Surface detection method for magnesium oxide single crystal target material preparation

Abstract

The invention relates to the field of target material surface detection, in particular to a magnesium oxide single crystal target material preparation surface detection method, which comprises the following steps: acquiring a surface topography image of each detection sub-region and synchronously acquiring component spectrum data of the corresponding detection sub-region; removing continuous background noise from the component spectrum data, extracting characteristic X-ray peak intensity and recording the characteristic X-ray peak intensity as component parameters; extracting morphology parameters from the surface morphology image, and performing spatial registration on the morphology parameters and the component parameters to obtain a morphology-component fusion feature vector; comparing and analyzing the morphology-component fusion feature vector with a preset pollution particle feature library and a micro-bulge feature library to obtain a similarity probability, and obtaining a defect type judgment result of the detection sub-region according to the similarity probability; mapping the judgment result to a three-dimensional coordinate of the surface of the target material, marking the positions and confidence degrees of the pollution particles and the micro-bulges, and generating a detection report; and the accuracy and comprehensiveness of surface defect detection of the magnesium oxide single crystal target material are effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of surface detection for the preparation of target materials, and specifically to a method for surface detection in the preparation of magnesium oxide single crystal target materials. Background Art

[0002] During the preparation of magnesium oxide single crystal target materials, the surface quality of the target material plays a crucial role in subsequent application performance. Magnesium oxide single crystal target materials are widely used in high-precision fields such as semiconductor manufacturing and optical coating. Surface defects will directly affect the performance and reliability of products. Therefore, precise and comprehensive detection of the surface during the preparation of magnesium oxide single crystal target materials, timely discovery, and accurate identification of surface defects have become key links in ensuring the quality of target materials.

[0003] Currently, for the detection of the surface of magnesium oxide single crystal target materials, the method mainly used is a combination of scanning electron microscopy (SEM) and energy-dispersive X-ray spectroscopy (EDS). SEM is used to obtain the morphological image of the target material surface, and EDS is used to collect surface composition spectrum data. However, in the actual detection process, for weak signal defects such as tiny contaminant particles (e.g., <10nm metal particles), there is a serious problem of missed detection. Since the signals generated by tiny contaminant particles in the EDS spectrum are extremely weak and are easily masked by noise, it is difficult to accurately extract their characteristic information. In the defect identification process, differentiating tiny contaminant particles from micro-protrusions highly depends on the experience of the detection personnel, which makes it difficult to ensure the consistency of detection results. This problem of poor detection consistency not only increases the uncertainty of target material quality assessment but also may cause some defective target materials to flow into subsequent production processes, affecting the quality and performance of the final products, and bringing economic losses and market risks to enterprises.

[0004] Therefore, there is an urgent need for a method for surface detection in the preparation of magnesium oxide single crystal target materials, which effectively improves the detection ability for tiny contaminant particles, reduces the missed detection rate of weak signal defects, enhances the consistency and accuracy of detection, and provides a reliable guarantee for the quality control of magnesium oxide single crystal target materials. Summary of the Invention

[0005] (1) Technical Problems to be Solved The purpose of the present invention is to provide a method for surface detection in the preparation of magnesium oxide single crystal target materials, to solve the problem that it is difficult to automatically distinguish between tiny contaminant particles and micro-protrusion defects on the surface of magnesium oxide single crystal target materials under weak signal and high noise interference.

[0006] (2) Technical Solutions To achieve the above purpose, the present invention provides a method for surface detection in the preparation of magnesium oxide single crystal target materials, and the method includes: S1. Divide the surface of the target into multiple detection sub-regions, and determine the priority scanning order of each detection sub-region; obtain the surface topography image of each detection sub-region through a scanning electron microscope, and simultaneously collect the compositional spectrum data of the corresponding detection sub-region through an energy dispersive X-ray spectrometer.

[0007] S2. After performing multi-frame stacking processing on the compositional spectrum data, remove the continuous background noise by non-linear least squares fitting and extract the characteristic X-ray peak intensity, which is denoted as the compositional parameter; extract the topography parameter from the surface topography image, and perform spatial registration of the topography parameter and the compositional parameter to obtain a topography-composition fusion feature vector; the topography parameter includes the micro-region height gradient, the local surface curvature radius, and the edge sharpness parameter.

[0008] S3. Compare and analyze the topography-composition fusion feature vector with a preset contamination particle feature library and a micro-protrusion feature library to obtain a similarity probability, and obtain the defect type discrimination result of the detection sub-region according to the similarity probability.

[0009] S4. Map the discrimination result to the three-dimensional coordinates of the target surface, mark the positions and confidence levels of the contamination particles and micro-protrusions, and generate a detection report.

[0010] Further, the method for removing the continuous background noise by non-linear least squares fitting after performing multi-frame stacking processing on the compositional spectrum data includes: Perform multi-frame stacking processing on the compositional spectrum data by the weighted average method. The number of stacked frames for the multi-frame stacking processing is dynamically adjusted according to the priority of the current detection sub-region. When the priority of the detection sub-region is higher than the preset priority threshold, gradually increase the number of stacked frames through an adaptive algorithm until the signal-to-noise ratio improvement rate of the stacked compositional spectrum data is greater than the preset improvement rate threshold.

[0011] Construct a composite background model including a polynomial basis function and an exponential decay basis function, perform global fitting on the multi-frame stacked compositional spectrum data by non-linear least squares to obtain the weight coefficients of the polynomial basis function and the exponential decay basis function, substitute the weight coefficients into the composite background model to obtain the continuous background noise covering the full energy range; remove the continuous background noise from the multi-frame stacked compositional spectrum data to obtain the denoised compositional spectrum data.

[0012] According to the preset characteristic energy range of the target element, the sliding window method is used to screen candidate characteristic peaks that meet the local maximum condition and have a full width at half maximum within the preset width range in the denoised component spectrum data. The peak area integral value is calculated based on the energy channel range of each candidate characteristic peak as the intensity of the characteristic X-ray peak, and the left and right energy boundaries of each candidate characteristic peak are dynamically calibrated. The dynamic calibration is to determine the starting point of the peak boundary according to the change trend of the first derivative of the count rate of the energy channel corresponding to the candidate characteristic peak. If the correlation coefficient of the count rates of adjacent energy channels is lower than the preset correlation threshold, it is determined as a pseudo characteristic peak formed by isolated noise points and is removed.

[0013] Further, the method for extracting the morphological parameters from the surface morphology image and spatially registering the morphological parameters with the component parameters to obtain the morphology-component fusion feature vector includes: Performing three-dimensional morphological reconstruction on the surface morphology image of each detection sub-region to obtain a height distribution matrix; calculating the height gradient modulus of each pixel point within a preset neighborhood window according to the height distribution matrix as the micro-region height gradient, and calculating the local surface curvature radius through quadratic surface fitting; extracting the pixel gray-scale change rate along the edge contour of the surface morphology image, and taking the ratio of the maximum value of the gray-scale change rate to the gray-scale variance of the adjacent region as the edge sharpness parameter.

[0014] Normalizing the micro-region height gradient, local surface curvature radius, and edge sharpness parameter according to the pixel coordinates of the detection sub-region to obtain a morphological feature vector; establishing a mapping relationship between the image pixel coordinates of the detection sub-region and the energy channel position of the component spectrum data synchronously according to the scanning paths of the scanning electron microscope and the energy dispersive X-ray spectrometer; according to the mapping relationship between the image pixel coordinates of the detection sub-region and the energy channel position of the component spectrum data, splicing the characteristic X-ray peak intensity and the morphological feature vector of the corresponding detection sub-region according to the dimension to obtain a morphology-component fusion feature vector.

[0015] Further, the method for establishing a mapping relationship between the image pixel coordinates of the detection sub-region and the energy channel position of the component spectrum data synchronously according to the scanning paths of the scanning electron microscope and the energy dispersive X-ray spectrometer includes: Determining the synchronization mark between the pixel coordinates of the surface morphology image and the acquisition time of the component spectrum data of each detection sub-region according to the corresponding relationship between the scanning step of the scanning electron microscope and the sampling time interval of the energy dispersive X-ray spectrometer; extracting the real-time coordinate data of the stage position encoder of the scanning electron microscope according to the synchronization mark, and combining the calibration parameters of the energy channel and the spatial position of the energy dispersive X-ray spectrometer to establish a geometric transformation matrix from the image pixel coordinates to the energy channel position.

[0016] Decompose the geometric transformation matrix into a translation vector and a rotation matrix component, and optimize the parameters of the translation vector and the rotation matrix component by the least squares method to minimize the spatial deviation between the central pixel coordinates of the surface topography image and the central position of the energy channel of the compositional spectrum data in the same detection sub-region; apply the optimized geometric transformation matrix to all detection sub-regions to achieve a point-by-point mapping between the image pixel coordinates and the energy channel positions.

[0017] Further, the method for comparing and analyzing the topography-composition fusion feature vector with a preset contaminant particle feature library and a micro-protrusion feature library to obtain a similarity probability, and obtaining a defect type discrimination result for the detection sub-region includes: After standardizing the topography-composition fusion feature vector, calculate the first kernel space similarity with each sample feature vector in the contaminant particle feature library and the second kernel space similarity with each sample feature vector in the micro-protrusion feature library according to the kernel function mapping.

[0018] Estimate the probability distribution of the first kernel space similarity in the contaminant particle category and the probability distribution of the second kernel space similarity in the micro-protrusion category respectively through probability density functions to obtain the first posterior probability that the current topography-composition fusion feature vector belongs to the contaminant particle and the second posterior probability that it belongs to the micro-protrusion; obtain the probability ratio of the first posterior probability to the second posterior probability, and when the probability ratio exceeds a preset probability ratio threshold, determine that the defect type is a contaminant particle, otherwise determine that the defect type is a micro-protrusion; the similarity probability includes the first posterior probability and the second posterior probability.

[0019] Further, the method for estimating the probability distribution of the first kernel space similarity in the contaminant particle category and the probability distribution of the second kernel space similarity in the micro-protrusion category respectively through probability density functions to obtain the first posterior probability that the current topography-composition fusion feature vector belongs to the contaminant particle and the second posterior probability that it belongs to the micro-protrusion includes: Calculate the kernel function bandwidth parameter according to the covariance matrix eigenvalues of all sample feature vectors in the preset contaminant particle feature library and micro-protrusion feature library, and dynamically adjust the bandwidth parameter according to the distribution sparsity of the sample feature vectors in the contaminant particle feature library and the micro-protrusion feature library.

[0020] Use the Gaussian kernel function to map the first kernel space similarity and the second kernel space similarity to a high-dimensional space respectively, and perform non-parametric kernel density estimation on the kernel space similarities of all samples in the contaminant particle feature library and the micro-protrusion feature library to generate the kernel density distribution function of the contaminant particle category and the kernel density distribution function of the micro-protrusion category respectively.

[0021] Substitute the first kernel space similarity of the current morphology-composition fusion feature vector into the kernel density distribution function of the contaminated particle category to calculate its probability density value as the first posterior probability, and substitute the second kernel space similarity into the kernel density distribution function of the micro-protrusion category to calculate its probability density value as the second posterior probability.

[0022] Further, the method for dynamically adjusting the bandwidth parameter according to the distribution sparsity of the sample feature vectors in the contaminated particle feature library and the micro-protrusion feature library includes: Calculate the average Mahalanobis distance between samples in each feature library as a distribution sparsity metric according to the spatial distributions of all sample feature vectors in the preset contaminated particle feature library and micro-protrusion feature library; generate the bandwidth parameter adjustment coefficients for the contaminated particle feature library and the micro-protrusion feature library respectively according to the sparsity ratio between the distribution sparsity metric and the preset reference sparsity.

[0023] When the distribution sparsity metric is greater than the preset reference sparsity, the bandwidth parameter adjustment coefficient is the square root of the sparsity ratio, otherwise it is the square value of the sparsity ratio; multiply the bandwidth parameter adjustment coefficient by the initial bandwidth parameter to obtain the dynamically adjusted kernel function bandwidth parameter, and apply it to the construction of the kernel density distribution functions of the contaminated particle category and the micro-protrusion category respectively.

[0024] Further, the method for mapping the discrimination result to the three-dimensional coordinates of the target surface, marking the positions and confidence levels of the contaminated particles and micro-protrusions, and generating a detection report includes: Establish a mapping matrix between the center points of the detection sub-regions and the three-dimensional coordinate system of the target surface according to the division order of the detection sub-regions and the spatial position relationship of the scanning path; convert the positions of the contaminated particles and micro-protrusions in the defect type discrimination result into three-dimensional space coordinates on the target surface according to the mapping matrix, and attach the corresponding similarity probability as a confidence score to each defect three-dimensional coordinate.

[0025] Perform spatial clustering analysis on all defect three-dimensional coordinates to statistically analyze the defect distribution density, spatial aggregation characteristics, and weighted average of the confidence scores of the contaminated particles and micro-protrusions. At the same time, calculate the component proportion weights of different elements in the contaminated particle region and the micro-protrusion region according to the characteristic X-ray peak intensity; integrate the defect distribution density, spatial aggregation characteristics, component proportion weights, and the weighted average of the corresponding confidence scores into a structured data table, and generate a detection report including a defect spatial position distribution map, an element composition weight distribution heat map, and a confidence level hierarchical annotation map in combination with the three-dimensional coordinate mapping result; correct the statistical result of the defect distribution density through the weighted average algorithm with the weighted average of the confidence scores.

[0026] Further, the method for correcting the statistical result of the defect distribution density through the weighted average algorithm includes: Calculate the weight factors of the contamination particles and micro-protrusions in the detection sub-region according to the confidence scores attached to the three-dimensional coordinates of each defect. The weight factor is the ratio of the confidence score of the current defect coordinate to the sum of the confidence scores of all similar defects; perform a weighted sum on the defect distribution density according to the weight factor to obtain the mean value of the confidence-weighted defect distribution density.

[0027] Compare and analyze the mean value of the defect distribution density and the preset confidence interval with the unweighted defect distribution density, generate a corrected defect distribution density statistical result and embed it in the structured data table of the detection report.

[0028] Further, the method for converting the positions of the contamination particles and micro-protrusions in the defect type discrimination result into three-dimensional space coordinates on the target surface according to the mapping matrix includes: Extract the detection sub-region numbers marked as contamination particles or micro-protrusions in the discrimination result according to the division index number of the detection sub-region. According to the corresponding relationship between the detection sub-region numbers stored in the mapping matrix and the three-dimensional coordinates, map the detection sub-region numbers corresponding to each defect to the three-dimensional space coordinates on the target surface.

[0029] (3) Beneficial effects Compared with the prior art, the beneficial effects of the present invention are: 1. By combining a scanning electron microscope to obtain surface topography images and an energy dispersive X-ray spectrometer to collect compositional spectrum data, information is extracted from two dimensions of topography (micro-region height gradient, radius of curvature, edge sharpness) and composition (characteristic X-ray peak intensity), and a topography-composition fusion feature vector is formed. By comparing and analyzing with a preset contamination particle feature library and micro-protrusion feature library, the defect type discrimination result of the detection sub-region can be accurately obtained, effectively improving the accuracy and comprehensiveness of the surface defect detection of the magnesium oxide single crystal target, and avoiding missed detection or misjudgment that may be caused by a single detection method.

[0030] 2. In terms of compositional spectrum data processing, multi-frame superposition processing is adopted and the number of superposed frames is dynamically adjusted to improve the signal-to-noise ratio according to the priority of the detection sub-region; a composite background model is constructed and the continuous background noise is subtracted by fitting using the non-linear least squares method. At the same time, the candidate characteristic peaks are dynamically calibrated and the pseudo-characteristic peaks are removed. It can significantly improve the quality of the compositional spectrum data, remove noise interference, so as to more accurately extract the characteristic X-ray peak intensity, provide a reliable data basis for subsequent defect discrimination, and improve the reliability of the detection result.

[0031] 3. By establishing the mapping relationship between the pixel coordinates of the detected sub-region image and the energy channel position of the component spectrum data, as well as the mapping matrix between the center point of the detected sub-region and the three-dimensional coordinate system of the target surface, the defect type discrimination result is accurately mapped to the three-dimensional coordinates on the target surface, realizing the accurate marking of the defect position. At the same time, a detection report including the defect spatial position distribution map, the element composition weight distribution heat map, and the confidence level stratified annotation map is generated, which can intuitively display the detection results, facilitate a comprehensive understanding of information such as the distribution, composition, and confidence level of the defects on the target surface, and provide strong support for subsequent process improvement and quality control. Description of the Drawings

[0032] Figure 1 It is a flowchart of a surface detection method for a magnesium oxide single crystal target in Embodiment 1 of the present invention. Detailed Embodiments

[0033] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0034] Before giving examples, it is necessary to elaborate on the application scenario of the concept of the present invention. The present invention is a surface detection method for a magnesium oxide single crystal target, which is applied to the problems that it is difficult to automatically distinguish between tiny contaminant particles (such as <10nm metal particles) and micro-protrusion defects on the surface of a magnesium oxide single crystal target under weak signals and high noise interference, and the traditional detection relies on manual experience resulting in poor result consistency, effectively improving the accuracy and comprehensiveness of the defect detection on the surface of a magnesium oxide single crystal target, and providing strong support for the quality control of a magnesium oxide single crystal target.

[0035] Embodiment 1: As Figure 1 shown, this embodiment provides a surface detection method for a magnesium oxide single crystal target, and the method includes: S1. Divide the surface of the target into multiple detection sub-regions, and determine the priority scanning order of each detection sub-region; obtain the surface topography image of each detection sub-region through a scanning electron microscope, and simultaneously collect the compositional spectrum data of the corresponding detection sub-region through an energy dispersive X-ray spectrometer; in the embodiments of this specification, the surface of the target is divided into multiple detection sub-regions by a uniform division method. The number of detection sub-regions depends on the specific business scenario and is not specifically limited here. In a specific embodiment, the surface of a 100 mm × 100 mm magnesium oxide single crystal target is uniformly divided into 10 × 10 detection sub-regions, and each sub-region is 10 mm × 10 mm. Each sub-region included in the multiple detection sub-regions is set with a priority scanning order. Specifically, the priority scanning order is determined according to the pre-analysis result of the surface roughness of the target. Specifically, the pre-analysis result of the surface roughness of the target can be determined according to production experience. Since defects are likely to occur in the edge region due to machining stress concentration, the detection sub-regions in the edge region are given a higher priority; the central region is relatively stable and is given a lower priority. When detecting the surface of the target, a Zeiss GeminiSEM 500 scanning electron microscope equipped with an Oxford X-MaxN 80 energy dispersive X-ray spectrometer is used. In each detection sub-region, the working voltage of the scanning electron microscope is set to 10 kV, the working distance is 10 mm, and the magnification is 8000× to obtain the surface topography image; at the same time, the energy dispersive X-ray spectrometer synchronously obtains the compositional spectrum data with a collection time of 120 seconds, and the scanning path sequentially completes the scanning of all detection sub-regions according to an "S" shape route.

[0036] S2. After performing multi-frame superposition processing on the compositional spectrum data, remove the continuous background noise by non-linear least squares fitting and extract the characteristic X-ray peak intensity, which is denoted as a compositional parameter; extract the topography parameter from the surface topography image, and perform spatial registration on the topography parameter and the compositional parameter to obtain a topography-composition fusion feature vector; the topography parameter includes the micro-region height gradient, the local surface curvature radius, and the edge sharpness parameter.

[0037] S3. Compare and analyze the topography-composition fusion feature vector with a preset pollution particle feature library and a micro-protrusion feature library to obtain a similarity probability, and obtain a defect type discrimination result of the detection sub-region according to the similarity probability.

[0038] S4. Map the discrimination result to the three-dimensional coordinates of the target surface, mark the positions and confidence levels of the pollution particles and micro-protrusions, and generate a detection report.

[0039] The method of removing the continuous background noise by non-linear least squares fitting and extracting the characteristic X-ray peak intensity after performing multi-frame superposition processing on the compositional spectrum data includes: Perform multi-frame superposition processing on the component spectrum data by the weighted average method. The number of superposed frames in the multi-frame superposition processing is dynamically adjusted according to the priority of the current detection sub-region. When the priority of the detection sub-region is higher than the preset priority threshold, the number of superposed frames is gradually increased through an adaptive algorithm until the signal-to-noise ratio improvement rate of the superposed component spectrum data is greater than the preset improvement rate threshold; for each detection sub-region, 8 frames of original component spectrum data are first collected for superposition processing. Taking the high-priority edge detection sub-region as an example, according to the priority threshold (set to 7, full score 10), it is determined that the priority of this region is 8, higher than the threshold. Therefore, the number of superposed frames is gradually increased to 12 frames. At this time, the signal-to-noise ratio improvement rate reaches 35%, exceeding the preset 30% threshold, and the multi-frame superposition is completed.

[0040] Construct a composite background model that includes polynomial basis functions and exponential decay basis functions. Obtain the weight coefficients of the polynomial basis functions and exponential decay basis functions through non-linear least squares fitting of the component spectrum data after multi-frame superposition processing. Substitute the weight coefficients into the composite background model to obtain continuous background noise covering the entire energy range; remove the continuous background noise from the component spectrum data after multi-frame superposition processing to obtain denoised component spectrum data; the constructed composite background model includes a cubic polynomial basis function and two exponential decay basis functions , where E represents the energy value. Solve the optimal values of the parameters through non-linear least squares to minimize the fitting error between the composite model and the superposed component spectrum data. Among them, is the constant term of the polynomial basis function, representing the baseline shift of the background noise near zero energy; is the coefficient of the first-order term of energy, fitting the linear change trend of the background noise with energy; is the coefficient of the second-order term of energy, fitting the parabolic change of the background noise; is the coefficient of the third-order term of energy, fitting the high-order non-linear change of the background noise. , is the amplitude coefficient of the exponential decay basis function, used to control the initial intensity of the background noise; , is the decay rate coefficient of the exponential decay basis function, determining the speed at which the background noise decreases with increasing energy. The composite background model can effectively fit the background noise characteristics in different energy intervals, especially suitable for processing the complex background noise of magnesium oxide single crystal targets.

[0041] According to the preset characteristic energy range of the target element, the sliding window method is used to screen candidate characteristic peaks that meet the local maximum condition and have a full width at half maximum within a preset width interval in the denoised component spectrum data. The peak area integral value is calculated based on the energy channel interval of each candidate characteristic peak as the characteristic X-ray peak intensity, and the left and right energy boundaries of each candidate characteristic peak are dynamically calibrated. The dynamic calibration determines the starting point of the peak boundary according to the change trend of the first derivative of the count rate of the energy channel corresponding to the candidate characteristic peak. If the correlation coefficient of the count rates of adjacent energy channels is lower than the preset correlation threshold, it is determined as a pseudo-characteristic peak formed by isolated noise points and is excluded; the count rate is the number of X-ray photons detected by the energy dispersive X-ray spectrometer per unit time in each energy channel. When screening candidate characteristic peaks in the denoised component spectrum data, the sliding window width is set to 50 eV, and the screening conditions include: determination of local maximum (the count rate of the current point is higher than the count rates of the points in the 3 energy channels on the left and right respectively), and the full width at half maximum is within the preset interval of 40 - 200 eV. For common metal contamination elements such as Fe, Cu, and Al on the surface of the magnesium oxide single crystal target, energy channel intervals of ±50 eV are set near their characteristic energies (Fe Kα 6.4 keV, Cu Kα 8.0 keV, Al Kα ~1.5 keV), and the peak area integral value is calculated as the characteristic X-ray peak intensity. During the dynamic calibration process, the starting point of the peak boundary is determined by calculating the change trend of the first derivative of the count rate of the energy channel of each candidate characteristic peak. For example, for the Fe Kα characteristic peak, by calculating the correlation coefficient of the count rates of adjacent energy channels in the interval of 6.35 - 6.45 keV, if it is lower than the preset correlation threshold of 0.65, it is determined as a pseudo-characteristic peak and excluded, effectively avoiding the interference of false peaks formed by noise fluctuations.

[0042] The method for extracting the morphology parameters from the surface morphology image and spatially registering the morphology parameters with the composition parameters to obtain the morphology-composition fusion feature vector includes: Perform three-dimensional topography reconstruction on the surface topography images of each detection sub-region to obtain a height distribution matrix; calculate the magnitude of the height gradient of each pixel point within a preset neighborhood window according to the height distribution matrix as the micro-region height gradient, and calculate the local surface curvature radius through quadratic surface fitting; extract the pixel gray-scale change rate along the edge contour of the surface topography image, and take the ratio of the maximum gray-scale change rate to the gray-scale variance of the adjacent region as the edge sharpness parameter; perform three-dimensional topography reconstruction on the SEM image, and use the multi-light-source stereoscopic imaging algorithm combined with shadow information analysis to construct a height distribution matrix with a pixel resolution of 1024×1024. Within a neighborhood window of 9×9 pixels, calculate the height gradient vector of the central pixel point relative to the surrounding pixel points, and take the magnitude of the height gradient vector as the micro-region height gradient value. Fit a quadratic surface z = ax² + by² + cxy + dx + ey + f to each central pixel point and its neighborhood pixels by the least squares method, and calculate the principal curvature according to the fitting coefficients a, b, and c , and take the curvature radius as the local surface curvature radius. Along the edge contour of the SEM image, calculate the pixel gray-scale change rate maxgrad perpendicular to the edge direction, and at the same time calculate the gray-scale variance var of each 2.5-pixel region (a total of 5 pixels) on both sides of the edge, and take the edge sharpness parameter s = maxgrad / var.

[0043] Normalize the micro-region height gradient, local surface curvature radius, and edge sharpness parameter according to the pixel coordinates of the detection sub-region to obtain a topography feature vector; establish a mapping relationship between the image pixel coordinates of the detection sub-region and the energy channel position of the compositional spectrum data synchronously according to the scanning paths of the scanning electron microscope and the energy-dispersive X-ray spectrometer; according to the mapping relationship between the image pixel coordinates of the detection sub-region and the energy channel position of the compositional spectrum data, splice the characteristic X-ray peak intensity and the topography feature vector of the corresponding detection sub-region according to the dimension to obtain a topography-composition fusion feature vector. To achieve the spatial registration of the topography parameters and the composition parameters, use the synchronous control interface of the scanning electron microscope and the energy-dispersive X-ray spectrometer to record the acquisition time of each pixel point and the corresponding energy channel position during the scanning process. Establish a mapping relationship between the image pixel coordinates (x, y) and the energy channel position (E, n) of the compositional spectrum data according to the synchronous signal, where E represents the energy value and n represents the channel number.

[0044] The method for synchronously establishing the mapping relationship between the image pixel coordinates of the detection sub-region and the energy channel position of the compositional spectrum data according to the scanning paths of the scanning electron microscope and the energy-dispersive X-ray spectrometer includes: Determine the synchronization mark between the pixel coordinates of the surface topography image and the acquisition time of the compositional spectrum data for each detection sub-region according to the correspondence between the scanning step of the scanning electron microscope and the sampling time interval of the energy-dispersive X-ray spectrometer; extract the real-time coordinate data of the stage position encoder of the scanning electron microscope according to the synchronization mark, and combine the calibration parameters of the energy channels and spatial positions of the energy-dispersive X-ray spectrometer to establish a geometric transformation matrix from image pixel coordinates to energy channel positions; determine the synchronization mark according to the correspondence between the scanning step of the scanning electron microscope (assumed to be 10 nm) and the sampling time interval of the energy-dispersive X-ray spectrometer (assumed to be 10 μs). For example, when the scanning electron microscope completes the scanning of a pixel point, a synchronization pulse signal is triggered, and after receiving the synchronization pulse signal, the energy-dispersive X-ray spectrometer starts to collect the compositional spectrum data at the current position. By extracting the real-time coordinate data (x, y, z) of the stage position encoder of the scanning electron microscope and combining the calibration parameters of the energy channels and spatial positions of the energy-dispersive X-ray spectrometer, a geometric transformation matrix T is established. The geometric transformation matrix can be expressed as , where is the translation vector [tx, ty, tz], and R is the rotation matrix, which contains the rotation angle parameters [θx, θy, θz].

[0045] Decompose the geometric transformation matrix into translation vector and rotation matrix components, and optimize the parameters of the translation vector and rotation matrix components by the least squares method to minimize the spatial deviation between the central pixel coordinates of the surface topography image and the central position of the energy channel of the compositional spectrum data in the same detection sub-region; apply the optimized geometric transformation matrix to all detection sub-regions to achieve a point-by-point mapping between image pixel coordinates and energy channel positions. To optimize the parameters of the geometric transformation matrix, the least squares method is used. Select 10 feature points, record their pixel coordinates in the surface topography image and the energy channel positions in the compositional spectrum data respectively, and construct an error function , where pi is the image pixel coordinate and qi is the corresponding energy channel position. Iteratively optimize the parameters by the gradient descent method to minimize the error function and obtain the optimal geometric transformation matrix to achieve an accurate mapping between pixel coordinates and energy channel positions.

[0046] The method of comparing the morphology-composition fusion feature vector with the preset pollution particle feature library and micro-protrusion feature library to obtain the similarity probability and obtaining the defect type discrimination result of the detection sub-region includes: After standardizing the morphology-composition fusion feature vector, calculate the first kernel space similarity between it and each sample feature vector in the pollution particle feature library according to the kernel function mapping, and calculate the second kernel space similarity between it and each sample feature vector in the micro-protrusion feature library; standardize the morphology-composition fusion feature vector, and convert the feature values of each dimension to the interval [0,1]. Select the radial basis function kernel , set the parameter γ to 0.1, and calculate the kernel space similarity between the current feature vector and each sample in the feature library. Calculate the first kernel space similarity between the current feature vector and each sample in the pollution particle feature library (including 100 typical samples) through the kernel function mapping , and the second kernel space similarity between it and each sample in the micro-protrusion feature library (also including 100 typical samples) .

[0047] Estimate the probability distribution of the first kernel space similarity in the pollution particle category and the probability distribution of the second kernel space similarity in the micro-protrusion category respectively through the probability density function, and obtain the first posterior probability that the current morphology-composition fusion feature vector belongs to the pollution particle and the second posterior probability that it belongs to the micro-protrusion; obtain the probability ratio of the first posterior probability to the second posterior probability, and when the probability ratio exceeds the preset probability ratio threshold, determine that the defect type is a pollution particle, otherwise determine that the defect type is a micro-protrusion; the similarity probability includes the first posterior probability and the second posterior probability. Estimate respectively through the probability density function in the pollution particle category and in the micro-protrusion category. For example, for a suspicious defect detected in the edge area of the target, the average value of the first kernel space similarity between the calculated morphology-composition fusion feature vector and the pollution particle feature library is 0.85, and the average value of the second kernel space similarity with the micro-protrusion feature library is 0.35. The posterior probability that the feature vector belongs to the pollution particle is obtained through kernel density estimation =0.78, and the posterior probability that it belongs to the micro-protrusion =0.22. Calculate the probability ratio / =3.55, which exceeds the preset probability ratio threshold of 2.5, so it is determined that the defect is a pollution particle.

[0048] The method of estimating the probability distribution of the first kernel space similarity in the pollution particle category and the probability distribution of the second kernel space similarity in the micro-protrusion category respectively through the probability density function, and obtaining the first posterior probability that the current morphology-composition fusion feature vector belongs to the pollution particle and the second posterior probability that it belongs to the micro-protrusion includes: Calculate the kernel function bandwidth parameter according to the eigenvalues of the covariance matrix of all sample feature vectors in the preset pollution particle feature library and micro-protrusion feature library, and dynamically adjust the bandwidth parameter according to the distribution sparsity of the sample feature vectors in the pollution particle feature library and micro-protrusion feature library; Calculate the kernel function bandwidth parameter according to the eigenvalues of the covariance matrix of all sample feature vectors in the pollution particle feature library and micro-protrusion feature library. For the pollution particle feature library, the average eigenvalue of the covariance matrix is 0.15, and the initial bandwidth parameter = 0.5×0.15^(0.5)=0.194; For the micro-protrusion feature library, the average eigenvalue of the covariance matrix is 0.18, and the initial bandwidth parameter = 0.5×0.18^(0.5)=0.212.

[0049] Use the Gaussian kernel function to map the first kernel space similarity and the second kernel space similarity to a high-dimensional space respectively, and perform non-parametric kernel density estimation on the kernel space similarities of all samples in the pollution particle feature library and micro-protrusion feature library to generate the kernel density distribution function of the pollution particle category and the kernel density distribution function of the micro-protrusion category respectively; Substitute the first kernel space similarity of the current morphology-composition fusion feature vector into the kernel density distribution function of the pollution particle category to calculate its probability density value as the first posterior probability, and substitute the second kernel space similarity into the kernel density distribution function of the micro-protrusion category to calculate its probability density value as the second posterior probability. Use the Gaussian kernel function to perform non-parametric kernel density estimation on the kernel space similarities of all samples in the pollution particle feature library and micro-protrusion feature library, and generate the kernel density distribution function of the pollution particle category and the kernel density distribution function of the micro-protrusion category respectively. For the current morphology-composition fusion feature vector, calculate the first kernel space similarity = 0.85, substitute it into to obtain the probability density value = 0.78 as the first posterior probability; Calculate the second kernel space similarity = 0.35, substitute it into to obtain the probability density value = 0.22 as the second posterior probability.

[0050] The method for dynamically adjusting the bandwidth parameter according to the distribution sparsity of the sample feature vectors in the pollution particle feature library and micro-protrusion feature library includes: Calculate the average Mahalanobis distance between samples in each feature library as a distribution sparsity metric according to the spatial distributions of all sample feature vectors in the preset pollution particle feature library and micro-protrusion feature library respectively; generate the bandwidth parameter adjustment coefficient for each of the pollution particle feature library and the micro-protrusion feature library according to the sparsity ratio between the distribution sparsity metric and the preset reference sparsity; for each pair of sample feature vectors xi and xj in the pollution particle feature library, calculate their Mahalanobis distance , where Σ is the sample covariance matrix. Take the average of the Mahalanobis distances of all sample pairs to obtain the distribution sparsity metric of the pollution particle feature library = 2.8; similarly calculate the distribution sparsity metric of the micro-protrusion feature library = 2.3.

[0051] When the distribution sparsity metric is greater than the preset reference sparsity, the bandwidth parameter adjustment coefficient is the square root of the sparsity ratio, otherwise it is the square value of the sparsity ratio; multiply the bandwidth parameter adjustment coefficient by the initial bandwidth parameter to obtain the dynamically adjusted kernel function bandwidth parameter, and apply it to the construction of the kernel density distribution functions for the pollution particle category and the micro-protrusion category respectively; the reference sparsity is obtained by statistically calculating the logarithm of the determinant of the covariance matrix of the sample feature vectors in the preset pollution particle feature library and micro-protrusion feature library. Reference sparsity is determined by the logarithm of the determinant of the covariance matrix of the sample feature vectors in the feature library, specifically as , where n is the dimension of the feature vector. Calculate the sparsity ratio , . When the sparsity ratio is greater than 1, the bandwidth parameter adjustment coefficient is the square root of the sparsity ratio; when the sparsity ratio is less than 1, the bandwidth parameter adjustment coefficient is the square value of the sparsity ratio. Therefore, the bandwidth parameter adjustment coefficient of the pollution particle feature library = 1.058; the bandwidth parameter adjustment coefficient of the micro-protrusion feature library = 0.846. Finally, the dynamically adjusted kernel function bandwidth parameters are , .

[0052] The method of mapping the discrimination result to the three-dimensional coordinates of the target surface, marking the positions and confidences of pollution particles and micro-protrusions and generating a detection report includes: Establish a mapping matrix between the center points of the detection sub-regions and the three-dimensional coordinate system of the target surface according to the division order of the detection sub-regions and the spatial position relationship of the scanning path; convert the positions of the contamination particles and micro-protrusions in the defect type discrimination result into three-dimensional spatial coordinates on the target surface according to the mapping matrix, and convert the corresponding similarity probability into a confidence score and attach it to each defect three-dimensional coordinate; establish a mapping matrix M between the center points of the detection sub-regions and the three-dimensional coordinate system of the target surface. For a target surface of 100mm×100mm, it is divided into 10×10 detection sub-regions, and each sub-region center point corresponds to a three-dimensional coordinate (X, Y, Z). For example, the three-dimensional coordinate of the center point of the detection sub-region located at the position (3, 4) mapped to the target surface is (35mm, 45mm, 0mm). According to the discrimination result, associate the defect type (contamination particle or micro-protrusion) with the detection sub-region number where it is located, and convert it into three-dimensional spatial coordinates on the target surface through the mapping matrix M. For example, a defect judged to be a contamination particle is located in the detection sub-region (3, 4), and according to the mapping matrix, its three-dimensional coordinate is (35mm, 45mm, 0mm), and at the same time, the corresponding similarity probability (such as 0.78) is converted into a confidence score of 78.

[0053] Perform spatial clustering analysis on the three-dimensional coordinates of all defects to statistically calculate the weighted average of the defect distribution density, spatial aggregation characteristics, and confidence score of pollution particles and micro-protrusions. At the same time, calculate the component proportion weights of different elements in the pollution particle region and the micro-protrusion region according to the characteristic X-ray peak intensity; integrate the defect distribution density, spatial aggregation characteristics, component proportion weights, and the weighted average of the corresponding confidence scores into a structured data table, and generate a detection report including a defect spatial position distribution map, an element composition weight distribution heat map, and a confidence level stratified annotation map by combining the structured data table with the three-dimensional coordinate mapping result; correct the statistical result of the defect distribution density by the weighted average algorithm using the weighted average of the confidence scores. Perform spatial clustering analysis on the three-dimensional coordinates of all defects using the DBSCAN algorithm with parameters set as eps = 5mm (clustering radius) and minPts = 3 (minimum number of points) to statistically calculate the defect distribution density of pollution particles and micro-protrusions. For example, in the edge region of the target (0 - 10mm range), the density of pollution particles is 3.6 per cm², and the density of micro-protrusions is 1.2 per cm²; in the central region of the target, the density of pollution particles is 0.8 per cm², and the density of micro-protrusions is 0.5 per cm². At the same time, calculate the component proportion weights of different elements in the pollution particle region and the micro-protrusion region according to the characteristic X-ray peak intensity. For example, in the pollution particle region, the proportion of Fe element is 15.3%, the proportion of Cu element is 8.7%, and the proportion of Al element is 5.2%; in the micro-protrusion region, the main components are Mg (70.5%) and O (29.2%), and almost no metal elements are contained. Integrate the defect distribution density, spatial aggregation characteristics, component proportion weights, and the weighted average of the corresponding confidence scores into a structured table, and generate three visualization charts: a defect spatial position distribution map, using a scatter plot, with different colors representing different types of defects; an element composition weight distribution heat map, using a color gradient to represent the distribution concentration of different elements; a confidence level stratified annotation map, dividing the confidence level into three layers: high, medium, and low, and marking them with red, yellow, and green respectively.

[0054] The method for correcting the statistical result of the defect distribution density by the weighted average algorithm includes: The weight factors for contaminants and micro-bumps within the detection sub-region are calculated based on the confidence scores assigned to each defect's three-dimensional coordinates. The weight factor is the ratio of the confidence score for the current defect coordinate to the sum of the confidence scores for all similar defects. The defect distribution density is weighted and summed according to the weight factors to obtain the confidence-weighted mean defect distribution density. For example, for a contaminant defect with a confidence score of 78, if the sum of the confidence scores for all contaminants within the region is 300, the weight factor for this defect is w = 78 / 300 = 0.26. The defect distribution density is weighted and summed according to the weight factors. Raw statistics show a contaminant particle density of 3.6 particles / cm² at the target edge. After factoring in confidence weighting, the weighted average density is 3.2 particles / cm².

[0055] The defect distribution density mean and the preset confidence interval are compared and analyzed with the unweighted defect distribution density to generate a revised defect distribution density statistical result and embed it into the structured data table of the test report. The preset confidence interval is [2.4, 4.0] pieces / cm², with a confidence level of 95%. The unweighted 3.6 pieces / cm² is within the confidence interval but deviates towards the upper limit, indicating that the defect density in the high-confidence area is slightly lower than the average level. Finally, in the test report, the revised contamination particle distribution density is expressed as 3.2±0.8 pieces / cm². This weighted average method can more accurately reflect the actual situation of the defect distribution and reduce the interference of low-confidence judgment results on statistical analysis.

[0056] The method of converting the contamination particles and micro-protrusion positions in the defect type discrimination result into three-dimensional spatial coordinates of the target surface according to the mapping matrix includes: Extract the detection sub-region numbers marked as contaminated particles or micro-protrusions in the discrimination results according to the division index numbers of the detection sub-regions. According to the corresponding relationship between the detection sub-region numbers stored in the mapping matrix and the three-dimensional coordinates, map the detection sub-region numbers corresponding to each defect to the three-dimensional space coordinates on the surface of the target material. Establish a mapping matrix to store the corresponding relationship between the detection sub-region numbers and the three-dimensional coordinates. For a standard 100mm×100mm magnesium oxide single crystal target, its surface is usually evenly divided into 10×10 detection sub-regions, and the size of each sub-region is 10mm×10mm. The mapping matrix M is in the form of a two-dimensional array, and M[i][j] represents the three-dimensional coordinates (X, Y, Z) corresponding to the detection sub-region (i, j), where i and j respectively represent the index numbers of the detection sub-region in the row and column directions, and the value range is from 1 to 10. The coordinate calculation formula is X = (i - 0.5)×10mm, Y = (j - 0.5)×10mm, Z = 0mm (indicating the plane of the target material surface). For example, the three-dimensional coordinates corresponding to the detection sub-region (1, 1) (i.e., the first region in the lower left corner) are (5mm, 5mm, 0mm); the three-dimensional coordinates corresponding to the detection sub-region (10, 10) (i.e., the last region in the upper right corner) are (95mm, 95mm, 0mm).

[0057] Finally, it should be noted that although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for surface detection of a magnesium oxide single crystal target material, characterized in that The method includes: Dividing the surface of the target into multiple detection sub-regions, and determining the priority scanning order of each detection sub-region; obtaining the surface topography image of each detection sub-region through a scanning electron microscope, and simultaneously collecting the compositional spectrum data of the corresponding detection sub-region through an energy dispersive X-ray spectrometer; After performing multi-frame superposition processing on the compositional spectrum data, removing the continuous background noise by non-linear least squares fitting and extracting the characteristic X-ray peak intensity, which is denoted as the compositional parameter; extracting the topography parameter from the surface topography image, and performing spatial registration on the topography parameter and the compositional parameter to obtain a topography-composition fusion feature vector; the topography parameter includes the micro-region height gradient, the local surface curvature radius, and the edge sharpness parameter; Comparing and analyzing the topography-composition fusion feature vector with a preset contamination particle feature library and a micro-protrusion feature library to obtain a similarity probability, and obtaining a defect type discrimination result of the detection sub-region according to the similarity probability; Mapping the discrimination result to the three-dimensional coordinates of the target surface, marking the positions and confidence levels of the contamination particles and micro-protrusions, and generating a detection report.

2. The surface detection method for preparing a magnesium oxide single crystal target according to claim 1, characterized in that The method of removing the continuous background noise by non-linear least squares fitting and extracting the characteristic X-ray peak intensity after performing multi-frame superposition processing on the compositional spectrum data includes: Performing multi-frame superposition processing on the compositional spectrum data by the weighted average method, and dynamically adjusting the number of superposed frames of the multi-frame superposition processing according to the priority of the current detection sub-region. When the priority of the detection sub-region is higher than the preset priority threshold, the number of superposed frames is gradually increased by an adaptive algorithm until the signal-to-noise ratio improvement rate of the compositional spectrum data after superposition is greater than the preset improvement rate threshold; Constructing a composite background model including a polynomial basis function and an exponential decay basis function, globally fitting the compositional spectrum data after multi-frame superposition processing by non-linear least squares to obtain the weight coefficients of the polynomial basis function and the exponential decay basis function, substituting the weight coefficients into the composite background model to obtain the continuous background noise covering the full energy range; removing the continuous background noise from the compositional spectrum data after multi-frame superposition processing to obtain the denoised compositional spectrum data; According to the preset characteristic energy range of the target element, using the sliding window method to screen candidate characteristic peaks that meet the local maximum condition and have a full width at half maximum within a preset width interval in the denoised compositional spectrum data, and calculating the peak area integral value as the characteristic X-ray peak intensity according to the energy channel interval of each candidate characteristic peak, and dynamically calibrating the left energy boundary and the right energy boundary of each candidate characteristic peak. The dynamic calibration determines the starting point of the peak boundary according to the change trend of the first derivative of the counting rate of the energy channel corresponding to the candidate characteristic peak. If the correlation coefficient of the counting rates of adjacent energy channels is lower than the preset correlation threshold, it is determined as a pseudo-characteristic peak formed by isolated noise points and is excluded.

3. A surface detection method for preparing a magnesium oxide single crystal target according to claim 1, characterized in that, The method of extracting the topography parameter from the surface topography image and performing spatial registration on the topography parameter and the compositional parameter to obtain a topography-composition fusion feature vector includes: Perform three-dimensional topography reconstruction on the surface topography image of each detection sub-region to obtain a height distribution matrix; calculate the magnitude of the height gradient of each pixel point within a preset neighborhood window based on the height distribution matrix as the micro-region height gradient, and calculate the local surface curvature radius through quadratic surface fitting; extract the pixel gray-scale change rate along the edge contour of the surface topography image, and use the ratio of the maximum gray-scale change rate to the gray-scale variance of the adjacent region as the edge sharpness parameter; Normalize the micro-region height gradient, local surface curvature radius, and edge sharpness parameter according to the pixel coordinates of the detection sub-region to obtain a topography feature vector; establish a mapping relationship between the image pixel coordinates of the detection sub-region and the energy channel position of the compositional spectrum data synchronously according to the scanning paths of the scanning electron microscope and the energy-dispersive X-ray spectrometer; according to the mapping relationship between the image pixel coordinates of the detection sub-region and the energy channel position of the compositional spectrum data, splice the characteristic X-ray peak intensity and the topography feature vector of the corresponding detection sub-region according to the dimension to obtain a topography-composition fusion feature vector.

4. A surface detection method for preparing a magnesium oxide single crystal target according to claim 3, characterized in that, The method for establishing the mapping relationship between the image pixel coordinates of the detection sub-region and the energy channel position of the compositional spectrum data synchronously according to the scanning paths of the scanning electron microscope and the energy-dispersive X-ray spectrometer includes: Determine the synchronization mark between the surface topography image pixel coordinates of each detection sub-region and the acquisition time of the compositional spectrum data according to the corresponding relationship between the scanning step of the scanning electron microscope and the sampling time interval of the energy-dispersive X-ray spectrometer; extract the real-time coordinate data of the stage position encoder of the scanning electron microscope according to the synchronization mark, and combine the calibration parameters of the energy channel and the spatial position of the energy-dispersive X-ray spectrometer to establish a geometric transformation matrix from the image pixel coordinates to the energy channel position; Decompose the geometric transformation matrix into a translation vector and a rotation matrix component, and optimize the parameters of the translation vector and the rotation matrix component by the least squares method to minimize the spatial deviation between the central pixel coordinates of the surface topography image and the central position of the energy channel of the compositional spectrum data in the same detection sub-region; apply the optimized geometric transformation matrix to all detection sub-regions to realize the point-by-point mapping between the image pixel coordinates and the energy channel position.

5. A surface detection method for preparing a magnesium oxide single crystal target according to claim 1, characterized in that The method for comparing and analyzing the topography-composition fusion feature vector with a preset pollution particle feature library and micro-protrusion feature library to obtain a similarity probability, and obtaining the defect type discrimination result of the detection sub-region according to the similarity probability includes: After performing normalization processing on the topography-composition fusion feature vector, calculate the first kernel space similarity with each sample feature vector in the pollution particle feature library and the second kernel space similarity with each sample feature vector in the micro-protrusion feature library according to the kernel function mapping; Estimate the probability distributions of the first kernel space similarity in the category of contaminated particles and the second kernel space similarity in the category of micro-protrusions through probability density functions respectively, to obtain the first posterior probability that the current morphology-composition fusion feature vector belongs to contaminated particles and the second posterior probability that it belongs to micro-protrusions; obtain the probability ratio of the first posterior probability to the second posterior probability, and determine that the defect type is contaminated particles when the probability ratio exceeds a preset probability ratio threshold, otherwise determine that the defect type is micro-protrusions; the similarity probability includes the first posterior probability and the second posterior probability.

6. The surface detection method for preparing a magnesium oxide single crystal target according to claim 5, characterized in that The method for estimating the probability distributions of the first kernel space similarity in the category of contaminated particles and the second kernel space similarity in the category of micro-protrusions through probability density functions respectively, to obtain the first posterior probability that the current morphology-composition fusion feature vector belongs to contaminated particles and the second posterior probability that it belongs to micro-protrusions includes: Calculate the kernel function bandwidth parameter according to the covariance matrix eigenvalues of all sample feature vectors in the preset contaminated particle feature library and micro-protrusion feature library, and dynamically adjust the bandwidth parameter according to the distribution sparsity of the sample feature vectors in the contaminated particle feature library and micro-protrusion feature library; Use the Gaussian kernel function to map the first kernel space similarity and the second kernel space similarity to a high-dimensional space respectively, and perform non-parametric kernel density estimation on the kernel space similarities of all samples in the contaminated particle feature library and micro-protrusion feature library to generate the kernel density distribution function of the contaminated particle category and the kernel density distribution function of the micro-protrusion category respectively; Substitute the first kernel space similarity of the current morphology-composition fusion feature vector into the kernel density distribution function of the contaminated particle category to calculate its probability density value as the first posterior probability, and substitute the second kernel space similarity into the kernel density distribution function of the micro-protrusion category to calculate its probability density value as the second posterior probability.

7. A surface detection method for preparing a magnesium oxide single crystal target according to claim 6, characterized in that, The method for dynamically adjusting the bandwidth parameter according to the distribution sparsity of the sample feature vectors in the contaminated particle feature library and micro-protrusion feature library includes: Calculate the average Mahalanobis distance between samples in each feature library as the distribution sparsity metric according to the spatial distributions of all sample feature vectors in the preset contaminated particle feature library and micro-protrusion feature library respectively; generate the bandwidth parameter adjustment coefficients for the contaminated particle feature library and micro-protrusion feature library respectively according to the sparsity ratio of the distribution sparsity metric to the preset reference sparsity; When the distribution sparsity metric is greater than the preset reference sparsity, the bandwidth parameter adjustment coefficient is the square root of the sparsity ratio, otherwise it is the square value of the sparsity ratio; multiply the bandwidth parameter adjustment coefficient by the initial bandwidth parameter to obtain the dynamically adjusted kernel function bandwidth parameter, and apply it to the construction of the kernel density distribution functions of the contaminated particle category and micro-protrusion category respectively.

8. A surface detection method for preparing a magnesium oxide single crystal target according to claim 1, characterized in that, The method for mapping the discrimination result to the three-dimensional coordinates of the target surface, marking the positions and confidences of contaminated particles and micro-protrusions and generating a detection report includes: Establish a mapping matrix between the center points of the detection sub-regions and the three-dimensional coordinate system of the target surface according to the division order of the detection sub-regions and the spatial position relationship of the scanning path; according to the mapping matrix, convert the positions of the contamination particles and micro-protrusions in the defect type discrimination result into three-dimensional spatial coordinates on the target surface, and convert the corresponding similarity probability into a confidence score and append it to each defect three-dimensional coordinate; Perform spatial clustering analysis on all defect three-dimensional coordinates to statistically analyze the defect distribution density, spatial aggregation characteristics, and weighted average of the confidence scores of the contamination particles and micro-protrusions. At the same time, calculate the component proportion weights of different elements in the contamination particle region and the micro-protrusion region according to the characteristic X-ray peak intensity; integrate the defect distribution density, spatial aggregation characteristics, component proportion weights, and the weighted average of the corresponding confidence scores into a structured data table, and generate a detection report including a defect spatial position distribution map, an element composition weight distribution heat map, and a confidence level stratified annotation map in combination with the three-dimensional coordinate mapping result; correct the statistical result of the defect distribution density by the weighted average algorithm using the weighted average of the confidence scores.

9. A method for surface detection of a magnesium oxide single crystal target material according to claim 8, characterized in that The method for correcting the statistical result of the defect distribution density by the weighted average algorithm includes: Calculate the weight factor of the contamination particles and micro-protrusions in the detection sub-region according to the confidence score appended to each defect three-dimensional coordinate, where the weight factor is the ratio of the confidence score of the current defect coordinate to the sum of the confidence scores of all similar defects; perform weighted summation on the defect distribution density according to the weight factor to obtain the confidence-weighted defect distribution density mean value; Compare and analyze the defect distribution density mean value and the preset confidence interval with the unweighted defect distribution density, generate the corrected defect distribution density statistical result, and embed it in the structured data table of the detection report.

10. A method for surface detection of a magnesium oxide single crystal target material according to claim 8, characterized in that, The method for converting the positions of the contamination particles and micro-protrusions in the defect type discrimination result into three-dimensional spatial coordinates on the target surface according to the mapping matrix includes: Extract the detection sub-region numbers marked as contamination particles or micro-protrusions in the discrimination result according to the division index number of the detection sub-region, and map the detection sub-region number corresponding to each defect to the three-dimensional spatial coordinates on the target surface according to the corresponding relationship between the detection sub-region number and the three-dimensional coordinates stored in the mapping matrix.

Images