Method for detecting sample uniformity in backscattered electron imaging

Abstract

Embodiments of the present application relate to a detection method based on sample uniformity in backscattered electron imaging, comprising: preprocessing the backscattered image to obtain a gray mean value corrected image; inputting the gray mean value corrected image into a trained model to obtain a color segmentation map; the color segmentation map includes multiple categories of sample particles; analyzing each category of sample particle to obtain feature data of each category of sample particle; and calculating the deposition uniformity of the sample particles and the deposition uniformity between the sample particles in the backscattered image according to the feature data of each category of sample particle. The detection method can realize fast and accurate phase discrimination and quantitative analysis, facilitate product qualification judgment and material optimization, and does not require manual intervention in the entire process, and is suitable for large-scale, fast-flow sample detection.

Description

Technical Field

[0001] This invention relates to the field of materials analysis technology, and in particular to a method for detecting sample uniformity based on backscattered electron imaging. Background Technology

[0002] Backscattered electron imaging (BSE) is a microscopy technique based on the interaction between electrons and the sample, generating backscattered electrons. Its main characteristic is its ability to reflect the contrast differences between regions with different atomic numbers within the sample. The intensity of the backscattered electrons is directly proportional to the atomic number in the sample; the higher the atomic number, the stronger the backscattered electron signal, resulting in a brighter area in the image. Conversely, areas with lower atomic numbers appear as darker areas. The main advantage of BSE images is their ability to provide information on the distribution of different phases within the sample. Particularly in the analysis of complex mixtures, BSE images clearly demonstrate the distribution of different components through differences in grayscale contrast, making them suitable for the compositional analysis of alloys, composite materials, and multiphase compounds.

[0003] In these mixed phases, the differentiation of different components has a decisive impact on the macroscopic properties of the material. Accurate differentiation of various components can achieve several beneficial results, including: a. Material performance optimization: By understanding the distribution and interfacial characteristics of different phases, the composition and preparation process of the material can be optimized, thereby improving its mechanical, thermal, or electrical properties; b. Failure analysis: In material failure analysis, the distribution information of each phase in the mixture can reveal the causes of material failure, such as interfacial cracks, stress concentration caused by phase transformation, etc.; c. Quality control: For materials in industrial production, BSE imaging can serve as a rapid detection method to check the consistency of materials and phase distribution during production, ensuring product quality.

[0004] In backscattered electron imaging (BSE), the gray value of each pixel represents the electron scattering intensity of that region, thus reflecting the relative atomic number information of that region. Therefore, gray-level differences in BSE images can serve as an important basis for phase analysis.

[0005] Current techniques for analyzing the morphology and elemental distribution of silicon-carbon anode materials mainly rely on scanning electron microscopy combined with energy-powered electron spectroscopy. Analyzing the distribution of elements within the material often requires lengthy area scans, line scans, or even spot scans, which is time-consuming, consumes a significant amount of electron microscope time, and makes it difficult to determine the uniformity of all particles within the observation range.

[0006] Existing BSE image analysis methods typically include the following steps: a. Sample pretreatment (e.g., surface polishing under electron microscopy, coating with a conductive layer, etc.); b. Obtaining images of the sample using BSE imaging to observe differences in chemical composition in different regions; c. Performing manual or semi-automated region selection and analysis on the obtained images, such as grayscale difference analysis. These traditional methods mainly rely on grayscale differences in images to infer the uniformity of material composition and microstructure, lacking automated and quantitative analysis methods, and often requiring significant manual intervention and judgment, thus lacking a certain degree of objectivity. Summary of the Invention

[0007] The purpose of this invention is to address the shortcomings of existing technologies by providing a method for detecting sample uniformity in backscattered electron imaging. This method can objectively and quantitatively evaluate the deposition uniformity of sample particles.

[0008] To achieve the above objectives, the present invention provides a method for detecting sample uniformity in backscattered electron imaging, the method comprising:

[0009] The backscattered image is preprocessed to obtain an image with grayscale mean correction;

[0010] The image after grayscale mean correction is input into the trained model to obtain a color segmentation map; the color segmentation map includes sample particles of multiple categories;

[0011] Analyze the particles of each category of samples to obtain the characteristic data of each category of sample particles;

[0012] Based on the characteristic data of each category of sample particles, the deposition uniformity of sample particles and the deposition uniformity between sample particles in the backscattered image are calculated.

[0013] Preferably, before inputting the gray-scale mean-corrected image into the trained model, the detection method further includes training the model, specifically including:

[0014] Step 1: Identify the particles and gaps between particles in the backscattered image based on the image after grayscale mean correction and the preset recognition algorithm parameters; mark the sample particles and the background of the gaps between particles in multiple preset grayscale ranges to obtain a color marking map; mark the background of the gaps between particles and a preset number of sample particles in each preset grayscale range.

[0015] Step 2: Based on the color marker map, classify the particles and gaps between particles in the backscattered image to obtain a color segmentation map;

[0016] Step 3: Adjust the preset recognition algorithm parameters, and repeat Step 1 and Step 2 until the target particles occupy a preset percentage of the area in an image after grayscale mean correction and the segmentation accuracy reaches a preset accuracy value, thereby obtaining the trained model.

[0017] More preferably, the plurality of preset grayscale ranges are specifically four, namely (55-75), (75-95), (95-115) and (115-155).

[0018] More preferably, the preset quantity is 4.

[0019] More preferably, the preset recognition algorithm parameters include one or more of the following: mean, variance, edge, and Gabor features.

[0020] More preferably, the preset recognition algorithm parameters also include film thickness.

[0021] More preferably, based on the color-coded image, the particles and gaps between particles in the backscattered image are classified, specifically as follows:

[0022] A random forest classifier is used to classify the particles and gaps between particles in the backscattered image.

[0023] Preferably, the backscattered image is randomly acquired using a field emission scanning electron microscope (FET), with the FET having a magnification of 1000x or 2000x; and the randomly acquired region is at least 10.

[0024] Preferably, the preprocessing specifically includes:

[0025] The backscattered image is format-converted and its grayscale value is normalized so that the average grayscale value of the backscattered image is a preset grayscale value; the preset grayscale value is 90.

[0026] Preferably, the characteristic data of each category of sample particles includes area, mean gray level, standard deviation of gray level, peak height and peak width in the gray level distribution histogram.

[0027] In summary, the sample uniformity detection method based on backscattered electron imaging provided by this invention utilizes the principle of backscattered electron imaging and intelligent image processing algorithms to intelligently identify sample particles in backscattered images. By optimizing algorithm parameters, it analyzes different types of particles in the sample and classifies and statistically analyzes phases with similar grayscale distributions. This enables rapid and accurate phase differentiation and quantitative analysis. In conclusion, this application achieves efficient and objective quantitative evaluation of material distribution uniformity through intelligent identification and analysis of sample particles in backscattered images using intelligent algorithms. This facilitates product qualification rate judgment and material optimization. The entire process requires no manual intervention and is suitable for large-scale, high-speed sample detection. Attached Figure Description

[0028] Figure 1 A flowchart of a method for detecting sample uniformity in backscattered electron imaging provided in an embodiment of the present invention;

[0029] Figure 2 This diagram illustrates the intelligent recognition process and its effects in the sample uniformity detection method based on backscattered electron imaging provided in this embodiment of the invention. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0031] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0032] This invention provides a method for detecting sample uniformity in backscattered electron imaging, the process of which is as follows: Figure 1 As shown, it includes the following steps:

[0033] Step 110: Preprocess the backscattered image to obtain an image with grayscale mean correction;

[0034] Specifically, backscattered images can be obtained by randomly acquiring samples using a field emission scanning electron microscope (FET), reflecting the morphology and composition of the sample surface. The FET can have a magnification of 1000x or 2000x. At least 10 random acquisition areas are required.

[0035] Furthermore, the preprocessing specifically involves: converting the format of the backscattered image and normalizing its grayscale value so that the average grayscale value of the backscattered image is a preset grayscale value.

[0036] The format conversion process involves converting the color information of each pixel in the backscattered image into a single grayscale value, i.e., converting it into an 8-bit grayscale image, which simplifies the image processing.

[0037] Generally, there is a certain positive proportional relationship between atomic number and image grayscale value. For example, when the sample is a silicon-carbon material, a higher grayscale value indicates a higher silicon content, and a lower grayscale value indicates a higher carbon content. Therefore, this application needs to use atomic number to distinguish the causes affecting uniformity and also needs to normalize the grayscale value.

[0038] First, the average grayscale value of all pixels in the converted image needs to be calculated. Then, based on the difference between this average grayscale value and a preset grayscale value, a linear transformation is performed on the grayscale value of each pixel in the image so that the average grayscale value of the image equals the preset grayscale value. This ensures that images obtained under different shooting conditions (such as different lighting conditions, exposure times, etc.) maintain consistent grayscale distribution characteristics, facilitating subsequent image analysis and processing.

[0039] This is an example, not a limitation; the default grayscale level can be 90.

[0040] Before performing step 120, this application also includes a model training process. It should be noted that the model training in this application can be completed in ImageJ software, and the detailed steps are as follows:

[0041] Step S1: Identify the particles and gaps between particles in the backscattered image based on the image after grayscale mean correction and the preset recognition algorithm parameters, and mark the sample particles and the background of the gaps between particles in multiple preset grayscale ranges to obtain a color marking map.

[0042] Specifically, a randomly selected image after grayscale mean correction can be imported into ImageJ software and then identified using a preset recognition algorithm. In this application, the preset recognition algorithm may include the algorithm of the Trainable Weka Segmentation (TWS) intelligent recognition module and the Membrane Thickness control recognition algorithm. The algorithm parameters in the Trainable Weka Segmentation (TWS) intelligent recognition module may specifically include mean, variance, edge characteristics, Gabor features, etc.

[0043] The mean parameter reflects the average brightness of a local area and is suitable for distinguishing the overall grayscale difference between particles and the background. It is mainly achieved by calculating the mean grayscale of the neighborhood (such as a 3×3 window) of each pixel in the image and establishing a threshold division to initially separate particles from gaps.

[0044] Variance parameter: Variance reflects the degree of change in local grayscale, which is suitable for distinguishing between textured particles and smooth backgrounds. In granular areas, the grayscale value changes significantly and the variance is high. Gap areas usually have lower variance. Calculate the variance of the neighborhood of each pixel and set a threshold for division.

[0045] Edge parameters can be used to identify particle boundaries, especially when particles are dense or have complex boundaries. Edge information is determined by calculating the gray-scale gradient.

[0046] Gabor features capture directional texture information in images, making them suitable for identifying particles with specific texture orientations. Gabor filters with different orientations and frequencies (e.g., 0°, 45°, 90°, and 135°) are used to convolve the image. The filtering results in each direction highlight the fine structure of the particles, such as stripes or layered features. By combining the response values ​​in each direction, the distribution area of ​​the particles relative to the background is determined.

[0047] In complex samples, a single parameter may not fully reflect the particle characteristics. Therefore, parameter combinations are often used to improve the model's adaptability to complex textures and boundary regions. For example:

[0048] Combining mean and variance: First, use the mean parameter to coarsely divide the granular regions, and then use the variance parameter to optimize the boundary.

[0049] Combining mean and Gabor features: the mean distinguishes the overall gray level, while Gabor features identify subtle textures.

[0050] Edges combined with other parameters: use edge parameters to correct the segmentation results of complex boundary regions.

[0051] Because the samples may contain subtle structural and textural features, the parameters need to be adjusted accordingly for the grayscale values ​​of pixels and the features of the image during model training. Therefore, the algorithm parameters are preferably a combination of several parameters from the TWS intelligent recognition module, including mean, variance, edge, and Gabor features.

[0052] Furthermore, the identification of membrane structures in the sample is crucial for extracting complex boundary regions. MembraneThickness controls the sensitivity of the identification algorithm to the continuity of the boundary region, making it suitable for extracting membrane structures with limited width. Membrane thickness can be understood as the smallest unit of algorithm sensitivity, i.e., a pixel. When the membrane thickness is too small (e.g., 1-2 pixels), the sensitivity of the segmentation algorithm needs to be increased; when the membrane thickness is large (e.g., 3-5 pixels), the algorithm's tolerance to width variations needs to be enhanced. In this application, to improve the segmentation effect and achieve a preset accuracy value, a membrane thickness of 1 pixel is preferred.

[0053] There are four preset grayscale ranges: (55-75), (75-95), (95-115), and (115-155).

[0054] The background between particles and each preset grayscale range are marked with a preset number of sample particles. Specifically, the preset number can be four.

[0055] like Figure 2 The middle image is a color-coded image after being recognized by the TWS intelligent recognition module.

[0056] Step 2: Based on the color marker map, classify the particles and gaps between particles in the backscattered image to obtain a color segmentation map;

[0057] Specifically, after identifying particles and gaps in backscattered images, a random forest classifier is preferred to classify these particles and gaps to improve stability and accuracy when processing high-dimensional features (grayscale features, texture features, geometric features, and spatial features, etc.). Since backscattered images contain multiple dimensions of features (grayscale, texture, geometry, etc.), the random forest classifier, through its feature selection mechanism, can effectively utilize relevant features and reduce the influence of redundant features. For sample particles and gaps with complex textures, boundaries, and distributions, the random forest classifier can improve classification accuracy through ensemble learning. The feature distribution of backscattered images may be non-linear (e.g., the relationship between particle shape and grayscale features), and the random forest classifier can approximate complex non-linear relationships through the combination of multiple trees. Under a random sampling mechanism, the random forest classifier is insensitive to parameter selection, resulting in a more stable model.

[0058] Different colors are used to cover different types of particles, thus creating a color segmentation map with particles of different colors, such as... Figure 2 The third image is the color segmentation map. Each preset grayscale range is labeled as a category, namely class1, class2, class3, and class4, and the background area between particles is labeled as class0. This is how the color segmentation map is obtained.

[0059] Step 3: Adjust the preset recognition algorithm parameters, and repeat Step 1 and Step 2 until the target particles occupy a preset percentage of the area in an image after grayscale mean correction and the segmentation accuracy reaches the preset accuracy value, thereby obtaining the trained model.

[0060] Specifically, the preset percentage can be above 95%, and the preset precision value can be above 94%.

[0061] Step 120: Input the image after grayscale mean correction into the trained model to obtain the color segmentation map;

[0062] The color segmentation diagram includes multiple categories of sample particles and the background between the particles, specifically four categories of sample particles in this application.

[0063] Step 130: Analyze the particles of each category of samples to obtain the characteristic data of each category of sample particles;

[0064] Specifically, the ImageJ software's Analyze Particles algorithm is used to analyze the sample particles of each category. The ROI manager for each category of sample particles is called and recorded sequentially, and the marked sample particles are statistically recorded. The characteristic data of each category of sample particles may include area, mean gray level, standard deviation of gray level, peak height and peak width in the gray level distribution histogram.

[0065] Step 140: Based on the characteristic data of each type of sample particle, calculate the deposition uniformity of the sample particles in the backscattered image and the deposition uniformity between sample particles.

[0066] Specifically, the formula for calculating the deposition uniformity of sample particles is as follows:

[0067]

[0068] Among them, A i σ is the total area of ​​sample particles in each class i (i = 1-4); i It is the standard deviation of gray level of sample particles in each class i (i = 1-4); μ i It is the average gray value of each class of sample particles in class i (i = 1-4).

[0069] The formula for calculating the deposition uniformity between sample particles is as follows:

[0070]

[0071] Among them, H i It is the peak height of sample particle i under a certain category in the gray-scale distribution histogram; W i It is the half-peak width of sample particle i under a certain category in the grayscale histogram. The certain category can be class1, class2, class3, or class4.

[0072] Analyzing deposition uniformity using characteristic data of sample particles improves the objectivity of the analysis. This automated and quantitative analysis greatly reduces the inaccuracy of the analysis results caused by human intervention and judgment.

[0073] It should be noted that the above steps only list the data processing procedure for one backscattered image. When processing backscattered images from 10 randomly collected regions, the data obtained from processing each image are summed and the average value is taken.

[0074] For example, the deposition uniformity of sample particles from multiple backscattered images is specifically calculated as follows:

[0075]

[0076] in, is the deposition uniformity of sample particles in the backscattered image of the i-th region, and n is the number of regions where the sample was randomly collected. In this application, n = 10.

[0077] The calculation of the uniformity between particles from multiple backscattered images is as follows:

[0078]

[0079] in, is the particle uniformity of a certain type of particles in the i-th region, and n is the number of regions where the sample was randomly collected. In this application, n = 10.

[0080] In summary, the sample uniformity detection method based on backscattered electron imaging provided by this invention utilizes the principle of backscattered electron imaging and intelligent image processing algorithms to intelligently identify sample particles in backscattered images. By optimizing algorithm parameters, it analyzes different types of particles in the sample and classifies and statistically analyzes phases with similar grayscale distributions. This enables rapid and accurate phase differentiation and quantitative analysis. In conclusion, this application achieves efficient and objective quantitative evaluation of material distribution uniformity through intelligent identification and analysis of sample particles in backscattered images using intelligent algorithms. This facilitates product qualification rate judgment and material optimization. The entire process requires no manual intervention and is suitable for large-scale, high-speed sample detection.

[0081] The sample uniformity detection method based on backscattered electron imaging provided by this invention can be applied to the analysis of electrode materials and alloy materials of energy storage devices such as supercapacitors, lithium-ion batteries, sodium-ion batteries and dye-sensitized batteries.

[0082] To better understand the technical solution provided by the present invention, the following example illustrates the specific process of applying the sample uniformity detection method based on backscattered electron imaging provided in the above embodiments of the present invention, using silicon-carbon material as an example.

[0083] Example 1

[0084] The first step was to set the magnification of the field emission scanning electron microscope to 1000x, and to randomly select 10 different regions of the silicon-carbon sample for image acquisition, resulting in 10 backscattered images.

[0085] The second step is to convert the backscattered image into an 8-bit grayscale image and perform normalization processing so that the average grayscale value of the backscattered image is 90, resulting in 10 images with grayscale mean correction.

[0086] The third step is to randomly select an image after grayscale mean correction for model training.

[0087] First, the image was imported into ImageJ software, and the Trainable Weka Segmentation intelligent algorithm was used to identify silicon-carbon particles and interparticle gaps in the silicon-carbon sample, resulting in a color-coded image. Four grayscale ranges were selected: (55-75), (75-95), (95-115), and (115-155). Within each grayscale range, four silicon-carbon particles were labeled as class-1, class-2, class-3, and class-4, respectively. Similarly, four interparticle gaps were selected on the background and labeled as class-0. The fine structure and texture features of the silicon-carbon sample were identified using a combination of mean and Gabor features, with the film thickness set to 1 pixel.

[0088] Then, a random forest classifier is used to classify the silicon-carbon particles and interparticle gaps in the backscattered image to obtain a color segmentation map. Each preset grayscale interval is labeled as a class, namely class1, class2, class3, and class4, and the background part of the interparticle gap is labeled as class0. In this way, the color segmentation map is obtained.

[0089] Then, the parameters are adjusted and the model is trained until 95% of the area in the image is selected and the accuracy reaches 94%, thus obtaining the trained model.

[0090] The fourth step is to input the remaining 9 grayscale mean-corrected images into the trained model to obtain the color segmentation map.

[0091] The fifth step involves analyzing the sample particles of each category using the Analyze Particles algorithm in ImageJ software, sequentially calling and recording the ROI manager for each category of sample particles, and statistically recording the marked sample particles.

[0092] The sixth step involves calculating the deposition uniformity of the sample particles and the deposition uniformity between sample particles based on the characteristic data of each category of silicon-carbon particles in the 10 backscattered images.

[0093] In this example, when calculating the uniformity of interparticle deposition, class 3 (with grayscale values ​​in the range of (95, 115)) with a relatively high proportion of silicon-carbon particles was selected for calculation and analysis.

[0094] The results show This indicates that the silicon distribution in the sample is highly uniform across all categories; among them, the silicon-carbon particle area of ​​class 3 is 74%, indicating that the silicon-carbon particle area of ​​this category is relatively large, the sample deposition is relatively uniform, and the silicon content is low. This uniformity indicates that the deposition uniformity of the same type of particles in the tested silicon-carbon material is relatively high, and the deposition uniformity of the silicon-carbon particles of the largest proportion is relatively good.

[0095] Comparative Example 1

[0096] Step 1-Step 2 are the same as in Example 1.

[0097] Third, set Membrane Thickness to 2 pixels.

[0098] Comparative Example 2

[0099] Step 1-Step 2 are the same as in Example 1.

[0100] Third, set Membrane Thickness to 3 pixels.

[0101] Comparative Example 3

[0102] Step 1-Step 2 are the same as in Example 1.

[0103] The third step involves having two silicon-carbon particles in each grayscale range.

[0104] Comparative Example 4

[0105] Step 1-Step 2 are the same as in Example 1.

[0106] The third step involves having 8 silicon-carbon particles in each grayscale range.

[0107] The results show that the accuracy of Example 1 is 94%, indicating that when the film thickness is 1 pixel, the film edge is clear and the detail processing is accurate. When the sampling amount of each grayscale interval is 4 pixels, the training time is controllable.

[0108] Comparative Examples 1 and 2 adjusted the parameters for different film thicknesses. In Comparative Example 2, the accuracy dropped to 88%, and the film edges became blurred, resulting in the loss of some details. In Comparative Example 3, the accuracy dropped to 84%, and image noise significantly interfered with the segmentation results.

[0109] Comparative examples 3 and 4 show adjustments to the sampling amount for each grayscale interval. Comparative example 3 has a shorter model training time and less accurate segmentation results, especially in complex texture regions, with an accuracy of 85%. Comparative example 4, while achieving an accuracy of 92%, doubles the model training time, leading to a significant increase in computational resource consumption, particularly in large-scale image datasets.

[0110] Therefore, this application only uses the model trained in Example 1 to process other images and perform subsequent calculations.

[0111] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for detecting sample uniformity in backscattered electron imaging, characterized in that, The detection method includes: The backscattered image is preprocessed to obtain an image with grayscale mean correction; The image after grayscale mean correction is input into the trained model to obtain a color segmentation map; the color segmentation map includes sample particles of multiple categories; Analyze the particles of each category of samples to obtain the characteristic data of each category of sample particles; Based on the characteristic data of each category of sample particles, the deposition uniformity of sample particles and the deposition uniformity between sample particles in the backscattered image are calculated.

2. The detection method according to claim 1, characterized in that, Before inputting the gray-scale mean-corrected image into the trained model, the detection method further includes training the model, specifically including: Step 1: Identify the particles and gaps between particles in the backscattered image based on the image after grayscale mean correction and the preset recognition algorithm parameters; mark the sample particles and the background of the gaps between particles in multiple preset grayscale ranges to obtain a color marking map; mark the background of the gaps between particles and a preset number of sample particles in each preset grayscale range. Step 2: Based on the color marker map, classify the particles and gaps between particles in the backscattered image to obtain a color segmentation map; Step 3: Adjust the preset recognition algorithm parameters, and repeat Step 1 and Step 2 until the target particles occupy a preset percentage of the area in an image after grayscale mean correction and the segmentation accuracy reaches a preset accuracy value, thereby obtaining the trained model.

3. The detection method according to claim 2, characterized in that, The multiple preset grayscale ranges are specifically four, namely (55-75), (75-95), (95-115) and (115-155).

4. The detection method according to claim 2, characterized in that, The preset quantity is 4.

5. The detection method according to claim 2, characterized in that, The preset recognition algorithm parameters include one or more of the following: mean, variance, edge, and Gabor features.

6. The detection method according to claim 2, characterized in that, The preset recognition algorithm parameters also include film thickness.

7. The detection method according to claim 2, characterized in that, Based on the color-coded image, the particles and interparticle gaps in the backscattered image are classified as follows: A random forest classifier is used to classify the particles and gaps between particles in the backscattered image.

8. The detection method according to claim 1, characterized in that, The backscattered images are randomly acquired using a field emission scanning electron microscope (FET), with a magnification of 1000x or 2000x; the randomly acquired regions are at least 10.

9. The detection method according to claim 1, characterized in that, The preprocessing specifically includes: The backscattered image is format-converted and its grayscale value is normalized so that the average grayscale value of the backscattered image is a preset grayscale value; the preset grayscale value is 90.

10. The detection method according to claim 1, characterized in that, The characteristic data of each category of sample particles include area, mean gray level, standard deviation of gray level, peak height and peak width in the gray level distribution histogram.

Images