On-line detection system for finishing surface of photomask substrate

Abstract

The application relates to the technical field of photo mask substrate finishing, in particular to a photo mask substrate finishing surface online detection system, which comprises a multi-source collection module, a data preprocessing module and the like.The multi-source collection module receives micro-morphology data, optical performance data and positioning data of a photo mask substrate through an integrated sensing unit and is used for executing a multi-dimensional data synchronous collection operation; the data preprocessing module executes time-space alignment, abnormal value elimination, noise suppression and standardization operations based on the collected multi-source data. The integrated collection probe is used to verify the collection synchronism by using a synchronous precision checking formula, multi-source synchronous collection of the micro-morphology, optical performance and positioning data is realized, time-space deviation of the detection data is eliminated, cross-dimension data accurate correlation is established, multi-source data is cooperatively pretreated, the quality of the original data is improved, and a reliable foundation is laid for subsequent detection and analysis.

Description

Technical Field

[0001] This invention relates to the field of photomask substrate finishing technology, specifically to an online inspection system for the surface of photomask substrate finishing. Background Technology

[0002] As a core material in semiconductor lithography, the microscopic morphology precision and optical performance parameters of the photomask substrate directly determine the chip lithography resolution and product yield.

[0003] Generally, the microstructure and optical performance testing of photomask substrates are performed in stages using discrete equipment, which has significant technical drawbacks: First, discrete testing leads to asynchronous multi-dimensional data acquisition, making it difficult to establish a precise correlation between morphological features and optical performance due to unavoidable time deviations and spatial misalignments. Second, the lack of a cross-dimensional collaborative calibration mechanism makes single-dimensional test results susceptible to interference from environmental noise and system errors, resulting in insufficient reliability. Third, the test parameters are mostly fixed configurations, which cannot be adapted to photomask substrates of different materials and thicknesses, resulting in poor versatility. Fourth, the testing and finishing equipment are disconnected, and the test results cannot be fed back to the processing system in real time to optimize parameters, leading to a high defect rate. Fifth, the test data is stored in a scattered manner, lacking full-process traceability capabilities, making it difficult to iteratively optimize the test model and parameters through historical data.

[0004] In summary, an online inspection system for the fine-machining surface of photomask substrates is needed to solve the above problems. Summary of the Invention

[0005] The purpose of this invention is to provide an online inspection system for the fine-machining surface of photomask substrates, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: This invention proposes an online inspection system for the fine-machining surface of a photomask substrate, comprising: A multi-source acquisition module receives microscopic morphology data, optical performance data, and positioning data of a photomask substrate through an integrated sensing unit, and is used to perform multi-dimensional data synchronous acquisition operations. The data preprocessing module performs time-space alignment, outlier removal, noise suppression, and standardization operations based on the collected multi-source data. The dual-dimensional analysis module performs microscopic morphology feature extraction and defect identification, optical performance parameter calculation and uniformity analysis based on the preprocessed data. A cross-dimensional calibration module, which uses a morphology-optical correlation model to perform deviation calibration and confidence verification operations for two types of detection results; The output feedback module is used to perform multi-format output of test results, accuracy evaluation, and feedback of parameters of the finishing equipment, so as to realize closed-loop linkage between testing and processing.

[0007] Preferably, it further includes: The dynamic adaptation and adjustment module is used to dynamically adjust the sensor acquisition parameters, feature extraction thresholds, and calibration model coefficients according to the material type, thickness specifications, and accuracy requirements of the photomask substrate to adapt to different detection scenarios.

[0008] Preferably, it further includes: The data storage and traceability module is used to persistently store the collected raw data, preprocessing results, test analysis reports, calibration parameters and equipment feedback records, and supports accurate data retrieval and full-process traceability by substrate number and test time.

[0009] Based on the online inspection system for the fine-machining surface of photomask substrates, this invention also proposes an inspection method, comprising the following steps: S1. The microstructure, optical performance and real-time positioning data of the photomask substrate are collected synchronously through the multi-source acquisition module to obtain a multi-dimensional raw data set; S2. The data preprocessing module performs temporal and spatial alignment on the raw data, removes outlier data points, suppresses environmental noise interference, and generates a standardized dataset. S3. The dual-dimensional analysis module extracts microscopic morphological features from standardized data and identifies defect types and sizes, calculates core optical performance parameters, and analyzes uniformity distribution. S4. The cross-dimensional calibration module corrects the systematic bias of the two types of detection results based on the morphology-optical correlation model, calculates the comprehensive detection confidence, and screens qualified detection data. S5. The output feedback module verifies the accuracy of the calibrated test results, outputs test results in multiple formats, and simultaneously sends parameter adjustment instructions to the finishing equipment. S6. Record the entire process data and operating parameters from steps S1 to S5, optimize the detection configuration through the dynamic adaptation adjustment module, and iteratively update the associated model based on historical data to improve detection adaptability and accuracy.

[0010] Preferably, the implementation process of step S1 is as follows: An integrated atomic force microscope sensing unit, a deep ultraviolet spectroscopy sensing unit, and a laser positioning unit are combined to form an integrated acquisition probe. Set the acquisition parameters: atomic force microscope scanning resolution ≤ 0.01 nm, spectral sensing unit covers the 193 nm-248 nm deep ultraviolet band, and laser positioning unit positioning error ≤ 0.1 μm; Based on the trigger signal output by the positioning unit, the topography and optical sensing unit are synchronously acquired. Data transmission adopts real-time Ethernet with a transmission delay of ≤10ms. It supports two modes: continuous scanning acquisition and fixed-point verification acquisition, and automatically switches according to the substrate processing progress. The synchronization accuracy verification formula is used to verify the acquisition synchronization, as shown in equation (1): (1); In the formula, To ensure synchronous acquisition accuracy, For time synchronization error, For spatial synchronization error, For the collection period, For the reference positioning accuracy; when If the synchronization is deemed successful, the acquisition parameters will be readjusted.

[0011] Preferably, the implementation process of step S2 is as follows: Time alignment uses UTC timestamps for unified calibration, and spatial alignment establishes a mapping relationship between the topography detection area and the optical detection area based on laser positioning coordinates. An improved local outlier factor algorithm is used to remove outliers from the morphological data, as shown in equation (2): (2); In the formula, For data points Enhanced anomaly score, For the number of nearest neighbors, for Local density, For the first Local density of nearest neighbors for With the The distance between the nearest neighbors; when When an anomaly is identified, its location is marked. The Savitzky-Golay smoothing algorithm is used to suppress optical data noise and preserve characteristic peak signals; The topographic data is standardized to the [0,1] interval, and the optical data is converted into standardized performance indicators to generate a standardized dataset that is consistent in time and space.

[0012] Preferably, the implementation process of step S3 is as follows: Microscopic morphology feature extraction includes arithmetic mean roughness Ra, maximum contour height Rz, defect type and size parameters. Defect identification is achieved through morphological corrosion expansion and threshold segmentation, distinguishing three types of defects: scratches, particles and dents. The calculation of optical performance parameters includes transmittance at a specific wavelength, refractive index uniformity, and polarization state deviation. The core parameters are obtained through spectral fitting and interference fringe analysis. The uniformity of parameter spatial distribution is analyzed using the formula for the uniformity deviation of optical performance, as shown in equation (3): (3); In the formula, For optical performance uniformity deviation, This is the set of optical parameters at the target wavelength. The maximum value of the parameter. For the minimum value of the parameter, The mean of the parameters; Preliminary criteria: Ra≤0.5nm, transmittance≥99.5%, If the percentage is ≤0.3%, and all criteria are met, the sample is marked as a preliminary qualified sample.

[0013] Preferably, the implementation process of step S4 is as follows: A gradient-enhanced tree-based morphology-optical correlation model was constructed to establish the mapping relationship between defect area, depth and transmittance attenuation. Based on the model, the optical performance test results are corrected, and the calibration value is calculated using the cross-dimensional collaborative calibration formula, as shown in equation (4): (4); In the formula, These are the calibrated optical measurement values. These are the original optical detection values. The correlation coefficient, The diameter of the morphological defect. For defect depth, The area to be detected; The value is preset to 0.01-0.05 depending on the substrate material. The reliability of the detection is evaluated using the comprehensive detection confidence formula, as shown in equation (5): (5); In the formula, To assess the overall confidence level, This is a weighting coefficient, ranging from 0.4 to 0.6, which is dynamically adjusted according to the defect density. Confidence level for morphology detection. For optical detection confidence level; when ≥ The test results should be retained. < This will trigger a re-collection and analysis of the data in that area; Output the calibrated microstructure test report and optical performance test report.

[0014] Preferably, the implementation process of step S5 is as follows: Accuracy verification employs two criteria: repeatability measurement error for morphology detection ≤ 0.02 nm, and comparison error with standard samples for optical detection ≤ 0.1%. Anomalies are classified into minor anomalies and major anomalies. A minor anomaly is when a single indicator exceeds the acceptable range but does not affect the use of the product. A major anomaly is when any indicator is seriously out of standard or the defect density is ≥3 defects / cm². Outputs CSV format inspection reports, JSON format real-time data, visual defect distribution maps, and optical performance heat maps; provides a TCP / IP interface for interface with the control system of precision machining equipment. Feedback is sent to the finishing equipment regarding the substrate offset and processing power correction value, with an adjustment response time of ≤200ms.

[0015] Preferably, the implementation process of step S6 is as follows: Record all process data, including acquisition parameters, preprocessing parameters, test results, calibration coefficients, equipment feedback instructions, and execution effects; The parameters of the correlation model are optimized weekly based on historical test data, and the feature extraction thresholds and pass standards are updated monthly based on test data of different substrate types. Establish a performance monitoring mechanism to display the detection accuracy, data acquisition success rate, and interface response time in real time. When any indicator fails to meet the preset standard three times in a row, the dynamic adaptation and adjustment module will be automatically triggered to perform full parameter optimization.

[0016] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention achieves multi-source synchronous acquisition of microscopic morphology, optical performance, and positioning data by using an integrated acquisition probe and verifying the acquisition synchronization using a synchronous accuracy verification formula. This eliminates spatiotemporal deviations in detection data, establishes accurate cross-dimensional data correlations, and corrects systematic deviations in two types of detection results by constructing a gradient-enhanced tree-based morphology-optical correlation model and cross-dimensional collaborative calibration. It also filters qualified data based on a comprehensive detection confidence threshold. Furthermore, by setting a dynamic adaptation adjustment module, it dynamically adjusts acquisition parameters, feature extraction thresholds, and calibration model coefficients according to the material type, thickness specifications, and accuracy requirements of the photomask substrate, thus adapting to multiple types of detection. To enhance the system's versatility and adaptability, the detection output and feedback module provides real-time feedback to the finishing equipment regarding substrate offset adjustments and processing power corrections. This establishes a closed-loop linkage between detection and processing, enabling timely optimization of processing parameters and reducing substrate defect rates. The data storage and traceability module persistently stores the entire process of detection data. Combined with an iterative mechanism that optimizes the correlation model weekly and updates the detection standards monthly, the system achieves full-process traceability of detection data and continuously improves detection accuracy. Furthermore, by improving the local outlier factor algorithm to remove outliers and the Savitzky-Golay algorithm to suppress noise, the system performs collaborative preprocessing on multi-source data, improving the quality of raw data and laying a reliable foundation for subsequent detection and analysis. Attached Figure Description

[0017] Figure 1 The topology diagram of the online inspection system for the fine-machining surface of the photomask substrate of the present invention is shown; Figure 2 The flowchart of the online detection method for the integrated microstructure and optical properties of the finely processed surface of the photomask substrate of the present invention is shown. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1, please refer to Figure 1 This invention proposes an online inspection system for the fine-processed surface of a photomask substrate, comprising: The multi-source acquisition module receives microscopic morphology data, optical performance data, and positioning data of the photomask substrate through an integrated sensing unit, and is used to perform multi-dimensional data synchronous acquisition operations. The data preprocessing module performs time-space alignment, outlier removal, noise suppression, and standardization operations based on the collected multi-source data. The dual-dimensional analysis module performs microscopic morphology feature extraction and defect identification, optical performance parameter calculation and uniformity analysis based on the preprocessed data. The cross-dimensional calibration module uses a morphology-optical correlation model to perform deviation calibration and confidence verification operations for two types of detection results. The output feedback module is used to perform multi-format output of test results, accuracy evaluation, and parameter feedback of finishing equipment, so as to realize closed-loop linkage between testing and processing. In this embodiment, it should also be noted that it further includes: The dynamic adaptation and adjustment module is used to dynamically adjust the sensor acquisition parameters, feature extraction thresholds, and calibration model coefficients according to the material type, thickness specifications, and accuracy requirements of the photomask substrate, so as to adapt to different detection scenarios. In this embodiment, it should also be noted that it further includes: The data storage and traceability module is used to persistently store the collected raw data, preprocessing results, test analysis reports, calibration parameters and equipment feedback records. It supports accurate data retrieval and full-process traceability by substrate number and test time. Example 2, please refer to Figure 2 Based on the system, in practical applications, the detection method of the present invention based on the online inspection system for the fine processing surface of photomask substrates specifically includes the following steps: S1. The microstructure, optical performance and real-time positioning data of the photomask substrate are collected synchronously through the multi-source acquisition module to obtain a multi-dimensional raw data set; In this embodiment, it should also be noted that the implementation process of step S1 is as follows: An integrated atomic force microscope sensing unit, a deep ultraviolet spectroscopy sensing unit, and a laser positioning unit are combined to form an integrated acquisition probe. Set the acquisition parameters: atomic force microscope scanning resolution ≤ 0.01 nm, spectral sensing unit covers the 193 nm-248 nm deep ultraviolet band, and laser positioning unit positioning error ≤ 0.1 μm; Based on the trigger signal output by the positioning unit, the topography and optical sensing unit are synchronously acquired. Data transmission adopts real-time Ethernet with a transmission delay of ≤10ms. It supports two modes: continuous scanning acquisition and fixed-point verification acquisition, and automatically switches according to the substrate processing progress. The synchronization accuracy verification formula is used to verify the acquisition synchronization, as shown in equation (1): (1); In the formula, To ensure synchronous acquisition accuracy, For time synchronization error, For spatial synchronization error, For the collection period, For the reference positioning accuracy; when If synchronization is deemed successful, otherwise the acquisition parameters are readjusted. S2. The data preprocessing module performs temporal and spatial alignment on the raw data, removes outlier data points, suppresses environmental noise interference, and generates a standardized dataset. In this embodiment, it should also be noted that the implementation process of step S2 is as follows: Time alignment uses UTC timestamps for unified calibration, and spatial alignment establishes a mapping relationship between the topography detection area and the optical detection area based on laser positioning coordinates. An improved local outlier factor algorithm is used to remove outliers from the morphological data, as shown in equation (2): (2); In the formula, For data points Enhanced anomaly score, For the number of nearest neighbors, for Local density, For the first Local density of nearest neighbors for With the The distance between the nearest neighbors; when When an anomaly is identified, its location is marked. The Savitzky-Golay smoothing algorithm is used to suppress optical data noise and preserve characteristic peak signals; The topographic data is standardized to the [0,1] interval, and the optical data is converted into standardized performance indicators to generate a standardized dataset that is consistent in time and space. S3. The dual-dimensional analysis module extracts microscopic morphological features from standardized data and identifies defect types and sizes, calculates core optical performance parameters, and analyzes uniformity distribution. In this embodiment, it should also be noted that the implementation process of step S3 is as follows: Microscopic morphology feature extraction includes arithmetic mean roughness Ra, maximum contour height Rz, defect type and size parameters. Defect identification is achieved through morphological corrosion expansion and threshold segmentation, distinguishing three types of defects: scratches, particles and dents. The calculation of optical performance parameters includes transmittance at a specific wavelength, refractive index uniformity, and polarization state deviation. The core parameters are obtained through spectral fitting and interference fringe analysis. The uniformity of parameter spatial distribution is analyzed using the formula for the uniformity deviation of optical performance, as shown in equation (3): (3); In the formula, For optical performance uniformity deviation, This is the set of optical parameters at the target wavelength. The maximum value of the parameter. For the minimum value of the parameter, The mean of the parameters; Preliminary criteria: Ra≤0.5nm, transmittance≥99.5%, If the concentration is ≤0.3%, and all criteria are met, the sample is marked as a preliminary qualified sample. S4. The cross-dimensional calibration module corrects the systematic bias of the two types of detection results based on the morphology-optical correlation model, calculates the comprehensive detection confidence, and screens qualified detection data. In this embodiment, it should also be noted that the implementation process of step S4 is as follows: A gradient-enhanced tree-based morphology-optical correlation model was constructed to establish the mapping relationship between defect area, depth and transmittance attenuation. Based on the model, the optical performance test results are corrected, and the calibration value is calculated using the cross-dimensional collaborative calibration formula, as shown in equation (4): (4); In the formula, These are the calibrated optical measurement values. These are the original optical detection values. The correlation coefficient, Where is the diameter of the morphological defect, and is the depth of the defect. The area to be detected; The value is preset to 0.01-0.05 depending on the substrate material. The reliability of the detection is evaluated using the comprehensive detection confidence formula, as shown in equation (5): (5); In the formula, To assess the overall confidence level, This is a weighting coefficient, ranging from 0.4 to 0.6, which is dynamically adjusted according to the defect density. Confidence level for morphology detection. For optical detection confidence level; when ≥ The test results should be retained. < This will trigger a re-collection and analysis of the data in that area; Output the calibrated microstructure inspection report and optical performance inspection report; S5. The output feedback module verifies the accuracy of the calibrated test results, outputs test results in multiple formats, and simultaneously sends parameter adjustment instructions to the finishing equipment. In this embodiment, it should also be noted that the implementation process of step S5 is as follows: Accuracy verification employs two criteria: repeatability measurement error for morphology detection ≤ 0.02 nm, and comparison error with standard samples for optical detection ≤ 0.1%. Anomalies are classified into minor anomalies and major anomalies. A minor anomaly is when a single indicator exceeds the acceptable range but does not affect the use of the product. A major anomaly is when any indicator is seriously out of standard or the defect density is ≥3 defects / cm². Outputs CSV format inspection reports, JSON format real-time data, visual defect distribution maps, and optical performance heat maps; provides a TCP / IP interface for interface with the control system of precision machining equipment. Feedback of substrate adjustment offset and processing power correction value to the finishing equipment, with an adjustment response time ≤200ms; S6. Record the full process data and operating parameters of steps S1 to S5, optimize the detection configuration through the dynamic adaptation adjustment module, and iteratively update the correlation model by combining historical data to improve detection adaptability and accuracy. In this embodiment, it should also be noted that the implementation process of step S6 is as follows: Record all process data, including acquisition parameters, preprocessing parameters, test results, calibration coefficients, equipment feedback instructions, and execution effects; The parameters of the correlation model are optimized weekly based on historical test data, and the feature extraction thresholds and pass standards are updated monthly based on test data of different substrate types. Establish a performance monitoring mechanism to display the detection accuracy, data acquisition success rate, and interface response time in real time. When any indicator fails to meet the preset standard three times in a row, the dynamic adaptation and adjustment module will be automatically triggered to perform full parameter optimization.

[0020] Through the above steps, this invention achieves multi-source synchronous acquisition of microscopic morphology, optical performance, and positioning data by using an integrated acquisition probe and verifying acquisition synchronicity with a synchronous accuracy verification formula. This eliminates spatiotemporal deviations in detection data, establishes precise cross-dimensional data correlations, and corrects systematic deviations in two types of detection results by constructing a gradient-enhanced tree-based morphology-optical correlation model and cross-dimensional collaborative calibration. It also filters qualified data based on a comprehensive detection confidence threshold. A dynamic adaptation adjustment module dynamically adjusts acquisition parameters, feature extraction thresholds, and calibration model coefficients according to the material type, thickness specifications, and accuracy requirements of the photomask substrate, adapting to multiple detection scenarios and improving system versatility. The detection output and feedback module provides real-time feedback to the finishing equipment on substrate adjustment offset and processing power correction values, establishing a closed-loop linkage between detection and processing. This optimizes processing parameters and reduces substrate defect rates. A data storage and traceability module persistently stores the entire process of detection data. Combined with an iterative mechanism that optimizes the correlation model weekly and updates detection standards monthly, the invention achieves full-process traceability of detection data and continuously improves detection accuracy.

[0021] Meanwhile, by improving the local outlier factor algorithm to remove outliers and the Savitzky-Golay algorithm to suppress noise, collaborative preprocessing of multi-source data was performed to improve the quality of the original data and lay a reliable foundation for subsequent detection and analysis.

[0022] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An online inspection system for the surface finish of a photomask substrate, characterized in that, include: A multi-source acquisition module receives microscopic morphology data, optical performance data, and positioning data of a photomask substrate through an integrated sensing unit, and is used to perform multi-dimensional data synchronous acquisition operations. The data preprocessing module performs time-space alignment, outlier removal, noise suppression, and standardization operations based on the collected multi-source data. The dual-dimensional analysis module performs microscopic morphology feature extraction and defect identification, optical performance parameter calculation and uniformity analysis based on the preprocessed data. A cross-dimensional calibration module, which uses a morphology-optical correlation model to perform deviation calibration and confidence verification operations for two types of detection results; The output feedback module is used to perform multi-format output of test results, accuracy evaluation, and feedback of parameters of the finishing equipment, so as to realize closed-loop linkage between testing and processing.

2. The online inspection system for the fine-machining surface of a photomask substrate according to claim 1, characterized in that, Also includes: The dynamic adaptation and adjustment module is used to dynamically adjust the sensor acquisition parameters, feature extraction thresholds, and calibration model coefficients according to the material type, thickness specifications, and accuracy requirements of the photomask substrate to adapt to different detection scenarios.

3. The online inspection system for the fine-machining surface of a photomask substrate according to claim 2, characterized in that, Also includes: The data storage and traceability module is used to persistently store the collected raw data, preprocessing results, test analysis reports, calibration parameters and equipment feedback records, and supports accurate data retrieval and full-process traceability by substrate number and test time.

4. The detection method of the online inspection system for the fine-finished surface of a photomask substrate according to any one of claims 1-3, characterized in that, Includes the following steps: S1. The microstructure, optical performance and real-time positioning data of the photomask substrate are collected synchronously through the multi-source acquisition module to obtain a multi-dimensional raw data set; S2. The data preprocessing module performs temporal and spatial alignment on the raw data, removes outlier data points, suppresses environmental noise interference, and generates a standardized dataset. S3. The dual-dimensional analysis module extracts microscopic morphological features from standardized data and identifies defect types and sizes, calculates core optical performance parameters, and analyzes uniformity distribution. S4. The cross-dimensional calibration module corrects the systematic bias of the two types of detection results based on the morphology-optical correlation model, calculates the comprehensive detection confidence, and screens qualified detection data. S5. The output feedback module verifies the accuracy of the calibrated test results, outputs test results in multiple formats, and simultaneously sends parameter adjustment instructions to the finishing equipment. S6. Record the entire process data and operating parameters from steps S1 to S5, optimize the detection configuration through the dynamic adaptation adjustment module, and iteratively update the associated model based on historical data to improve detection adaptability and accuracy.

5. The method according to claim 4, characterized in that, The implementation process of step S1 is as follows: An integrated atomic force microscope sensing unit, a deep ultraviolet spectroscopy sensing unit, and a laser positioning unit are combined to form an integrated acquisition probe. Set the acquisition parameters: atomic force microscope scanning resolution ≤ 0.01 nm, spectral sensing unit covers the 193 nm-248 nm deep ultraviolet band, and laser positioning unit positioning error ≤ 0.1 μm; Based on the trigger signal output by the positioning unit, the topography and optical sensing unit are synchronously acquired. Data transmission adopts real-time Ethernet with a transmission delay of ≤10ms. It supports two modes: continuous scanning acquisition and fixed-point verification acquisition, and automatically switches according to the substrate processing progress. The synchronization accuracy verification formula is used to verify the acquisition synchronization, as shown in equation (1): （1）； In the formula, To ensure synchronous acquisition accuracy, This represents time synchronization error, and this represents spatial synchronization error. For the collection period, The baseline positioning accuracy is used; if the synchronization is deemed satisfactory, otherwise the acquisition parameters are readjusted.

6. The method according to claim 5, characterized in that, The implementation process of step S2 is as follows: Time alignment uses UTC timestamps for unified calibration, and spatial alignment establishes a mapping relationship between the topography detection area and the optical detection area based on laser positioning coordinates. An improved local outlier factor algorithm is used to remove outliers from the morphological data, as shown in equation (2): （2）； In the formula, For data points The enhanced anomaly score is the number of nearest neighbors. for Local density, For the first Local density of nearest neighbors for With the The distance between the nearest neighbors; when When an anomaly is identified, its location is marked. The Savitzky-Golay smoothing algorithm is used to suppress optical data noise and preserve characteristic peak signals; The topographic data is standardized to the [0,1] interval, and the optical data is converted into standardized performance indicators to generate a standardized dataset that is consistent in time and space.

7. The method according to claim 6, characterized in that, The implementation process of step S3 is as follows: Microscopic morphology feature extraction includes arithmetic mean roughness Ra, maximum contour height Rz, defect type and size parameters. Defect identification is achieved through morphological corrosion expansion and threshold segmentation, distinguishing three types of defects: scratches, particles and dents. The calculation of optical performance parameters includes transmittance at a specific wavelength, refractive index uniformity, and polarization state deviation. The core parameters are obtained through spectral fitting and interference fringe analysis. The uniformity of parameter spatial distribution is analyzed using the formula for the uniformity deviation of optical performance, as shown in equation (3): （3）； In the formula, For optical performance uniformity deviation, This is the set of optical parameters at the target wavelength. The maximum value of the parameter. Let be the minimum value of the parameter, and be the mean value of the parameter. Preliminary criteria: Ra≤0.5nm, transmittance≥99.5%, If the percentage is ≤0.3%, and all criteria are met, the sample is marked as a preliminary qualified sample.

8. The method according to claim 7, characterized in that, The implementation process of step S4 is as follows: A gradient-enhanced tree-based morphology-optical correlation model was constructed to establish the mapping relationship between defect area, depth and transmittance attenuation. Based on the model, the optical performance test results are corrected, and the calibration value is calculated using the cross-dimensional collaborative calibration formula, as shown in equation (4): （4）； In the formula, These are the calibrated optical measurement values. These are the original optical detection values. The correlation coefficient, The diameter of the morphological defect. For defect depth, The area to be detected; The value is preset to 0.01-0.05 depending on the substrate material. The reliability of the detection is evaluated using the comprehensive detection confidence formula, as shown in equation (5): （5）； In the formula, To assess the overall confidence level, This is a weighting coefficient, ranging from 0.4 to 0.6, which is dynamically adjusted according to the defect density. Confidence level for morphology detection. For optical detection confidence level; when ≥ The test results should be retained. < This will trigger a re-collection and analysis of the data in that area; Output the calibrated microstructure test report and optical performance test report.

9. The method according to claim 8, characterized in that, The implementation process of step S5 is as follows: Accuracy verification employs two criteria: repeatability measurement error for morphology detection ≤ 0.02 nm, and comparison error with standard samples for optical detection ≤ 0.1%. Anomalies are classified into minor anomalies and major anomalies. A minor anomaly is when a single indicator exceeds the acceptable range but does not affect the use of the product. A major anomaly is when any indicator is seriously out of standard or the defect density is ≥3 defects / cm². Outputs CSV format inspection reports, JSON format real-time data, visual defect distribution maps, and optical performance heat maps; provides a TCP / IP interface for interface with the control system of precision machining equipment. Feedback is sent to the finishing equipment regarding the substrate offset and processing power correction value, with an adjustment response time of ≤200ms.

10. The method according to claim 9, characterized in that, The implementation process of step S6 is as follows: Record all process data, including acquisition parameters, preprocessing parameters, test results, calibration coefficients, equipment feedback instructions, and execution effects; The parameters of the correlation model are optimized weekly based on historical test data, and the feature extraction thresholds and pass standards are updated monthly based on test data of different substrate types. Establish a performance monitoring mechanism to display the detection accuracy, data acquisition success rate, and interface response time in real time. When any indicator fails to meet the preset standard three times in a row, the dynamic adaptation and adjustment module will be automatically triggered to perform full parameter optimization.

Images