Optical element subsurface defect detection method and system

Abstract

The application relates to the technical field of optical detection, and discloses an optical element subsurface defect detection method and system. The method comprises the following steps: performing full-area scanning on an optical element to obtain a backscattering light signal; performing light filtering on the backscattering light signal to obtain a depth polarization scattering light signal group; performing coherent light compensation enhancement on the depth polarization scattering light signal group to obtain an enhanced polarization scattering light signal group; analyzing the enhanced polarization scattering light signal group in terms of light signal characteristic parameters, and performing multi-level fusion on light intensity distribution characteristic parameters, phase shift characteristic parameters and polarization state deflection degree characteristic parameters corresponding to the analyzed signal group to obtain a comprehensive deviation vector; comparing and analyzing the comprehensive deviation vector with preset standard parameter data to determine defect anomaly positioning data; performing full-attribute standardization judgment on the optical element to obtain a subsurface defect detection result; and the application can improve the accuracy of optical element subsurface defect detection.

Description

Technical Field

[0001] This invention relates to the field of optical inspection technology, and in particular to a method and system for detecting subsurface defects in optical components. Background Technology

[0002] Existing technologies have significant shortcomings in the signal acquisition and preprocessing stages of subsurface defect detection in optical components. They fail to employ a narrow-linewidth polarized laser full-area scanning method combined with spatial confocal filtering to acquire scattered light signals, making it difficult to accurately capture the weak light signals corresponding to subsurface defects. Furthermore, the lack of a systematic processing flow for multi-channel beam splitting, polarization beam splitting, and depth feature filtering prevents the effective separation of effective signal components at different depths, wavelengths, and polarization directions, resulting in a high proportion of noise and interference in the signal and insufficient reliability of the basic data provided for subsequent defect feature extraction.

[0003] Existing technologies have significant shortcomings in the feature analysis and judgment of subsurface defects in optical components. They fail to perform multi-level fusion of characteristics such as light intensity distribution, phase shift, and polarization state deflection of the optical signal, relying solely on single feature parameters for defect judgment, which makes it difficult to comprehensively reflect the multidimensional attributes of defects. Furthermore, they lack a precise comparison mechanism with standard parameter data, making it impossible to accurately pinpoint defect anomalies. The absence of a standardized judgment process covering all attributes hinders fine-grained identification of defect types, resulting in insufficient accuracy and comprehensiveness in defect detection, and failing to meet the high-precision detection requirements for subsurface defects in optical components. Summary of the Invention

[0004] This invention provides a method and system for detecting subsurface defects in optical components to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a method for detecting subsurface defects in optical components, comprising:

[0006] S1. Using a narrow-linewidth polarized laser, the optical element is scanned across its entire area to obtain the backscattered light signal of the optical element;

[0007] S2. The backscattered light signal is split and filtered to obtain the depth-polarized scattered light signal group of the optical element;

[0008] S3. Perform coherent light compensation enhancement on the deep polarization scattered light signal group to obtain the enhanced polarization scattered light signal group of the optical element;

[0009] S4. Perform optical signal characteristic parameter analysis on the enhanced polarization scattered light signal group, and perform multi-level fusion on the light intensity distribution characteristic quantity, phase shift characteristic quantity, and polarization state deflection degree characteristic quantity corresponding to the analyzed signal group to obtain the comprehensive deviation vector of the optical element;

[0010] S5. Compare and analyze the comprehensive deviation vector with the preset standard parameter data to determine the defect anomaly location data of the optical element;

[0011] S6. Based on the defect anomaly location data, perform full attribute standardization judgment on the optical element to obtain the subsurface defect detection result of the optical element.

[0012] In a preferred embodiment, the step of using a narrow-linewidth polarized laser to perform a full-area scan of the optical element to obtain the backscattered light signal of the optical element includes:

[0013] A continuous laser beam output from a narrow-linewidth polarized laser is received, and the continuous laser beam is used to irradiate the internal test area of ​​an optical element to obtain the initial backscattered light of the optical element.

[0014] Spatial confocal filtering is performed on the initial backscattered light to obtain the effective scattered light of the optical element;

[0015] The effective scattered light is photoelectrically converted, and the converted electrical signal is subjected to multi-level gain adjustment to obtain the backscattered light signal of the optical element.

[0016] In a preferred embodiment, the step of performing beam splitting and filtering on the backscattered light signal to obtain the depth-polarized scattered light signal group of the optical element includes:

[0017] The backscattered light signal is split into multiple channels in parallel to obtain the band-scattered light signal of the optical element;

[0018] The scattered light signal in the aforementioned band is polarized and split to obtain the orthogonal polarization component signal of the optical element;

[0019] The orthogonal polarization component signal is subjected to depth feature filtering to obtain the effective signal components of the optical element;

[0020] The depth layer identifier, band identifier, and polarization direction identifier in the effective signal components are structurally recombined to obtain the depth polarization scattered light signal group of the optical element.

[0021] In a preferred embodiment, the step of performing coherent light compensation enhancement on the deep polarization scattered light signal group to obtain the enhanced polarization scattered light signal group of the optical element includes:

[0022] Extract the initial amplitude data and initial phase data from the deep polarization scattered light signal group;

[0023] The initial phase data is subjected to phase consistency calibration to obtain the calibrated phase data of the optical element;

[0024] The initial amplitude data and the calibrated phase data are coherently coupled to obtain the coherent compensation coefficient of the optical element;

[0025] The coherence compensation coefficient and the initial amplitude data are weighted and fused to obtain the compensated amplitude data of the optical element;

[0026] The compensated amplitude data and the calibrated phase data are combined to form a signal, and the combined signal is normalized to obtain the enhanced polarization scattered light signal group of the optical element.

[0027] In a preferred embodiment, the step of analyzing the optical signal characteristic parameters of the enhanced polarization scattered light signal group includes:

[0028] Waveform feature deconstruction of the enhanced polarization scattered light signal group yields the light intensity distribution characteristic of the optical element;

[0029] Phase difference analysis is performed on the enhanced polarization scattered light signal group to obtain the phase shift characteristic quantity of the optical element;

[0030] The polarization state of the enhanced polarized scattered light signal group is calculated to obtain the polarization state deflection characteristic of the optical element.

[0031] In a preferred embodiment, the step of multi-level fusion of the intensity distribution characteristics, phase shift characteristics, and polarization state deflection characteristics corresponding to the analyzed signal group to obtain the comprehensive deviation vector of the optical element includes:

[0032] By aligning the light intensity distribution characteristic, the phase shift characteristic, and the polarization state deflection characteristic in spatial positions, the spatial mapping relationship of the optical element is obtained.

[0033] Based on the spatial mapping relationship, the correlation calculation is performed on the light intensity distribution characteristic, the phase shift characteristic, and the polarization state deflection characteristic to obtain the correlation characteristic value of the optical element. The calculation formula for the correlation characteristic value is as follows: ;

[0034] In the formula, For the associated feature value, Let be the light intensity distribution characteristic quantity under the said spatial mapping relationship. The maximum value of the light intensity distribution characteristic quantity. The phase offset feature quantity under the said spatial mapping relationship, This is the maximum value of the phase offset feature. The polarization state deflection characteristic quantity under the said spatial mapping relationship, This represents the maximum value of the polarization state deflection characteristic quantity;

[0035] The associated feature values ​​are normalized and weighted to obtain the comprehensive deviation component of the optical element;

[0036] The combined deviation components are arranged in a vector arrangement to obtain the combined deviation vector of the optical element.

[0037] In a preferred embodiment, the step of comparing and analyzing the integrated deviation vector with preset standard parameter data to determine the defect anomaly location data of the optical element includes:

[0038] Obtain preset standard parameter data, which is the standard deviation vector corresponding to the pre-collected defect-free optical element;

[0039] The comprehensive deviation vector is compared element by element with the preset standard deviation vector to determine the deviation difference data of the optical element;

[0040] Based on the deviation difference data, determine the candidate locations of the anomalies in the optical element;

[0041] The abnormal candidate locations are merged to obtain the abnormal regions of the optical element;

[0042] The boundary contour and center point of the abnormal region on the optical element are extracted to obtain the defect anomaly location data of the optical element.

[0043] In a preferred embodiment, the step of performing full-attribute normalization judgment on the optical element based on the defect anomaly localization data to obtain the subsurface defect detection result of the optical element includes:

[0044] Based on the defect anomaly localization data, local signal segments and local deviation sub-vectors in the enhanced polarized scattered light signal group and the comprehensive deviation vector are extracted simultaneously.

[0045] The local signal segment is deconstructed using multidimensional attributes to obtain the instantaneous light intensity fluctuation value, instantaneous phase jump value, and instantaneous polarization distortion value of the local signal segment;

[0046] Gradient analysis is performed on the local deviation sub-vector to obtain the deviation change rate and deviation change direction of the local deviation sub-vector;

[0047] The instantaneous light intensity fluctuation value, the instantaneous phase jump value, the instantaneous polarization distortion value, the deviation change rate, and the deviation change direction are combined to obtain the full attribute feature vector of the optical element.

[0048] The defect type confidence distribution of the optical element is obtained by performing hierarchical mapping on the full attribute feature vector;

[0049] By comprehensively judging the confidence distribution of the defect types, the subsurface defect detection results of the optical element are obtained.

[0050] In a preferred embodiment, the step of performing hierarchical mapping on the full-attribute feature vector to obtain the defect type confidence distribution of the optical element includes:

[0051] The full-attribute feature vector is subjected to type classification identification to obtain the defect category of the full-attribute feature vector;

[0052] Based on the aforementioned defect categories, the full-attribute feature vector is decomposed into fine-grained features to obtain the key feature components of the full-attribute feature vector.

[0053] The key feature components and the defect anomaly localization data are correlated and fused to obtain the feature fit of the optical element;

[0054] The confidence level of the feature fit is evaluated to obtain the confidence component of the optical element;

[0055] The confidence components are hierarchically aggregated to obtain the defect type confidence distribution of the optical element.

[0056] To address the above problems, the present invention also provides a subsurface defect detection system for optical components, the system comprising:

[0057] The signal acquisition module is used to perform a full-area scan of the optical element using a narrow-linewidth polarized laser to obtain the backscattered light signal of the optical element.

[0058] The beam splitting and filtering module is used to split and filter the backscattered light signal to obtain the depth polarization scattered light signal group of the optical element.

[0059] An illumination compensation module is used to perform coherent light compensation enhancement on the deep polarization scattered light signal group to obtain the enhanced polarization scattered light signal group of the optical element.

[0060] The parameter fusion module is used to analyze the optical signal characteristic parameters of the enhanced polarization scattered light signal group, and to perform multi-level fusion of the light intensity distribution characteristic, phase shift characteristic, and polarization state deflection characteristic corresponding to the analyzed signal group to obtain the comprehensive deviation vector of the optical element.

[0061] The vector comparison module is used to compare and analyze the comprehensive deviation vector with preset standard parameter data to determine the defect and anomaly location data of the optical element.

[0062] The defect detection module is used to perform full-attribute standardization judgment on the optical element based on the defect anomaly location data, and obtain the subsurface defect detection result of the optical element.

[0063] Compared with the prior art, the present invention has the following beneficial effects:

[0064] 1. This invention provides high-quality data support for the detection of subsurface defects in optical components through precise optical signal acquisition and enhancement processing. A narrow-linewidth polarized laser is used to scan the entire area of ​​the optical component. Backscattered light signals are obtained through spatial confocal filtering, photoelectric conversion, and gain adjustment. Deeply polarized scattered light signals are obtained through multi-channel beam splitting, polarization beam splitting, and depth feature filtering. These signals are then enhanced with coherent light compensation to improve signal strength and stability, ensuring that defect-related optical signal characteristics are clearly distinguishable, thus laying a reliable foundation for subsequent defect analysis.

[0065] 2. This invention significantly improves the accuracy and comprehensiveness of subsurface defect detection in optical components by leveraging multi-dimensional feature fusion and standardized judgment. It analyzes the enhanced polarization scattered light signal group to obtain light intensity distribution, phase shift, and polarization state deflection characteristics, which are then fused at multiple levels to generate a comprehensive deviation vector. This vector is compared with standard parameter data to determine defect anomaly location. Finally, through the combination of full-attribute features and hierarchical mapping, precise defect type identification is achieved, comprehensively capturing the location, range, and attribute information of subsurface defects, thus meeting the high-precision and refined requirements of optical component defect detection. Attached Figure Description

[0066] Figure 1 This is a schematic flowchart of a method for detecting subsurface defects in optical components according to an embodiment of the present invention;

[0067] Figure 2 A functional block diagram of an optical element subsurface defect detection system provided in an embodiment of the present invention;

[0068] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0069] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0070] This application provides a method for detecting subsurface defects in optical components. The execution entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method for detecting subsurface defects in optical components can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0071] Reference Figure 1 The diagram shown is a flowchart illustrating a method for detecting subsurface defects in optical components according to an embodiment of the present invention. In this embodiment, the method for detecting subsurface defects in optical components includes:

[0072] S1. Using a narrow-linewidth polarized laser, the optical element is scanned across its entire area to obtain the backscattered light signal of the optical element;

[0073] In this embodiment of the invention, the step of using a narrow-linewidth polarized laser to perform a full-area scan of the optical element to obtain the backscattered light signal of the optical element includes:

[0074] A continuous laser beam output from a narrow-linewidth polarized laser is received, and the continuous laser beam is used to irradiate the internal test area of ​​an optical element to obtain the initial backscattered light of the optical element.

[0075] Spatial confocal filtering is performed on the initial backscattered light to obtain the effective scattered light of the optical element;

[0076] The effective scattered light is photoelectrically converted, and the converted electrical signal is subjected to multi-level gain adjustment to obtain the backscattered light signal of the optical element.

[0077] After startup, the narrow-linewidth polarization laser continuously outputs a laser beam with good monochromaticity and stable polarization direction. The laser transmission line uses low-loss optical fiber to guide the continuous laser beam to the optical scanning device. The optical scanning device drives the laser emission port to move along a preset path through a mechanical transmission mechanism to ensure that the continuous laser beam can fully cover the internal test area of ​​the optical element. After the laser beam irradiates the material in the test area inside the optical element, it interacts with defects and impurities in the material to produce scattering. The backscattered light is the initial backscattered light of the optical element. The data source of this initial backscattered light is the scattering result of the interaction between the continuous laser beam and the material in the test area inside the optical element.

[0078] A spatial confocal filter is set up along the transmission path of the initial backscattered light. The device consists of a pinhole aperture and a focusing lens. The focusing lens converges the initial backscattered light to the center of the pinhole aperture. The pinhole aperture only allows scattered light propagating along the optical axis to pass through, while stray light deviating from the optical axis is blocked by the pinhole aperture. In this way, scattered light from the test point inside the optical element is filtered out, and interference components such as ambient stray light and reflected light from the surface of the optical element are eliminated to obtain the effective scattered light of the optical element. The data of the effective scattered light comes from the result of the initial backscattered light after being filtered by the spatial confocal filter.

[0079] The effectively scattered light is guided to a photoelectric conversion device, which is a photomultiplier tube. When the effectively scattered light shines on the photosensitive surface of the photomultiplier tube, the photosensitive surface converts the photons into corresponding photoelectrons. The photoelectrons are amplified step by step by the multiplier inside the device to form a weak electrical signal. This weak electrical signal is then input to a multi-stage gain adjustment circuit. The first-stage gain circuit uses a low-noise amplifier to initially amplify the weak electrical signal. The second-stage gain circuit adaptively adjusts the amplification factor according to the intensity of the effectively scattered light to ensure that the electrical signal is within the appropriate range for subsequent processing. After two stages of gain adjustment, an electrical signal with stable amplitude and a signal-to-noise ratio that meets the requirements is obtained. This electrical signal is the backscattered light signal of the optical element. The data source of this backscattered light signal is the processing result of the effectively scattered light after photoelectric conversion and multi-stage gain adjustment.

[0080] The beneficial effects are that the narrow-linewidth polarized laser beam has excellent monochromaticity and polarization stability, and the full-area illumination can completely capture the scattered signal of the test area inside the optical element, providing comprehensive raw data for detection.

[0081] Spatial confocal filtering effectively eliminates interference such as ambient stray light and reflected light from component surfaces, retaining only the scattered light from the internal test point, thus significantly improving signal purity.

[0082] Photoelectric conversion transforms optical signals into easily processed electrical signals. Multi-level gain adjustment adaptively amplifies weak defect signals and stabilizes their amplitude, ensuring that the signal is within a suitable processing range and improving resolvability.

[0083] The entire acquisition process accurately captures the weak light signals corresponding to subsurface defects, solving the problem of insufficient signal reliability in traditional acquisition processes, and laying a high-quality data foundation for subsequent spectral filtering and feature analysis.

[0084] S2. The backscattered light signal is split and filtered to obtain the depth-polarized scattered light signal group of the optical element;

[0085] In this embodiment of the invention, the step of performing beam splitting and filtering on the backscattered light signal to obtain the depth-polarized scattered light signal group of the optical element includes:

[0086] The backscattered light signal is split into multiple channels in parallel to obtain the band-scattered light signal of the optical element;

[0087] The scattered light signal in the aforementioned band is polarized and split to obtain the orthogonal polarization component signal of the optical element;

[0088] The orthogonal polarization component signal is subjected to depth feature filtering to obtain the effective signal components of the optical element;

[0089] The depth layer identifier, band identifier, and polarization direction identifier in the effective signal components are structurally recombined to obtain the depth polarization scattered light signal group of the optical element.

[0090] The backscattered light signal originates from the effective scattered light after photoelectric conversion and multi-level gain adjustment. This signal is then fed into a multi-channel parallel beam splitter, which contains multiple independent beam splitters, each corresponding to a fixed wavelength range. After the backscattered light signal enters the device, light signals of different wavelengths are refracted to different channels by the corresponding beam splitters. In this way, the backscattered light signal containing multiple wavelength components is separated into multiple single-band light signals. The single-band light signal output by each channel is the band scattered light signal of the optical element. The data source of this band scattered light signal is the result of the backscattered light signal being separated by the multi-channel parallel beam splitter.

[0091] The band-scattered light signal is introduced into a polarization beam splitter, the core of which is a polarization beam splitter prism. This prism can separate light signals according to the polarization direction of the light. When the band-scattered light signal is incident perpendicularly to the polarization beam splitter prism, the polarized light parallel to the polarization axis of the prism will be directly transmitted through the prism, forming one polarization component signal. The polarized light perpendicular to the polarization axis of the prism will be reflected by the prism to a preset direction, forming another polarization component signal. These two light signals with mutually perpendicular polarization directions are the orthogonal polarization component signals of the optical element. The data source of this orthogonal polarization component signal is the result of the band-scattered light signal being separated by the polarization beam splitter.

[0092] The orthogonal polarization component signal is transmitted to the depth feature filtering device. This device has a built-in filter template corresponding to the depth layer of the optical element. The filter template has a specific transmission rule preset according to the scattered light characteristics of different depth layers of the optical element. Only signals that match the scattering characteristics of the target depth layer are allowed to pass through. For example, the scattered light signal in the shallow region of the optical element has a specific intensity distribution pattern. The filter template will retain the signal that conforms to the pattern and block the interference signal that does not conform to the depth layer characteristics. The signal obtained after filtering is the effective signal component of the optical element. The data source of the effective signal component is the result of the orthogonal polarization component signal being filtered by the depth feature filtering device.

[0093] The effective signal components include depth layer identifiers, band identifiers, and polarization direction identifiers. The depth layer identifier is assigned by the depth feature filter based on the correspondence of the filter template and is directly associated with the specific depth position of the optical element. The band identifier is determined by the channel attributes of the multi-channel parallel beam splitter and corresponds to the single band information of the signal. The polarization direction identifier is determined by the separation direction of the polarization beam splitter and distinguishes between two orthogonal polarization states. These identifiers and their corresponding effective signal components are integrated and arranged in a unified structural format so that each signal component can clearly identify its corresponding depth, band, and polarization direction information through the identifier. The integrated signal set is the depth polarization scattered light signal group of the optical element. The data source of this depth polarization scattered light signal group is the structured recombination result of the effective signal components and each identifier.

[0094] The beneficial effects are that multi-channel parallel beam splitting enables the separation of signals in different bands, accurately extracts the scattered light characteristics of each band, and lays the foundation for subsequent targeted analysis.

[0095] Polarization beam splitting acquires orthogonal polarization component signals, captures the influence of defects on the polarization state of light, enriches the signal dimensions, and improves defect identification.

[0096] Depth feature filtering filters out effective signals at the target depth layer, eliminates irrelevant depth interference, and focuses on defect information in specific subsurface regions.

[0097] Structured recombination integrates depth, band, and polarization identifiers to form standardized signal groups, facilitating subsequent feature parameter analysis and correlation calculation, and improving the systematicness and efficiency of the detection process.

[0098] The entire process effectively separates effective signal components from different dimensions, reduces the proportion of noise and interference, solves the problem of insufficient signal reliability in traditional methods, and provides high-quality data support for defect detection.

[0099] S3. Perform coherent light compensation enhancement on the deep polarization scattered light signal group to obtain the enhanced polarization scattered light signal group of the optical element;

[0100] In this embodiment of the invention, the step of performing coherent light compensation enhancement on the deep polarization scattered light signal group to obtain the enhanced polarization scattered light signal group of the optical element includes:

[0101] Extract the initial amplitude data and initial phase data from the deep polarization scattered light signal group;

[0102] The initial phase data is subjected to phase consistency calibration to obtain the calibrated phase data of the optical element;

[0103] The initial amplitude data and the calibrated phase data are coherently coupled to obtain the coherent compensation coefficient of the optical element;

[0104] The coherence compensation coefficient and the initial amplitude data are weighted and fused to obtain the compensated amplitude data of the optical element;

[0105] The compensated amplitude data and the calibrated phase data are combined to form a signal, and the combined signal is normalized to obtain the enhanced polarization scattered light signal group of the optical element.

[0106] The deep polarization scattered light signal group originates from the structured recombination results of effective signal components and each identifier. This signal group is input into a signal analysis device, which obtains initial amplitude data by identifying the fluctuation intensity of each signal component in the signal group. The fluctuation intensity directly reflects the magnitude of the signal. At the same time, it obtains initial phase data by capturing the vibration phase changes of the signal components. The vibration phase changes reflect the phase state of the signal during propagation. The initial amplitude data and the initial phase data together constitute the core feature data of the deep polarization scattered light signal group. The source of the initial amplitude data and the initial phase data is the result of the deep polarization scattered light signal group being analyzed by the signal analysis device.

[0107] The initial phase data is imported into the phase calibration system, and phase consistency calibration is achieved using a difference compensation algorithm based on a standard reference phase. The specific implementation of this algorithm is as follows: The system has a built-in reference phase standard generated by a high-stability light source and standard optical components. This standard is a time series function corresponding to the scanning position. For each initial phase data point, the difference between it and the reference standard at the same time point is calculated. Then, this difference is eliminated through linear compensation, i.e., the initial phase data is subtracted from the difference to obtain the corrected phase data. If residual high-frequency noise exists, it can be further processed using lock-in amplification technology: the initial phase data and the reference standard are mixed, and then a low-pass filter is used to extract a stable phase difference signal, which is then used to correct the original phase. The phase data obtained after calibration is the calibrated phase data of the optical components. The data source of this calibrated phase data is the result of comparing and correcting the initial phase data using the phase calibration system.

[0108] The initial amplitude data and the calibrated phase data are input into a coherent coupling device, which achieves coherent coupling based on a complex amplitude synthesis model of coherent light superposition. Specifically, the complex amplitude of the optical signal corresponding to the initial amplitude is superimposed with the ideal complex amplitude constructed from standard parameters, and the coherent compensation coefficient is calculated through the interference term. This coefficient quantifies the amount of compensation for signal attenuation and phase distortion, and its magnitude depends on the degree of matching between the initial amplitude and the ideal amplitude, as well as the phase deviation.

[0109] The coherent compensation coefficient and initial amplitude data are input into the weighted fusion module, which uses an adaptive weight determination algorithm for weighted fusion. The weights are dynamically adjusted based on the signal-to-noise ratio (SNR), calculated as the ratio of signal power to noise power. The weight function is designed so that when the SNR is high, the weights approach 1, and the fusion result depends more on the original amplitude; when the SNR is low, the weights approach 0, and the fusion result depends more on the compensation coefficients. Specific weight values ​​are calculated using a logistic function, the parameters of which are set empirically, for example, setting a reference SNR threshold of 20 dB and adjusting the curve slope to make the weight changes gradual. The weighted fused compensated amplitude data is the result of combining the original amplitude and compensation coefficients according to the weights. If the scenario requirements are simple, a fixed weight can also be used for fusion, for example, a weight of 0.7, which can be adjusted according to the actual detection accuracy requirements. The source of this compensated amplitude data is the result of the coherent compensation coefficient and initial amplitude data being superimposed and calculated by the weighted fusion module using the aforementioned adaptive algorithm.

[0110] The compensated amplitude data and calibrated phase data are input into a signal synthesis device, which resynthesizes the complete optical signal according to the wave characteristics of the optical signal. The synthesized signal is then input into a normalization processing unit, which uses a minimum-maximum normalization algorithm to adjust the signal amplitude range to a uniform standard range, such as between 0 and 1, eliminating amplitude differences and interference between different signal components. The normalized signal set constitutes the enhanced polarization scattered light signal group of the optical element. The data source for this enhanced polarization scattered light signal group is the result of signal synthesis and normalization of the compensated amplitude data and calibrated phase data.

[0111] The beneficial effects are that the initial amplitude and phase data are extracted, the core characteristics of the deep polarized scattered light are fully preserved, and comprehensive data support is provided for subsequent compensation and enhancement.

[0112] Phase consistency calibration eliminates phase distortion and noise interference, ensuring the accuracy and stability of phase data and improving signal reliability.

[0113] Coherent coupling generates targeted coherent compensation coefficients, accurately quantifying the degree of signal attenuation and distortion, and providing a scientific basis for amplitude compensation.

[0114] The weighted fusion dynamically balances the original amplitude and the compensation coefficient, adaptively optimizes the signal amplitude, and enhances the identifiability of weak defect signals.

[0115] Signal synthesis and normalization integrate amplitude and phase information, unify the signal amplitude range, and eliminate scale differences between signals from different channels and depths, providing a standardized signal with a high signal-to-noise ratio for subsequent feature parameter analysis.

[0116] S4. Perform optical signal characteristic parameter analysis on the enhanced polarization scattered light signal group, and perform multi-level fusion on the light intensity distribution characteristic quantity, phase shift characteristic quantity, and polarization state deflection degree characteristic quantity corresponding to the analyzed signal group to obtain the comprehensive deviation vector of the optical element;

[0117] In this embodiment of the invention, the step of performing optical signal feature parameter analysis on the enhanced polarization scattered light signal group includes:

[0118] Waveform feature deconstruction of the enhanced polarization scattered light signal group yields the light intensity distribution characteristic of the optical element;

[0119] Phase difference analysis is performed on the enhanced polarization scattered light signal group to obtain the phase shift characteristic quantity of the optical element;

[0120] The polarization state of the enhanced polarized scattered light signal group is calculated to obtain the polarization state deflection characteristic of the optical element.

[0121] The multi-level fusion of the intensity distribution characteristics, phase shift characteristics, and polarization state deflection characteristics corresponding to the analyzed signal group yields the comprehensive deviation vector of the optical element, including:

[0122] By aligning the light intensity distribution characteristic, the phase shift characteristic, and the polarization state deflection characteristic in spatial positions, the spatial mapping relationship of the optical element is obtained.

[0123] Based on the spatial mapping relationship, the correlation calculation is performed on the light intensity distribution characteristic, the phase shift characteristic, and the polarization state deflection characteristic to obtain the correlation characteristic value of the optical element. The calculation formula for the correlation characteristic value is as follows: ;

[0124] In the formula, For the associated feature value, Let be the light intensity distribution characteristic quantity under the said spatial mapping relationship. The maximum value of the light intensity distribution characteristic quantity. The phase offset feature quantity under the said spatial mapping relationship, This is the maximum value of the phase offset feature. The polarization state deflection characteristic quantity under the said spatial mapping relationship, This represents the maximum value of the polarization state deflection characteristic quantity;

[0125] The associated feature values ​​are normalized and weighted to obtain the comprehensive deviation component of the optical element;

[0126] The combined deviation components are arranged in a vector arrangement to obtain the combined deviation vector of the optical element.

[0127] The enhanced polarization scattered light signal group is derived from the result of signal synthesis and normalization of compensated amplitude data and calibrated phase data. This signal group is input into a waveform deconstruction system. The system collects the amplitude changes of each signal component point by point to depict a complete signal waveform curve. Then, it analyzes the key parts of the waveform, such as peaks, valleys, and flat sections. Peaks reflect the strongest amplitude state of the signal, valleys reflect the weakest amplitude level of the signal, and flat sections represent the region where the signal amplitude is stable. By integrating the distribution of these key parts of the waveform, characteristic data that can comprehensively reflect the signal intensity change law is obtained. This characteristic data is the light intensity distribution characteristic quantity of the optical element. The data source of this light intensity distribution characteristic quantity is the result of the enhanced polarization scattered light signal group after analysis by the waveform deconstruction system.

[0128] The orthogonal polarization component signals in the enhanced polarization scattered light signal group are introduced into a phase difference analysis device. This device first extracts the phase information of the two orthogonal polarization signals, and then finds the phase difference between them by comparing the phase states of the two signals at the same time point. The trend of the phase difference over the entire signal period is analyzed. For example, the phase difference gradually increases in a certain period of time and remains stable in another period of time. By recording these patterns of change and specific phase differences, characteristic data that can characterize the phase relationship between the two polarization signals is formed. This characteristic data is the phase shift characteristic of the optical element. The data source of this phase shift characteristic is the result of the comparative analysis of the orthogonal polarization component signals in the enhanced polarization scattered light signal group by the phase difference analysis device.

[0129] The enhanced polarization scattered light signal group is input into the polarization state calculation system. This system has a built-in polarization state detection module. By detecting the distribution of the vibration direction and amplitude of the signal in different directions, the initial polarization state of the signal is determined. Then, the detected polarization state is compared with the preset standard polarization state to analyze the deflection difference between the two. For example, if the vibration direction of the signal is tilted at a certain angle from the standard polarization direction, the system quantifies the degree of tilt and the distribution change of the vibration amplitude to obtain characteristic data that can accurately reflect the change of the signal polarization state. This characteristic data is the polarization state deflection characteristic quantity of the optical element. The data source of this polarization state deflection characteristic quantity is the result of the enhanced polarization scattered light signal group being detected and compared by the polarization state calculation system.

[0130] The intensity distribution characteristic quantity is derived from the result of waveform deconstruction analysis of the enhanced polarization scattered light signal group. The phase shift characteristic quantity is derived from the result of phase difference analysis of the orthogonal polarization component signals in the enhanced polarization scattered light signal group. The polarization state deflection characteristic quantity is derived from the result of polarization state calculation system detection and comparison of the enhanced polarization scattered light signal group. These three characteristic quantities are input into the spatial position alignment system. This system uses the physical structure coordinates of the optical element as a reference. By identifying the specific position information of the optical element corresponding to each characteristic quantity, it adjusts the coordinate mapping relationship of each characteristic quantity so that the three characteristic quantities correspond one-to-one in the same spatial coordinate system. For example, a certain intensity value in the intensity distribution characteristic quantity, the corresponding phase difference value in the phase shift characteristic quantity, and the corresponding deflection degree in the polarization state deflection characteristic quantity are all accurately associated with the same position point of the optical element. The correspondence obtained after coordinate adjustment and matching is the spatial mapping relationship of the optical element. The data source of this spatial mapping relationship is the result of coordinate matching of the three characteristic quantities by the spatial position alignment system.

[0131] Based on spatial mapping relationships, the light intensity distribution characteristics, phase shift characteristics, and polarization state deflection characteristics are imported into a correlation calculation device. This device analyzes the mutual influence between light intensity changes and phase shifts, as well as the correlation between the two and polarization state deflection, according to the correspondence between the three characteristics at the same spatial location. For example, when the light intensity increases at a certain location, the phase shift shows a specific trend, and the polarization state deflection also changes accordingly. By capturing this inherent correlation and quantifying and integrating it, data that reflects the synergistic change characteristics of the three characteristics is obtained. This data is the correlation characteristic value of the optical element. The data source of this correlation characteristic value is the result of the three characteristics being quantified and integrated by the correlation calculation device based on spatial mapping relationships.

[0132] The purpose of setting normalization weighting here is to unify the associated feature values ​​to the standard range of 0 to 1, eliminate the numerical scale differences of associated feature values ​​at different spatial locations, and further highlight the severity of defects at different spatial locations. This provides a unified dimension and directly comparable basic data for the subsequent construction of the comprehensive deviation vector. Although the associated feature values ​​have been normalized by dividing by the maximum value of each feature, the associated feature values ​​at different spatial locations may still have a large numerical range, making them unsuitable for direct quantitative evaluation of comprehensive deviation. Furthermore, the impact of defects in different regions of the optical element on the overall quality has different weights, so weighting is needed to highlight the deviation in key regions.

[0133] The specific normalized weighted algorithm adopts a combination of "minimum-maximum normalization weight allocation", with the following steps: First, perform minimum-maximum normalization on the correlation feature values ​​of all spatial locations. Specifically, first find the minimum and maximum values ​​among all correlation feature values, then subtract the minimum value from the correlation feature value of each location, and then divide by the difference between the maximum and minimum values. This calculation maps all correlation feature values ​​to the interval between 0 and 1, ensuring that the correlation feature values ​​of different spatial locations are comparable. Second, perform weighted processing on the normalized correlation feature values ​​to obtain the comprehensive deviation component. Specifically, multiply the normalized correlation feature values ​​by a weight factor. The weighting factors are clearly defined and quantifiable, and are mainly divided into two preset methods, which can be selected according to actual testing needs: one is fixed weight, which presets weight coefficients for different spatial locations based on the application scenario and quality standards of the optical components. For example, the central area of ​​the optical component is the critical working area, and the weighting factor is set to 1.2, while the weighting factor for the edge area is set to 0.8, highlighting the deviation impact of the critical area; the other is dynamic weight, which is dynamically adjusted according to the signal-to-noise ratio of the enhanced polarized scattered light signal group. The higher the signal-to-noise ratio, the stronger the signal reliability, and the larger the weighting factor at the corresponding position. Specifically, it can be obtained through linear mapping: the weighting factor equals 0.5 plus the signal-to-noise ratio divided by the global maximum signal-to-noise ratio and then multiplied by 0.5, ensuring that the weight allocation matches the signal reliability and improving the accuracy of the comprehensive deviation component.

[0134] After the associated feature values ​​are input into the normalization weighting module, the module first performs the above minimum-maximum normalization operation, then completes the weighting calculation according to the preset weight factors, and finally uses the normalized weighted result as the comprehensive deviation component. The data source of the comprehensive deviation component is the result of the associated feature values ​​processed by the normalization weighting module according to the above algorithm.

[0135] The comprehensive deviation components are input into a vector arrangement device. This device arranges the comprehensive deviation components corresponding to each position according to the spatial position order of the optical elements, from one end of the optical elements to the other, forming an ordered vector data set. The position of each component in the vector corresponds one-to-one with the spatial position of the optical elements, which can intuitively reflect the comprehensive deviation of different positions of the optical elements. This ordered vector data set is the comprehensive deviation vector of the optical elements. The data source of the comprehensive deviation vector is the result of the comprehensive deviation components being arranged in spatial order by the vector arrangement device.

[0136] The beneficial effects are that waveform feature deconstruction accurately extracts light intensity distribution characteristics, fully presenting the signal intensity variation law, and providing a basic intensity basis for defect detection. Phase difference analysis captures phase shift characteristics, reflecting the influence of defects on the light phase, and supplementing the limitations of single light intensity analysis.

[0137] Polarization state calculation yields polarization deflection characteristics, uncovering anomalous changes in optical polarization and enriching the dimensions of defect features. Three-dimensional feature parameters comprehensively cover the core attributes of optical signals, overcoming the limitations of traditional single-feature detection. This provides multi-dimensional and comprehensive feature data support for subsequent multi-level fusion, improving the comprehensiveness and accuracy of defect identification.

[0138] Spatial alignment establishes a unified coordinate system, ensuring that the three types of feature quantities accurately correspond to the same location of the optical element, laying a spatially consistent foundation for multi-dimensional feature fusion. Correlation calculation integrates the three-dimensional features of light intensity, phase, and polarization through quantization formulas, comprehensively capturing the combined impact of defects on the optical signal and overcoming the limitations of single features.

[0139] Normalized weighting eliminates differences in feature scales while highlighting the defect weights in key areas, thus improving the rationality and relevance of deviation assessment. Vector arrangement forms a structured comprehensive deviation vector, intuitively presenting the deviation distribution across the entire optical component, providing accurate and comprehensive quantitative basis for subsequent comparison with standard parameters and defect localization.

[0140] S5. Compare and analyze the comprehensive deviation vector with the preset standard parameter data to determine the defect anomaly location data of the optical element;

[0141] In this embodiment of the invention, the step of comparing and analyzing the comprehensive deviation vector with preset standard parameter data to determine the defect anomaly location data of the optical element includes:

[0142] Obtain preset standard parameter data, which is the standard deviation vector corresponding to the pre-collected defect-free optical element;

[0143] The comprehensive deviation vector is compared element by element with the preset standard deviation vector to determine the deviation difference data of the optical element;

[0144] Based on the deviation difference data, determine the candidate locations of the anomalies in the optical element;

[0145] The abnormal candidate locations are merged to obtain the abnormal regions of the optical element;

[0146] The boundary contour and center point of the abnormal region on the optical element are extracted to obtain the defect anomaly location data of the optical element.

[0147] The preset standard parameter data is the standard deviation vector, which originates from the inspection and data accumulation of a large number of defect-free optical components. Using the same inspection process as the optical component under test, the defect-free optical component undergoes a series of processes including optical signal acquisition, feature parameter analysis, and multi-level fusion to obtain the deviation vector corresponding to each defect-free optical component. These deviation vectors are then statistically analyzed, and their common characteristics and average levels are extracted to form a standard deviation vector that represents the defect-free state. Specifically, the comprehensive deviation vectors of no less than 100 defect-free optical components from the same batch are collected, and the mean and standard deviation at each spatial location are calculated. The standard deviation vector is then calculated as the mean. This standard deviation vector is pre-stored in the inspection system's database and can be directly retrieved through a data retrieval interface. The data source for this preset standard parameter data is the statistical result of defect-free optical components processed using the same inspection process.

[0148] The comprehensive deviation vector originates from the result of arranging the comprehensive deviation components in spatial order using a vector arrangement device. This comprehensive deviation vector is then compared with a preset standard deviation vector using an element-by-element comparison device. This device, based on the correspondence between the elements in the two vectors, sequentially compares the comprehensive deviation component at each spatial location with the corresponding element in the standard deviation vector, calculating the degree of difference between the two to obtain deviation difference data. The data formed by integrating the difference records from all locations constitutes the deviation difference data of the optical element. This deviation difference data originates from the comparison between the comprehensive deviation vector and the preset standard deviation vector using the element-by-element comparison device.

[0149] The deviation difference data is input into the anomaly location determination system. This system has a built-in difference threshold standard, which is set based on the deviation fluctuation range of defect-free optical components and is used to distinguish between normal and abnormal deviations. The threshold is set as follows: based on the detection data of the aforementioned 100 defect-free optical components, the standard deviation of each location is calculated. The threshold is set equal to a confidence coefficient multiplied by the standard deviation. The confidence coefficient is generally set to 3, corresponding to a 99.7% confidence interval, and can be adjusted according to the actual detection sensitivity requirements; for example, a value of 2 can be used in high-sensitivity scenarios. The system judges whether the deviation difference data at each location exceeds the threshold. If the deviation difference data at a certain location exceeds the threshold, it indicates that the deviation at that location exceeds the normal fluctuation range of a defect-free state, and the location is determined to be an anomaly candidate location. For example, if the deviation difference data in a certain area of ​​the optical component is significantly higher than the threshold, all corresponding locations in that area are marked as anomaly candidate locations. The data source for these anomaly candidate locations is the result of the deviation difference data being judged by the anomaly location determination system's threshold.

[0150] Anomaly candidate locations are processed by region merging using a spatial clustering algorithm based on connected component analysis. The specific rules are as follows: Anomaly candidate locations are mapped to a two-dimensional spatial grid, with each grid cell corresponding to a scan point. A neighborhood radius is set, for example, two scan steps. Using the eight-neighbor connectivity criterion, interconnected anomaly candidate locations are merged into the same region, i.e., the spatial distance is less than or equal to the neighborhood radius. After merging, isolated regions with an area smaller than the preset minimum defect area, such as four scan points, are removed, forming complete and continuous anomaly regions. For example, if the scan step size is 20 micrometers, the neighborhood radius is set to 40 micrometers, and anomaly candidate locations with a spatial distance less than or equal to 40 micrometers are merged into the same region. The data source for this anomaly region is the result of connected component analysis and integration of the anomaly candidate locations.

[0151] The data from the anomalous area is imported into a localization feature extraction device. This device analyzes the spatial distribution of the anomalous area, determines its edge boundaries, delineates its boundary contour, clarifies its coverage area on the optical element, and calculates the geometric center position of the anomalous area, which is its center point. The boundary contour reflects the shape and size of the anomalous area, while the center point accurately points to its core location. Integrating the boundary contour and center point data yields the defect anomaly localization data for the optical element. This defect anomaly localization data originates from the analysis results of the anomalous area by the localization feature extraction device.

[0152] The beneficial effects are that, using the standard deviation vector of defect-free components as a benchmark, a unified and reliable comparison reference is provided, ensuring the objectivity and accuracy of deviation judgment. Element-by-element comparison accurately captures deviation differences across the entire range, comprehensively reflecting the degree of anomaly at each position of the optical component and avoiding the omission of potential defects.

[0153] By using threshold-based screening of abnormal candidate locations, the suspected defect area is identified, which narrows down the scope and reduces redundant calculations for subsequent accurate localization.

[0154] Region merging eliminates isolated false anomalies, forming continuous and complete anomaly regions, thus improving the completeness and accuracy of defect localization. Extracting boundary contours and center points accurately pinpoints the spatial extent, core location, and morphological characteristics of defects, providing precise spatial location data for subsequent full-attribute determination.

[0155] S6. Based on the defect anomaly location data, perform full attribute standardization judgment on the optical element to obtain the subsurface defect detection result of the optical element.

[0156] In this embodiment of the invention, the step of performing full-attribute normalization judgment on the optical element based on the defect anomaly localization data to obtain the subsurface defect detection result of the optical element includes:

[0157] Based on the defect anomaly localization data, local signal segments and local deviation sub-vectors in the enhanced polarized scattered light signal group and the comprehensive deviation vector are extracted simultaneously.

[0158] The local signal segment is deconstructed using multidimensional attributes to obtain the instantaneous light intensity fluctuation value, instantaneous phase jump value, and instantaneous polarization distortion value of the local signal segment;

[0159] Gradient analysis is performed on the local deviation sub-vector to obtain the deviation change rate and deviation change direction of the local deviation sub-vector;

[0160] The instantaneous light intensity fluctuation value, the instantaneous phase jump value, the instantaneous polarization distortion value, the deviation change rate, and the deviation change direction are combined to obtain the full attribute feature vector of the optical element.

[0161] The defect type confidence distribution of the optical element is obtained by performing hierarchical mapping on the full attribute feature vector;

[0162] By comprehensively judging the confidence distribution of the defect types, the subsurface defect detection results of the optical element are obtained.

[0163] The step of performing hierarchical mapping on the full-attribute feature vector to obtain the defect type confidence distribution of the optical element includes:

[0164] The full-attribute feature vector is subjected to type classification identification to obtain the defect category of the full-attribute feature vector;

[0165] Based on the aforementioned defect categories, the full-attribute feature vector is decomposed into fine-grained features to obtain the key feature components of the full-attribute feature vector.

[0166] The key feature components and the defect anomaly localization data are correlated and fused to obtain the feature fit of the optical element;

[0167] The confidence level of the feature fit is evaluated to obtain the confidence component of the optical element;

[0168] The confidence components are hierarchically aggregated to obtain the defect type confidence distribution of the optical element.

[0169] The defect anomaly location data originates from the analysis of the anomaly region using a location feature extraction device. This data clearly defines the boundary contour and center point of the anomaly region on the optical element. Based on this location information, the signal portion corresponding to the anomaly region is extracted from the enhanced polarization scattered light signal group; this signal portion is the local signal segment. Simultaneously, the combination of deviation components corresponding to the spatial location of the anomaly region is extracted from the comprehensive deviation vector; this combination is the local deviation sub-vector. The data source for the local signal segment is the extracted signal corresponding to the anomaly region from the enhanced polarization scattered light signal group, and the data source for the local deviation sub-vector is the extracted deviation component corresponding to the anomaly region from the comprehensive deviation vector.

[0170] Local signal segments are input into a multi-dimensional attribute deconstruction system. This system monitors the instantaneous changes in signal amplitude in real time, capturing the intensity fluctuations of the signal within a very short period, thus obtaining instantaneous light intensity fluctuation values. It also tracks abrupt changes in signal phase, recording the jump values ​​of phase in states without continuous transitions, thus obtaining instantaneous phase jump values. Furthermore, it analyzes abnormal changes in the signal polarization direction, quantifying the degree to which the polarization state deviates from the normal range, thus obtaining instantaneous polarization distortion values. These values ​​reflect the abnormal attributes of the local signal segments from three dimensions: intensity, phase, and polarization. The data for instantaneous light intensity fluctuation values, instantaneous phase jump values, and instantaneous polarization distortion values ​​originate from the results of the local signal segments analyzed by the multi-dimensional attribute deconstruction system.

[0171] The local deviation subvector is imported into a gradient analysis device. This device calculates the difference between adjacent deviation components and, combined with the spatial distance between the adjacent components, obtains the change in deviation per unit distance, which is the deviation change rate. By comparing the numerical relationship between the deviation components before and after the change, it determines whether the deviation is increasing or decreasing, which is the deviation change direction. The deviation change rate and deviation change direction together reflect the dynamic change characteristics of the local deviation subvector. The data for the deviation change rate and deviation change direction come from the results of the local deviation subvector calculated and compared by the gradient analysis device.

[0172] The instantaneous light intensity fluctuation value, instantaneous phase jump value, instantaneous polarization distortion value, deviation change rate, and deviation change direction are input into the feature combination module. This module arranges and integrates these feature data in a fixed order to form an ordered data set containing all feature information. Each feature data occupies a fixed position in the set and can completely retain its own attribute information and interrelationships. This ordered data set is the full attribute feature vector of the optical element. The data source of this full attribute feature vector is the result of arranging and integrating various feature data by the feature combination module.

[0173] The full-attribute feature vector is input into a hierarchical mapping system. This system incorporates a machine learning-based classification and recognition mechanism, with a core multi-layer neural network classifier used to achieve hierarchical mapping of the full-attribute feature vector. The detailed design of this classifier is as follows: The number of nodes in the input layer equals the dimension of the full-attribute feature vector, i.e., 5 nodes, corresponding to instantaneous light intensity fluctuation, instantaneous phase jump, instantaneous polarization distortion, deviation change rate, and deviation change direction, respectively. Two hidden layers are set, with 16 nodes in the first layer and 8 nodes in the second layer. Linear rectified units are used as the activation function for both layers. The number of nodes in the output layer equals the preset number of defect categories. For example, scratches, bubbles, and impurities correspond to 3 output nodes. A normalized exponential function is used as the activation function, outputting the probability of each category. Training data consists of no fewer than 1000 labeled optical element inspection samples. Each sample contains a full-attribute feature vector and manually labeled defect categories. The samples cover optical elements of different materials and processes to ensure diversity. The training parameters were set as follows: cross-entropy loss was used as the loss function; the adaptive moments estimation was selected as the optimizer, with an initial learning rate of 0.001, which decayed to 0.9 every 50 training epochs; the batch size was set to 32; the number of training epochs was 200, and early stopping was used to prevent overfitting, i.e., training was stopped if the validation set loss did not decrease for 10 consecutive epochs; L2 regularization and random dropout techniques were added to improve generalization ability. Five-fold cross-validation was used to evaluate the model performance and ensure the model's stability under different data partitions. The classifier was used to calculate the matching degree between the full attribute feature vector and the features of each defect type, outputting a three-dimensional probability vector, which is the defect type confidence distribution. The data source for this defect type confidence distribution is the result of comparison and calculation of the full attribute feature vector through a hierarchical mapping system.

[0174] The confidence distribution of defect types is input into a comprehensive discrimination device. This device selects the defect type with the highest confidence value as the primary judgment criterion, while also using the confidence values ​​of other defect types for auxiliary verification. If the difference between the highest and second-highest confidence values ​​is greater than a preset difference threshold, such as 0.3, the defect type corresponding to the highest confidence value is directly determined to be a subsurface defect type of the optical element. If the confidence values ​​of multiple defect types are relatively close, with a difference less than or equal to 0.3, a secondary discrimination is performed by combining key features from the full-attribute feature vector. For example, it checks whether instantaneous light intensity fluctuations and phase jumps conform to the typical characteristics of a specific defect, ultimately clarifying whether the optical element has subsurface defects and their specific types. The resulting judgment is the subsurface defect detection result of the optical element, and the data source for this subsurface defect detection result is the result of the defect type confidence distribution analyzed and verified by the comprehensive discrimination device.

[0175] The full-attribute feature vector originates from the result of arranging and integrating various feature data through a feature combination module. This vector is input into a type classification identification system, which uses a multi-layer neural network classifier trained as described above to achieve type classification identification. The classifier has built-in preset feature templates for various defect categories, constructed based on statistical parameters such as the mean and variance of features for various defect categories in the training samples. The system inputs the full-attribute feature vector into the classifier, and through the forward propagation calculation, outputs the preliminary matching probability of the vector for each defect category. The category with the highest matching probability and above a preset probability threshold is selected as the defect category of the full-attribute feature vector, for example, 0.5. For example, if the instantaneous light intensity fluctuation and polarization distortion features in the full-attribute feature vector highly match the impurity defect template, the classifier outputs that this category has the highest matching probability and exceeds 0.5, thus classifying it as an impurity defect category. The data source for this defect category is the result of comparing the full-attribute feature vector with the defect category feature templates through the type classification identification system.

[0176] Based on the determined defect categories, the full-attribute feature vectors are imported into a fine-grained feature decomposition device. The fine-grained feature decomposition rules of this device are determined based on the feature importance analysis of the defect categories. The random forest algorithm is used to rank the feature importance of the full-attribute feature vectors in the training samples, and the top three features contributing the most to distinguishing the subcategories of the defect category are selected as key feature components. For example, for scratch defects, the feature importance ranking is: instantaneous phase jump value contributes the most, followed by deviation change rate, and then instantaneous light intensity fluctuation value. Therefore, the key feature components are instantaneous phase jump value, deviation change rate, and instantaneous light intensity fluctuation value. Key feature components can be the original features themselves or linear combinations of the original features, such as multiplying the instantaneous phase jump value by the deviation change rate, or adding the instantaneous light intensity fluctuation value to half of the deviation change rate. The combined weights are determined by fitting the training samples. The device is specifically designed to extract the specific data of these three types of features or their preset linear combination results to form a feature set that can reflect the differences between different subcategories under the major category, such as shallow scratches and deep scratches. This feature set is the key feature component of the full attribute feature vector. The data source of this key feature component is the result of the full attribute feature vector being separated by a fine-grained feature decomposition device according to the feature importance rules of the major defect category.

[0177] The defect anomaly location data originates from the analysis of the anomaly area by a location feature extraction device. Key feature components are input into the correlation and fusion module along with this location data. This module uses a cosine similarity algorithm to calculate the feature fit. Specifically, the key feature components are constructed as one feature vector, and the standard features corresponding to the defect anomaly location data (features extracted from normal or anomaly standard samples of the location area, such as the mean feature extracted from a defect-free area) are constructed as another feature vector. The cosine similarity is calculated by dividing the dot product of the two vectors by the product of their magnitudes. The similarity value ranges from 0 to 1; the closer the value is to 1, the higher the feature fit between the key feature component and the location data. For example, if a strong phase jump displayed by the key feature component highly matches the standard feature corresponding to the anomaly center point in a certain area of ​​the location data, the calculated cosine similarity is close to 0.95. This similarity value represents the feature fit of the optical element, and the data source for this feature fit is the result of the correlation and fusion module calculating the cosine similarity algorithm between the key feature components and the defect anomaly location data.

[0178] The feature fit is input into the confidence assessment system. This system's confidence assessment criterion is constructed based on the mapping relationship between feature fit and confidence, using the logistic function as the mapping function to map the feature fit (0 to 1 interval) calculated by cosine similarity to a confidence component within the 0 to 1 interval. The logistic function converts the feature fit into confidence through a preset steepness parameter and a bias parameter. The higher the feature fit, the closer the mapped confidence component is to 1. For example, when the feature fit is 0.8, the mapped confidence component is approximately 0.88. This value is the confidence component of the optical element, and the data source for this confidence component is the result of the feature fit being determined by the confidence assessment system using the logistic function.

[0179] The confidence scores are imported into a hierarchical aggregation device, which categorizes and organizes the confidence scores for each subcategory according to the sub-categories under the major defect category. For example, the confidence scores for different subcategories under the same major defect category are aggregated to form a distribution data containing the confidence scores of all subcategories under that major defect category. This distribution data clearly shows the comparison of the confidence levels of defects in each subcategory, i.e., the confidence distribution of defect types for optical components. The data source for this confidence distribution of defect types is the result of the confidence scores being categorized and organized by the hierarchical aggregation device according to the sub-categories.

[0180] The beneficial effects include: accurately extracting local signals and deviation sub-vectors in abnormal areas, focusing on core defect information, reducing interference from irrelevant data, and improving the specificity of judgment. Multi-dimensional attribute deconstruction and gradient analysis comprehensively uncover the optical signal characteristics and dynamic changes in deviations of defects, enriching the dimensions of defect attributes and overcoming the limitations of single-feature judgment.

[0181] Feature combinations form a full-attribute feature vector, integrating multi-dimensional key information about defects to provide comprehensive data support for accurate identification. Hierarchical mapping and confidence distribution quantify the degree of defect type matching, and combined with comprehensive discrimination, achieve accurate classification of defect types, improving the scientific nature of the judgment.

[0182] The standardized judgment process throughout the entire process ensures consistency and accuracy in the judgment of different optical components and different defect types, meeting the needs of high-precision detection.

[0183] Type classification quickly identifies major defect categories, narrowing the scope of judgment and improving the efficiency and relevance of defect classification. Fine-grained feature decomposition extracts key distinguishing features, accurately capturing the core differences between different defect types and providing support for accurate identification.

[0184] Key features are correlated and fused with location data, and the spatial location information of defects is combined to enhance feature matching and improve the accuracy of judgment. Confidence assessment quantifies the degree of feature fit, providing objective quantitative basis for defect type determination and reducing subjective errors.

[0185] The hierarchical aggregation forms a complete confidence distribution, clearly presenting the matching probability of various defects, providing a comprehensive and intuitive reference for comprehensive judgment, and ensuring the scientificity and reliability of defect type determination.

[0186] like Figure 2 The diagram shown is a functional block diagram of an optical element subsurface defect detection system provided in an embodiment of the present invention.

[0187] The optical element subsurface defect detection system described in this invention can be installed in an electronic device. Depending on the functions implemented, the optical element subsurface defect detection system may include a signal acquisition module, a beam splitting and filtering module, an illumination compensation module, a parametric fusion module, a vector comparison module, and a defect detection module. The module described in this invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and are stored in the memory of the electronic device.

[0188] In this embodiment, the functions of each module / unit are as follows:

[0189] The signal acquisition module is used to perform a full-area scan of the optical element using a narrow-linewidth polarized laser to obtain the backscattered light signal of the optical element.

[0190] The beam splitting and filtering module is used to perform beam splitting and filtering on the backscattered light signal to obtain the depth polarization scattered light signal group of the optical element;

[0191] The illumination compensation module is used to perform coherent light compensation enhancement on the deep polarization scattered light signal group to obtain the enhanced polarization scattered light signal group of the optical element.

[0192] The parameter fusion module is used to analyze the optical signal characteristic parameters of the enhanced polarization scattered light signal group, and to perform multi-level fusion of the light intensity distribution characteristic, phase shift characteristic, and polarization state deflection characteristic corresponding to the analyzed signal group to obtain the comprehensive deviation vector of the optical element.

[0193] The vector comparison module is used to compare and analyze the comprehensive deviation vector with preset standard parameter data to determine the defect and anomaly location data of the optical element.

[0194] The defect detection module is used to perform full-attribute standardization judgment on the optical element based on the defect anomaly location data, and obtain the subsurface defect detection result of the optical element.

[0195] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0196] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0197] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0198] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0199] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0200] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for detecting subsurface defects in optical components, characterized in that, The method includes: S1. Using a narrow-linewidth polarized laser, the optical element is scanned across its entire area to obtain the backscattered light signal of the optical element; S2. The backscattered light signal is split and filtered to obtain the depth-polarized scattered light signal group of the optical element; S3. Perform coherent light compensation enhancement on the deep polarization scattered light signal group to obtain the enhanced polarization scattered light signal group of the optical element, including: Extract the initial amplitude data and initial phase data from the deep polarization scattered light signal group; The initial phase data is subjected to phase consistency calibration to obtain the calibrated phase data of the optical element; The initial amplitude data and the calibrated phase data are coherently coupled to obtain the coherent compensation coefficient of the optical element; The coherence compensation coefficient and the initial amplitude data are weighted and fused to obtain the compensated amplitude data of the optical element; The compensated amplitude data and the calibrated phase data are combined into a signal, and the combined signal is normalized to obtain the enhanced polarization scattered light signal group of the optical element. S4. Perform optical signal characteristic parameter analysis on the enhanced polarization scattered light signal group, and perform multi-level fusion on the light intensity distribution characteristic quantity, phase shift characteristic quantity, and polarization state deflection degree characteristic quantity corresponding to the analyzed signal group to obtain the comprehensive deviation vector of the optical element, including: Waveform feature deconstruction of the enhanced polarization scattered light signal group yields the light intensity distribution characteristic of the optical element; Phase difference analysis is performed on the enhanced polarization scattered light signal group to obtain the phase shift characteristic quantity of the optical element; The polarization state of the enhanced polarized scattered light signal group is calculated to obtain the polarization state deflection characteristic of the optical element; By aligning the light intensity distribution characteristic, the phase shift characteristic, and the polarization state deflection characteristic in spatial positions, the spatial mapping relationship of the optical element is obtained. Based on the spatial mapping relationship, the correlation calculation is performed on the light intensity distribution characteristic, the phase shift characteristic, and the polarization state deflection characteristic to obtain the correlation characteristic value of the optical element. The calculation formula for the correlation characteristic value is as follows: ； In the formula, For the associated feature value, Let be the light intensity distribution characteristic quantity under the said spatial mapping relationship. The maximum value of the light intensity distribution characteristic quantity. The phase offset feature quantity under the said spatial mapping relationship, This is the maximum value of the phase offset feature. The polarization state deflection characteristic quantity under the said spatial mapping relationship, This represents the maximum value of the polarization state deflection characteristic quantity; The associated feature values ​​are normalized and weighted to obtain the comprehensive deviation component of the optical element; The combined deviation components are arranged in a vector arrangement to obtain the combined deviation vector of the optical element; S5. Compare and analyze the comprehensive deviation vector with the preset standard parameter data to determine the defect anomaly location data of the optical element; S6. Based on the defect anomaly location data, perform full-attribute normalization judgment on the optical element to obtain the subsurface defect detection results of the optical element, including: Based on the defect anomaly localization data, local signal segments and local deviation sub-vectors in the enhanced polarized scattered light signal group and the comprehensive deviation vector are extracted simultaneously. The local signal segment is deconstructed using multidimensional attributes to obtain the instantaneous light intensity fluctuation value, instantaneous phase jump value, and instantaneous polarization distortion value of the local signal segment; Gradient analysis is performed on the local deviation sub-vector to obtain the deviation change rate and deviation change direction of the local deviation sub-vector; The instantaneous light intensity fluctuation value, the instantaneous phase jump value, the instantaneous polarization distortion value, the deviation change rate, and the deviation change direction are combined to obtain the full attribute feature vector of the optical element. The defect type confidence distribution of the optical element is obtained by performing hierarchical mapping on the full attribute feature vector; By comprehensively judging the confidence distribution of the defect types, the subsurface defect detection results of the optical element are obtained.

2. The method for detecting subsurface defects in optical components as described in claim 1, characterized in that, The method of using a narrow-linewidth polarized laser to perform a full-area scan of the optical element to obtain the backscattered light signal of the optical element includes: A continuous laser beam output from a narrow-linewidth polarized laser is received, and the continuous laser beam is used to irradiate the internal test area of ​​an optical element to obtain the initial backscattered light of the optical element. Spatial confocal filtering is performed on the initial backscattered light to obtain the effective scattered light of the optical element; The effective scattered light is photoelectrically converted, and the converted electrical signal is subjected to multi-level gain adjustment to obtain the backscattered light signal of the optical element.

3. The method for detecting subsurface defects in optical components as described in claim 1, characterized in that, The step of performing beam splitting and filtering on the backscattered light signal to obtain the depth-polarized scattered light signal group of the optical element includes: The backscattered light signal is split into multiple channels in parallel to obtain the band-scattered light signal of the optical element; The scattered light signal in the aforementioned band is polarized and split to obtain the orthogonal polarization component signal of the optical element; The orthogonal polarization component signal is subjected to depth feature filtering to obtain the effective signal components of the optical element; The depth layer identifier, band identifier, and polarization direction identifier in the effective signal components are structurally recombined to obtain the depth polarization scattered light signal group of the optical element.

4. The method for detecting subsurface defects in optical components as described in claim 1, characterized in that, The step of comparing and analyzing the comprehensive deviation vector with preset standard parameter data to determine the defect anomaly location data of the optical element includes: Obtain preset standard parameter data, which is the standard deviation vector corresponding to the pre-collected defect-free optical element; The comprehensive deviation vector is compared element by element with the preset standard deviation vector to determine the deviation difference data of the optical element; Based on the deviation difference data, determine the candidate locations of the anomalies in the optical element; The abnormal candidate locations are merged to obtain the abnormal regions of the optical element; The boundary contour and center point of the abnormal region on the optical element are extracted to obtain the defect anomaly location data of the optical element.

5. The method for detecting subsurface defects in optical components as described in claim 1, characterized in that, The step of performing hierarchical mapping on the full-attribute feature vector to obtain the defect type confidence distribution of the optical element includes: The full-attribute feature vector is subjected to type classification identification to obtain the defect category of the full-attribute feature vector; Based on the aforementioned defect categories, the full-attribute feature vector is decomposed into fine-grained features to obtain the key feature components of the full-attribute feature vector. The key feature components and the defect anomaly localization data are correlated and fused to obtain the feature fit of the optical element; The confidence level of the feature fit is evaluated to obtain the confidence component of the optical element; The confidence components are hierarchically aggregated to obtain the defect type confidence distribution of the optical element.

6. A subsurface defect detection system for optical components, characterized in that, The system for implementing the subsurface defect detection method for optical elements according to claim 1 includes: The signal acquisition module is used to perform a full-area scan of the optical element using a narrow-linewidth polarized laser to obtain the backscattered light signal of the optical element. The beam splitting and filtering module is used to split and filter the backscattered light signal to obtain the depth polarization scattered light signal group of the optical element. An illumination compensation module is used to perform coherent light compensation enhancement on the deep polarization scattered light signal group to obtain the enhanced polarization scattered light signal group of the optical element. The parameter fusion module is used to analyze the optical signal characteristic parameters of the enhanced polarization scattered light signal group, and to perform multi-level fusion of the light intensity distribution characteristic, phase shift characteristic, and polarization state deflection characteristic corresponding to the analyzed signal group to obtain the comprehensive deviation vector of the optical element. The vector comparison module is used to compare and analyze the comprehensive deviation vector with preset standard parameter data to determine the defect and anomaly location data of the optical element. The defect detection module is used to perform full-attribute standardization judgment on the optical element based on the defect anomaly location data, and obtain the subsurface defect detection result of the optical element.

Images