment Tool to convert the image into a binary image by selecting an appropriate threshold. This step will divide the pixel values ​​into two categories: black and white, based on the lightness and darkness of the stripes in the image.

[0036] By adjusting the threshold value, the binary effect of controlling the stripes can be achieved;

[0037] After completing the binarization, use the pixel value inversion function provided by the software to swap the black and white pixel values;

[0038] The darker stripes will become white, and the lighter background will become black;

[0039] By inverting the pixel values, the morphology and structure of the stripes can be better highlighted, making the binary image more clearly visible.

[0040] Preferably, the step six at least includes the following steps:

[0041] The improved eight-neighborhood algorithm is used to extract the image backbone;

[0042] The eight-neighborhood algorithm is a classic marking algorithm, which aims to identify and mark aggregated pixels in an image to form a trunk area.

[0043] Then use the erosion algorithm, which iteratively reduces the edge pixels in the main area of ​​the image in each step until the main line is obtained;

[0044] After the trunk extraction is completed, the white trunk lines can be displayed on the black background;

[0045] This means setting the pixel values ​​of the main lines to white (255), while the pixel values ​​of the background remain black (0);

[0046] Against the black background, the main lines will be more clearly visible, highlighting the structure in the image;

[0047] This backbone line image is used to help us better analyze the morphology and connectivity of lattice structures or other specific targets.

[0048] Preferably, the step seven at least includes the following steps:

[0049] Compile a script to extract information about the main lines. The key point is that for each line segmented from a connected domain, a new picture should be generated to count them separately.

[0050] Mark the pixels of the main lines on the newly generated image, use the linear fitting method for these marked pixels, and use a straight line to fit the direction of the main lines, so that as many marked pixels as possible fall on the fitted straight line or on both sides of the straight line;

[0051] For the straight line obtained, its length and slope are counted. In the subsequent processing, the length in pixels is converted into data in nanometers according to the scale, and the slope is converted into an angle.

[0052] By performing statistical analysis on these data, we can obtain information such as the length distribution and angle distribution of the main lines in the lattice structure, thereby better understanding the morphology and structural characteristics of the lattice.

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] The present invention extracts the lattice structure of conductive carbon black and collects relevant data, and uses a high-resolution transmission electron microscope as the original sample for analysis. Compared with traditional methods, the image processing method of the present invention has higher flexibility and automation level, and the operation process is simple;

[0055] First, by writing a Python script and using the sliding window method, the original image can be segmented into multiple small image blocks, which increases the flexibility of processing and adapts to conductive carbon black lattice structures of different sizes and shapes;

[0056] Secondly, the contrast between the lattice structure and the image background can be enhanced through Fourier-inverse Fourier transform and grayscale processing, making the lattice structure more visible, which is conducive to subsequent extraction and analysis. In addition, the lattice structure stripes can be enhanced for a second time by using threshold adjustment and coloring processing;

[0057] Through binarization and trunk extraction algorithm, the trunk lines of the conductive carbon black lattice structure can be extracted;

[0058] The linear fitting method is used for the marked main pixel points. A straight line is used to fit the direction of the main line, so that as many marked pixel points as possible fall on the straight line generated by the fitting or on both sides of the straight line;

[0059] For the straight lines obtained, their lengths and slopes are counted. In the subsequent processing, the lengths in pixels are converted into data in nanometers according to the scale, and the slopes are converted into angles. By statistically analyzing these data, information such as the length distribution and angle distribution of the main lines in the lattice structure is obtained, so as to better understand the morphology and structural characteristics of the lattice.

[0060] In summary, this image processing method can efficiently and accurately extract the lattice structure of conductive carbon black, providing a reliable basis for subsequent analysis and research. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0062] Figure 1 This is the HRTEM image of conductive carbon black;

[0063] Figure 2 It is a cropped image with a pixel value of 512*512;

[0064] Figure 3 The image was processed by Fourier-inverse Fourier transformation, grayscale, and coloring the stripes using Adobe Photoshop to enhance the stripes.

[0065] Figure 4 It is the image obtained after trunk extraction of the image after inverting the pixel values;

[0066] Figure 5 It is a histogram obtained after data extraction of the image after the backbone is extracted. DETAILED DESCRIPTION

[0067] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0068] A method for extracting the lattice structure of conductive carbon black comprises at least the following steps:

[0069] Step 1: Prepare a dispersion liquid with conductive carbon black and ethanol in a certain ratio, perform a high-resolution transmission electron microscope test after sample preparation, obtain an electron microscope original image and perform a precursor treatment to obtain a pre-processed image;

[0070] Prepare a Python script to process the original image. The original image is generally a tif format file with a pixel value of 4096*4096. In basic processing, the original image needs to be cropped to generate a 512*512 pixel image and converted to jpg format.

[0071] Step 2: Extract features from the preprocessed image obtained in step 1 through a trained convolutional neural network to enhance the contrast between the lattice structure stripes and the background;

[0072] Convolutional neural networks are trained using HRTEM images with clear lattice structures. Convolutional neural networks (CNNs) are a deep learning model or a multi-layer perceptron similar to an artificial neural network, which is often used to analyze visual images. They include the following parts: input layer: input image and other information; convolution layer: used to extract the underlying features of the image; pooling layer: to prevent overfitting and reduce the data dimension; fully connected layer: summarizes the underlying features and information of the image obtained by the convolution layer and the pooling layer; output layer: obtains the result with the highest probability based on the information of the fully connected layer.

[0073] Determine the convolution kernel parameters through supervised learning methods and further process the image;

[0074] Since the input image has three channels, convolution calculation will be performed on a per-channel basis, which will restore the original shape and texture of the image and enhance the contrast between the lattice structure and the background.

[0075] Step 3: The image extracted from the features in step 2 is combined using different weights to enhance the contrast between the main lattice part and the image background to obtain a grayscale processed image;

[0076] For the three obtained three-channel images, appropriate weights are selected to combine them to obtain a result image with enhanced contrast between the main lattice part and the background;

[0077] The color distribution histograms of the three images are exported before combination, and the final weights are determined according to the overall light and dark ratio of the obtained distribution histograms.

[0078] Step 4: Using Adobe Photoshop to adjust the threshold of the grayscale image obtained in step 3, and color the stripes to enhance the stripes for a second time;

[0079] Use Adobe Photoshop software for further processing;

[0080] Using Photoshop's threshold adjustment tool, according to the image features and needs, by adjusting the threshold value, the stripes of the lattice structure are separated to form a more obvious contrast with the image background, and the color processing function of Photoshop is further used to enhance the stripes for a second time;

[0081] By choosing the right color and adjusting the color parameters, you can give the striped part a more vivid color.

[0082] Step 5: Binarize the image after enhancing the stripes in step 4 and invert the pixel values;

[0083] Use the Threshold Adjustment Tool to convert the image into a binary image by selecting an appropriate threshold. This step will divide the pixel values ​​into two categories: black and white, based on the lightness and darkness of the stripes in the image.

[0084] By adjusting the threshold value, the binary effect of controlling the stripes can be achieved;

[0085] After completing the binarization, use the pixel value inversion function provided by the software to swap the black and white pixel values;

[0086] The darker stripes will become white, and the lighter background will become black;

[0087] By inverting the pixel values, the morphology and structure of the stripes can be better highlighted, making the binary image more clearly visible.

[0088] Step 6: In step 5, the eight-neighborhood labeling algorithm is used to extract the main frame of the image after the pixel value is inverted, and the corrosion algorithm is further used to remove the main edge pixels, and finally the white main line is displayed on the black background;

[0089] The improved eight-neighborhood algorithm is used to extract the image backbone;

[0090] The eight-neighborhood algorithm is a classic marking algorithm, which aims to identify and mark aggregated pixels in an image to form a trunk area.

[0091] Then use the erosion algorithm, which iteratively reduces the edge pixels in the main area of ​​the image in each step until the main line is obtained;

[0092] After the trunk extraction is completed, the white trunk lines can be displayed on the black background;

[0093] This means setting the pixel values ​​of the main lines to white (255), while the pixel values ​​of the background remain black (0);

[0094] Against the black background, the main lines will be more clearly visible, highlighting the structure in the image;

[0095] This backbone line image is used to help us better analyze the morphology and connectivity of lattice structures or other specific targets.

[0096] Step 7: Use the linear fitting algorithm to calculate the length and angle of the main lines of the lattice structure on the image obtained in step 6, and collect and calculate the statistical data;

[0097] Compile a script to extract information about the main lines. The key point is that for each line segmented from a connected domain, a new picture should be generated to count them separately.

[0098] Mark the pixels of the main lines on the newly generated image, use the linear fitting method for these marked pixels, and use a straight line to fit the direction of the main lines, so that as many marked pixels as possible fall on the fitted straight line or on both sides of the straight line;

[0099] For the straight line obtained, its length and slope are counted. In the subsequent processing, the length in pixels is converted into data in nanometers according to the scale, and the slope is converted into an angle.

[0100] By performing statistical analysis on these data, we can obtain information such as the length distribution and angle distribution of the main lines in the lattice structure, thereby better understanding the morphology and structural characteristics of the lattice.

[0101] Step 8: Repeat the steps to piece the cropped images together to complete the extraction process of the original tif image.

[0102] The precursor treatment in step 1 at least includes the following steps:

[0103] First, compile a Python script to crop the original image;

[0104] Using the sliding window method, a window is slid on the image with a specified step size, the image is cropped into 512*512 pixel image blocks, and the image blocks within the window are extracted.

[0105] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A method for extracting the lattice structure of conductive carbon black, characterized in that: At least the following steps are included: Step 1: Prepare a dispersion of conductive carbon black and ethanol in a certain ratio. After sample preparation, perform a high-resolution transmission electron microscope test to obtain an original electron microscope image and perform precursor treatment to obtain a pre-processed image, that is, compile a Python script to process the original image. The original image is a tif format file with a pixel value of 4096*4096. In the basic processing, the original image needs to be cropped to generate a 512*512 pixel image and convert the image to jpg format; The precursor treatment comprises at least the following steps: First, compile a Python script to crop the original image; Use the sliding window method to slide the window on the image with a specified step size, crop the image into 512*512 pixel image blocks, and extract the image blocks within the window; Step 2: Extract features from the preprocessed image obtained in step 1 through a trained convolutional neural network to enhance the contrast between the lattice structure stripes and the background; Step 3: The image extracted from the features in step 2 is combined using different weights to enhance the contrast between the main lattice part and the image background to obtain a grayscale processed image; Step 4: Using Adobe Photoshop to adjust the threshold of the grayscale image obtained in step 3, and color the stripes to enhance them twice; Step 5: Binarize the image after enhancing the stripes in step 4 and invert the pixel values; Step 6: In step 5, the eight-neighborhood labeling algorithm is used to extract the main frame of the image after the pixel values ​​are inverted, and the corrosion algorithm is further used to remove the main edge pixels, and finally the white main lines are displayed on the black background; Step 7: Use the linear fitting algorithm to calculate the length and angle of the main lines of the lattice structure on the image obtained in step 6, and collect and calculate the statistical data; Step 8: Repeat the above steps and piece the cropped images together to complete the extraction process of the original tif image.

2. A method for extracting the lattice structure of conductive carbon black according to claim 1, characterized in that: The step 2 at least comprises the following steps: Use HRTEM images with clear lattice structures to train a convolutional neural network; Determine the convolution kernel parameters through supervised learning methods and further process the image; Since the input image has three channels, convolution calculation will be performed on a per-channel basis, which will restore the original shape and texture of the image and enhance the contrast between the lattice structure and the background.

3. A method for extracting the lattice structure of conductive carbon black according to claim 2, characterized in that: The step 3 at least comprises the following steps: For the three obtained three-channel images, appropriate weights are selected to combine them to obtain a result image with enhanced contrast between the main lattice part and the background; The color distribution histograms of the three images are exported before combination, and the final weights are determined according to the overall light and dark ratio of the obtained distribution histograms.

4. The method for extracting the lattice structure of conductive carbon black according to claim 1, characterized in that: The step 4 at least comprises the following steps: Use Adobe Photoshop software for further processing; Using Photoshop's threshold adjustment tool, according to the image features and needs, by adjusting the threshold value, the stripes of the lattice structure are separated to form a more obvious contrast with the image background, and the color processing function of Photoshop is further used to enhance the stripes for a second time; By choosing the right color and adjusting the color parameters, you can give the striped part a more vivid color.

5. The method for extracting the lattice structure of conductive carbon black according to claim 1, characterized in that: The step 5 at least comprises the following steps: Use the Threshold Adjustment Tool to convert the image into a binary image by selecting an appropriate threshold. This step will divide the pixel values ​​into two categories: black and white, based on the lightness and darkness of the stripes in the image. By adjusting the threshold value, the binary effect of controlling the stripes can be achieved; After completing the binarization, use the pixel value inversion function provided by the software to swap the black and white pixel values; The originally darker stripes will become white, and the originally lighter background will become black; By inverting the pixel values, the morphology and structure of the stripes can be better highlighted, making the binary image more clearly visible.

6. The method for extracting the lattice structure of conductive carbon black according to claim 1, characterized in that: The step 6 at least comprises the following steps: The improved eight-neighborhood algorithm is used to extract the image backbone; Then use the erosion algorithm, which iteratively reduces the edge pixels in the main area of ​​the image in each step until the main line is obtained; After the trunk extraction is completed, the white trunk lines can be displayed on the black background; This means setting the pixel value of the main line to white, i.e. 255, while the pixel value of the background remains black, i.e. 0; Against the black background, the main lines will be more clearly visible, highlighting the structure in the image.

7. The method for extracting the lattice structure of conductive carbon black according to claim 1, characterized in that: The step 7 at least includes the following steps: Compile a script to extract information about the main lines. The key point is that for each line segmented from a connected domain, a new picture should be generated to count them separately. Mark the pixels of the main lines on the newly generated image, use the linear fitting method for these marked pixels, and use a straight line to fit the direction of the main lines, so that as many marked pixels as possible fall on the fitted straight line or on both sides of the straight line; For the straight line obtained, its length and slope are counted. In the subsequent processing, the length in pixels is converted into data in nanometers according to the scale, and the slope is converted into an angle. By performing statistical analysis on these data, we can obtain information on the length distribution and angle distribution of the main lines in the lattice structure, thereby better understanding the morphology and structural characteristics of the lattice.

Images