Method and system for analyzing and calculating aggregation condition of analyte in chemical composition

Abstract

The invention relates to the technical field of chemical analysis, and particularly discloses a method and a system for analyzing and calculating the aggregation condition of an analyte in a chemical composition, and the analysis method comprises the following steps: obtaining a chemical image of the analyte in a control sample and a chemical image of the analyte in at least one test sample, carrying out Gaussian blur on each chemical image, and calculating the aggregation condition of the analyte in the chemical composition; carrying out brightness calibration on the obtained Gaussian blurred image; converting the Gaussian blurred image after brightness calibration into a binary image; converting the binary image into a one-dimensional array; counting continuous lengths when predetermined pixels appear in the one-dimensional array and frequencies corresponding to the continuous lengths, and generating a histogram of the frequencies and the continuous lengths; fitting the histogram by using an exponential Gaussian distribution function to obtain an exponential attenuation parameter of the fitted exponential Gaussian distribution function; and calculating the aggregation index of the test sample according to the index attenuation parameters corresponding to the test sample and the control sample. According to the technical scheme, the aggregation condition of the analyte can be quantitatively determined.

Description

Technical Field

[0001] The present disclosure generally relates to the field of chemical analysis technology, and more specifically, to a method and system for analyzing and calculating the aggregation of analytes in a chemical composition. Background Art

[0002] In chemical analysis, evaluating the aggregation of particulate matter is crucial for understanding the physical properties and chemical behavior of materials. For example, in the quality control of products such as cosmetics, pharmaceuticals, and chemical products, the aggregation of particles affects the stability, appearance, and performance of the products. Especially in the field of cosmetics, products are usually provided in the form of emulsions or suspensions. In emulsions, especially in color cosmetic products, a high concentration of particulate matter is usually required to provide the desired color of the product. During the manufacturing process of these cosmetics, high-energy dispersion techniques are needed to break up aggregated particulate matter and reduce the particle size to ensure that most particles reach the original particle state to achieve the uniformity of the cosmetics.

[0003] Traditionally, the aggregation of particulate matter is usually analyzed by a drawdown fineness gauge or by observation under an optical microscope. However, a drawdown fineness gauge is mainly used for measuring the fineness of paint and ink particles. Due to the subjectivity of the user's operation and judgment criteria, it can generally only be used for rough measurement. An optical microscope can be used to directly observe the aggregation of particulate matter. However, due to the limited size of the observation area at high magnification and the fact that the microscope optical path is easily blocked by the particles being observed, it is difficult to obtain an accurate analysis result regarding the aggregation. With the development of chemical analysis technology, chemical imaging technology has been used in the field of cosmetic product development and in the assessment of pigment component aggregation. However, currently, the aggregation of particulate matter can only be roughly estimated by observing chemical images, and the aggregation of particulate matter cannot be quantitatively analyzed. Summary of the Invention

[0004] To solve the problems in the related art, the present disclosure provides a method and system for analyzing and calculating the aggregation of analytes in a chemical composition.

[0005] According to one aspect of the present invention, the present disclosure provides an analysis method for the aggregation of analytes in a chemical composition, including:

[0006] Obtaining a chemical image of the analyte in a control sample and chemical images of the analyte in at least one test sample, where the control sample and the test sample are chemical compositions;

[0007] For each chemical image, performing Gaussian blur on the chemical image using a preset Gaussian kernel to obtain a Gaussian blurred image;

[0008] Performing brightness calibration on the Gaussian blurred image;

[0009] Convert the Gaussian blurred image after brightness calibration into a binary image;

[0010] Convert the binary image into a one-dimensional array;

[0011] Statistically calculate the continuous length when a predetermined pixel appears in the one-dimensional array and the frequency corresponding to each continuous length, and generate a histogram of frequency and continuous length;

[0012] Use an exponential Gaussian distribution function to fit the histogram to obtain the exponential decay parameter of the fitted exponential Gaussian distribution function;

[0013] Calculate the aggregation index of the test sample according to the exponential decay parameter corresponding to the test sample and the exponential decay parameter corresponding to the control sample.

[0014] In a possible implementation manner, the kernel size of the preset Gaussian kernel is greater than or equal to 3×3 and less than 61×61.

[0015] In a possible implementation manner, the calculation formula of the standard deviation σ of the preset Gaussian kernel includes:

[0016] σ = 0.3×(kernel size / 2 - 1) + 0.8;

[0017] Alternatively, the standard deviation σ of the preset Gaussian kernel is greater than or equal to 2% of the kernel size and less than or equal to 49% of the kernel size. Preferably, the standard deviation σ is greater than or equal to 10% of the kernel size and less than or equal to 40% of the kernel size. More preferably, the standard deviation σ is greater than or equal to 20% of the kernel size and less than or equal to 30% of the kernel size.

[0018] In a possible implementation manner, the brightness calibration of the Gaussian blurred image includes:

[0019] Perform brightness calibration on the Gaussian blurred image of the control sample to obtain the control brightness of the Gaussian blurred image of the control sample after brightness calibration;

[0020] Perform brightness calibration on the Gaussian blurred image of the test sample according to the control brightness, so that the difference between the brightness of the Gaussian blurred image of the test sample after brightness calibration and the control brightness is within a predetermined range.

[0021] In a possible implementation manner, the predetermined range includes less than 100% of the control brightness. Preferably, it is less than 10%. More preferably, it is less than 5%. Even more preferably, it is less than 1%.

[0022] In a possible implementation manner, the conversion of the binary image into a one-dimensional array includes:

[0023] Traverse each pixel of the binary image in raster mode and convert the binary image into a one-dimensional array.

[0024] In one possible implementation, the predetermined pixel is a pixel representing the presence of the analyte.

[0025] In one possible implementation, calculating the aggregation index of the test sample according to the exponential decay parameter of the test sample and the exponential decay parameter of the control sample includes:

[0026] Calculate the aggregation index A of the test sample according to the following formula:

[0027]

[0028] where Tau sample is the exponential decay parameter of the test sample, and Tau control is the exponential decay parameter of the control sample.

[0029] In one possible implementation, the chemical image includes an SEM-EDX image.

[0030] In one possible implementation, the chemical image is a grayscale chemical image.

[0031] In one possible implementation, if the chemical image is a color image, before performing brightness calibration on the Gaussian blurred image, the method further includes:

[0032] Convert the Gaussian blurred image from a color image to a grayscale image.

[0033] According to another aspect of the present invention, the present disclosure provides an analysis system for the aggregation of an analyte in a chemical composition, including:

[0034] An image acquisition module configured to acquire a chemical image of the analyte in a control sample and a chemical image of the analyte in at least one test sample, where the control sample and the test sample are chemical compositions;

[0035] A Gaussian blur module configured to perform Gaussian blur on each chemical image using a preset Gaussian kernel to obtain a Gaussian blurred image;

[0036] A brightness calibration module configured to perform brightness calibration on the Gaussian blurred image;

[0037] A binary conversion module configured to convert the brightness-calibrated Gaussian blurred image into a binary image;

[0038] A fitting module, configured to convert the binarized image into a one-dimensional array; count the continuous lengths when a predetermined pixel appears in the one-dimensional array and the frequency corresponding to each continuous length, generate a histogram of the frequency and the continuous length; use an exponential Gaussian distribution function to fit the histogram to obtain the exponential decay parameter of the fitted exponential Gaussian distribution function.

[0039] A calculation module, configured to calculate the aggregation index of the test sample according to the exponential decay parameter corresponding to the test sample and the exponential decay parameter corresponding to the control sample.

[0040] In a possible implementation manner, the kernel size of the preset Gaussian kernel is greater than or equal to 3×3 and less than 61×61.

[0041] The calculation formula of the standard deviation σ of the preset Gaussian kernel includes:

[0042] σ = 0.3×(kernel size / 2 - 1) + 0.8;

[0043] The standard deviation σ of the preset Gaussian kernel is greater than or equal to 2% of the kernel size and less than or equal to 49% of the kernel size.

[0044] In a possible implementation manner, the brightness calibration module is specifically configured to:

[0045] A first calibration sub-module, configured to perform brightness calibration on the Gaussian blurred image of the control sample to obtain the control brightness of the Gaussian blurred image of the control sample after brightness calibration;

[0046] A second calibration sub-module, configured to perform brightness calibration on the Gaussian blurred image of the test sample according to the control brightness, so that the difference between the brightness of the Gaussian blurred image of the test sample after brightness calibration and the control brightness is within a predetermined range.

[0047] In a possible implementation manner, the predetermined range includes less than 100% or 10% of the control brightness.

[0048] In a possible implementation manner, the part in the fitting module that converts the binarized image into a one-dimensional array is configured to:

[0049] Traverse each pixel of the binarized image in a raster pattern and convert the binarized image into a one-dimensional array.

[0050] In a possible implementation manner, the predetermined pixel is a pixel representing the presence of the analyte.

[0051] In a possible implementation manner, the calculation module is specifically configured to:

[0052] Calculate the aggregation index A of the test sample according to the following formula:

[0053]

[0054] where Tau sample is the exponential decay parameter of the test sample, and Tau control is the exponential decay parameter of the control sample.

[0055] In a possible implementation, the chemical image includes an SEM-EDX image.

[0056] In a possible implementation, the chemical image is a grayscale chemical image.

[0057] In a possible implementation, if the chemical image is a color image, the analysis system further includes:

[0058] A color conversion module configured to convert the Gaussian blurred image from a color image to a grayscale image before the brightness calibration module calibrates the brightness of the Gaussian blurred image.

[0059] According to one aspect of the present invention, the present disclosure provides a method for calculating the aggregation of an analyte in a chemical composition, which includes:

[0060] Obtain a chemical image of the analyte in a test sample, where the test sample is a chemical composition;

[0061] Perform Gaussian blur on the chemical image using a preset Gaussian kernel to obtain a Gaussian blurred image;

[0062] Perform brightness calibration on the Gaussian blurred image;

[0063] Convert the brightness-calibrated Gaussian blurred image into a binary image;

[0064] Convert the binary image into a one-dimensional array;

[0065] Statistically count the continuous length when a predetermined pixel appears in the one-dimensional array and the frequency corresponding to each continuous length, and generate a histogram of frequency and continuous length;

[0066] Fit the histogram through an exponential Gaussian distribution function to determine the average value and standard deviation in the physical size of the test sample.

[0067] In a possible implementation, the continuous length is the continuous pixel length, and the step of fitting the histogram through an exponential Gaussian distribution function to determine the average value and standard deviation in the physical size of the test sample includes:

[0068] Fitting the histogram with an exponential Gaussian distribution function to obtain the mean and standard deviation of the fitted exponential Gaussian distribution function;

[0069] According to the correspondence between pixels and physical dimensions, convert the mean and standard deviation of the fitted exponential Gaussian distribution function into the mean and standard deviation corresponding to the physical dimensions.

[0070] In a possible implementation manner, the continuous length is a continuous physical length, and the statistics of the continuous length when a predetermined pixel appears in the one-dimensional array includes:

[0071] Statistical the continuous pixel length when a predetermined pixel appears in the one-dimensional array;

[0072] According to the correspondence between pixels and physical dimensions, convert the continuous pixel length into a continuous physical length.

[0073] According to one aspect of the present invention, the present disclosure provides a calculation system for the aggregation of analytes in a chemical composition, which includes:

[0074] An acquisition module configured to acquire a chemical image of an analyte in a test sample, where the test sample is a chemical composition:

[0075] A blurring module configured to perform Gaussian blurring on the chemical image using a preset Gaussian kernel to obtain a Gaussian blurred image;

[0076] A calibration module configured to perform brightness calibration on the Gaussian blurred image;

[0077] A conversion module configured to convert the Gaussian blurred image after brightness calibration into a binary image;

[0078] A dimension conversion module configured to convert the binary image into a one-dimensional array;

[0079] A statistical module configured to count the continuous length when a predetermined pixel appears in the one-dimensional array and the frequency corresponding to each continuous length, and generate a histogram of frequency and continuous length;

[0080] A physical dimension module configured to fit the histogram with an exponential Gaussian distribution function to determine the mean and standard deviation in the physical dimensions of the test sample.

[0081] In a possible implementation manner, the continuous length is a continuous pixel length, and the physical dimension module is configured to:

[0082] Fitting the histogram with an exponential Gaussian distribution function to obtain the mean and standard deviation of the fitted exponential Gaussian distribution function;

[0083] According to the correspondence between pixels and physical dimensions, the mean and standard deviation of the fitted exponential Gaussian distribution function are converted into the mean and standard deviation corresponding to the physical dimensions.

[0084] In a possible implementation manner, the continuous length is a continuous physical length, and the statistical module is configured to:

[0085] Statistically count the continuous pixel length when a predetermined pixel appears in the one-dimensional array;

[0086] According to the correspondence between pixels and physical dimensions, the continuous pixel length is converted into a continuous physical length.

[0087] According to the technical solution provided by the present disclosure, Gaussian blur, brightness calibration, image binarization, one-dimensional array conversion, generation of a histogram of the occurrence frequency and continuous length of a predetermined pixel representing an analyte, and fitting of an exponential Gaussian distribution function can be performed based on chemical images of a control sample and a test sample containing the same analyte, so as to obtain an exponential decay parameter corresponding to the test sample and an exponential decay parameter corresponding to the control sample. Furthermore, based on this, the aggregation index of the analyte in the test sample relative to the control sample can be calculated, and the aggregation situation of the analyte in the test sample can be accurately and quantitatively determined;

[0088] In addition, by performing Gaussian blur, brightness calibration, image binarization, one-dimensional array conversion, generation of a histogram of the occurrence frequency and continuous length of a predetermined pixel representing an analyte, and fitting of an exponential Gaussian distribution function on the chemical image of the test sample, the mean and standard deviation in physical dimensions of the test sample are obtained, and the mean and standard deviation in physical dimensions can clearly characterize the aggregation situation of the analyte in the test sample.

[0089] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0090] Combined with the drawings, through the following detailed description of non-limiting embodiments, other features, objectives, and advantages of the present disclosure will become more obvious. In the drawings:

[0091] Figure 1 A flowchart of an analysis method for the aggregation situation of an analyte in a chemical composition provided by an embodiment of the present disclosure is shown.

[0092] Figure 2 A structural block diagram of an analysis system for the aggregation situation of an analyte in a chemical composition according to an embodiment of the present disclosure is shown.

[0093] Figure 3 A flowchart of a calculation method for the aggregation situation of an analyte in a chemical composition provided by an embodiment of the present disclosure is shown.

[0094] Figure 4 The structural block diagram of a computing system for calculating the analyte aggregation in a chemical composition provided by an embodiment of the present disclosure is shown.

[0095] Figure 5 The illustration showing the chemical images corresponding to the three samples provided in Example 1 is shown.

[0096] Figure 6 The illustration showing the grayscale images of the Gaussian blurred images corresponding to the three samples provided in Example 1 is shown.

[0097] Figure 7 The illustration showing the binary images corresponding to the three samples provided in Example 1 is shown.

[0098] Figure 8 The illustration showing the fitting curve graphs of the exponential Gaussian distribution functions corresponding to the three samples provided in Example 1 is shown.

[0099] Figure 9 The illustration showing the chemical images corresponding to the two samples provided in Example 2.0 is shown.

[0100] Figure 10 The illustration showing the grayscale images of the Gaussian blurred images corresponding to the two samples provided in Example 2.0 is shown.

[0101] Figure 11 The illustration showing the binary images corresponding to the two samples provided in Example 2.0 is shown.

[0102] Figure 12 The illustration showing the fitting curve graphs of the exponential Gaussian distribution functions corresponding to the two samples provided in Example 2.0 is shown.

[0103] Figure 13 The illustration showing the grayscale images of the Gaussian blurred images corresponding to the two samples provided in Example 2.1 is shown.

[0104] Figure 14 The illustration showing the binary images corresponding to the two samples provided in Example 2.1 is shown.

[0105] Figure 15 The illustration showing the fitting curve graphs of the exponential Gaussian distribution functions corresponding to the two samples provided in Example 2.1 is shown.

[0106] Figure 16 The illustration showing the grayscale images of the Gaussian blurred images corresponding to the two samples provided in Example 2.2 is shown.

[0107] Figure 17 The illustration showing the binary images corresponding to the two samples provided in Example 2.2 is shown.

[0108] Figure 18Illustration showing the fitting curve graphs of the exponential Gaussian distribution functions corresponding to the two samples provided in Example 2.2.

[0109] Figure 19 Illustration showing the binary image corresponding to the test sample provided in Example 3.0.

[0110] Figure 20 Illustration showing the fitting curve graphs of the exponential Gaussian distribution functions corresponding to the test sample provided in Example 3.0.

[0111] Figure 21 Illustration showing the binary image corresponding to the test sample provided in Example 3.1.

[0112] Figure 22 Illustration showing the fitting curve graphs of the exponential Gaussian distribution functions corresponding to the test sample provided in Example 3.1.

[0113] Figure 23 Illustration showing the binary image corresponding to the test sample provided in Example 3.2.

[0114] Figure 24 Illustration showing the fitting curve graphs of the exponential Gaussian distribution functions corresponding to the test sample provided in Example 3.2.

[0115] Figure 25 Illustration showing the binary image corresponding to the test sample provided in Example 3.3.

[0116] Figure 26 Illustration showing the fitting curve graphs of the exponential Gaussian distribution functions corresponding to the test sample provided in Example 3.3.

[0117] Figure 27 Illustration showing a chemical image of titanium dioxide particulate matter provided in Example 4.

[0118] Figure 28 Showing Figure 27 Illustrations of the grayscale images of the chemical image shown in three directions.

[0119] Figure 29 Showing Figure 28 Illustrations of the binary images of the grayscale images in the three directions shown.

[0120] Figure 30 Illustrations showing the fitting curve graphs of the exponential Gaussian distribution functions corresponding to the three directions respectively.

[0121] Figure 31 Illustration showing the grayscale images corresponding to the two samples provided in Comparative Example 1.

[0122] Figure 32Illustration of the binary images corresponding to the two samples provided in Comparative Example 1.

[0123] Figure 33 Illustration of the fitting curve graph of the exponential Gaussian distribution function corresponding to the two samples provided in Comparative Example 1.

[0124] Figure 34 Illustration of the histogram of the FFT-processed chemical image provided in Comparative Example 2.

[0125] Figure 35 Illustration of the histogram of the Gaussian blurred image after FFT processing provided in Comparative Example 2.

[0126] Figure 36 Illustration of the histogram of the grayscale image after FFT processing provided in Comparative Example 2.

[0127] Figure 37 Illustration of the fitting curve of the exponential Gaussian distribution function fitted by the histogram of the continuous physical length and the count of the continuous physical length provided in Example 5. Detailed Description of the Invention

[0128] Hereinafter, exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings.

[0129] In the present disclosure, Gaussian blur is an image processing technique used to reduce noise and details in an image. It works by applying a Gaussian function to the image, such that the value of each pixel in the image is replaced by the weighted average of its neighboring pixels. The weights during weighting are determined by the Gaussian distribution. Generally, the influence of neighboring pixels is greater and the weights are higher, while the influence of farther pixels is smaller and the weights are lower.

[0130] In the present disclosure, the Exponential Gaussian Distribution Function is a probability distribution in statistics. It is a variant of the Gaussian distribution and adjusts its shape by introducing an exponential factor. This distribution can be used to model data with long-tail characteristics. The Exponential Gaussian Distribution Function contains the following three parameters:

[0131] μ (Mu): Location parameter, similar to the mean of the Gaussian distribution, which determines the position of the peak of the distribution;

[0132] σ (Sigma): Scale parameter, similar to the standard deviation of the Gaussian distribution, which determines the spread or width of the distribution;

[0133] T (Tau): Exponential decay parameter, representing the decay rate of the exponential tail. A larger Tau value results in a slower decay of the tail, while a smaller Tau value results in a faster decay.

[0134] In the present disclosure, a chemical image generally refers to an image obtained by chemical imaging techniques, and the chemical image obtained by chemical imaging techniques can provide information on the spatial distribution of the chemical composition of a sample. Different from traditional images that only show the physical appearance of a sample, a chemical image provides information on the composition or concentration of specific chemical components at different positions within the sample.

[0135] Figure 1 The flowchart of an analysis method for the aggregation of an analyte in a chemical composition provided by an embodiment of the present disclosure is shown. As Figure 1 shown, the analysis method includes the following steps S101 - S108:

[0136] In step S101, a chemical image of the analyte in a control sample and chemical images of the analyte in at least one test sample are obtained, where the control sample and the test sample are chemical compositions;

[0137] In step S102, for each chemical image, the chemical image is subjected to Gaussian blur using a preset Gaussian kernel to obtain a Gaussian blurred image;

[0138] In step S103, the brightness of the Gaussian blurred image is calibrated;

[0139] In step S104, the brightness - calibrated Gaussian blurred image is converted into a binary image;

[0140] In step S105, the binary image is converted into a one - dimensional array;

[0141] In step S106, the continuous length when a predetermined pixel appears in the one - dimensional array and the frequency corresponding to each continuous length are statistically counted to generate a histogram of frequency and continuous length;

[0142] In step S107, the histogram is fitted using an exponential Gaussian distribution function to obtain the exponential decay parameter of the fitted exponential Gaussian distribution function;

[0143] In step S108, according to the exponential decay parameter corresponding to the test sample and the exponential decay parameter corresponding to the control sample, the aggregation index of the test sample is calculated.

[0144] In a possible implementation manner, the analysis method provided in this embodiment is mainly used for analyzing the aggregation of an analyte in a chemical composition. The chemical composition can be a chemical composition in various fields such as personal care products, pharmaceuticals, and foods. The chemical composition can be in various forms such as liquid, colloid, powder, semi - solid, etc. For example, it can be toothpaste, emulsion, cream, shampoo, loose powder, powder cake, and concentrated solutions and suspensions of high - molecular polymers, etc. The analyte refers to the particulate matter in the chemical composition.

[0145] In a possible implementation, both the control sample and the test sample are chemical compositions. The same analyte is contained in both the control sample and the test sample. The distribution of the analyte in the control sample is relatively uniform, and the aggregation situation can be known. Based on the control sample, the aggregation situation of the analyte in one or more test samples can be analyzed.

[0146] In a possible implementation, a chemical imaging device can be used to photograph the analyte in the control sample and the analyte in the test sample respectively to obtain corresponding chemical images. A chemical imaging device refers to a device that can obtain chemical images using chemical imaging technology. For example, a scanning electron microscope - energy dispersive X-ray spectroscopy (SEM-EDX) device, an SEM-EDX image obtained by combining the high-resolution imaging ability of a scanning electron microscope (SEM) and the elemental analysis ability of energy dispersive X-ray spectroscopy (EDX); an X-ray fluorescence imaging device, which can use X-rays to excite fluorescence in the sample to obtain an elemental distribution image of the sample; a secondary ion mass spectrometry device, which analyzes the secondary ions generated by exposing the sample to a high-energy ion beam to obtain a chemical image of the analyte in the sample; a Fourier transform infrared spectroscopy imaging device, a Fourier transform infrared spectrometer image that can obtain spatial distribution information of different chemical components in the sample by measuring the absorption or emission of infrared light by the sample; and so on. Preferably, the chemical image in this implementation can be an SEM-EDX image.

[0147] In a possible implementation, for each chemical image, whether it is the chemical image of the control sample or the chemical image of any test sample, a preset Gaussian kernel is used for Gaussian blurring to obtain a Gaussian blurred image. The Gaussian kernel can be preset according to the required degree of blurring. Preferably, the kernel size of the Gaussian kernel is greater than or equal to 3×3 and less than 61×61.

[0148] In a possible implementation, the standard deviation σ of the Gaussian kernel can be determined by the kernel size, and the value of the standard deviation σ determines the distribution of weights within the kernel. When the value of σ is low, the Gaussian distribution is narrow, which means that most of the weights are concentrated near the center of the kernel. Therefore, the pixels near the center of the kernel have a greater impact on the final blurred pixel value, while the pixels far from the center of the kernel have little impact. If the value of the standard deviation σ is too low relative to the kernel size, it may result in inefficient use of the outer-layer pixels of the kernel because their weights are very small. When the value of σ is high, the Gaussian distribution is wide, and the weights are distributed more evenly across the entire kernel. The pixels far from the center of the kernel have a greater impact on the generated blurred pixels. This results in a stronger blur effect, with more extensive smoothing of both details and larger structures. If the value of the standard deviation σ is too high for the kernel size, the weights may become too uniform, causing the kernel to behave like a simple mean filter (average filter).

[0149] Therefore, preferably, the value of the standard deviation σ can be generated by the following formula: σ = 0.3×(kernel size / 2 - 1) + 0.8; for example, when the kernel size is 18×18, the standard deviation σ of the preset Gaussian kernel = 0.3×(18 / 2 - 1) + 0.8 = 3.2. The value of the standard deviation σ calculated in this way can ensure that the Gaussian distribution is well represented within the kernel, thus achieving a balanced and effective blur effect.

[0150] In another possible implementation, the standard deviation σ of the preset Gaussian kernel can also be selected within a predetermined range. Preferably, the standard deviation σ of the preset Gaussian kernel is greater than or equal to 2% of the kernel size and less than or equal to 49% of the kernel size. For example, when the kernel size is 30×30, the standard deviation σ of the preset Gaussian kernel is greater than or equal to 2% of 30 and less than or equal to 49% of 30. In this way, the value of the standard deviation σ of the Gaussian kernel is restricted within a certain range, which can effectively blur the image and ensure that the blurred image is sufficiently suitable for subsequent analysis. Preferably, the standard deviation σ is greater than or equal to 10% of the kernel size and less than or equal to 40% of the kernel size. More preferably, the standard deviation σ is greater than or equal to 20% of the kernel size and less than or equal to 30% of the kernel size.

[0151] In a possible implementation, the chemical image can be a grayscale chemical image. A grayscale image is a special image that only contains grayscale (light and dark) information and no color information. Alternatively, the chemical image can be a color chemical image. In this case, after obtaining the Gaussian blurred image, it is necessary to convert the Gaussian blurred image from a color image to a grayscale image.

[0152] Preferably, the grayscale image in the present disclosure is an 8-bit grayscale image. Here, "8-bit" means that the pixel values in the grayscale image are represented by 8-bit binary numbers, with 2 8= 256 possibilities, and the pixel value of each pixel is an integer between 0 (black) and 255 (white). Of course, the grayscale image in the present disclosure is not limited to an 8-bit grayscale image. As long as it can be achieved, the grayscale image can also be a grayscale image of any bit, such as a 1-bit grayscale image, a 2-bit grayscale image, a 4-bit grayscale image, or a 16-bit grayscale image, etc. Taking the 16-bit grayscale image as an example, the pixel value of the 16-bit grayscale image is represented by a 16-bit binary number, and each pixel has 2 蹈 possibilities, so the range of grayscale values is 0 - 65535.

[0153] It should be noted that the grayscale value reflects the signal intensity of the spatial distribution of the analyte. Usually, one end of the color scale refers to zero intensity, and the other end refers to the highest intensity. For example, for the above 8-bit grayscale image, the grayscale value 0 can represent the highest signal intensity in the grayscale image, which can be shown as black in the grayscale image; while the grayscale value 255 can represent the lowest signal intensity in the grayscale image, which can be shown as white in the grayscale image. It is also possible to represent the exactly same grayscale image by assigning 0 to the lowest intensity and 255 to the highest intensity, which has no impact on subsequent analysis.

[0154] In a possible implementation, the Gaussian blurred image can be brightness calibrated so that the image brightness of the Gaussian blurred images of the control sample and the test sample is at the same level, so as to subsequently analyze the aggregation of the analyte in the test sample with the control sample as a reference at the same brightness level.

[0155] In a possible implementation, a binarization technique can be adopted to convert the brightness-calibrated Gaussian blurred image into a binarized image. Binarization refers to converting a grayscale image into an image that only contains two colors (usually black and white). For example, a threshold can be set, and pixels with grayscale values higher than the threshold are set to one color (usually white), while all pixels lower than the threshold are set to another color (usually black). Taking the grayscale image as an 8-bit grayscale image as an example, the range of grayscale values of each pixel in the 8-bit grayscale image is 0 - 255. Binarization refers to the process of making the grayscale value of each pixel in the pixel matrix of the grayscale image become 0 (black) or 255 (white), that is, the grayscale value in the binarized image is only 0 or 255, and the whole binarized image only looks black and white.

[0156] In a possible implementation, after obtaining the binary image, the pixels in the binary image can be traversed in a predetermined order, such as row by row or column by column, to convert the binary image into a one-dimensional array. There are two values in the converted one-dimensional array representing the black and white pixels in the binary image. Effectively, during the conversion, the pixels representing the presence of the analyte in the binary image can be converted into the value 1 in the one-dimensional array, while the pixels without the analyte are converted into the value 0 in the one-dimensional array. Of course, it is also possible to convert the pixels representing the presence of the analyte in the binary image into the value 0 in the one-dimensional array, and the pixels without the analyte into the value 1 in the one-dimensional array. In this implementation, the former numerical representation method is selected to ensure better visibility of the data.

[0157] In a possible implementation, the predetermined pixel is the pixel representing the presence of the analyte. For example, when converting the pixels representing the presence of the analyte in the binary image into the value 1 in the one-dimensional array and the pixels without the analyte into the value 0 in the one-dimensional array, the continuous length when the value 1 continuously appears in the one-dimensional array and the frequency corresponding to each continuous length can be counted. Here, the continuous length can be the continuous pixel length or the continuous physical length. The continuous pixel length is the length in units of pixels, and the continuous physical length is the length in physical dimensions such as micrometers. One bit of data in the one-dimensional data corresponds to one pixel, and each pixel has a corresponding physical size (which can be obtained by measuring the length of the image).

[0158] For example, assuming the one-dimensional data is [11110011110000111011101110111101], the continuous pixel length and the frequency corresponding to each continuous pixel length can be counted as follows: the frequency corresponding to the continuous pixel length of 4 pixels is 3 times, the frequency of the continuous pixel length of 3 pixels appears 4 times, and the frequency of the continuous pixel length of 1 pixel appears 1 time; the continuous length can also be the continuous physical length. At this time, according to the corresponding physical size of each pixel, the continuous pixel length can be converted into the continuous physical length. Assuming that the corresponding physical size of each pixel is a micrometers, the continuous physical length and the frequency corresponding to each continuous physical length are counted according to the above one-dimensional data as follows: the frequency of the continuous physical length of 4a micrometers is 3 times, the frequency of the continuous physical length of 5a micrometers appears 4 times, and the frequency of the continuous physical length of 1a micrometers appears 1 time. After counting the frequency corresponding to each continuous length, a histogram of the frequency and the continuous length can be generated.

[0159] In a possible implementation, after generating the histogram, an exponential Gaussian distribution function can be used to fit the histogram to obtain a fitted exponential Gaussian distribution function. The fitted exponential Gaussian distribution function corresponds to three parameters μ, σ, and τ. In this implementation, the exponential decay parameter of the test sample and the exponential decay parameter T of the control sample can be compared to calculate the aggregation index of the test sample.

[0160] In a possible implementation, calculating the aggregation index of the test sample according to the exponential decay parameter of the test sample and the exponential decay parameter of the control sample includes:

[0161] Calculating the aggregation index A of the test sample according to the following formula:

[0162]

[0163] where Tau sample is the exponential decay parameter of the control sample, and Tau control is the exponential decay parameter of the test sample.

[0164] In a possible implementation, calibrating the brightness of the Gaussian blurred image includes:

[0165] Calibrating the brightness of the Gaussian blurred image of the control sample to obtain the control brightness of the Gaussian blurred image of the control sample after brightness calibration;

[0166] Calibrating the brightness of the Gaussian blurred image of the test sample according to the control brightness so that the difference between the brightness of the Gaussian blurred image of the test sample after brightness calibration and the control brightness is within a predetermined range.

[0167] In this implementation, it is necessary to calibrate the image brightness of the Gaussian blurred images corresponding to the control sample and the test sample to ensure the consistency in brightness between the images corresponding to the control sample and the test sample, so that the image data is comparable.

[0168] In this implementation, the brightness of the Gaussian blurred image of the control sample can be calibrated to a predetermined brightness range according to a preset brightness range, and then the Gaussian blurred image of the test sample can be calibrated according to the control brightness of the calibrated Gaussian blurred image of the control sample, so that the difference between the brightness of the calibrated test sample and the control brightness is within a predetermined range. For example, the predetermined range is less than or equal to 20, etc.

[0169] It should be noted here that the original chemical image has a black background color. When performing Gaussian blur, brightness calibration, and binarization on the chemical image, in order to ensure better visibility, black and white inversion can be performed to make the background color white. In the binarized image, black pixels represent the analyte. At this time, if the grayscale image is an 8-bit grayscale image (the brightness range of an 8-bit grayscale image is 0 - 255), the preset brightness range can be between 10 and 254. Preferably, the brightness range can be between 60 and 254. This range is selected to ensure that the image after brightness calibration contains sufficient features while filtering out unnecessary background noise generated by the chemical image detector. If the grayscale image is a 64-bit grayscale image (the brightness range of an 8-bit grayscale image is 0 - 65535), the preset brightness range can be correspondingly scaled to be between 2570 and 65278, preferably between 15420 and 65278. Of course, if it is a grayscale image of other bits, the brightness range can be other values, which will not be exemplified one by one here. If black and white inversion is not performed and the background color becomes black, and white pixels in the binarized image represent the analyte, then the preset brightness range will be lower.

[0170] In a possible implementation, the difference between the brightness of the calibrated test sample and the control brightness is less than or equal to 10% of the control brightness. For example, if the control brightness is 157, then the brightness of the Gaussian blurred image of the calibrated test sample needs to be within [157 - 157×10%, 157 + 157×10%], that is, within [141.3, 172.7].

[0171] In a possible implementation, the conversion of the binarized image into a one-dimensional array includes:

[0172] Traverse each pixel of the binarized image in a raster pattern and convert the binarized image into a one-dimensional array.

[0173] In this implementation, the dispersion of particulate matter, especially in the cosmetic field, usually involves high-shear and high-energy dispersion processes. These processes can effectively break up aggregates of the high aspect ratio type. Therefore, after the complete processing sequence, the particulate matter remaining in the formulation will only contain aggregates with an aspect ratio close to 1. When the aspect ratio is close to 1, the size of the particulate matter is the same in any direction.

[0174] In addition, these analytes are randomly distributed in the sample and thus exist in random directions. The analysis method provided by the present disclosure, when analyzing the entire chemical image, usually processes and quantifies thousands of particulate matter analytes. The randomness of the analyte distribution and the large number of analyte particles eliminate the adverse effects brought by the geometric shape / direction / aspect ratio of the analytes.

[0175] Therefore, whether rasterization is performed in the horizontal direction, vertical direction, or even any direction, the resulting exponential decay parameters are similar.

[0176] In this embodiment, the raster pattern refers to the scanning path along which pixels in an image are scanned or sampled in image processing. The scanning path of this raster pattern is usually a path that scans from left to right and from top to bottom. By using the raster pattern, each pixel of the binary image can be scanned along the path from left to right and from top to bottom, and a one-dimensional array is used to record the values (such as 0 or 1) corresponding to the scanned pixels. Of course, this raster pattern can also be scanned along the path from top to bottom and from left to right, or scanned sequentially from any other angle.

[0177] Figure 2 The structural block diagram of an analysis system for analyzing the aggregation of an analyte in a chemical composition according to an embodiment of the present disclosure is shown. Among them, this device can be implemented as part or all of an electronic device through software, hardware, or a combination of both. As Figure 2 shown, the analysis system 200 includes an image acquisition module 201, a Gaussian blur module 202, a brightness calibration module 203, a binary conversion module 204, a fitting module 205, and a calculation module 206.

[0178] The image acquisition module 201 is configured to acquire a chemical image of the analyte in a control sample and chemical images of the analyte in at least one test sample, where the control sample and the test sample are chemical compositions;

[0179] The Gaussian blur module 202 is configured to perform Gaussian blur on each chemical image using a preset Gaussian kernel to obtain a Gaussian blurred image;

[0180] The brightness calibration module 203 is configured to perform brightness calibration on the Gaussian blurred image;

[0181] The binary conversion module 204 is configured to convert the Gaussian blurred image after brightness calibration into a binary image;

[0182] The fitting module 205 is configured to convert the binary image into a one-dimensional array; count the consecutive lengths when a predetermined pixel appears in the one-dimensional array and the frequency corresponding to each consecutive length, generate a histogram of the frequency and the consecutive length; use an exponential Gaussian distribution function to fit the histogram to obtain the exponential decay parameter of the fitted exponential Gaussian distribution function;

[0183] The calculation module 206 is configured to calculate the aggregation index of the test sample according to the exponential decay parameter corresponding to the test sample and the exponential decay parameter corresponding to the control sample.

[0184] In a possible implementation, the analysis system provided in this embodiment is mainly used to analyze the aggregation of analytes in a chemical composition, which can be a chemical composition in various fields such as personal care products, pharmaceuticals, and foods, and can be various non-solid compositions such as liquids, colloids, powders, and semi-solids.

[0185] In a possible implementation, both the control sample and the test sample are chemical compositions. The same analyte is contained in both the control sample and the test sample. The distribution of the analyte in the control sample is relatively uniform. Therefore, based on the control sample, the aggregation of the analyte in one or more test samples can be analyzed.

[0186] In a possible implementation, the image acquisition module 201 can be a chemical imaging device. By separately photographing the analyte in the control sample and the analyte in the test sample, corresponding chemical images are obtained. Alternatively, the image acquisition module 201 can obtain the corresponding chemical images taken by the chemical imaging device. A chemical imaging device refers to a device that can obtain chemical images using chemical imaging technology. Preferably, the chemical image in this embodiment can be an SEM-EDX image.

[0187] In a possible implementation, whether it is the chemical image of the control sample or the chemical image of any test sample, the Gaussian blur module 202 performs Gaussian blur on each chemical image using a preset Gaussian kernel to obtain a Gaussian blurred image. The Gaussian kernel can be preset according to the required degree of blur. Preferably, the kernel size of the Gaussian kernel is greater than or equal to 3×3 and less than 61×61.

[0188] In a possible implementation, the standard deviation σ of the Gaussian kernel can be determined by the kernel size. The value of the standard deviation σ determines the distribution method of the weights within the kernel. When the value of σ is low, the Gaussian distribution is narrow, which means that most of the weights are concentrated near the center of the kernel. Therefore, the pixels near the center of the kernel have a greater impact on the final blurred pixel value, while the pixels far from the center of the kernel have little impact. If the value of σ is too low relative to the kernel size, it may lead to low utilization efficiency of the outer-layer pixels of the kernel because their weights are very small. When the value of σ is high, the Gaussian distribution is wide, and the weights are more evenly distributed throughout the kernel. The pixels far from the center of the kernel have a greater impact on the generated blurred pixels. This results in a stronger blur effect, with more extensive smoothing of both details and larger structures. If the value of σ is too high for the kernel size, the weights may become too uniform, making the behavior of the kernel similar to that of a simple mean filter (average filter).

[0189] Therefore, preferably, the standard deviation σ value can be generated through the following formula: σ = 0.3×(kernel size / 2 - 1) + 0.8; for example, when the kernel size is 18×18, the standard deviation σ of the preset Gaussian kernel = 0.3×(18 / 2 - 1) + 0.8 = 3.2. The standard deviation σ value calculated in this way can ensure that the Gaussian distribution is well represented within the kernel, thus achieving a balanced and effective blurring effect.

[0190] In another possible implementation, the standard deviation σ of the preset Gaussian kernel can also be selected within a predetermined range. Preferably, the standard deviation σ of the preset Gaussian kernel is greater than or equal to 2% of the kernel size and less than or equal to 49% of the kernel size. For example, when the kernel size is 30×30, the standard deviation σ of the preset Gaussian kernel is greater than or equal to 2% of 30 and less than or equal to 49% of 30. In this way, the standard deviation σ value of the Gaussian kernel is restricted within a certain range, which can effectively blur the image and ensure that the blurred image is sufficiently suitable for subsequent analysis.

[0191] In one possible implementation, the chemical image can be a grayscale chemical image. A grayscale image is a special image that only contains grayscale (light and dark) information and no color information. Alternatively, the chemical image can be a color chemical image. In this case, after obtaining the Gaussian blurred image, it is necessary to convert the Gaussian blurred image from a color image to a grayscale image.

[0192] Preferably, the grayscale image in the present disclosure is an 8-bit grayscale image. Here, "8-bit" means that the pixel value in the grayscale image is represented by 8-bit binary numbers, and there are 2 8 = 256 possibilities. The pixel value of each pixel is an integer between 0 (black) and 255 (white). Of course, the grayscale image in the present disclosure is not limited to an 8-bit grayscale image. As long as it can be achieved, the grayscale image can also be a grayscale image of any bit, such as a 1-bit grayscale image, a 2-bit grayscale image, a 4-bit grayscale image, or a 16-bit grayscale image, etc. Taking a 16-bit grayscale image as an example, the pixel value of the 16-bit grayscale image is represented by 16-bit binary numbers, and each pixel has 2 蹈 possibilities. Therefore, the range of grayscale values is 0 - 65535.

[0193] In one possible implementation, the brightness calibration module 203 can perform brightness calibration on the Gaussian blurred image so that the image brightness of the Gaussian blurred images of the control sample and the test sample is at the same level, so as to subsequently analyze the aggregation situation of the analyte in the test sample based on the control sample at the same brightness level.

[0194] In a possible implementation, the binary conversion module 204 can adopt a binarization technique to convert the Gaussian blurred image after brightness calibration into a binary image. Binarization refers to the process of converting a grayscale image into an image that contains only two colors (usually black and white). For example, a threshold can be set, and all pixels above this threshold are set to one color (usually white), while all pixels below the threshold are set to another color (usually black). Taking an 8-bit grayscale image as an example, the grayscale value range of each pixel in the 8-bit grayscale image is 0 to 255. Binarization refers to the process of changing the grayscale value of each pixel in the pixel matrix of the grayscale image to 0 (black) or 255 (white), that is, the grayscale value in the binary image is only 0 or 255, and the entire binary image only appears black and white.

[0195] In a possible implementation, after obtaining the binary image, the fitting module 205 can traverse the pixels in the binary image in a predetermined order, such as row by row or column by column, and convert the binary image into a one-dimensional array. There are two values in the converted one-dimensional array representing the black and white pixels in the binary image. Effectively, during the conversion, the pixels representing the presence of the analyte in the binary image can be converted into the value 1 in the one-dimensional array, while the pixels without the analyte are converted into the value 0 in the one-dimensional array. Of course, it is also possible to convert the pixels representing the presence of the analyte in the binary image into the value 0 in the one-dimensional array, while the pixels without the analyte are converted into the value 1 in the one-dimensional array. In this implementation, the former numerical representation method is selected to ensure better visibility of the data.

[0196] In a possible implementation, the predetermined pixel is the pixel representing the presence of the analyte. For example, when converting the pixels representing the presence of the analyte in the binary image into the value 1 in the one-dimensional array and the pixels without the analyte into the value 0 in the one-dimensional array, the fitting module 205 can count the continuous length when the value 1 appears in the one-dimensional array and the frequency corresponding to each continuous length.

[0197] The continuous length here can be the continuous pixel length or the continuous physical length. The continuous pixel length is the length in units of pixels, and the continuous physical length is the length in physical dimensions such as micrometers. One bit of data in the one-dimensional data corresponds to one pixel, and each pixel has a corresponding physical size (which can be obtained by measuring the length of the image).

[0198] In a possible implementation, after generating the histogram, the fitting module 205 can fit the histogram using an exponential Gaussian distribution function to obtain a fitted exponential Gaussian distribution function. The fitted exponential Gaussian distribution function corresponds to three parameters μ, σ, and τ. In this implementation, only the exponential decay parameter T is required. The calculation module 206 can compare the exponential decay parameter of the test sample with that of the control sample to calculate the aggregation index of the test sample.

[0199] In a possible implementation, the calculation module 206 is specifically configured to:

[0200] Calculate the aggregation index A of the test sample according to the following formula:

[0201]

[0202] where Tau sample is the exponential decay parameter of the control sample, and Tau control is the exponential decay parameter of the test sample.

[0203] In a possible implementation, the brightness calibration module 203 is specifically configured to:

[0204] A first calibration sub-module, configured to perform brightness calibration on the Gaussian blurred image of the control sample to obtain the control brightness of the Gaussian blurred image of the control sample after brightness calibration;

[0205] A second calibration sub-module, configured to perform brightness calibration on the Gaussian blurred image of the test sample according to the control brightness, so that the difference between the brightness of the Gaussian blurred image of the test sample after brightness calibration and the control brightness is within a predetermined range.

[0206] In this implementation, it is necessary to calibrate the image brightness of the Gaussian blurred images corresponding to the control sample and the test sample to ensure the brightness consistency between the images corresponding to the control sample and the test sample, so that the image data is comparable.

[0207] In this implementation, the first calibration sub-module can, according to a preset

[0208] range, calibrate the brightness of the Gaussian blurred image of the control sample to a predetermined brightness range, and then calibrate the Gaussian blurred image of the test sample according to the control brightness of the calibrated Gaussian blurred image of the control sample, so that the difference between the calibrated brightness of the test sample and the control brightness is within a predetermined range. For example, the predetermined range is less than or equal to 20, etc.

[0209] It should be noted here that the original chemical image has a black background color. When performing Gaussian blur, brightness calibration, and binarization on the chemical image, in order to ensure better visibility, black-and-white inversion can be performed so that black pixels in the binarized image represent the analyte. At this time, if the grayscale image is an 8-bit grayscale image (the brightness range of an 8-bit grayscale image is 0 - 255), then the preset brightness range can be between 10 and 254. Preferably, the brightness range can be between 60 and 254. This range is selected to ensure that the image after brightness calibration contains sufficient features while filtering out unnecessary background noise generated by the chemical image detector; if the grayscale image is a 64-bit grayscale image (the brightness range of an 8-bit grayscale image is 0 - 65535), the preset brightness range can be proportionally changed to between 2570 and 65278, preferably between 15420 and 65278; of course, if it is a grayscale image of other bits, the brightness range can be other values, which will not be exemplified one by one here. If black-and-white inversion is not performed and white pixels in the binarized image represent the analyte, then the preset brightness range will be lower.

[0210] In a possible implementation manner, the difference between the brightness of the calibrated test sample and the control brightness is less than or equal to 10% of the control brightness. For example, if the control brightness is 157, then the brightness of the Gaussian blurred image of the calibrated test sample needs to be within [157 - 157×10%, 157 + 157×10%], that is, within [141.3, 172.7].

[0211] In a possible implementation manner, the part in the fitting module 205 that converts the binarized image into a one-dimensional array is configured as:

[0212] Traverse each pixel of the binarized image in raster mode and convert the binarized image into a one-dimensional array.

[0213] In this implementation manner, the raster mode refers to the scanning path in image processing where pixels in the image are scanned or sampled. The scanning path of this raster mode is usually a scanning path from left to right and from top to bottom. By using the raster mode, each pixel of the binarized image can be scanned along the scanning path from left to right and from top to bottom, and the values corresponding to the scanned pixels (such as 0 or 1) are recorded using a one-dimensional array. Of course, this raster mode can also be scanned along the scanning path from top to bottom and from left to right, or scanned sequentially from any other angle.

[0214] The present disclosure also provides a calculation method for the aggregation situation of analytes in a chemical composition. Figure 3 The flowchart showing a calculation method for the aggregation situation of analytes in a chemical composition provided by an embodiment of the present disclosure is as follows. Figure 3As shown, the calculation method includes the following steps S301 - S307:

[0215] In step S301, obtain the chemical image of the analyte in the test sample, where the test sample is a chemical composition;

[0216] In step S302, perform Gaussian blur on the chemical image using a preset Gaussian kernel to obtain a Gaussian blurred image;

[0217] In step S303, perform brightness calibration on the Gaussian blurred image;

[0218] In step S304, convert the brightness - calibrated Gaussian blurred image into a binary image;

[0219] In step S305, convert the binary image into a one - dimensional array;

[0220] In step S306, count the continuous length when a predetermined pixel appears in the one - dimensional array and the frequency corresponding to each continuous length, and generate a histogram of frequency and continuous length;

[0221] In step S307, fit the histogram through an exponential Gaussian distribution function to determine the average value and standard deviation in terms of physical size of the test sample.

[0222] In a possible implementation, the calculation method provided in this implementation is mainly used for quantitatively calculating the aggregation situation of the analyte in the chemical composition. The chemical composition can be a chemical composition in various fields such as personal care products, pharmaceuticals, and foods. The chemical composition can be in various forms such as liquid, colloid, powder, semi - solid, etc. The analyte is the particulate matter in the chemical composition.

[0223] In a possible implementation, the process of processing the chemical image of the analyte in the test sample in this implementation is similar to the process described in the analysis method of the aggregation situation of the analyte in the above - mentioned chemical composition. For the interpretation and description of the technical terms and technical features involved in this implementation, reference can be made to the interpretation and description of the above - mentioned method implementation, which will not be elaborated here.

[0224] In a possible implementation, after obtaining the one - dimensional array, when counting the continuous length when a predetermined pixel appears in the one - dimensional array and the frequency corresponding to each continuous length, there are the following two situations:

[0225] One case is that: the continuous length to be counted is the continuous pixel length. For example, assuming the one-dimensional data is [11110011110000111011101110111101], the continuous pixel length and the frequency corresponding to each continuous pixel length can be counted as follows: the frequency corresponding to the continuous pixel length of 4 pixels is 3 times, the frequency of the continuous pixel length of 3 pixels is 4, and the frequency of the continuous pixel length of 1 pixel is 1 time. At this time, by fitting the histogram of the continuous pixel length and the frequency with an exponential Gaussian distribution function, the mean and standard deviation of the fitted exponential Gaussian distribution function are the mean and standard deviation corresponding to the continuous pixel length. In order to accurately characterize the physical size and its distribution of the analyte, according to the corresponding relationship between the pixel and the physical size, the mean and standard deviation of the fitted exponential Gaussian distribution function can be converted into the mean and standard deviation corresponding to the physical size. For example, if the corresponding relationship between the pixel and the physical size is that each pixel has a corresponding physical size of a microns, then the mean μ pixels and standard deviation σ pixels of the fitted exponential Gaussian distribution function are converted into the mean μ * a microns corresponding to the physical size and the standard deviation σ * a microns corresponding to the physical size.

[0226] Another case is that: the continuous length to be counted is the continuous physical length. At this time, the continuous pixel length when a predetermined pixel appears in the one-dimensional array can be counted; according to the corresponding relationship between the pixel and the physical size, the continuous pixel length is converted into the continuous physical length. For example, assuming the one-dimensional data is [11110011110000111011101110111101], the continuous pixel length and the frequency corresponding to each continuous pixel length can be counted as follows: the frequency corresponding to the continuous pixel length of 4 pixels is 3 times, the frequency of the continuous pixel length of 3 pixels is 4, and the frequency of the continuous pixel length of 1 pixel is 1 time; then, according to the fact that each pixel has a corresponding physical size, the continuous pixel length can be converted into the continuous physical length. Assuming that each pixel has a corresponding physical size of a microns, the continuous physical length and the frequency corresponding to each continuous physical length are counted according to the above one-dimensional data as follows: the frequency of the continuous physical length of 4a microns is 3 times, the frequency of the continuous physical length of 5a microns is 4, and the frequency of the continuous physical length of 1a microns is 1 time. After counting the frequency corresponding to each continuous physical length, a histogram of the frequency and the continuous physical length can be generated. By fitting the histogram of the frequency and the continuous physical length with an exponential Gaussian distribution function, the standard deviation and mean of the exponential Gaussian distribution function are the mean and standard deviation of the analyte in the physical size of the test sample.

[0227] The present disclosure also provides a calculation system for the aggregation situation of an analyte in a chemical composition. Figure 4The structural block diagram of a computing system for calculating the aggregation situation of analytes in a chemical composition provided by an embodiment of the present disclosure is shown. As Figure 4 shown, the computing system 400 includes:

[0228] An acquisition module 401, configured to acquire a chemical image of an analyte in a test sample, where the test sample is a chemical composition;

[0229] A blur module 402, configured to perform Gaussian blur on the chemical image using a preset Gaussian kernel to obtain a Gaussian blurred image;

[0230] A calibration module 403, configured to perform brightness calibration on the Gaussian blurred image;

[0231] A conversion module 404, configured to convert the brightness-calibrated Gaussian blurred image into a binary image; and convert the binary image into a one-dimensional array;

[0232] A statistics module 405, configured to count the continuous length when a predetermined pixel appears in the one-dimensional array and the frequency corresponding to each continuous length, and generate a histogram of frequency and continuous length;

[0233] A physical size module 406, configured to determine the average value and standard deviation of the physical size of the analyte in the test sample by fitting the histogram with an exponential Gaussian distribution function.

[0234] In a possible implementation manner, the continuous length is the continuous pixel length, and the physical size module 406 is configured to:

[0235] Fit the histogram with an exponential Gaussian distribution function to obtain the average value and standard deviation of the fitted exponential Gaussian distribution function;

[0236] According to the correspondence between pixels and physical sizes, convert the average value and standard deviation of the fitted exponential Gaussian distribution function into the average value and standard deviation corresponding to the physical size.

[0237] In a possible implementation manner, the continuous length is the continuous physical length, and the statistics module 405 is configured to:

[0238] Count the continuous pixel length when a predetermined pixel appears in the one-dimensional array;

[0239] According to the correspondence between pixels and physical sizes, convert the continuous pixel length into a continuous physical length.

[0240] The technical terms and technical features mentioned in the implementation of this system are the same as or similar to those mentioned in the above method implementation. For the explanations and descriptions of the technical terms and technical features involved in this system, reference can be made to the explanations and descriptions of the above method implementation, which will not be elaborated here.

[0241] The analysis scheme for the aggregation situation of analytes in the chemical composition provided by this disclosure can quantitatively analyze the aggregation situation of analytes in the chemical composition, which cannot be achieved based on the existing scheme. The following will be illustrated with multiple examples and comparative examples.

[0242] First, use Example 1 to prove that the analysis scheme of this disclosure can quantitatively analyze the aggregation situation of analytes in the chemical composition.

[0243] Example 1:

[0244] To prove the analysis method provided by this disclosure, the test samples can be distinguished by the obvious aggregation differences generated by different processing techniques, and a quantitative aggregation index can be calculated for the test samples.

[0245] The analyte is iron. There is one control sample and two test samples. These three samples have exactly the same composition, but they are prepared according to different dispersion techniques:

[0246] Control sample: The raw material is directly dispersed in oil (without further dispersion treatment);

[0247] Test sample 1: After the raw material is dispersed in oil, it is dispersed using moderate shear energy;

[0248] Test sample 2: After the raw material is dispersed in oil, it is dispersed using a high-energy dispersion technique (which can disperse the raw material better compared to moderate shear energy).

[0249] Obtain the chemical images corresponding to these three samples, Figure 5 The illustration showing the chemical images corresponding to the three samples provided in Example 1, where, Figure 5 the upper image in is the chemical image corresponding to the control sample, Figure 5 the middle image in is the chemical image corresponding to Test sample 1, Figure 5 the lower image in is the chemical image corresponding to Test sample 2. Based only on Figure 5 the chemical images shown in, it is easy to observe that the dispersion of iron in the control sample is poor, while Test sample 1 and Test sample 2 show better dispersion compared to the control sample. In addition, Test sample 2 has a better improvement in dispersion compared to Test sample 1. Nevertheless, except for these qualitative judgments, no more information can be obtained.

[0250] Based on the method provided by the present disclosure, first, a Gaussian blur is performed on the chemical image using a Gaussian kernel with a kernel size of 3×3 and a standard deviation σ = 0.3×(kernel size / 2 - 1) + 0.8 = 0.3×(3 / 2 - 1) + 0.8 = 0.95, and the Gaussian blurred image is converted from a color image to a grayscale image; wherein, Figure 6 A diagram showing the grayscale images of the Gaussian blurred images corresponding to the three samples provided in Example 1, wherein, Figure 6 The upper image in [the diagram] is the grayscale image of the Gaussian blurred image corresponding to the control sample, Figure 6 The middle image in [the diagram] is the grayscale image of the Gaussian blurred image corresponding to Test Sample 1, Figure 6 The lower image in [the diagram] is the grayscale image of the Gaussian blurred image corresponding to Test Sample 2.

[0251] Then, the grayscale images of the Gaussian blurred images of these three samples are calibrated to the same level of brightness. Since all the test samples have the same composition, the grayscale images can be calibrated to the same level of brightness. The brightness of the grayscale image of the Gaussian blurred image of the control sample can be calibrated to 147 first. Taking 147 as a reference, the brightness of the grayscale image of the Gaussian blurred image of Test Sample 2 compared to that of Test Sample 1 can be calibrated to 150 and 154.

[0252] Then, the brightness-calibrated Gaussian blurred image is converted into a binary image, wherein, Figure 7 A diagram showing the binary images corresponding to the three samples provided in Example 1, wherein, Figure 7 The upper image in [the diagram] is the binary image corresponding to the control sample, Figure 7 The middle image in [the diagram] is the binary image corresponding to Test Sample 1, Figure 7 The lower image in [the diagram] is the binary image corresponding to Test Sample 2.

[0253] Subsequently, the above binary image is converted into a one-dimensional array (1 in the array represents the black pixels in the binary image, and 0 represents the white pixels in the binary image). The continuous pixel lengths when the black pixels appear in the one-dimensional array and the frequencies corresponding to each continuous pixel length are counted, a histogram of the frequency and the continuous pixel length is generated, and the histogram is fitted using an exponential Gaussian distribution function. Figure 8 A diagram showing the fitting curve graphs of the exponential Gaussian distribution functions corresponding to the three samples provided in Example 1, Figure 8 The upper image in [the diagram] is the fitting curve graph of the exponential Gaussian distribution function corresponding to the control sample, Figure 8 The middle image in [the diagram] is the fitting curve graph of the exponential Gaussian distribution function corresponding to Test Sample 1, Figure 8 The lower image in [the diagram] is the fitting curve graph of the exponential Gaussian distribution function corresponding to Test Sample 2. Figure 8The abscissa of each graph is the continuous pixel length, with the unit of pixel, and the ordinate is the normalized count, which is the count of the occurrence frequency of the continuous pixel length; the blue data points in the graph are the occurrence frequencies of each continuous pixel length corresponding to the statistically sampled samples, and the red line in the graph is the curve of the fitted exponential Gaussian distribution function.

[0254] All three parameters of the exponential Gaussian distribution function corresponding to these three samples are shown in Table 1 below:

[0255] μ σ τ <![CDATA[R 2 > Control sample 34.7 18.8 13.1 0.98 Test sample 1 24.7 12.1， 9.0 0.99 Test sample 2 20.8 9.0 7.1 0.99

[0256] Table 1

[0257] Note that although all three parameters of the exponential Gaussian distribution function can be obtained, only the τ (Tau) value needs to be obtained in this disclosure. The R 2 value is used to measure the proportion of the variance in the dependent variable that can be predicted from the independent variable; the R 2 value ranges from 0 to 1, where 1 represents a perfect fit and 0 means no variance in the data is explained. The R 2 value listed in Table 1 indicates that the curve fitted in this Example 1 has a very high goodness of fit.

[0258] Based on the τ (Tau) value of the exponential decay parameter obtained from the fitting, it can be easily quantified as follows:

[0259] Test sample 1: Aggregation index = (9.0 / 13.1 - 1) * 100% = -31%;

[0260] Test sample 2: Aggregation index = (7.1 / 13.1 - 1) * 100% = -46%.

[0261] It can be seen from the calculated aggregation index that compared with the control sample, the aggregation index of test sample 1 decreased by 31% and is more dispersed; compared with the control sample, the aggregation index of test sample 2 decreased by 46% and is further dispersed; compared with test sample 1, the aggregation index of test sample 2 decreased by 15%, and test sample 2 is more dispersed than test sample 1. Thus, the aggregation situation of the analyte in the test sample is correctly quantitatively reflected by this aggregation index.

[0262] Second, use Example 2.0, Example 2.1, and Example 2.2 to prove that different kernel sizes have little impact on the calculation results of the aggregation index, and appropriate results can be calculated.

[0263] Example 2.0:

[0264] The control sample is a cosmetic emulsion prepared on a laboratory scale, and the test sample is a cosmetic emulsion prepared on a commercial production scale; whether these two production scales will lead to different aggregation situations of particulate matter is of particular interest to process engineers.

[0265] The analyte is aluminum; although aluminum is not typically a directly used ingredient in cosmetic formulations, aluminum hydroxide is commonly present on the surface of certain types of pigments (including titanium dioxide and iron oxide). Therefore, using aluminum as the analyte can effectively evaluate the aggregation of all particulate matter in cosmetic emulsions.

[0266] Obtain the chemical images corresponding to these two samples. Figure 9 A diagram showing the chemical images corresponding to the two samples provided in Example 2.0, where Figure 9 the upper image in is the chemical image corresponding to the control sample, Figure 9 and the lower image in is the chemical image corresponding to the test sample. Based solely on Figure 9 the chemical images shown in, it is difficult to observe which sample, the control or the test, has a worse aggregation of aluminum.

[0267] Based on the method provided in this disclosure, first perform Gaussian blur on the chemical images using a Gaussian kernel with a kernel size of 5×5 and a standard deviation σ = 0.3×(5 / 2 - 1) + 0.8 = 0.3×(5 / 2 - 1) + 0.8 = 1.25, and convert the Gaussian-blurred image from a color image to a grayscale image; where Figure 10 A diagram showing the grayscale images of the Gaussian-blurred images corresponding to the two samples provided in Example 2.0, Figure 10 the upper image in is the grayscale image of the Gaussian-blurred image corresponding to the control sample, Figure 10 and the lower image in is the grayscale image of the Gaussian-blurred image corresponding to the test sample.

[0268] Then, calibrate the grayscale images of the Gaussian-blurred images of these two samples to the same level of brightness. Since the compositions of the two samples are the same, the grayscale images can be calibrated to the same level of brightness. First, calibrate the brightness of the grayscale image of the Gaussian-blurred image of the control sample to 207. Taking 207 as a reference, the brightness of the grayscale image of the Gaussian-blurred image of the test sample can be calibrated to 201. The difference of 6 between the two is less than or equal to 10% of the control brightness of the control sample, which is 20.7.

[0269] Then, convert the Gaussian-blurred image with calibrated brightness to a binary image, where Figure 11 A diagram showing the binary images corresponding to the two samples provided in Example 2.0, where Figure 11 the upper image in is the binary image corresponding to the control sample, Figure 11 and the lower image in is the binary image corresponding to the test sample.

[0270] Subsequently, the above binary image is converted into a one-dimensional array, the continuous length when black pixels appear in the one-dimensional array and the frequency corresponding to each continuous length are counted, a histogram of frequency and continuous length is generated, and the histogram is fitted using an exponential Gaussian distribution function. Figure 12 The illustration shows the fitting curve diagrams of the exponential Gaussian distribution functions corresponding to two samples provided by Example 2.0. Figure 12 The upper image in [reference] is the fitting curve diagram of the exponential Gaussian distribution function corresponding to the control sample. Figure 12 The lower image in [reference] is the fitting curve diagram of the exponential Gaussian distribution function corresponding to the test sample. The exponential decay parameter T and the fitting R of the exponential Gaussian distribution functions corresponding to these two samples 2 are shown in Table 2 below:

[0271] τ <![CDATA[R 2 > Control sample 6.6 0.99 Test sample 8.8 0.99

[0272] Table 2

[0273] The R values listed in Table 2 2 indicate that the curves fitted in this Example 2.0 have a very high goodness of fit. According to the obtained exponential decay parameter τ (Tau) values, it can be easily quantified that:

[0274] The aggregation index of the test sample = (8.8 / 6.6 - 1) * 100% = 33%;

[0275] From the calculated aggregation index, it can be seen that compared with the control sample prepared on a laboratory scale, the aggregation index of the test sample prepared on a large-scale commercial production scale has increased by 33%, and there is a 33% aggregation deterioration.

[0276] Example 2.1:

[0277] In Example 2.1, the control sample, test sample, and analyte are the same as those in Example 2.0, and the same chemical images are obtained. Only the kernel size of the Gaussian kernel during Gaussian blur is 15×15, and the standard deviation σ = 0.3×(kernel size / 2 - 1) + 0.8 = 0.3×(15 / 2 - 1) + 0.8. The Gaussian blurred image is converted from a color image to a grayscale image. Among them, Figure 13 The illustration shows the grayscale images of the Gaussian blurred images corresponding to two samples provided by Example 2.1. Among them, Figure 13 the upper image in [reference] is the grayscale image of the Gaussian blurred image corresponding to the control sample, Figure 13 and the lower image in [reference] is the grayscale image of the Gaussian blurred image corresponding to the test sample.

[0278] Then, calibrate the grayscale images of the Gaussian blurred images of these two samples to the same level of brightness. Since all the test samples have the same composition, the grayscale images can be calibrated to the same level of brightness. First, calibrate the brightness of the grayscale image of the Gaussian blurred image of the control sample to 243. Taking 243 as a reference, calibrate the brightness of the grayscale image of the Gaussian blurred image of the test sample to 232. The difference of 11 between the two is less than or equal to 10% of the control brightness 243 of the control sample, that is, 24.3.

[0279] Then, convert the Gaussian blurred image after brightness calibration into a binary image, where, Figure 14 The figure shows the binary images corresponding to the two samples provided in Example 2.1, where, Figure 14 The upper image in is the binary image corresponding to the control sample, Figure 14 The lower image in is the binary image corresponding to the test sample.

[0280] Subsequently, convert the above binary image into a one-dimensional array, count the continuous length when black pixels appear in the one-dimensional array and the frequency corresponding to each continuous length, generate a histogram of frequency and continuous length, and fit the histogram using an exponential Gaussian distribution function. Figure 15 The figure shows the fitting curve diagram of the exponential Gaussian distribution function corresponding to the two samples provided in Example 2.1. Figure 15 The upper image in is the fitting curve diagram of the exponential Gaussian distribution function corresponding to the control sample, Figure 15 The lower image in is the fitting curve diagram of the exponential Gaussian distribution function corresponding to the test sample. The exponential decay parameter T and the fitting R 2 values of the exponential Gaussian distribution functions corresponding to these two samples are shown in Table 3 below:

[0281] τ <![CDATA[R 2 > Control sample 6.9 0.96 Test sample 9.5 0.95

[0282] Table 3

[0283] The R 2 values listed in Table 3 indicate that the curves fitted in this Example 2.1 have a very high goodness of fit. According to the fitted exponential decay parameter τ (Tau) values, it can be easily quantified that:

[0284] The aggregation index of the test sample = (9.5 / 6.9 - 1) * 100% = 38%;

[0285] Due to the different kernel sizes of 5×5 and standard deviations during Gaussian blurring, the aggregation index calculated in Example 2.1 is different from that in Example 2.0, but the calculation results are still valid, indicating that the dispersion state of the test sample is worse than that of the control sample and the aggregation situation is more serious.

[0286] By comparing Example 2.0 and Example 2.1, it can be seen that different kernel sizes have little impact on the calculation results of the aggregation index, and appropriate results can be obtained in both cases.

[0287] In addition, it should be noted that different from the blurred image display after Gaussian blur under normal circumstances, in this Example 2.1, after Gaussian blur, brightness calibration and binary image processing are performed, Figure 14 the area where the analyte is located (i.e., the black pixel area) in the shown binary image is clearer than the original chemical image.

[0288] Example 2.2:

[0289] In Example 2.2, the control samples, test samples, and analytes are the same as those in Example 2.0, and the same chemical images are obtained. Only the kernel size of the Gaussian kernel during Gaussian blur is 21×21, and the standard deviation σ = 0.3×(kernel size / 2 - 1)+0.8 = 0.3×(21 / 2 - 1)+0.8. The Gaussian blurred image is converted from a color image to a grayscale image; among them, Figure 16 the figure shows the grayscale images of the Gaussian blurred images corresponding to the two samples provided in Example 2.2, Figure 16 the upper image in it is the grayscale image of the Gaussian blurred image corresponding to the control sample, Figure 16 and the lower image in it is the grayscale image of the Gaussian blurred image corresponding to the test sample.

[0290] Then, the grayscale images of the Gaussian blurred images of these two samples are calibrated to the same level of brightness. Since all test samples have the same composition, the grayscale images can be calibrated to the same level of brightness. The brightness of the grayscale image of the Gaussian blurred image of the control sample can be calibrated to 249 first. Taking 249 as a reference, the brightness of the grayscale image of the Gaussian blurred image of the test sample can be calibrated to 239. The difference of 10 between the two is less than or equal to 10% of the control brightness 249 of the control sample, that is, 24.9.

[0291] Then, the Gaussian blurred image after brightness calibration is converted into a binary image, among which, Figure 17 the figure shows the binary images corresponding to the two samples provided in Example 2.2, Figure 17 the upper image in it is the binary image corresponding to the control sample, Figure 17 and the lower image in it is the binary image corresponding to the test sample.

[0292] Subsequently, the above binary image is converted into a one-dimensional array, the continuous length when black pixels appear in the one-dimensional array and the frequency corresponding to each continuous length are counted, a histogram of frequency and continuous length is generated, and the histogram is fitted using an exponential Gaussian distribution function. Figure 18Illustration showing the fitting curve graphs of the exponential Gaussian distribution functions corresponding to the two samples provided in Example 2.2 Figure 18 The upper image in Figure 18 is the illustration of the fitting curve of the exponential Gaussian distribution function corresponding to the control sample, 2 and the lower image in

[0293] τ <![CDATA[R 2 > Control sample 6.7 0.94 Test sample 9.7 0.95

[0294] is the illustration of the fitting curve of the exponential Gaussian distribution function corresponding to the test sample. The exponential decay parameters T and fitting R

[0295] corresponding to these two samples are shown in Table 4 below: 2 The R

[0296] values listed in Table 4 indicate that the curves fitted in Example 2.1 have a very high goodness of fit. Based on the exponential decay parameter τ (Tau) values obtained from the fitting, it can be easily quantified that:

[0297] Aggregation index of the test sample = (9.7 / 6.7 - 1) * 100% = 45%;

[0298] Although this aggregation index is different from that in Example 2.0, the calculation result is still valid, indicating that the dispersion state of the test sample is worse and the aggregation is more serious than that of the control sample.

[0299] By comparing Example 2.0, Example 2.1, and Example 2.2, it can be seen that different core sizes have little effect on the calculation results of the aggregation index, and appropriate results can be calculated. Figure 17 In addition, it should be noted that different from the blurred image display after Gaussian blur under normal circumstances, in Example 2.2, after Gaussian blur, brightness calibration and binary image processing,

[0300] the region where the analyte is located (i.e., the black pixel region) in the binary image shown is clearer than the original chemical image.

[0301] Third, Examples 3.0, 3.1, 3.2, and 3.3 are used to prove that different standard deviations have little effect on the calculation results of the aggregation index, and appropriate results can be calculated.

[0302] The test sample in Example 3.0 is the test sample in Example 2.2. The same chemical image as the test sample in Example 2.2 is obtained. The core size of the Gaussian kernel during Gaussian blur is the same, both 21×21. Only the standard deviation σ = 25% * core size = 5.25. After converting the Gaussian blurred image from a color image to a grayscale image, brightness calibration, and binary processing, a binary image is obtained; among them,Figure 19 A diagram showing the binary image corresponding to the test sample provided by Example 3.0. Subsequently, the above binary image is converted into a one-dimensional array, the continuous length when black pixels appear in the one-dimensional array and the frequency corresponding to each continuous length are counted, a histogram of frequency and continuous length is generated, and the histogram is fitted using an exponential Gaussian distribution function. Figure 20 A diagram showing the fitting curve graph of the exponential Gaussian distribution function corresponding to the test sample provided by Example 3.0. Figure 19 It can be seen that the aggregation of the analyte is properly displayed in the binary image, and Figure 20 the exponential Gaussian distribution of the continuous length is correctly displayed, and appropriate results can be calculated therefrom.

[0303] Example 3.1:

[0304] The test sample of Example 3.1 is the test sample in Example 2.2. The same chemical image as the test sample in Example 2.2 is obtained. The kernel size of the Gaussian kernel during Gaussian blur is the same, both 21×21, only the standard deviation σ = 10% * kernel size = 2.1. After converting the Gaussian blurred image from a color image to a grayscale image, performing brightness calibration and binaryzation, a binary image is obtained; among them, Figure 21 A diagram showing the binary image corresponding to the test sample provided by Example 3.1. Subsequently, the above binary image is converted into a one-dimensional array, the continuous length when black pixels appear in the one-dimensional array and the frequency corresponding to each continuous length are counted, a histogram of frequency and continuous length is generated, and the histogram is fitted using an exponential Gaussian distribution function. Figure 22 A diagram showing the fitting curve graph of the exponential Gaussian distribution function corresponding to the test sample provided by Example 3.1. Figure 21 It can be seen that the aggregation of the analyte is properly displayed in the binary image, and Figure 22 the exponential Gaussian distribution of the continuous length is correctly displayed, and appropriate results can be calculated therefrom.

[0305] Example 3.2:

[0306] The test sample of Example 3.2 is the test sample in Example 2.2. The same chemical image as the test sample in Example 2.2 is obtained. The kernel size of the Gaussian kernel during Gaussian blur is the same, both 21×21, only the standard deviation σ = 5% * kernel size = 1.05. After converting the Gaussian blurred image from a color image to a grayscale image, performing brightness calibration and binaryzation, a binary image is obtained; among them, Figure 23Illustration showing the binary image corresponding to the test sample provided in Example 3.2. Subsequently, the above binary image is converted into a one-dimensional array, the continuous lengths when black pixels appear in the one-dimensional array and the frequencies corresponding to each continuous length are counted, a histogram of frequency versus continuous length is generated, and the histogram is fitted using an exponential Gaussian distribution function. Figure 24 Illustration showing the fitting curve graph of the exponential Gaussian distribution function corresponding to the test sample provided in Example 3.2. From Figure 22 it can be seen that the aggregation of the analyte is properly shown in the binary image, and Figure 23 the exponential Gaussian distribution of the continuous lengths is correctly shown, and appropriate results can be calculated therefrom.

[0307] Example 3.3:

[0308] The test sample of Example 3.3 is the test sample in Example 2.2. The same chemical image as the test sample in Example 2.2 is obtained. The kernel size of the Gaussian kernel during Gaussian blur is the same, both being 21×21, and only the standard deviation σ = 2% * kernel size = 0.42. After converting the Gaussian blurred image from a color image to a grayscale image, performing brightness calibration and binarization, a binary image is obtained; where Figure 25 Illustration showing the binary image corresponding to the test sample provided in Example 3.3. Subsequently, the above binary image is converted into a one-dimensional array, the continuous lengths when black pixels appear in the one-dimensional array and the frequencies corresponding to each continuous length are counted, a histogram of frequency versus continuous length is generated, and the histogram is fitted using an exponential Gaussian distribution function. Figure 26 Illustration showing the fitting curve graph of the exponential Gaussian distribution function corresponding to the test sample provided in Example 3.3. From Figure 25 it can be seen that the aggregation of the analyte is properly shown in the binary image, and Figure 26 the exponential Gaussian distribution of the continuous lengths is correctly shown, and appropriate results can be calculated therefrom.

[0309] Fourth, use Example 4 to prove that measuring the continuous pixel length in one dimension can analyze the aggregation of analytes in a two-dimensional chemical image.

[0310] Example 4:

[0311] Figure 27 Illustration showing a chemical image of titanium dioxide particulate matter provided in Example 4. For Figure 27 the chemical image shown, Figure 28 Illustration showing Figure 27 the grayscale images of the chemical image shown in three directions. Among them, Figure 28 the upper image is Figure 27 Illustration showing the grayscale image of the chemical image shown in the horizontal direction. Figure 28The upper image is rotated by 90 degrees to obtain Figure 28 the middle image, that is, Figure 27 the illustration of the grayscale image of the chemical image shown in the vertical direction. Rotate the Figure 28 middle image by 45 degrees to obtain Figure 28 the lower image, that is, Figure 27 the illustration of the grayscale image of the chemical image shown in the 45-degree direction.

[0312] For the Figure 28 three grayscale images, perform Gaussian blur processing using an 11×11 Gaussian kernel and a standard deviation σ = 0.3×(kernel size / 2 - 1) + 0.8. Calibrate the image brightness of the Gaussian blur image to 190 and perform binarization to obtain a binarized image. Figure 29 Shown is Figure 28 the illustration of the binarized images of the grayscale images in the three directions shown. All are raster scanned in the order from left to right and from top to bottom Figure 29 the binarized image shown. Convert the above binarized image into a one-dimensional array (1 in the array represents the black pixels in the binarized image, and 0 represents the white pixels in the binarized image). Statistically analyze the continuous length when black pixels appear in the one-dimensional array and the frequency corresponding to each continuous length, generate a histogram of frequency and continuous length, and fit the histogram using an exponential Gaussian distribution function. Figure 30 Shown is the illustration of the fitting curves of the exponential Gaussian distribution functions corresponding to the three directions. All three parameters of the exponential Gaussian distribution functions corresponding to the three directions are shown in Table 5 below:

[0313]

[0314]

[0315] Table 5

[0316] As can be seen from Table 5, very good fitting effects are shown for the rasterized data in each direction, and the decay parameter T (Tau) is almost the same, hardly affecting the calculation of the aggregation index. This proves that the present disclosure can analyze the aggregation situation of analytes in two-dimensional chemical images by measuring the one-dimensional continuous pixel length.

[0317] Fifth, compare Comparative Example 1 with Example 2.0 above to prove the necessity of the Gaussian blur step.

[0318] Comparative Example 1:

[0319] The control sample, test sample, and analyte are the same as those in Example 2.0. Obtain the same chemical image, but do not perform the Gaussian blur step to obtain the grayscale image of the chemical image. Figure 31Illustration showing the grayscale images corresponding to the two samples provided in Comparative Example 1, Figure 31 where the upper image is the grayscale image corresponding to the control sample, Figure 31 and the lower image is the grayscale image corresponding to the test sample.

[0320] Then, the grayscale images of these two samples are calibrated to the same level of brightness. Since all the test samples have the same composition, the grayscale images can be calibrated to the same level of brightness. The brightness of the grayscale image of the control sample can be calibrated to 203 first. Taking 203 as a reference, the brightness of the grayscale image of the test sample can be calibrated to 206. The difference of 3 between the two is less than or equal to 10% of the control brightness 203 of the control sample, that is, 20.3.

[0321] Then, the Gaussian blurred image after brightness calibration is converted into a binary image, where, Figure 32 Illustration showing the binary images corresponding to the two samples provided in Comparative Example 1, Figure 32 where the upper image is the binary image corresponding to the control sample, Figure 32 and the lower image is the binary image corresponding to the test sample.

[0322] Subsequently, the above binary image is converted into a one-dimensional array, the continuous length when black pixels appear in the one-dimensional array and the frequency corresponding to each continuous length are counted, a histogram of frequency vs. continuous length is generated, and the histogram is fitted using an exponential Gaussian distribution function. Figure 33 Illustration showing the fitting curve diagrams of the exponential Gaussian distribution functions corresponding to the two samples provided in Comparative Example 1, Figure 33 where the upper image is the fitting curve diagram of the exponential Gaussian distribution function corresponding to the control sample, Figure 33 and the lower image is the fitting curve diagram of the exponential Gaussian distribution function corresponding to the test sample. The exponential decay parameters T and fitting R 2 values of the exponential Gaussian distribution functions corresponding to these two samples are shown in Table 6 below:

[0323] τ <![CDATA[R 2 > Control sample 4.7 1.0 Test sample 4.8 0.98

[0324] Table 6

[0325] The R 2 values listed in Table 6 indicate that the curve fitting effect of the control sample in this Comparative Example 1 is very poor. According to the exponential decay parameter τ (Tau) values obtained by fitting, it can be easily quantified that:

[0326] Aggregation index of the test sample = (4.8 / 4.7 - 1) * 100% = 2%; Obviously, the difference of 2% greatly underestimates the difference between the two samples. Therefore, the Gaussian blur step was not performed and the analysis method could not proceed.

[0327] Sixth, Comparative Example 2 is used to prove that using traditional image processing methods cannot present a meaningful difference between the control sample and the test sample. Traditional image processing techniques, such as, most commonly, Fourier Transform (FT) or Fast Fourier Transform (FFT), can also be used to generate histograms, identify the transition from high-contrast regions to low-contrast regions, or identify regions with rich features (high frequencies) and regions mainly without features (low frequencies). Therefore, Fast Fourier Transform is used for image processing in Comparative Example 2.

[0328] Comparative Example 2:

[0329] This Comparative Example 2 is compared with Example 2.2.

[0330] Use FFT to transform the binary image obtained in Example 2.2 Figure 17 shown from the spatial domain to the frequency domain, generating a histogram of frequency and amplitude, Figure 34 showing a diagram of the FFT-processed chemical image histogram provided by Comparative Example 2, as Figure 34 shown, the two histograms of the control sample and the test sample look similar, and it can be identified from the histogram that there is no significant difference between the control sample and the test sample. And it has been well determined in Example 2.2 that the test sample shows more aggregation compared with the control sample. Therefore, the commonly used Fourier transform image analysis method cannot provide sufficient discrimination between the test sample and the control sample.

[0331] Use FFT to transform the Gaussian blurred image obtained in Example 2.2 Figure 16 shown from the spatial domain to the frequency domain, generating a histogram of frequency and amplitude, Figure 35 showing a diagram of the FFT-processed Gaussian blurred image histogram provided by Comparative Example 2, as Figure 35 shown, the two histograms of the control sample and the test sample look similar, and it can be identified from the histogram that there is no significant difference between the control sample and the control sample.

[0332] Convert the original chemical image obtained in Example 2.2, which is Figure 8 into a grayscale image, without Gaussian blurring, and directly use FFT to transform the image from the spatial domain to the frequency domain, generating a histogram of frequency and amplitude, Figure 36 showing a diagram of the FFT-processed grayscale image histogram provided by Comparative Example 2, as Figure 36 shown, the two histograms of the control sample and the test sample look similar, and it can be identified from the histogram that there is no significant difference between the control sample and the test sample.

[0333] ByFigure 35 and Figure 36 It can be seen that when only the grayscale image of the original chemical image or the grayscale image of Gaussian blur is used, the Fourier transform of the one-dimensional array generated by rasterizing the chemical image also fails to show a meaningful difference between the test sample and the control sample.

[0334] Therefore, it can be seen from Comparative Example 2 that it is impossible to quantify the aggregation of particulate matter by processing chemical images through Fourier transform.

[0335] Seventh, Example 5 proves that the calculation scheme for the aggregation of analytes in the chemical composition of the present disclosure can quantitatively calculate the aggregation of analytes in the chemical composition.

[0336] Example 5:

[0337] The sample analyzed in Example 5 is the control sample in Example 2.2, and the control sample in Example 2.2, the same chemical image, Gaussian blur image, binary image, and fitting curve are obtained. Figure 18 In the illustrated fitting curve diagram, the x-axis is the continuous pixel length. It can be measured that the conversion ratio between the pixel and the physical size is approximately 0.39 microns / pixel, and this conversion ratio can be determined by the magnification factor when taking the chemical image of the sample. Based on this conversion ratio, the Figure 18 continuous pixel length on the x-axis can be easily converted into a continuous physical length. Figure 37 shows a diagram of the fitting curve of the exponential Gaussian distribution function fitted by the continuous physical length and the histogram of the count of this continuous physical length provided in Example 5; Table 7 below shows Figure 37 the parameter values of the fitting curve of the exponential Gaussian distribution function shown:

[0338] μ [μm] σ [μm] Control sample 23 8.6

[0339] Table 7

[0340] The average value μ in Table 7 is used to represent that the average size of the analytes in this control sample is 23 microns, and the standard deviation σ is 8.6 microns, indicating the degree of distribution of the sizes of the analytes. Through this average value μ and standard deviation σ, the aggregation of analytes in this chemical composition can be quantitatively characterized.

[0341] The above description is only the preferred embodiments of the present disclosure and the description of the applied technical principles. Those skilled in the art should understand that the scope of the invention involved in the present disclosure is not limited to the technical solutions formed by the specific combination of the above technical features, and should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the inventive concept. For example, the technical solutions formed by mutually replacing the above features with the technical features (but not limited to) having similar functions disclosed in the present disclosure.

Claims

1. A method for analyzing the aggregation of an analyte in a chemical composition, comprising: acquiring a chemical image of an analyte in a control sample and a chemical image of the analyte in at least one test sample, the control sample and the test sample being a chemical composition; For each chemical image, performing Gaussian blurring on the chemical image using a preset Gaussian kernel to obtain a Gaussian blurred image; Performing brightness calibration on the Gaussian blurred image; Convert the brightness-calibrated Gaussian blurred image into a binary image; Convert the binary image into a one-dimensional array; Counting the continuous lengths when predetermined pixels appear in the one-dimensional array and the frequency corresponding to each continuous length, and generating a histogram of the frequency and the continuous length; Fitting the histogram using an exponential Gaussian distribution function to obtain an exponential decay parameter of the fitted exponential Gaussian distribution function; The aggregation index of the test sample is calculated according to the exponential decay parameter corresponding to the test sample and the exponential decay parameter corresponding to the control sample.

2. The analytical method according to claim 1, wherein: The kernel size of the preset Gaussian kernel is greater than or equal to 3×3 and less than 61×61.

3. The analysis method according to claim 1, wherein The calculation formula of the preset standard deviation σ of the Gaussian kernel includes: σ = 0.3 × (kernel size / 2-1) + 0.8; Alternatively, the standard deviation σ of the preset Gaussian kernel is greater than or equal to 2% and less than or equal to 49% of the kernel size.

4. The analysis method according to claim 1, wherein The step of performing brightness calibration on the Gaussian blurred image comprises: Performing brightness calibration on the Gaussian blurred image of the control sample to obtain a control brightness of the Gaussian blurred image of the control sample after brightness calibration; The Gaussian blurred image of the test sample is brightness calibrated according to the control brightness, so that the difference between the brightness of the Gaussian blurred image of the test sample after brightness calibration and the control brightness is within a predetermined range.

5. The analytical method according to claim 4, wherein: The predetermined range includes less than 100% or 10% of the control brightness.

6. The analysis method according to claim 1, wherein The step of converting the binary image into a one-dimensional array comprises: Each pixel of the binary image is traversed in raster mode to convert the binary image into a one-dimensional array.

7. The analytical method according to claim 1, wherein: The predetermined pixels are pixels representing the presence of the analyte.

8. The analysis method according to claim 1, wherein Calculating the aggregation index of the test sample according to the exponential decay parameter of the test sample and the exponential decay parameter of the control sample comprises: The aggregation index A of the test sample was calculated according to the following formula: Among them, Tau sample is the exponential decay parameter of the test sample, Tau control is the exponential decay parameter of the control sample.

9. The analysis method according to claim 1, wherein: The chemical images include scanning electron microscopy-energy dispersive X-ray spectroscopy SEM-EDX images.

10. The analysis method according to claim 1, wherein The chemical image is a grayscale chemical image.

11. The analysis method according to claim 1, wherein If the chemical image is a color image, before performing brightness calibration on the Gaussian blurred image, the method further includes: The Gaussian blurred image is converted from a color image to a grayscale image.

12. A system for analyzing the aggregation of analytes in a chemical composition, comprising: an image acquisition module configured to acquire a chemical image of an analyte in a control sample and a chemical image of the analyte in at least one test sample, the control sample and the test sample being a chemical composition; A Gaussian blur module is configured to perform Gaussian blur on each chemical image using a preset Gaussian kernel to obtain a Gaussian blurred image; A brightness calibration module, configured to perform brightness calibration on the Gaussian blurred image; A binary conversion module is configured to convert the Gaussian blurred image after brightness calibration into a binary image; A fitting module is configured to convert the binary image into a one-dimensional array; count the continuous lengths when predetermined pixels appear in the one-dimensional array and the frequency corresponding to each continuous length, and generate a histogram of the frequency and the continuous length; fit the histogram using an exponential Gaussian distribution function to obtain an exponential decay parameter of the fitted exponential Gaussian distribution function; The calculation module is configured to calculate the aggregation index of the test sample according to the exponential decay parameter corresponding to the test sample and the exponential decay parameter corresponding to the control sample.

13. The analysis system according to claim 12, wherein: The kernel size of the preset Gaussian kernel is greater than or equal to 3×3 and less than 61×61; and / or The calculation formula of the preset standard deviation σ of the Gaussian kernel includes: σ = 0.3 × (kernel size / 2-1) + 0.8; The standard deviation σ of the preset Gaussian kernel is greater than or equal to 2% and less than or equal to 49% of the kernel size; and / or The brightness calibration module is specifically configured as follows: A first calibration submodule is configured to perform brightness calibration on the Gaussian blurred image of the control sample to obtain a control brightness of the Gaussian blurred image of the control sample after brightness calibration; a second calibration submodule, configured to perform brightness calibration on the Gaussian blurred image of the test sample according to the control brightness, so that the difference between the brightness of the Gaussian blurred image of the test sample after brightness calibration and the control brightness is within a predetermined range, wherein the predetermined range includes less than 100% or 10% of the control brightness; and / or The part of the fitting module that converts the binary image into a one-dimensional array is configured as follows: Traversing each pixel of the binary image in a raster mode to convert the binary image into a one-dimensional array; and / or The predetermined pixel is a pixel representing the presence of the analyte; and / or The calculation module is specifically configured as follows: The aggregation index A of the test sample was calculated according to the following formula: Among them, Tau sample is the exponential decay parameter of the test sample, Tau control is the exponential decay parameter of the control sample; and / or The chemical image comprises a SEM-EDX image; and / or The chemical image is a grayscale chemical image; and / or If the chemical image is a color image, the analysis system further comprises: The color conversion module is configured to convert the Gaussian blurred image from a color image to a grayscale image before the brightness calibration module performs brightness calibration on the Gaussian blurred image.

14. A method for calculating the aggregation of an analyte in a chemical composition, comprising: acquiring a chemical image of an analyte in a test sample, the test sample being a chemical composition; Performing Gaussian blurring on the chemical image using a preset Gaussian kernel to obtain a Gaussian blurred image; Performing brightness calibration on the Gaussian blurred image; Convert the brightness-calibrated Gaussian blurred image into a binary image; Convert the binary image into a one-dimensional array: Counting the continuous lengths when predetermined pixels appear in the one-dimensional array and the frequency corresponding to each continuous length, and generating a histogram of the frequency and the continuous length; The mean and standard deviation of the physical dimensions in the test samples were determined by fitting the histogram with an exponential Gaussian distribution function.

15. The calculation method according to claim 14, wherein: The continuous length is a continuous pixel length, and fitting the histogram by an exponential Gaussian distribution function to determine the average value and standard deviation of the physical size of the test sample includes: Fitting the histogram by an exponential Gaussian distribution function to obtain an average value and a standard deviation of the fitted exponential Gaussian distribution function; According to the correspondence between pixels and physical dimensions, the mean value and standard deviation of the fitted exponential Gaussian distribution function are converted into the mean value and standard deviation corresponding to the physical dimensions.

16. The calculation method according to claim 14, wherein: The continuous length is a continuous physical length, and the counting of the continuous length when a predetermined pixel appears in the one-dimensional array includes: Counting the length of consecutive pixels when predetermined pixels appear in the one-dimensional array; According to the corresponding relationship between pixels and physical sizes, the continuous pixel lengths are converted into continuous physical lengths.

17. A system for calculating the aggregation of analytes in a chemical composition, comprising: an acquisition module configured to acquire a chemical image of an analyte in a test sample, the test sample being a chemical composition; A blur module is configured to perform Gaussian blur on the chemical image using a preset Gaussian kernel to obtain a Gaussian blurred image; A calibration module, configured to perform brightness calibration on the Gaussian blurred image; A conversion module, configured to convert the brightness-calibrated Gaussian blurred image into a binary image; A dimension conversion module, configured to convert the binary image into a one-dimensional array; A statistical module is configured to count the continuous lengths of the predetermined pixels in the one-dimensional array when they appear and the frequency corresponding to each continuous length, and generate a histogram of the frequency and the continuous length; The physical size module is configured to determine the average value and standard deviation of the physical size of the analytes in the test sample by fitting the histogram with an exponential Gaussian distribution function.

18. The computing system of claim 17, wherein: The continuous length is a continuous pixel length, and the physical size module is configured to: fit the histogram through an exponential Gaussian distribution function to obtain the average value and standard deviation of the fitted exponential Gaussian distribution function; convert the average value and standard deviation of the fitted exponential Gaussian distribution function into the average value and standard deviation corresponding to the physical size according to the corresponding relationship between pixels and physical size; or, The continuous length is a continuous physical length, and the statistical module is configured to: count the continuous pixel length when a predetermined pixel appears in the one-dimensional array; and convert the continuous pixel length into a continuous physical length according to the corresponding relationship between pixels and physical sizes.

Images