Scanning electron microscope image defect identification method for anti-radiation organic glass

Abstract

The invention provides a scanning electron microscope image defect recognition method for anti-radiation organic glass, and relates to the technical field of image defect recognizing.According to the method, the surface of an anti-radiation organic glass sample is divided and marked, a dynamic adjustment mechanism for multiple times of circulation is conducted, and the precision and reliability of defect recognition can be continuously optimized; on the basis of extraction and standardization processing of textural features and morphological features of a gray-level co-occurrence matrix, historical data analysis of different defect degrees is combined, a systematic defect supplement and feedback adjustment system is formed, and scientificity and practicability of defect recognition are improved; through effective defect area screening and standardized comparison of characteristic parameters, important technical support can be provided for quality control and application optimization of the radiation-proof organic glass.

Description

Technical Field

[0001] The invention relates to the technical field of image defect recognition, and in particular to a scanning electron microscope image defect recognition method for radiation-proof organic glass. Background Art

[0002] Scanning electron microscopy (SEM) plays an increasingly important role in materials science, nanotechnology and industrial applications. As a high-resolution imaging technology, SEM can observe and analyze the sample surface in detail, helping researchers and engineers to discover and study the microstructure and defect characteristics in the material. In particular, radiation-proof organic glass materials are used in the nuclear industry, aerospace and medical equipment, so the study of their composition and defects is particularly critical. At present, the technical methods for measuring the thermal expansion coefficient, glass transition temperature, crystal structure and diffraction characteristics of radiation-proof organic glass are relatively mature; however, how to efficiently and accurately identify and evaluate its microscopic defects is still a major problem facing the field of materials science.

[0003] In the prior art, the publication number is CN116953006A, and the name is a method for identifying and quantifying defects in scanning electron microscope images of casting materials. First, the scale of the scanning electron microscope image of the tensile fracture of the casting material is detected and identified; according to the scale size, the defects are detected and identified by using digital image processing methods and deep learning image recognition methods respectively; the identified defects are post-processed by noise reduction, expansion and other methods; according to the identified scale, the processed defects are quantified and counted. This method can accurately and quickly identify defects in the scanning electron microscope images of casting materials and realize full automation; in the deep learning model, targeted improvements and adjustments are made based on the characteristics of the defects to be identified to improve the model performance; it can accurately quantify defects, and greatly improves the processing speed compared with manual annotation calculations.

[0004] After dividing and marking the sample surface, traditional methods lack systematic data standardization processing and comprehensive analysis methods, resulting in the results of defect identification being greatly interfered by human factors. In addition, existing technologies often ignore the use of historical data of different defect levels and fail to establish an effective judgment threshold system, resulting in a lack of dynamic adjustment and feedback mechanism for defect identification results.

[0005] The above information disclosed in the above Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0006] The object of the present invention is to provide a method for identifying defects in a scanning electron microscope image of radiation-proof organic glass, so as to solve the problems raised in the above-mentioned background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A method for identifying defects in a scanning electron microscope image of radiation-proof organic glass, characterized in that the specific steps include:

[0009] Step S1: dividing and marking the surface area of ​​the radiation-proof organic glass sample to form a division sequence set; gradually increasing the test temperature of the surface area of ​​the radiation-proof organic glass sample, and measuring the thermal expansion coefficient and glass transition temperature of each division area in the division sequence set at different test temperatures;

[0010] Step S2: measuring the crystal structure and diffraction characteristics of each marked region in the divided sequence set to obtain the diffraction peak position and intensity data of each divided region;

[0011] Step S3: representing the diffraction peak position and intensity data of each marked area in the divided sequence set, as well as the thermal expansion coefficient and glass transition temperature of each marked area at different test temperatures through the same measurement data set, and performing standardization processing on the measurement data set to the same scale;

[0012] Step S4: analyzing the standardized measurement data set to comprehensively generate a first evaluation index for evaluating the defect degree of each marked area; and sequentially arranging the evaluation results of the defect degree of each marked area to obtain an evaluation set sequence;

[0013] Step S5: receiving the evaluation set sequence and screening the defective area, and obtaining the SEM image of each divided area in the divided sequence set, taking each screened defective area as an important defective area in the SEM image, and forming a screening sequence set;

[0014] Step S6: extracting texture features and morphological features from each important defect area in the screening sequence set, wherein the texture features include gray level co-occurrence matrix features, and the morphological features include shape factor and roundness, and performing standardization processing on the same scale on each parameter in the texture features and morphological features.

[0015] Step S7: acquiring historical scanning electron microscope images containing different defect degrees, and determining standard judgment thresholds of various parameters in texture features and morphological features under different defect degrees according to the defect recognition results of these historical scanning electron microscope images, thereby forming a judgment threshold set sequence;

[0016] Step S8: Compare and analyze the parameters in the texture features and morphological features of each important defect area in step S6 with the corresponding standard judgment thresholds in the judgment threshold set sequence to generate a feedback adjustment index, which is used to provide a dynamic adjustment feedback adjustment strategy for the screening sequence set;

[0017] Step S9: Repeat steps S4 to S8 until the feedback adjustment strategy is adjusted m1 times in a cycle, and finally generates a defect recognition result of the scanning electron microscope image.

[0018] Further, forming a partition sequence set, and measuring the thermal expansion coefficient and glass transition temperature of each partition area in the partition sequence set at different test temperatures, specifically includes:

[0019] Set the partition sequence set to {1, 2, ..., i, ..., n}, where i represents the index mark of the i-th partition area, and n represents the total number of partition areas;

[0020] Use the formula T m =T init +m×ΔT to calculate the test temperature each time it rises;

[0021] Among them, ΔT is the temperature increase value of each test, T m is the temperature of the mth test, T init is the initial temperature, m is the number of heating times;

[0022] The thermal expansion displacement dL is measured at each temperature point and the following thermal expansion coefficient is calculated:

[0023]

[0024] Where dL is the length change of thermal expansion displacement, L0 is the initial length, The temperature T of the i-th partition area at the m-th test m The coefficient of thermal expansion under

[0025] The glass transition temperature is identified by the thermal expansion curve, and the glass transition temperature of the i-th divided area is recorded as T g,i .

[0026] Furthermore, the crystal structure and diffraction characteristics of each marked area in the divided sequence set are measured to obtain the diffraction peak position and intensity data of each divided area, specifically including:

[0027] The crystal structure and diffraction characteristics are measured by using an X-ray diffractometer to obtain experimental data including diffraction peaks;

[0028] Based on the experiments conducted according to the partition sequence set {1, 2, ..., i, ..., n}, the diffraction peaks in the data of each measurement area are fitted using a Gaussian function to determine the positions and intensities of the diffraction peaks in each area;

[0029] The diffraction peak position of the i-th divided area is recorded as x 0,i ;

[0030] The diffraction peak intensity of the i-th divided area is recorded as I 0,i ;

[0031] Calculate the standard deviation σ of the Gaussian distribution curve to determine the diffraction peak width;

[0032] The formula for calculating the intensity at position x is as follows;

[0033]

[0034] where I(x) is the intensity at position x, I 0,i represents the maximum diffraction peak intensity of the ith divided area; x 0,i represents the center position of the diffraction peak of the ith divided area; σ is the standard deviation of the diffraction peak width.

[0035] Furthermore, the measurement data set is standardized to the same scale, including:

[0036] The diffraction peak position x of the i-th divided area 0,i , diffraction peak intensity I 0,i , thermal expansion coefficient and glass transition temperature T g,i Organize into a unified measurement data set and i∈{1,2,…,n};

[0037] Standardize the measurement data set to ensure that different data dimensions have the same scale;

[0038] The normalized diffraction peak position x 0,i , diffraction peak intensity I 0,i , thermal expansion coefficient and glass transition temperature T g,i Marked as x′ 0,i , I′ 0,i , T′ g,i .

[0039] Furthermore, the first evaluation index and the formation of the screening sequence set specifically include:

[0040] Define the first evaluation index of the i-th partition area as E i , the calculation formula is as follows:

[0041]

[0042] in, They are x′ 0,i , I′ 0,i , T′ g,i The average value of all n regions; η1 is the adjustment constant term, which is used to avoid When η is 0; and 0.11≤η1≤0.32;

[0043] Setting E i The valid value range of is (0,1);

[0044] If E i The closer the value is to 1, the greater the parameter difference in the area and the higher the degree of defect;

[0045] If E i The closer the value is to 0, the smaller the parameter difference in the area is and the lower the degree of defect is;

[0046] Calculate the first evaluation index E of each partition region in the partition sequence set {1, 2, ..., i, ..., n} i , press E to move each area i Sort from high to low to form an evaluation set sequence; and set the final screening threshold of the first evaluation index as

[0047] If the first evaluation index E of each partition region in the partition sequence set {1, 2, ..., i, ..., n} i The final screening threshold is the same for

[0048] Set the average index of the current partition sequence set to In the partition sequence set {1,2,…,i,…,n}, mark the first evaluation index greater than the average index The number of divided regions r1; mark the first evaluation index is less than the average index The number of divided areas r2;

[0049] On the basis of r1>r2, the final screening threshold for

[0050] Among them, η2 is the adjustment constant term; r1<n;

[0051] On the basis of r1<r2, the final screening threshold for

[0052] Among them, r2<n;

[0053] The evaluation set sequence above the final screening threshold The first evaluation indexes are screened out, and the marked areas corresponding to these first evaluation indices are re-marked to form a screening sequence set {1, 2, ..., k, ..., L}, where k represents the index mark of the kth important defect area, L represents the total number of important defect areas, and L<n.

[0054] Furthermore, each parameter in the texture feature and the morphological feature is standardized to the same scale, specifically including:

[0055] Calculate the gray level co-occurrence matrix of the kth important defect area, the gray level co-occurrence matrix includes contrast, energy, entropy and homogeneity;

[0056] The contrast of the kth important defect area is recorded as Ct k ; The energy of the kth important defect area is recorded as Ey k ; The entropy of the kth important defect area is recorded as Ep k ; The homogeneity of the kth important defect area is denoted as Hy k ; The shape factor of the kth important defect area is recorded as SF k ; The roundness of the kth important defect area is recorded as R k .

[0057] Furthermore, a judgment threshold set sequence is formed, which specifically includes:

[0058] Calculate the mean μ′ and standard deviation σ″ of the characteristic parameters of each defect degree in the kth important defect area; the characteristic parameters in the kth important defect area include contrast Ct k 、Energy k Ep k , Homogeneity Hy k , shape factor SF k , roundness R k ;

[0059] Use the following formula to determine the judgment threshold range:

[0060] Threshold=[μ′-2σ″,μ′+2σ″]

[0061] Threshold is the judgment threshold range of the characteristic parameter, which is used to judge the severity of the defect;

[0062] The characteristic parameter judgment thresholds under different defect levels form a judgment threshold set sequence, specifically:

[0063] {Ct k,z ,Ey k,z,Ep k,z ,Hy k,z ,SF k,z ,R k,z}for z∈{1,2,3}

[0064] Among them, Ct k,z represents the contrast threshold of the zth defect level under the kth important defect area;

[0065] Ey k,z represents the energy threshold of the zth defect level under the kth important defect area;

[0066] Ep k,z represents the entropy threshold of the zth defect level under the kth important defect area;

[0067] Hy k,z represents the homogeneity threshold of the zth defect level under the kth important defect area;

[0068] SF k,z represents the shape factor threshold of the zth defect level under the kth important defect area;

[0069] R k,z represents the roundness threshold of the zth defect level under the kth important defect area;

[0070] In z∈{1,2,3}, 1, 2, and 3 represent the first, second, and third levels, respectively. The higher the level, the higher the degree of defect. When 0<E i <0.35, z=1; when 0.35≤E i ≤0.66, z=2; when 0.66<E i When <1, z=3.

[0071] Further, a feedback adjustment index is generated, and the feedback adjustment index is used to provide a dynamic adjustment feedback adjustment strategy for the screening sequence set, specifically including:

[0072] Determine the defect level of each important defect area characteristic parameter in the screening sequence set {1, 2, ..., k, ..., L} respectively, and calculate the average ratio Rb between the characteristic parameter in the screening sequence set and the corresponding characteristic parameter judgment threshold. The formula is:

[0073]

[0074] The calculation formula of the feedback adjustment index is defined as follows:

[0075]

[0076] Among them, A adjis the feedback adjustment index, and its valid value range is (0.1, 0.9); α is the power of the exponential function, 1.1≤α≤2.1; η3 is the adjustment coefficient, 0.01≤η3≤0.65;

[0077] The feedback adjustment strategy is as follows:

[0078] When A adj When the value is in the range of (0.1, 0.4), it means that the defect characteristics are obvious and the final screening threshold needs to be lowered. To capture more defect areas;

[0079]

[0080] in, is the final screening threshold after adjustment;

[0081] When A adj When the value is in the range of [0.4,0.6], it means that the defect characteristics are general, and the final screening threshold remain unchanged;

[0082] When A adj When the value is in the range of (0.6, 0.9), it means that the defect characteristics are not obvious and the final screening threshold needs to be increased. To reduce false detection;

[0083]

[0084] The feedback adjustment strategy is adjusted m1 times in a cycle; the formula for determining m1 is as follows:

[0085]

[0086] Compared with the prior art, the beneficial effects of the present invention are as follows: by dividing and marking the surface of the radiation-proof organic glass sample and undergoing a dynamic adjustment mechanism of multiple cycles, the accuracy and reliability of defect identification can be continuously optimized; based on the extraction and standardized processing of texture features and morphological features of the gray-level co-occurrence matrix, combined with historical data analysis of different defect degrees, a systematic deficiency supplement and feedback adjustment system is formed, which improves the scientificity and practicality of defect identification; through effective defect area screening and standardized comparison of characteristic parameters, important technical support can be provided for the quality control and application optimization of radiation-proof organic glass. BRIEF DESCRIPTION OF THE DRAWINGS

[0087] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0088] Figure 1 It is a schematic diagram of the overall method flow of the present invention. DETAILED DESCRIPTION

[0089] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0090] Embodiment 1:

[0091] See also Figure 1 , the present invention provides a technical solution:

[0092] A method for identifying defects in a scanning electron microscope image of radiation-proof organic glass, comprising the following specific steps:

[0093] Step S1: In a thermomechanical analyzer, the surface area of ​​the radiation-proof organic glass sample is divided and marked to form a division sequence set; the test temperature of the surface area of ​​the radiation-proof organic glass sample is gradually increased, and the thermal expansion coefficient and glass transition temperature of each division area in the division sequence set at different test temperatures are measured;

[0094] Step S2: In an X-ray diffractometer, the crystal structure and diffraction characteristics of each marked region in the divided sequence set are measured to obtain the diffraction peak position and intensity data of each divided region;

[0095] Step S3: representing the diffraction peak position and intensity data of each marked area in the divided sequence set, as well as the thermal expansion coefficient and glass transition temperature of each marked area at different test temperatures through the same measurement data set, and performing standardization processing on the measurement data set to the same scale;

[0096] Step S4: analyzing the standardized measurement data set to comprehensively generate a first evaluation index for evaluating the defect degree of each marked area; and sequentially arranging the evaluation results of the defect degree of each marked area to obtain an evaluation set sequence;

[0097] Step S5: receiving the evaluation set sequence and screening the defective area, and obtaining the SEM image of each divided area in the divided sequence set, taking each screened defective area as an important defective area in the SEM image, and forming a screening sequence set;

[0098] Step S6: extracting texture features and morphological features from each important defect area in the screening sequence set, wherein the texture features include gray level co-occurrence matrix features, and the morphological features include shape factor and roundness, and performing standardization processing on the same scale on each parameter in the texture features and morphological features.

[0099] Step S7: acquiring historical scanning electron microscope images containing different defect degrees, and determining standard judgment thresholds of various parameters in texture features and morphological features under different defect degrees according to the defect recognition results of these historical scanning electron microscope images, thereby forming a judgment threshold set sequence;

[0100] Step S8: Compare and analyze the parameters in the texture features and morphological features of each important defect area in step S6 with the corresponding standard judgment thresholds in the judgment threshold set sequence to generate a feedback adjustment index, which is used to provide a dynamic adjustment feedback adjustment strategy for the screening sequence set;

[0101] Step S9: Repeat steps S4 to S8 until the feedback adjustment strategy is adjusted m1 times in a cycle, and finally generates a defect recognition result of the scanning electron microscope image.

[0102] It is further described that forming a partition sequence set and determining the thermal expansion coefficient and glass transition temperature of each partition area in the partition sequence set at different test temperatures specifically include:

[0103] Select a uniform radiation-proof organic glass sample with a size of 10mmx10mmx2mm;

[0104] Place the sample in an ultrasonic cleaner, add deionized water and neutral detergent, and clean for 5 minutes;

[0105] After wiping with dust-free paper, place in a drying oven and dry at 60°C for 30 minutes;

[0106] Using an optical microscope, magnify the sample 50 times and observe;

[0107] Grid division: Use a high-precision engraving knife and a microscope ruler to divide the sample into 5x5 rectangular grids, with each grid having a side length of 2 mm;

[0108] Number marking: Use a tiny paint pen or laser marking to mark the area as {1,2,...,i,...,25}.

[0109] Set the partition sequence set to {1, 2, ..., i, ..., n}, where i represents the index mark of the i-th partition area, and n represents the total number of partition areas;

[0110] For the thermomechanical analyzer setup:

[0111] Fix the radiation-proof organic glass sample on the TMA sample stage to ensure that it is flat and in good contact;

[0112] Software settings: In the control software, set the initial temperature to 25°C, the temperature increase value for each test is 10°C / min, and the range is 25°C to 250°C;

[0113] Preheat Stabilization: Preheat the sample for 5 minutes to stabilize the temperature.

[0114] For determination of coefficient of thermal expansion:

[0115] Use the formula T m =T init +m×ΔT to calculate the test temperature each time it rises;

[0116] Among them, ΔT is the temperature increase value of each test, where T m is the temperature of the mth test, T init is the initial temperature, m is the number of heating times;

[0117] The thermal expansion displacement dL is measured at each temperature point and the following thermal expansion coefficient is calculated:

[0118]

[0119] Where dL is the length change of thermal expansion displacement, L0 is the initial length, The temperature T of the i-th partition area at the m-th test m The coefficient of thermal expansion under

[0120] For the determination of glass transition temperature:

[0121] The glass transition temperature is identified by the thermal expansion curve, and the glass transition temperature of the i-th divided area is recorded as T g,i .

[0122] Record the test data of each area in an Excel table, as shown in Table 1 below:

[0123] Table 1

[0124]

[0125] For the glass transition temperature:

[0126] If at temperature point T g,iA glass transition is observed, with a glass transition temperature of 150°C in region 1 in the example data;

[0127] The glass transition temperature is the temperature point at which a material changes from a glassy state to a rubbery state, which is manifested as a sudden change in the thermal expansion curve.

[0128] Further description, the crystal structure and diffraction characteristics of each marked area in the divided sequence set are measured to obtain the diffraction peak position and intensity data of each divided area, specifically including:

[0129] The crystal structure and diffraction characteristics are measured by using an X-ray diffractometer to obtain experimental data including diffraction peaks;

[0130] X-ray diffraction measurements and data analysis are as follows:

[0131] Place the radiation-proof organic glass sample on the sample stage of the X-ray diffractometer;

[0132] Use a non-reflective fixture. In this embodiment, a low-reflective metal fixture is used to fix the radiation-proof organic glass sample to ensure that there is no movement.

[0133] Adjust the sample stage so that the surface of the radiation-proof organic glass sample is perpendicular to the X-ray beam; use a level or laser positioning device to ensure accurate positioning;

[0134] For the scanning parameter settings of X-ray diffractometer:

[0135] Start the control software of the X-ray diffractometer;

[0136] In the software interface, set the following parameters:

[0137] Wavelength: Input

[0138] Scanning angle range: set from 5° to 80°;

[0139] Step angle: set to 0.02°;

[0140] Scanning speed: set to 1° / min;

[0141] After confirming that the parameter settings are correct, save the settings.

[0142] Instrument Calibration:

[0143] Take a standard silicon wafer sample, place it on the sample stage and fix it;

[0144] Select calibration mode in the control software;

[0145] Start the calibration procedure and wait for the instrument to calibrate automatically;

[0146] After calibration, remove the silicon wafer and place the sample to be tested, making sure it remains in the same position.

[0147] According to the divided sequence set as {1,2,…,i,…,n}, a sequence of each measurement area is created in the software, and the measurement order is set;

[0148] Start the automatic scanning program and select the measurement sequence;

[0149] The X-ray diffractometer will scan each area in turn and record the data to ensure that the measurement conditions in each area are consistent;

[0150] After the scan is completed, open the data export interface of the control software;

[0151] Select the data of all measurement areas and choose to save it in CSV or TXT format;

[0152] Export the data and save it to the specified directory on the computer, and name the file "Sample Name_Date".

[0153] Open data processing software such as Origin or Excel;

[0154] Import the saved CSV or TXT data file into the data processing software;

[0155] In the data processing software, background correction was performed on the data of each region;

[0156] Use the background subtraction tool provided by the software and select the corresponding background model, such as linear background or polynomial background;

[0157] Subtract the background noise to obtain the corrected data;

[0158] Based on the experiments conducted according to the partition sequence set {1, 2, ..., i, ..., n}, the diffraction peaks in the data of each measurement area are fitted using a Gaussian function to determine the positions and intensities of the diffraction peaks in each area;

[0159] The diffraction peak position of the i-th divided area is recorded as x 0,i ; The diffraction peak position refers to the center position of the peak in the X-ray diffraction pattern, expressed as an angle (°2θ); this value reflects the specific spacing of the material's crystal structure and is an accurate value determined by Gaussian fitting;

[0160] The diffraction peak intensity of the i-th divided area is recorded as I 0,i ;

[0161] Calculate the standard deviation σ of the Gaussian distribution curve to determine the diffraction peak width;

[0162] Based on the peak shape in the diffraction pattern, the peak width (i.e., the full width at half maximum, FWHM) was preliminarily estimated;

[0163] Convert the full width at half maximum to standard deviation using the formula:

[0164]

[0165] Among them, FWHM (Full Width that Half Maximum) represents the full width at half height of the peak; in the peak graph, find the maximum height of the peak, and then determine half of the height. The difference in width on both sides is the FWHM; this parameter represents the width of the peak and is an important indicator of the peak shape characteristics;

[0166] σ is the standard deviation of the Gaussian distribution curve, which represents a standardized measure of the width of the peak, and a larger value indicates a wider peak;

[0167] ln2 is the natural logarithm function, ln2 represents the logarithm with base e;

[0168] is a constant, approximately equal to 2.3548, and is the coefficient that converts FWHM to standard deviation; this coefficient comes from the characteristics of the Gaussian distribution.

[0169] Start the fitting program, and the software will automatically adjust the parameters to obtain the best fitting result;

[0170] Record the diffraction peak position x in each fitting result 0,i and the diffraction peak intensity I 0,i .

[0171] The formula for calculating the intensity at position x is as follows;

[0172]

[0173] where I(x) is the intensity at position x, I 0,i represents the maximum diffraction peak intensity of the ith divided area; x 0,i represents the center position of the diffraction peak of the ith divided area; σ is the standard deviation of the diffraction peak width.

[0174] The measurement results of each area are entered into the table in sequence according to the number of experiments, forming the following Table 2:

[0175] Table 2

[0176]

[0177]

[0178] Through the above table data, accurate diffraction data can be obtained, and further defect identification and material analysis can be performed.

[0179] Further explanation: the measurement data set is standardized to the same scale, specifically including:

[0180] The diffraction peak position x of the i-th divided area 0,i , diffraction peak intensity I 0,i , thermal expansion coefficient and glass transition temperature T g,i Organize into a unified measurement data set and i∈{1,2,…,n};

[0181] Standardize the measurement data set to ensure that different data dimensions have the same scale;

[0182] Use the normalization formula: Where X is the original data, μ is the mean, and σ′ is the standard deviation;

[0183] The normalized diffraction peak position x 0,i , diffraction peak intensity I 0,i , thermal expansion coefficient and glass transition temperature T g,i Marked as x′ 0,i , I′ 0,i , T′ g,i .

[0184] Further explanation: the first evaluation index and forming the screening sequence set specifically include:

[0185] Define the first evaluation index of the i-th partition area as E i , the calculation formula is as follows:

[0186]

[0187] in, They are x′ 0,i , I′ 0,i , T′ g,i The average value of all n regions; η1 is the adjustment constant term, which is used to avoid is 0; and 0.11≤η1≤0.32; in this embodiment, η1 is 0.22; η1 is specifically determined by the expert group through adjustment based on experimental data;

[0188] Setting E i The valid value range of is (0,1);

[0189] If E i The closer the value is to 1, the greater the parameter difference in the area and the higher the degree of defect;

[0190] If E i The closer the value is to 0, the smaller the parameter difference in the area is and the lower the degree of defect is;

[0191] Calculate the first evaluation index E of each partition region in the partition sequence set {1, 2, ..., i, ..., n} i , press E to move each area i Sort from high to low to form an evaluation set sequence; and set the final screening threshold of the first evaluation index as

[0192] If the first evaluation index E of each partition region in the partition sequence set {1, 2, ..., i, ..., n} i The final screening threshold is the same for

[0193] Set the average index of the current partition sequence set to In the partition sequence set {1,2,…,i,…,n}, mark the first evaluation index greater than the average index The number of divided regions r1; mark the first evaluation index is less than the average index The number of divided areas r2;

[0194] On the basis of r1>r2, the final screening threshold for

[0195] Wherein, η2 is the adjustment constant term; r1<n, in this embodiment, η2 is set to 0.2; it is specifically determined by the expert group through adjustment based on experimental data;

[0196] On the basis of r1<r2, the final screening threshold for

[0197] Wherein, r2<n, and the value of η2 in this embodiment is 0.2; it is specifically determined by the expert group through adjustment based on experimental data;

[0198] The evaluation set sequence above the final screening threshold The first evaluation indexes are screened out, and the marked areas corresponding to these first evaluation indices are re-marked to form a screening sequence set {1, 2, ..., k, ..., L}, where k represents the index mark of the kth important defect area screened out, L represents the total number of important defect areas, and L<n.

[0199] Further explanation: the parameters in the texture features and morphological features are standardized to the same scale, specifically including:

[0200] Calculate the gray level co-occurrence matrix of the kth important defect area, the gray level co-occurrence matrix includes contrast, energy, entropy and homogeneity;

[0201] The contrast of the kth important defect area is recorded as Ct k :

[0202]

[0203] Among them, (i′-j′) is the index of the gray value; p k (i′, j′) is the element value of position (i′, j′) in the gray level co-occurrence matrix under the kth important defect area, which represents the joint probability of the occurrence of gray values ​​i′ and j′; It represents the sum of all the combinations of gray values ​​i′ and j′; the contrast Ct k The higher it is, the rougher the texture of the image;

[0204] The energy of the kth important defect area is denoted as Ey k :

[0205]

[0206] Among them, the energy value Ey k The larger it is, the more uniform the texture;

[0207] The entropy of the kth important defect area is denoted as Ep k :

[0208]

[0209] Among them, the entropy value Ep k The higher it is, the greater the complexity of the image;

[0210] The homogeneity of the kth important defect area is denoted as Hy k :

[0211]

[0212] Among them, the homogeneity value Hy k The higher it is, the more uniform the local grayscale is;

[0213] The shape factor of the kth important defect area is denoted as SF k :

[0214]

[0215] Among them, AS k is the area of ​​the kth important defect region, obtained by pixel counting; Pd k is the perimeter of the kth important defect area, obtained by edge detection; shape factor SF kThe closer it is to 1, the more regular the defect shape is;

[0216] The roundness of the kth important defect area is recorded as R k .

[0217]

[0218] Among them, Ld k is the major axis of the kth important defect area, obtained through morphological analysis; the closer the roundness is to 1, the closer the defect is to a circle.

[0219] Use the same normalization formula as for the measurement dataset normalization: Where X is the original data, μ is the mean, and σ′ is the standard deviation;

[0220] Z is the standardized eigenvalue, which ranges from (0, 1) to facilitate subsequent analysis.

[0221] Further explanation: forming a judgment threshold set sequence specifically includes:

[0222] Calculate the mean μ′ and standard deviation σ″ of the characteristic parameters of each defect degree in the kth important defect area; the characteristic parameters in the kth important defect area include contrast Ct k 、Energy k Ep k , Homogeneity Hy k , shape factor SF k , roundness R k ;

[0223] Use the following formula to determine the judgment threshold range:

[0224] Threshold=[μ′-2σ″,μ′+2σ″]

[0225] Threshold is the judgment threshold range of the characteristic parameter, which is used to judge the severity of the defect;

[0226] The characteristic parameter judgment thresholds under different defect levels form a judgment threshold set sequence, specifically:

[0227] {Ct k,z ,Ey k,z ,Ep k,z ,Hy k,z ,SF k,z ,R k,z}for z∈{1,2,3}

[0228] Among them, Ct k,z represents the contrast threshold of the zth defect level under the kth important defect area;

[0229] Ey k,z represents the energy threshold of the zth defect level under the kth important defect area;

[0230] Ep k,z represents the entropy threshold of the zth defect level under the kth important defect area;

[0231] Hy k,z represents the homogeneity threshold of the zth defect level under the kth important defect area;

[0232] SF k,z represents the shape factor threshold of the zth defect level under the kth important defect area;

[0233] R k,z represents the roundness threshold of the zth defect level under the kth important defect area;

[0234] In z∈{1,2,3}, 1, 2, and 3 represent the first, second, and third levels, respectively. The higher the level, the higher the degree of defect. When 0<E i <0.35, z=1; when 0.35≤E i ≤0.66, z=2; when 0.66<E i When <1, z=3.

[0235] Further explanation: generating a feedback adjustment index, which is used to provide a dynamically adjusted feedback adjustment strategy for the screening sequence set, specifically includes:

[0236] Determine the defect level of each important defect area characteristic parameter in the screening sequence set {1, 2, ..., k, ..., L} respectively, and calculate the average ratio Rb between the characteristic parameter in the screening sequence set and the corresponding characteristic parameter judgment threshold. The formula is:

[0237]

[0238] The calculation formula of the feedback adjustment index is defined as follows:

[0239]

[0240] Among them, A adj is the feedback adjustment index, ranging from (0.1, 0.9); α is the power of the exponential function, 1.1≤α≤2.1; it is used to increase the differentiation effect; η3 is the adjustment coefficient, 0.01≤η3≤0.65; η3 is used to ensure A adj The value range is (0.1, 0.9); in this embodiment, η3 is 0.42; the specific values ​​of α and η3 are determined by the expert group through experimental data;

[0241] The feedback adjustment strategy is as follows:

[0242] When A adj When the value is in the range of (0.1, 0.4), it means that the defect characteristics are obvious and the final screening threshold needs to be lowered. To capture more defect areas;

[0243]

[0244] in, is the final screening threshold after adjustment;

[0245] When A adj When the value is in the range of [0.4,0.6], it means that the defect characteristics are general, and the final screening threshold remain unchanged;

[0246] When A adj When the value is in the range of (0.6, 0.9), it means that the defect characteristics are not obvious and the final screening threshold needs to be increased. To reduce false detection;

[0247]

[0248] The feedback adjustment strategy is adjusted m1 times in a cycle; the formula for determining m1 is as follows:

[0249]

[0250] in, Indicates rounding up to an integer.

[0251] The final defect recognition results of the SEM image are generated, including:

[0252] Annotated image: Mark the defect screening sequence set {1,2,…,k,…,L} regions on the original SEM image;

[0253] Defect list: A list containing the location, size, shape, and other characteristic information of each defect;

[0254] Statistics: number of defects, average size, distribution, defect grade;

[0255] Feature analysis: The characteristic parameters of each defect area include contrast, energy, entropy and homogeneity.

[0256] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0257] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0258] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0259] The above embodiments are only used to help understand the method and core idea of ​​the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made to the present invention without departing from the principles of the present invention, and these improvements and modifications also fall within the scope of protection of the claims of the present invention.

Claims

1. A method for identifying defects in a scanning electron microscope image of radiation-proof organic glass, characterized in that: The specific steps include: Step S1: dividing and marking the surface area of ​​the radiation-proof organic glass sample to form a division sequence set; gradually increasing the test temperature of the surface area of ​​the radiation-proof organic glass sample, and measuring the thermal expansion coefficient and glass transition temperature of each division area in the division sequence set at different test temperatures; Step S2: measuring the crystal structure and diffraction characteristics of each marked region in the divided sequence set to obtain the diffraction peak position and intensity data of each divided region; Step S3: representing the diffraction peak position and intensity data of each marked area in the divided sequence set, as well as the thermal expansion coefficient and glass transition temperature of each marked area at different test temperatures through the same measurement data set, and performing standardization processing on the measurement data set to the same scale; Step S4: analyzing the standardized measurement data set to comprehensively generate a first evaluation index for evaluating the defect degree of each marked area; and sequentially arranging the evaluation results of the defect degree of each marked area to obtain an evaluation set sequence; Step S5: receiving the evaluation set sequence and screening the defective area, and obtaining the SEM image of each divided area in the divided sequence set, taking each screened defective area as an important defective area in the SEM image, and forming a screening sequence set; Step S6: extracting texture features and morphological features from each important defect area in the screening sequence set, wherein the texture features include gray level co-occurrence matrix features, and the morphological features include shape factor and roundness, and performing standardization processing on the same scale on each parameter in the texture features and morphological features. Step S7: acquiring historical scanning electron microscope images containing different defect degrees, and determining standard judgment thresholds of various parameters in texture features and morphological features under different defect degrees according to the defect recognition results of these historical scanning electron microscope images, thereby forming a judgment threshold set sequence; Step S8: Compare and analyze the parameters in the texture features and morphological features of each important defect area in step S6 with the corresponding standard judgment thresholds in the judgment threshold set sequence to generate a feedback adjustment index, which is used to provide a dynamic adjustment feedback adjustment strategy for the screening sequence set; Step S9: Repeat steps S4 to S8 until the feedback adjustment strategy is adjusted m1 times in a cycle, and finally generates a defect recognition result of the scanning electron microscope image.

2. The method for identifying defects in a scanning electron microscope image of radiation-proof organic glass according to claim 1, characterized in that: Forming a partition sequence set, and determining the thermal expansion coefficient and glass transition temperature of each partition area in the partition sequence set at different test temperatures, specifically including: Set the partition sequence set to {1, 2, ..., i, ..., n}, where i represents the index mark of the i-th partition area, and n represents the total number of partition areas; Use the formula T m =T init +m×ΔT to calculate the test temperature each time it rises; Among them, ΔT is the temperature increase value of each test, T m is the temperature of the mth test, T init is the initial temperature, m is the number of heating times; The thermal expansion displacement dL is measured at each temperature point and the following thermal expansion coefficient is calculated: Where dL is the length change of thermal expansion displacement, L0 is the initial length, The temperature T of the i-th partition area at the m-th test m The coefficient of thermal expansion under The glass transition temperature is identified by the thermal expansion curve, and the glass transition temperature of the i-th divided area is recorded as T g,i .

3. The method for identifying defects in a scanning electron microscope image of radiation-proof organic glass according to claim 2, characterized in that: The crystal structure and diffraction characteristics of each marked area in the divided sequence set are measured to obtain the diffraction peak position and intensity data of each divided area, including: The crystal structure and diffraction characteristics are measured by using an X-ray diffractometer to obtain experimental data including diffraction peaks; Based on the experiments conducted according to the partition sequence set {1, 2, ..., i, ..., n}, the diffraction peaks in the data of each measurement area are fitted using a Gaussian function to determine the positions and intensities of the diffraction peaks in each area; The diffraction peak position of the i-th divided area is recorded as x0,i; The diffraction peak intensity of the i-th divided area is recorded as I0,i; Calculate the standard deviation σ of the Gaussian distribution curve to determine the diffraction peak width; The formula for calculating the intensity at position x is as follows; Wherein, I(x) is the intensity at position x, I0,i represents the maximum diffraction peak intensity of the i-th divided area; x0,i represents the center position of the diffraction peak of the i-th divided area; σ is the standard deviation of the diffraction peak width.

4. The method for identifying defects in a scanning electron microscope image of radiation-proof organic glass according to claim 3, characterized in that: The measurement data sets are standardized to the same scale, including: The diffraction peak position x0,i, diffraction peak intensity I0,i, thermal expansion coefficient of the i-th divided area and glass transition temperature T g,i Organize into a unified measurement data set and i∈{1,2,…,n}; Standardize the measurement data set to ensure that different data dimensions have the same scale; The standardized diffraction peak position x0,i, diffraction peak intensity I0,i, thermal expansion coefficient and glass transition temperature T g,i Marked as x′ 0,i , I′ 0,i , T′ g,i .

5. The method for identifying defects in a scanning electron microscope image of radiation-proof organic glass according to claim 4, characterized in that: The first evaluation index and the formation of a screening sequence set specifically include: Define the first evaluation index of the i-th partition area as E i , the calculation formula is as follows: in, They are x′ 0,i , I′ 0,i , T′ g,i The average value of all n regions; η1 is the adjustment constant term, which is used to avoid When η is 0; and 0.11≤η1≤0.32; Setting E i The valid value range of is (0,1); If E i The closer the value is to 1, the greater the parameter difference in the area and the higher the degree of defect; If E i The closer the value is to 0, the smaller the parameter difference in the area is and the lower the degree of defect is; Calculate the first evaluation index E of each partition region in the partition sequence set {1, 2, ..., i, ..., n} i , press E to move each area i Sort from high to low to form an evaluation set sequence; and set the final screening threshold of the first evaluation index as If the first evaluation index E of each partition region in the partition sequence set {1, 2, ..., i, ..., n} i The final screening threshold is the same for Set the average index of the current partition sequence set to In the partition sequence set {1,2,…,i,…,n}, mark the first evaluation index greater than the average index The number of divided regions r1; mark the first evaluation index is less than the average index The number of divided areas r2; On the basis of r1>r2, the final screening threshold for Among them, η2 is the adjustment constant term; r1<n; On the basis of r1<r2, the final screening threshold for Among them, r2<n; The evaluation set sequence above the final screening threshold The first evaluation indexes are screened out, and the marked areas corresponding to these first evaluation indices are re-marked to form a screening sequence set {1, 2, ..., k, ..., L}, where k represents the index mark of the kth important defect area, L represents the total number of important defect areas, and L<n.

6. The method for identifying defects in a scanning electron microscope image of radiation-proof organic glass according to claim 5, characterized in that: The parameters in the texture features and morphological features are standardized to the same scale, including: Calculate the gray level co-occurrence matrix of the kth important defect area, the gray level co-occurrence matrix includes contrast, energy, entropy and homogeneity; The contrast of the kth important defect area is recorded as Ct k ; The energy of the kth important defect area is recorded as Ey k ; The entropy of the kth important defect area is recorded as Epk; the homogeneity of the kth important defect area is recorded as Hyk; the shape factor of the kth important defect area is recorded as SF k ; The roundness of the kth important defect area is recorded as R k .

7. The method for identifying defects in a scanning electron microscope image of radiation-proof organic glass according to claim 6, characterized in that: A judgment threshold set sequence is formed, specifically including: Calculate the mean μ′ and standard deviation σ″ of the characteristic parameters of each defect degree in the kth important defect area; the characteristic parameters in the kth important defect area include contrast Ct k 、Energy k Ep k , Homogeneity Hy k , shape factor SF k , roundness R k ; Use the following formula to determine the judgment threshold range: Threshold=[μ′-2σ″,μ′+2σ″] Threshold is the judgment threshold range of the characteristic parameter, which is used to judge the severity of the defect; The characteristic parameter judgment thresholds under different defect levels form a judgment threshold set sequence, specifically: {Ct k,z ,Ey k,z ,Ep k,z ,Hy k,z ,SF k,z ,R k,z }for z∈{1,2,3} Among them, Ct k,z represents the contrast threshold of the zth defect level under the kth important defect area; Ey k,z represents the energy threshold of the zth defect level under the kth important defect area; Ep k,z represents the entropy threshold of the zth defect level under the kth important defect area; Hy k,z represents the homogeneity threshold of the zth defect level under the kth important defect area; SF k,z represents the shape factor threshold of the zth defect level under the kth important defect area; R k,z represents the roundness threshold of the zth defect level under the kth important defect area; In z∈{1,2,3}, 1, 2, and 3 represent the first, second, and third levels, respectively. The higher the level, the higher the degree of defect. When 0<E i <0.35, z=1; when 0.35≤E i ≤0.66, z=2; when 0.66<E i When <1, z=3.

8. The method for identifying defects in a scanning electron microscope image of radiation-proof organic glass according to claim 7, characterized in that: Generate a feedback adjustment index, which is used to provide a dynamically adjusted feedback adjustment strategy for the screening sequence set, specifically including: Determine the defect level of each important defect area characteristic parameter in the screening sequence set {1, 2, ..., k, ..., L} respectively, and calculate the average ratio Rb between the characteristic parameter in the screening sequence set and the corresponding characteristic parameter judgment threshold. The formula is: The calculation formula of the feedback adjustment index is defined as follows: Among them, A adj is the feedback adjustment index, and its valid value range is (0.1, 0.9); α is the power of the exponential function, 1.1≤α≤2.1; η3 is the adjustment coefficient, 0.01≤η3≤0.65; The feedback adjustment strategy is as follows: When A adj When the value is in the range of (0.1, 0.4), it means that the defect characteristics are obvious and the final screening threshold needs to be lowered. To capture more defect areas; in, is the final screening threshold after adjustment; When A adj When the value is in the range of [0.4,0.6], it means that the defect characteristics are general, and the final screening threshold remain unchanged; When A adj When the value is in the range of (0.6, 0.9), it means that the defect characteristics are not obvious and the final screening threshold needs to be increased. To reduce false detection; The feedback adjustment strategy is adjusted m1 times in a cycle; the formula for determining m1 is as follows:

Images