A method for detecting the deformation of an optical reflection image on a lens surface

Abstract

The present application belongs to the technical field of deformation detection, and specifically relates to a lens surface optical reflection image deformation detection method, which comprises collecting the structural parameters and environmental light field data of the lens to be detected. The interference effect of the structural parameters on the mirror reflection light is analyzed to obtain mirror optical measurement data, the optical surface composite features are extracted from the mirror optical measurement data, and a surface topography reconstruction model is constructed. An error transfer function based on the coupling relationship between the film thickness change and the phase mutation is constructed. The parasitic light field interference is compensated in reverse based on the error transfer function, the real surface phase gradient field is extracted, the surface topography reconstruction model is input, and the data fusion calibration is performed in combination with the multi-directional reflection imaging viewing angle to obtain the deformation evaluation result. The present application can adapt to the detection requirements of different types of lenses, reduce the detection errors caused by parasitic reflection and multiple reflections, and improve the lens surface reflection image deformation detection precision.

Description

Technical Field

[0001] This invention belongs to the field of deformation detection technology, specifically a method for detecting the deformation of optical reflection images on the surface of a lens. Background Technology

[0002] In the field of optical lens manufacturing and quality inspection, the accurate detection of image distortion caused by reflection from the lens surface is a crucial step in ensuring the imaging quality of optical systems, and it is widely used in aerospace, microelectronics equipment, eyewear manufacturing, and many other fields. With the development of optical technology, the application of multi-layer composite coated lenses and aspherical lenses is becoming increasingly widespread, placing higher demands on the accuracy and comprehensiveness of testing.

[0003] Traditional methods for detecting lens reflection image deformation suffer from numerous technical bottlenecks, making it difficult to meet practical testing needs. First, traditional methods ignore parasitic light field interference generated by multiple reflections within the multilayer coating. This interference leads to spurious signals in the measured phase field, making it impossible to accurately distinguish between coating structure interference and the true deformation of the lens substrate, resulting in significant deviations in test results. Second, traditional detection methods only extract single-dimensional optical features, failing to comprehensively cover multi-scale information such as macroscopic curvature of the mirror surface, uniformity of the intermediate coating, and microscopic surface defects, resulting in missing feature information and difficulty in accurately characterizing the overall optical state of the mirror. Furthermore, traditional detection methods often employ single-view imaging, which is prone to detection blind spots due to the curved shape of the lens surface. Single-view distortion also leads to decreased detection accuracy in edge areas, making accurate assessment of the entire mirror surface deformation impossible. Finally, existing methods lack a correlation model between coating structure parameters and detection errors, making error tracing and compensation difficult.

[0004] Therefore, in order to address the problems of internal reflection interference, lack of single feature information, blind spots and distortion in single-view imaging in traditional testing, there is an urgent need for a method to detect the deformation of optical reflection images on the lens surface, so as to improve the detection accuracy and practicality and meet the quality inspection needs of modern optical lenses. Summary of the Invention

[0005] To overcome the shortcomings of existing technologies, this invention proposes a method for detecting the distortion of optical reflection images on a lens surface. This invention primarily addresses the problems of inaccurate detection due to internal reflections in traditional detection methods, information loss due to relying on single features, and blind spots and distortions in single-view imaging.

[0006] The technical solution adopted by this invention to solve its technical problem is: a method for detecting the optical reflection image distortion of a lens surface provided by this invention, comprising: The structural parameters and ambient light field data of the lens under test are collected. The structural parameters include substrate curvature, composite film thickness and interlayer refractive index distribution.

[0007] The optical measurement data of the mirror is obtained by analyzing the interference effect of structural parameters on the reflected light of the mirror using the multi-layer curved optical detection method. The optical surface composite features are obtained by using the variational mode decomposition algorithm to extract features from the optical surface measurement data, and a surface morphology reconstruction model is constructed based on the optical surface composite features.

[0008] Using Fresnel equations and ray tracing algorithms, the transmission and reflection coefficients of light at the interfaces of various media in the lens under test are calculated. Combined with the parasitic light field interference generated by multiple reflections inside the film, an error transfer function based on the coupling relationship between film thickness variation and phase change is constructed.

[0009] The parasitic light field interference is inversely compensated based on the error transfer function, the real surface phase gradient field is extracted, the surface morphology reconstruction model is input, and the deformation evaluation result is obtained by data fusion calibration combined with multi-directional reflection imaging perspective.

[0010] The present invention provides a method for detecting the optical reflection image distortion of a lens surface, which uses a multi-layer curvature optical method to obtain measured optical data of the lens surface, including the following steps: The incident angle range of light is determined based on the curvature of the substrate, and different reflective interfaces are distinguished by the thickness of the composite film to form multi-layer multi-path reflected beams. Effective reflected beams are selected by combining the interlayer refractive index distribution.

[0011] Using the thickness of the composite film as the basis for the optical path propagation distance, the actual propagation optical path of the effective reflected beam is calculated by combining the interlayer refractive index distribution. The total optical path difference between any two effective reflected beams is calculated by combining the optical path offset of the substrate curvature.

[0012] Based on the correspondence between the total optical path difference and the wavelength of monochromatic light, the theoretical phase difference between each effective reflected beam is calculated, and the characteristics of the interference fringe distribution are obtained.

[0013] The interference detection optical path of the lens under test is determined based on the distribution characteristics of the interference fringes, and the reflected light interference image of the lens under test is collected to extract the measured interference features.

[0014] The cause of fringe deviation is obtained by comparing the measured interference features with the theoretical interference features corresponding to the structural parameters, and then the reverse conversion is performed to obtain the measured data of mirror optics.

[0015] The present invention provides a method for detecting the optical reflection image distortion of a lens surface, the steps of which include obtaining the composite features of the optical surface are as follows: Based on the multi-scale distribution characteristics of specular optical errors, the parameters of the variational mode decomposition algorithm are set, and a variational mode decomposition constraint model for specular optical measured data is constructed. The specular optical measured data is input and iterative calculation is performed to obtain modal components at multiple different frequency scales.

[0016] Modal components are divided into corresponding modal types according to frequency scale, and specular optical detection features are extracted from the corresponding modal types.

[0017] Based on the optical detection characteristics of mirror surfaces, the correlation between curved surfaces, films, and micro-surfaces is integrated to extract composite features of energy proportion, scale correlation coupling, spatial distribution, and interference response as optical surface composite features.

[0018] The present invention provides a method for detecting the deformation of optical reflection images on a lens surface, the steps of which include obtaining modal components at multiple different frequency scales: Based on the components of mirror error, the modal data are divided and decomposed. A secondary penalty factor and a convergence termination condition are set. Based on the inherent fluctuation frequency range of mirror optical error, the initial center frequency range is preset to obtain the variational mode decomposition algorithm parameters.

[0019] Using the measured data of mirror optics as the original signal for decomposition, the overall optical wave signal is defined as the linear superposition of multiple bandwidth-limited intrinsic mode components.

[0020] Based on minimizing the bandwidth of each modal component, a variational constraint objective is established. Hilbert transform is introduced to solve the single-sided spectrum of each optical wave mode. Combined with frequency domain hybrid constraints, the frequency distribution intervals of optical errors at different scales are distinguished, and a variational mode decomposition constraint model is constructed.

[0021] The system iteratively updates the intrinsic modal components and initial center frequency ranges of each order, and separates the curvature fluctuation signal, film thickness fluctuation signal, and refractive index fluctuation signal.

[0022] After each iteration, the convergence termination condition is determined. When the signal reconstruction error is less than the set threshold, the iteration operation is terminated immediately, and multiple modal components of different frequency scales after the decomposition of the specular optical measurement data are obtained.

[0023] The present invention provides a method for detecting the deformation of optical reflection images on a lens surface, wherein the steps of constructing a surface topography reconstruction model include: The composite features of optical surfaces are integrated into a multi-dimensional feature vector as input sample data. Simultaneously, the mirror three-dimensional surface topography data of the input sample data is collected, and the height values ​​of spatial coordinate points and the regional surface contour curves are used as the true values ​​of the model output labels.

[0024] The hierarchical logic of reconstruction is determined, and a model framework is constructed using hierarchical fitting and global fusion modeling methods. The mathematical correlation between optical surface composite features and surface morphology parameters is established, and morphology reconstruction sub-models are built in layers.

[0025] The loss function and iterative training parameters are determined based on optical morphology evaluation indicators. The morphology reconstruction sub-model is trained in combination with optical physical constraints until the preset accuracy is achieved in the morphology fitting of the corresponding level.

[0026] The superposition weights are determined based on the dominant role of optical surface composite features in different regions, and the weights are fused to obtain the surface morphology reconstruction model after training.

[0027] The present invention provides a method for detecting the optical reflection image distortion on the surface of a lens. The steps for calculating the transmission and reflection coefficients of light at the interfaces of various media of the lens under test include: All dielectric layers are arranged sequentially according to the order of light incidence, and the physical thickness, refractive index at the corresponding working wavelength and the position of the dielectric boundary interface of each layer are recorded. A three-dimensional rectangular coordinate system is established to distinguish the interface type and determine the incident light source properties and the termination condition of ray tracing.

[0028] Based on the initial propagation direction vector of the incident ray, the spatial coordinates of the ray reaching the first medium interface are calculated, the incident angle is solved using the vector dot product formula, and the refraction angle is solved using Snell's law to determine total internal reflection.

[0029] By substituting the incident angle and refraction angle into Fresnel's formula, the amplitude reflection and transmission coefficients of a single interface are calculated, and the transmission and reflection coefficients are derived.

[0030] The direction vector of the reflected ray is updated based on the reflection geometry, and the ray propagation path is updated by combining ray tracing. The transmission and reflection coefficients of the ray in each layer of the medium are recorded.

[0031] The present invention provides a method for detecting the deformation of optical reflection images on a lens surface, wherein the steps of constructing an error transfer function include: The parasitic light field formed by multiple reflections within a single film layer is analyzed, and the phase and amplitude of each parasitic light field are calculated to derive the total reflection coefficient.

[0032] The phase abrupt change is determined by the effect of film thickness variation on the phase, and the parasitic interference error is obtained by analyzing the effect of the phase abrupt change on the total reflectance coefficient of the parasitic optical field.

[0033] By using parasitic interference error as input and phase abrupt change as intermediate variable, a linear transmission model is established to analyze the transmission relationship from film thickness deviation to phase deviation and obtain the single film deviation.

[0034] The reflection coefficient deviation is obtained by combining the change in the equivalent reflection coefficient caused by a unit phase change with the total reflection coefficient.

[0035] The error function of a single film layer is obtained by mapping the single film deviation to the reflection coefficient deviation, and the error function of each film layer is superimposed to obtain the error propagation function.

[0036] The present invention provides a method for detecting the deformation of optical reflection images on a lens surface, wherein the step of extracting the phase gradient field of the real surface includes: Based on the error transfer function, the mapping relationship between the thickness deviation of each film layer and the parasitic phase perturbation is determined. Combined with the interference law of multiple reflection light fields, a global total parasitic model is constructed.

[0037] The global spatial phase data of the mirror under test is collected, and the continuous spatial distribution observation phase is obtained by unwrapping. The theoretical parasitic interference phase field under the current mirror state is calculated by substituting the film process parameters, film thickness distribution and interlayer refractive index distribution into the error transfer function.

[0038] The true phase is obtained by subtracting the theoretical parasitic interference phase field from the observed phase of the continuous spatial distribution. The true phase gradient in the x and y directions is solved by the spatial difference method, and then integrated to obtain the true surface phase gradient field.

[0039] The present invention provides a method for detecting the distortion of optical reflection images on a lens surface, the steps of which include obtaining the distortion assessment result are as follows: The real surface phase gradient field is aligned with global coordinates, split into orthogonal dual components, matched with the input dimension of the surface morphology reconstruction model, and solved using the global path integral algorithm to obtain the three-dimensional surface morphology from a single viewpoint.

[0040] Multi-view visual imaging data of the mirror under test from multiple spatial orientations and reflection imaging angles are collected, and the equivalent surface morphology observed at the corresponding angles is inferred to obtain multiple sets of morphology observation results from different perspectives.

[0041] Based on the incident angle, optical path propagation stability, and imaging signal-to-noise ratio, a fusion weight is assigned to each set of different viewpoint topography observation results. Using the three-dimensional surface topography under a single viewpoint as a benchmark, multiple sets of different viewpoint topography observation results are fused and weighted to obtain the calibrated surface topography field.

[0042] Using the designed morphology of the mirror target as a benchmark, and comparing it with the calibrated surface morphology field, the actual morphology deviation, peak-valley deformation deviation and root mean square surface deformation are calculated point by point.

[0043] The actual reflection and refraction angle of light across the entire domain is calculated based on the calibrated surface topography field. The comprehensive reflection image deformation is obtained by combining the peak-valley deformation deviation and the root mean square surface deformation.

[0044] Based on the preset lens testing standards, the deformation of the comprehensive reflection image is divided into multiple evaluation levels, and the deformation causes and deformation areas are distinguished to obtain the deformation evaluation results.

[0045] The present invention provides a method for detecting the deformation of optical reflection images on a lens surface. Using the three-dimensional surface morphology from a single viewpoint as a benchmark, the method fuses and corrects multiple sets of morphology observations from different viewpoints using a weighted fusion formula to obtain the calibrated surface morphology field. The formula is expressed as follows: In the formula, It is to calibrate the surface topography field. It is a three-dimensional surface morphology from a single viewpoint. It is the total number of perspectives. It is the first The fusion weight of each perspective It is the first The surface morphology obtained by inversion from a different perspective.

[0046] The beneficial effects of this invention are as follows: This invention significantly suppresses system noise and film interference through parasitic light field inverse compensation and theoretical-measured comparison, resulting in more accurate capture of minute deformations. Employing a hierarchical fitting and global fusion modeling strategy, combined with cross-validation of multi-view data, the final calibrated surface topography field is unaffected by local overexposure or shadow occlusion, ensuring extremely high reliability. It not only quantitatively calculates specific peak-valley deformation deviations and root-mean-square surface deformation, but also distinguishes the causes of deformation from specific regions based on deformation characteristics, providing invaluable reverse guidance for subsequent improvements in lens manufacturing processes. Attached Figure Description

[0047] The invention will now be further described with reference to the accompanying drawings.

[0048] Figure 1 This is a schematic flowchart of a method for detecting the deformation of optical reflection images on the surface of a lens, provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of the process for obtaining the transmission and reflection coefficients in a method for detecting the deformation of optical reflection images on the surface of a lens, provided in an embodiment of the present invention. Figure 3 This is a schematic diagram of the process for obtaining the true surface phase gradient field in a method for detecting the deformation of optical reflection images on a lens surface provided in an embodiment of the present invention. Detailed Implementation

[0049] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0050] like Figures 1 to 3 As shown in the figure, an embodiment of the present invention provides a method for detecting the optical reflection image distortion of a lens surface, comprising: The structural parameters and ambient light field data of the lens under test are collected. The structural parameters include substrate curvature, composite film thickness and interlayer refractive index distribution.

[0051] The optical measurement data of the mirror is obtained by analyzing the interference effect of structural parameters on the reflected light of the mirror using the multi-layer curved optical detection method. The optical surface composite features are obtained by using the variational mode decomposition algorithm to extract features from the optical surface measurement data, and a surface morphology reconstruction model is constructed based on the optical surface composite features.

[0052] The steps for obtaining specular optical measurement data using multi-layer refractive indexing include: The incident angle range of light is determined based on the curvature of the substrate, and different reflective interfaces are distinguished by the thickness of the composite film to form multi-layer multi-path reflected beams. Effective reflected beams are selected by combining the interlayer refractive index distribution.

[0053] The curvature parameters of the substrate are obtained by actual measurement, including the overall curvature radius of the mirror, local surface gradient, surface sagitta, and curvature distribution of the optically effective area. This clarifies whether the mirror is a spherical, aspherical, or freeform surface and determines the degree of surface bending at different locations.

[0054] The physical thickness of the composite film is accurately measured using a film thickness detection device. The actual physical thickness of each optical film layer is measured sequentially, and the deviation between the total thickness of the film layers and the thickness of a single layer is statistically analyzed to determine the order of film layer arrangement.

[0055] By using spectral detection and optical fitting, the interlayer refractive index distribution is determined, and the inherent refractive index of each film layer, the refractive index difference between adjacent film layers, and the refractive index difference between the film layer and the contact surface of the lens substrate are obtained.

[0056] The incident angle range of the incident light on the mirror surface is determined by combining the curvature of the substrate. The incident angle varies at different points on the curved surface, which directly changes the propagation direction and exit path of the reflected light.

[0057] Based on the physical thickness of the composite film, different reflective interfaces are distinguished, and the reflected light from the air and the upper surface of the outermost film, the reflected light from the interfaces between each film layer, and the reflected light from the contact surface between the innermost film and the lens substrate are separated in sequence to form a multi-layered, multi-path reflected light beam.

[0058] By combining the interlayer refractive index distribution, the intensity of beam reflection and the transmission ratio are determined. The greater the refractive index difference, the higher the intensity of the reflected light at the interface, and the higher the proportion of the effective beam energy participating in the interference, thus clarifying the coherent participation degree of each reflected beam. Effective reflected beams are then selected.

[0059] Using the thickness of the composite film as the basis for the optical path propagation distance, the actual propagation optical path of the effective reflected beam is calculated by combining the interlayer refractive index distribution. The total optical path difference between any two effective reflected beams is calculated by combining the optical path offset of the substrate curvature.

[0060] Using the physical thickness of the composite film as the basis for the optical path propagation distance, the propagation distance is directly determined by the actual thickness of the film when light passes perpendicularly or obliquely through a single-layer film or a multi-layer composite film. Substituting the interlayer refractive index distribution, the actual propagation optical path of each beam of reflected light from the interface is calculated inside the film and on the substrate surface according to the formula: optical path = refractive index of the medium × propagation distance.

[0061] Based on the correspondence between the total optical path difference and the wavelength of monochromatic light, the theoretical phase difference between each effective reflected beam is calculated, and the characteristics of the interference fringe distribution are obtained.

[0062] The density, orientation, curvature, and distribution pattern of interference fringes are considered as characteristics of interference fringe distribution.

[0063] The interference detection optical path of the lens under test is determined based on the distribution characteristics of the interference fringes, and the reflected light interference image of the lens under test is collected to extract the measured interference features.

[0064] Adjust the incident angle and illumination range of the coherent light source according to the curvature of the substrate to ensure that multi-layer interface reflection light can be formed in the entire curved area of ​​the mirror.

[0065] By matching the spectral response characteristics of the composite film, the detection light source band is selected so that the light can penetrate the entire film structure and excite all film interfaces to generate reflected beams.

[0066] The position of the imaging acquisition device in the fixed optical path is adjusted according to the curvature of the surface to eliminate the imaging field deviation caused by the tilt of the surface and ensure complete capture of the global interference image.

[0067] When the coherent light source is turned on, the light shines on the mirror under test. After reflection from the curved surface of the substrate and layered reflection from the interface of the multilayer composite film, multiple coherent beams superimpose and interfere, forming an actual interference fringe pattern in the imaging area.

[0068] Interference images are collected in sections, including the flat area at the center of the mirror, the edge area with large curvature, and the area with uneven film thickness, to fully record the real interference state at different structural locations.

[0069] Sample images are acquired multiple times in succession, and abnormal images caused by instantaneous fluctuations in the optical path are removed to obtain reflected light interference images.

[0070] The acquired interference images are processed by noise reduction, grayscale equalization, and fringe enhancement to clearly distinguish bright fringes, dark fringes, and fringe distortion areas.

[0071] Accurately extract the characteristics of measured interference fringes: actual fringe spacing, fringe bending offset, fringe breakage location, order of bright and dark fringes, and global phase fluctuation range.

[0072] The cause of fringe deviation is obtained by comparing the measured interference features with the theoretical interference features corresponding to the structural parameters, and then the reverse conversion is performed to obtain the measured data of mirror optics.

[0073] The theoretical interference features corresponding to the substrate curvature, film physical thickness, and interlayer refractive index distribution are compared one by one with the measured interference features extracted from the image.

[0074] By comparing the overall deviation of the fringes, the causes of the deviations can be determined as follows: Optical path deflection interference deviation caused by the actual curvature of the substrate deviating from the design value; optical path difference interference deviation caused by uneven actual thickness of the composite film; and phase difference interference deviation caused by the actual refractive index distribution between layers deviating from the standard value.

[0075] Based on the overall curvature of the interference fringes and the influence of substrate curvature interference, the actual surface shape accuracy, measured curvature error, and surface imaging deflection angle of the mirror are calculated.

[0076] Based on the interference order shift caused by reflection at the film interface, the actual physical thickness of the composite film, the uniformity of the thickness of a single layer, and the measured values ​​of the overall thickness deviation of the film can be deduced.

[0077] Based on the interference intensity and phase shift of the multilayer beams, the actual refractive index difference and refractive index gradient distribution between layers are calculated to obtain the optical uniformity data of the film.

[0078] By integrating the surface optical parameters, film geometric parameters, and film optical parameters, the actual measured optical indicators of the mirror are calculated, such as mirror reflection imaging distortion, reflection wavefront error, optical flatness, and consistency of reflection and transmission.

[0079] The steps to obtain optical surface composite features include: Based on the multi-scale distribution characteristics of specular optical errors, the parameters of the variational mode decomposition algorithm are set, and a variational mode decomposition constraint model for specular optical measured data is constructed. The specular optical measured data is input and iterative calculation is performed to obtain modal components at multiple different frequency scales.

[0080] The steps to obtain modal components at multiple different frequency scales include: Based on the components of mirror error, the modal data are divided and decomposed. A secondary penalty factor and a convergence termination condition are set. Based on the inherent fluctuation frequency range of mirror optical error, the initial center frequency range is preset to obtain the variational mode decomposition algorithm parameters.

[0081] Based on the composition of mirror errors, the modes are divided into: low-frequency overall surface deformation mode, mid-frequency composite film unevenness mode, and high-frequency microscopic surface defect mode. The reasonable number of modes is determined by combining the fluctuation complexity of measured data to ensure that the three core errors of curvature deviation, film thickness deviation, and refractive index fluctuation can be completely separated.

[0082] Set a secondary penalty factor: Select an appropriate penalty coefficient to balance the effect of modal component bandwidth constraint, avoid excessive mode compression due to excessively large parameters, and avoid modal spectrum aliasing due to excessively small parameters, and adapt to optically smoothed fluctuation data.

[0083] Using the measured data of mirror optics as the original signal for decomposition, the overall optical wave signal is defined as the linear superposition of multiple bandwidth-limited intrinsic mode components.

[0084] Based on minimizing the bandwidth of each modal component, a variational constraint objective is established. Hilbert transform is introduced to solve the single-sided spectrum of each optical wave mode. Combined with frequency domain hybrid constraints, the frequency distribution intervals of optical errors at different scales are distinguished, and a variational mode decomposition constraint model is constructed.

[0085] The system iteratively updates the intrinsic modal components and initial center frequency ranges of each order, and separates the curvature fluctuation signal, film thickness fluctuation signal, and refractive index fluctuation signal.

[0086] After each iteration, the convergence termination condition is determined. When the signal reconstruction error is less than the set threshold, the iteration operation is terminated immediately, and multiple modal components of different frequency scales after the decomposition of the specular optical measurement data are obtained.

[0087] Modal components are divided into corresponding modal types according to frequency scale, and specular optical detection features are extracted from the corresponding modal types.

[0088] Modal types may include: low-frequency modal components, which correspond to large-scale macroscopic optical characteristics such as overall curvature deviation of the substrate, large-scale surface deformation of the mirror, and overall distortion of global reflection.

[0089] Mid-frequency modal components: correspond to the optical characteristics of the mid-layer structure, such as uneven thickness fluctuations of the composite film, deviations in the distribution of refractive index gradient between layers, and optical differences at the film interface.

[0090] High-frequency modal components: correspond to microscopic surface optical features such as mirror roughness, surface micro-ripples, local micro-deformation, and coating micro-defects.

[0091] Extracting optical inspection features of the mirror surface includes: Low-frequency curvature-related features: extract modal mean, peak-valley deviation, spatial fluctuation amplitude, overall tilt, and surface distortion amplitude to characterize the overall curved optical state of the mirror substrate.

[0092] Mid-frequency film-related characteristics: Extract modal fluctuation variance, periodic fluctuation frequency, interlayer phase shift, thickness deviation dispersion, and refractive index uniformity coefficient to characterize the optical differences caused by the physical thickness and refractive index distribution of the composite film.

[0093] High-frequency microscopic surface features: Extract modal root mean square values, high-frequency impact amplitudes, microscopic fluctuation frequencies, and local defect fluctuation intensities to characterize the optical morphology of the mirror microscopic surface.

[0094] Based on the optical detection characteristics of mirror surfaces, the correlation between curved surfaces, films, and micro-surfaces is integrated to extract composite features of energy proportion, scale correlation coupling, spatial distribution, and interference response as optical surface composite features.

[0095] Energy proportion composite characteristics: Calculate the proportion of low-frequency, mid-frequency, and high-frequency signal energy in the total optical signal energy to intuitively determine the main sources of mirror optical errors, such as overall surface deformation / film unevenness / microscopic surface defects.

[0096] Scale-related coupling characteristics: Calculate the wave synchronization coefficient between low-frequency curvature mode and mid-frequency film mode, and the distortion correlation degree between mid-frequency film mode and high-frequency micro-mode, and analyze the linkage law between structural parameter deviation and surface optical defects.

[0097] Spatial distribution composite features: By combining the spatial distribution location of modes, the difference values ​​of optical features at different scales in the central and edge regions of the mirror are extracted to obtain the global non-uniform optical composite distribution features.

[0098] Interference response composite characteristics: Correlate each modal component with the interference fringe distortion data mentioned above to construct the corresponding coupling characteristics of modal fluctuation quantity and interference deformation quantity.

[0099] The steps for constructing a surface topography reconstruction model include: The composite features of optical surfaces are integrated into a multi-dimensional feature vector as input sample data. Simultaneously, the mirror three-dimensional surface topography data of the input sample data is collected, and the height values ​​of spatial coordinate points and the regional surface contour curves are used as the true values ​​of the model output labels.

[0100] A white light interferometer with a resolution better than 0.1 nm was used to scan the three-dimensional surface topography of the training sample lenses region by region, and the spatial coordinate height value z(x,y) of each sampling point was obtained in each region. For large curvature and aspherical lenses, a contact coordinate measuring machine (CMM) was used to scan the contour of typical cross sections of the mirror surface to obtain the regional surface contour curve. The above two types of measurement results were aligned with the coordinate system and interpolated and matched with the sampling point positions to be integrated into a three-dimensional ground truth dataset under a unified coordinate system, which was stored as the model training label.

[0101] The calibration process is as follows: (1) Calibration of reference sample: Using λ / 20 grade (λ=632.8nm) standard plane mirror and standard spherical mirror as reference samples, the measurement accuracy of white light interferometer and contact CMM is verified respectively, and it is confirmed that the surface shape measurement error of the two types of equipment is better than 5nm; (2) Coordinate system alignment: Using three or more high-precision reference points, the coordinate system of the white light interferometer, the coordinate system of the contact CMM and the coordinate system of the optical detection system are aligned through rigid body transformation to ensure that the translation error between each coordinate system is less than 1μm and the rotation error is less than 0.01°. (3) Sampling density matching: Ensure that the spatial sampling density of the true data is not lower than the spatial resolution of the optical surface composite features, and ensure that the feature input and the label output correspond one-to-one in spatial coordinates.

[0102] The hierarchical logic of reconstruction is determined, and a model framework is constructed using hierarchical fitting and global fusion modeling methods. The mathematical correlation between optical surface composite features and surface morphology parameters is established, and morphology reconstruction sub-models are built in layers.

[0103] The reconstruction layering logic is determined and corresponds one-to-one with the optical composite feature layers: overall surface morphology = macroscopic basic surface morphology of the substrate + mid-layer morphology disturbance caused by the composite film layer + surface micro-texture undulation morphology.

[0104] The model input is defined as the fused multi-scale optical composite feature set, and the output is the discrete point height values ​​of the global mirror surface and the continuous three-dimensional surface shape distribution results.

[0105] The overall approach to building the reconstruction model was determined, employing a hierarchical fitting + global fusion modeling method. First, single-scale features were used to fit the corresponding hierarchical morphology. Then, weighted fusion was used to achieve a complete global morphology reconstruction, conforming to the forming rules of optical mirror structures. The surface morphology reconstruction model adopted a hierarchical hybrid regression model architecture based on supervised learning.

[0106] The steps to establish mathematical correlations may include: combining the principle of optical reflection interference with the theory of surface geometry, establishing a mathematical mapping relationship between low-frequency curvature composite characteristics and the coefficients, radius of curvature, and spatial gradient of the substrate surface equation, and determining the basic fitting formula for macroscopic surface morphology.

[0107] Based on the optical propagation law of film layers, a conversion relationship is established between the composite characteristics of mid-frequency film layers and the local height offset of the surface caused by film thickness deviation and interlayer refractive index difference, so as to quantify the morphological distortion caused by the coating structure.

[0108] Based on the optical response laws of light scattering and micro-morphology, a corresponding calculation model is established for high-frequency micro-composite features and micro-morphological parameters such as surface roughness, micro-ripples, and local micro-dimples and protrusions.

[0109] A macroscopic substrate surface reconstruction sub-model was constructed using a hybrid modeling approach combining Zernike polynomial surface fitting and support vector regression (SVR, with a radial basis function (RBF) kernel). Low-frequency optical composite features were used as the sole input vector, encompassing six dimensions: low-frequency energy proportion, modal mean, peak-to-valley deviation, spatial fluctuation amplitude, overall tilt, and surface distortion amplitude. The model was fitted to obtain the standard basic surface profile of the mirror without coating or microscopic defects, restoring the original design curvature of the mirror and completing the overall large-scale morphological framework. The SVR regularization parameter C ranged from [1, 100], and its optimal value was determined through 5-fold cross-validation. The kernel function bandwidth γ was initialized heuristically and then optimized through grid search.

[0110] The Zernike polynomial expansion is as follows: In the formula, It is the macroscopic surface height. It is the radius. It's the angle. It is the radial order. It is the angular frequency. Zernike polynomial basis functions: radial polynomial Defined as: In the formula, It is the loop variable for summation. It is a radial coordinate. Power term.

[0111] In the application, the first 36 Zernike coefficients (corresponding to the ANSI Z80.28 standard) are selected to initially fit and obtain the basic surface. The SVR regularization parameter C is determined to have its optimal value in the range [1, 100] through 5-fold cross-validation, and the bandwidth of the RBF kernel function is... Optimized through grid search, the search range is: The step size is logarithmically uniform. The SVR output corrects the Zernike fitting residuals, improving fitting accuracy.

[0112] A sub-model for reconstructing the morphology of the film perturbation was constructed: a three-layer fully connected multilayer perceptron (MLP) was adopted, with the network layer structure as follows: input layer → fully connected layer (128 nodes, ReLU activation) → fully connected layer (64 nodes, ReLU activation) → output layer (fully connected, linear activation). The mid-frequency composite features of the film layers were used as the input vector, including five dimensions: modal fluctuation variance, periodic fluctuation frequency, interlayer phase shift, thickness deviation dispersion, and refractive index uniformity coefficient. The additional surface height undulations caused by the uneven thickness of the composite film layers and the differences in interlayer refractive index distribution were calculated, and the mid-layer morphology deviation field superimposed on the base surface was fitted. The optimizer used was Adam (initial learning rate lr=0.001, β1=0.9, β2=0.999), with a batch size of 32, a maximum number of iterations of 500, and an early stopping strategy (patience=20 rounds).

[0113] A sub-model for reconstructing microscopic surface textures was constructed using a lightweight one-dimensional convolutional neural network (1D-CNN). The network layer structure was as follows: input layer → convolutional layer (32 filters, kernel size 3, ReLU activation) → max pooling layer (stride 2) → convolutional layer (64 filters, kernel size 3, ReLU activation) → global average pooling layer → fully connected layer (32 nodes, ReLU activation) → output layer (linear activation). High-frequency microscopic optical composite features were used as the input vector, including four dimensions: modal root mean square value, high-frequency impact amplitude, microscopic fluctuation frequency, and local defect fluctuation intensity. This accurately fitted the micro-morphological distribution of micrometer- and nanometer-level undulations, texture patterns, and local defects on the mirror surface, supplementing the fine-grained morphological details.

[0114] The superposition weight is determined based on the dominant role of optical surface composite characteristics in different regions. ,satisfy The weights are dynamically allocated by the confidence scores of each sub-model in the corresponding region's morphology reconstruction: in regions dominated by large-scale curvature deformation of the substrate, Higher; in regions where film inhomogeneity causes significant deviations in the morphology of the middle layer, Higher; in regions dominated by microscopic surface defects, Relatively high.

[0115] The loss function and iterative training parameters are determined based on optical morphology evaluation indicators. The morphology reconstruction sub-model is trained in combination with optical physical constraints until the preset accuracy is achieved in the morphology fitting of the corresponding level.

[0116] Optical morphology evaluation indicators include three core losses: root mean square error of surface height, peak and valley error of surface shape, and overlap error of profile curve.

[0117] The formula for the root mean square error loss function of surface height is expressed as: In the formula, This represents the total number of spatial sampling points. For the first The model predicts the height value for each sampling point. For the first The true height of each sampling point It is the root mean square error loss function for surface height.

[0118] The formula for the surface peak-valley error loss function is expressed as follows: In the formula, These represent the maximum and minimum values ​​of the predicted height field, respectively. These represent the maximum and minimum values ​​of the true height field, respectively. It is the surface peak and valley value error loss function.

[0119] The formula for the contour curve overlap error loss function is expressed as follows: In the formula, To predict the average height, The true value is the field average height. It is the contour curve coincidence error loss function, with a value range of The closer a value is to 0, the higher the degree of overlap.

[0120] The formula for the comprehensive loss function is as follows: In the formula, This is a comprehensive loss function; recommended initial weights are: , , (The sum of the three is 1.0). , These are optical physical constraints; the aforementioned weighting coefficients can be adaptively adjusted during training based on the actual magnitude of each loss term.

[0121] Configure the model's iterative training parameters, set a reasonable learning rate, number of iterations, and convergence threshold, and set an early stopping mechanism to avoid overfitting of the model.

[0122] By incorporating optical and physical constraints, the curvature variation range, local height offset limit, and maximum value of micro-undulations of the reconstructed morphology are limited to ensure that the model output conforms to the physical laws of actual mirror production and forming, and to prevent the occurrence of distorted morphology data that does not conform to reality.

[0123] Optical physical constraint loss The curvature variation range, local height offset limit, and maximum amplitude of micro-undulations of the reconstructed morphology under comprehensive constraints are expressed by the following formula: In the formula, This represents the upper limit of the curvature variation range; The local height offset limit (determined by optical specifications, in nm); The maximum amplitude of micro-undulations (defined by the surface roughness level, in nm). The penalty coefficients for each physical constraint are 0.01, 0.05, and 0.05, respectively.

[0124] The superposition weights are determined based on the dominant role of optical surface composite features in different regions, and the weights are fused to obtain the surface morphology reconstruction model after training.

[0125] Using Fresnel equations and ray tracing algorithms, the transmission and reflection coefficients of light at the interfaces of various media in the lens under test are calculated. Combined with the parasitic light field interference generated by multiple reflections inside the film, an error transfer function based on the coupling relationship between film thickness variation and phase change is constructed.

[0126] The steps for calculating the transmission and reflection coefficients of light at the interfaces of various media in the lens under test include: All dielectric layers are arranged sequentially according to the order of light incidence, and the physical thickness, refractive index at the corresponding working wavelength and the position of the dielectric boundary interface of each layer are recorded. A three-dimensional rectangular coordinate system is established to distinguish the interface type and determine the incident light source properties and the termination condition of ray tracing.

[0127] Establish a three-dimensional rectangular coordinate system, solve the local unit normal vector of the interface point by point for the curved lens, and distinguish between planar interfaces and curvature-gradient curved surface interfaces.

[0128] Determine the properties of the incident light source: parallel monochromatic light, incident wavelength, incident angle range, and incident light polarization state.

[0129] Set the conditions for ray tracing to terminate: the light completely passes through the last layer of medium, total internal reflection occurs, or the light energy is lower than the set threshold.

[0130] Based on the initial propagation direction vector of the incident ray, the spatial coordinates of the ray reaching the first medium interface are calculated, the incident angle is solved using the vector dot product formula, and the refraction angle is solved using Snell's law to determine total internal reflection.

[0131] The angle of incidence can be solved using the vector dot product formula, which is expressed as: Obtained from inverse trigonometric functions: In the formula, It is the initial propagation direction vector of the incident ray. It is the unit normal vector of the medium interface. It is the angle of incidence.

[0132] The law of refraction is expressed as follows: Deformation to solve for the angle of refraction: In the formula, It is the refractive index of the incident medium. It is the refractive index of the exit-side medium. It is the angle of refraction.

[0133] Only when > There exists a critical angle, expressed by the formula: In the formula, It is the critical angle.

[0134] Judgment rule: If ≥ Total internal reflection occurs, at which point the transmission component is 0, and only the reflection correlation coefficient is calculated.

[0135] By substituting the incident angle and refraction angle into Fresnel's formula, the amplitude reflection and transmission coefficients of a single interface are calculated, and the transmission and reflection coefficients are derived.

[0136] The formula for s-polarized light with vibration direction perpendicular to the incident plane is expressed as: In the formula, It is the amplitude reflection coefficient of s-polarized light. It is the amplitude transmission coefficient of s-polarized light.

[0137] The formula for p-polarized light with vibration direction parallel to the plane of incidence is expressed as: In the formula, It is the amplitude reflection coefficient of p-polarized light. It is the amplitude transmission coefficient of p-polarized light.

[0138] The formula for the reflectivity of polarized light energy is expressed as: In the formula, It is the energy reflectivity of s-polarized light. It is the energy reflectivity of p-polarized light.

[0139] The formula for the energy transmittance of polarized light is expressed as: In the formula, It is the energy transmittance of s-polarized light. It is the energy transmittance of p-polarized light.

[0140] The direction vector of the reflected ray is updated based on the reflection geometry, and the ray propagation path is updated by combining ray tracing. The transmission and reflection coefficients of the ray in each layer of the medium are recorded.

[0141] The direction vector of the reflected ray is updated based on the reflection geometry, allowing the ray to continue propagating within the original medium, thus completing the intralayer round-trip reflection tracking.

[0142] Based on the relationship between the refraction angle and the interface normal, the spatial vector of the transmitted ray is updated, driving the ray to enter the next layer of medium.

[0143] For curved lenses, the local interface normal vector is refreshed in real time during the light's journey, and the incident angle is dynamically corrected to ensure that the tracking fits the true curvature of the substrate.

[0144] The steps to construct the error propagation function include: The parasitic light field formed by multiple reflections within a single film layer is analyzed, and the phase and amplitude of each parasitic light field are calculated to derive the total reflection coefficient.

[0145] Taking a single-layer film as an example, let's analyze all the optical field components involved in the interference: Component 1: Direct reflected light from the air-film interface.

[0146] Component 2: Light reflected through the film layer → film layer-substrate interface → secondary reflected light that then passes through the film layer again (first-order parasitic light).

[0147] Component 3: Reflection at the film-substrate interface → reflection at the film-air interface → reflection at the substrate interface (second-order parasitic light), and higher-order multiple reflection parasitic light.

[0148] Calculate the phase and amplitude of each parasitic optical field: The round-trip phase change of a first-order parasitic light is expressed by the following formula: In the formula, It is a round-trip phase change. It is the wavelength of the incident light. It is the first The refractive index of the film, It is the first The actual physical thickness of the film, It is light in the first The angle of refraction within the film.

[0149] Amplitude of first-order parasitic light: , ( The transmittance coefficient between air and the film layer. The reflectance coefficient of the film layer to the substrate. (This is the transmittance coefficient of the film layer to air).

[0150] Taking the three-layer structure of air-film-substrate as an example, the formula for the total reflectance is derived as follows: In the formula, It is the total reflectance. It is a phase factor. It is the air-to-film interface reflectance coefficient.

[0151] The phase abrupt change is determined by the effect of film thickness variation on the phase, and the parasitic interference error is obtained by analyzing the effect of the phase abrupt change on the total reflectance coefficient of the parasitic optical field.

[0152] The effect of film thickness variation on phase: When there is a deviation in film thickness, the phase thickness will change, as expressed by the formula: Wherein the phase deviation: In the formula, It is the total phase delay. It is the initial phase delay. It's a phase deviation. is the film thickness deviation, It refers to the design thickness.

[0153] Determination of phase abrupt change: When a change in film thickness leads to a phase change When the phase transition occurs at π, 2π, ..., a phase shift will occur, making... A jump in the sign or magnitude of the light field triggers a coherent constructive / destructive abrupt change in the optical field.

[0154] Disturbances in the parasitic optical field: Phase deviations can change the phase difference between the parasitic optical field and the main optical field at each order, resulting in a nonlinear change in the amplitude of the total optical field after superposition, thus forming parasitic interference errors.

[0155] By using parasitic interference error as input and phase abrupt change as intermediate variable, a linear transmission model is established to analyze the transmission relationship from film thickness deviation to phase deviation and obtain the single film deviation.

[0156] The formula for the linear transfer model is expressed as follows: Where the transfer coefficient is: In the formula, It is the transmission coefficient.

[0157] The reflection coefficient deviation is obtained by combining the change in the equivalent reflection coefficient caused by a unit phase change with the total reflection coefficient.

[0158] The formula for the reflection coefficient deviation is expressed as: In the formula, It's a deviation in the reflection coefficient. It is the derivative of the total reflection coefficient with respect to the phase.

[0159] The error function of a single film layer is obtained by mapping the single film deviation to the reflection coefficient deviation, and the error function of each film layer is superimposed to obtain the error propagation function.

[0160] The formula for the error function of a single film layer is as follows: In the formula, It is the error function of a single film layer. It is the derivative of the total reflection coefficient with respect to the phase.

[0161] The formula for the error propagation function, obtained by superimposing the error functions of each film layer, is as follows: In the formula, It is the error propagation function. It is the total number of floors.

[0162] The parasitic light field interference is inversely compensated based on the error transfer function, the real surface phase gradient field is extracted, the surface morphology reconstruction model is input, and the deformation evaluation result is obtained by data fusion calibration combined with multi-directional reflection imaging perspective.

[0163] The steps for extracting the phase gradient field of a real surface include: Based on the error transfer function, the mapping relationship between the thickness deviation of each film layer and the parasitic phase perturbation is determined. Combined with the interference law of multiple reflection light fields, a global total parasitic model is constructed.

[0164] The formula for the global total parasitic model is expressed as: In the formula, It is the total parasitic phase across the entire domain. It is a coupled superposition operator.

[0165] The global spatial phase data of the mirror under test is collected, and the continuous spatial distribution observation phase is obtained by unwrapping. The theoretical parasitic interference phase field under the current mirror state is calculated by substituting the film process parameters, film thickness distribution and interlayer refractive index distribution into the error transfer function.

[0166] The formula for the phase of continuous spatially distributed observations is expressed as: In the formula, It is the unwrapping phase. It is the original observation phase. These are spatial coordinates. This is the unpacking operation.

[0167] The theoretical formula for the parasitic interference phase field is expressed as: In the formula, It is the theoretical parasitic interference phase field. It is the film thickness deviation distribution function.

[0168] The true phase is obtained by subtracting the theoretical parasitic interference phase field from the observed phase of the continuous spatial distribution. The true phase gradient in the x and y directions is solved by the spatial difference method, and then integrated to obtain the true surface phase gradient field.

[0169] The formula for solving the true phase gradient in the x and y directions is expressed as: In the formula, It is the gradient of the true phase in the x-direction. It is the gradient of the true phase in the y-direction. It is the true phase distribution function. It is the derivative with respect to the x-coordinate. It is the derivative with respect to the y-coordinate.

[0170] The formula for the phase gradient field of a real surface is: In the formula, It is the real surface phase gradient field.

[0171] The steps to obtain the deformation assessment results include: The real surface phase gradient field is aligned with global coordinates, split into orthogonal dual components, matched with the input dimension of the surface morphology reconstruction model, and solved using the global path integral algorithm to obtain the three-dimensional surface morphology from a single viewpoint.

[0172] The formula for solving the 3D surface morphology from a single viewpoint using the global path integral algorithm is as follows: In the formula, It is a three-dimensional surface morphology from a single viewpoint. It is a path element.

[0173] Multi-view visual imaging data of the mirror under test from multiple spatial orientations and reflection imaging angles are collected, and the equivalent surface morphology observed at the corresponding angles is inferred to obtain multiple sets of morphology observation results from different perspectives.

[0174] Based on the incident angle, optical path propagation stability, and imaging signal-to-noise ratio, a fusion weight is assigned to each set of different viewpoint topography observation results. Using the three-dimensional surface topography under a single viewpoint as a benchmark, multiple sets of different viewpoint topography observation results are fused and weighted to obtain the calibrated surface topography field.

[0175] The formula for obtaining the fusion weights is expressed as: In the formula, It is the total number of perspectives. It is the first The fusion weights of each perspective.

[0176] The formula for obtaining the calibrated surface topography field is expressed as: In the formula, It is to calibrate the surface topography field. It is a three-dimensional surface morphology from a single viewpoint. It is the total number of perspectives. It is the first The fusion weight of each perspective It is the first The surface morphology obtained by inversion from a different perspective.

[0177] Using the designed morphology of the mirror target as a benchmark, and comparing it with the calibrated surface morphology field, the actual morphology deviation, peak-valley deformation deviation and root mean square surface deformation are calculated point by point.

[0178] The formula for calculating peak-valley deformation deviation is expressed as: In the formula, It is the peak-valley deformation deviation. It is the maximum value of the actual shape deviation across the entire region. It is the minimum value of the actual morphological deviation across the entire region. It is a deviation from the actual shape of the entire region.

[0179] The formula for calculating the root mean square surface deformation is expressed as: In the formula, It is the root mean square surface deformation. It is the area normalization coefficient. It is the square of the actual morphological deviation. It is an area micro-element.

[0180] The actual reflection and refraction angle of light across the entire domain is calculated based on the calibrated surface topography field. The comprehensive reflection image deformation is obtained by combining the peak-valley deformation deviation and the root mean square surface deformation.

[0181] The formula for calculating the actual reflection and refraction angle of light rays across the entire domain is expressed as: In the formula, It is the actual reflection and refraction angle of light across the entire region. It is the gradient operator.

[0182] Based on the preset lens testing standards, the deformation of the comprehensive reflection image is divided into multiple evaluation levels, and the deformation causes and deformation areas are distinguished to obtain the deformation evaluation results.

[0183] The assessment is divided into four levels: qualified, slight deformation, moderate deformation, and severe deformation.

[0184] Differentiate the causes of deformation: inherent deformation of the substrate during molding, micro-deformation due to post-process stress, and residual deformation from the process, to complete the auxiliary analysis for tracing the source of deformation.

[0185] The system statistically analyzes the global deformation distribution cloud map, the location of deformation extreme values, and the areas of high deformation concentration to form a visualized deformation distribution result.

[0186] In summary, this embodiment provides a method for detecting optical reflection image deformation on a lens surface. By analyzing the substrate curvature, film thickness, and interlayer refractive index, a pre-optical path analysis is conducted to deduce the optical response law from the physical structure level. This allows for accurate determination of the source of detection deviations, enabling reverse tracing of deformation errors from the detection end to the production process end. Variational mode decomposition separates high and low frequency signals, simultaneously extracting three types of composite features: macroscopic curved surfaces, intermediate coatings, and microscopic surfaces. This comprehensively covers optical defects across the entire lens surface, providing rich feature dimensions and high discriminative power. A layered fusion reconstruction model is built based on exclusive optical composite features, taking into account the basic curved surface morphology, film perturbation morphology, and microscopic texture morphology. This model can stably adapt to spherical, aspherical, freeform surfaces, and various multi-layer coated lenses, significantly improving morphology restoration fit. It outputs geometric deformation indicators such as peak-valley deviation and root mean square surface shape error, and also calculates actual imaging distortion parameters based on the reflection light path deflection law. The detection results closely match the actual optical performance of the lens, making the evaluation results more valuable for engineering applications. By using structural error modeling compensation, light field interference elimination, and multi-view data fault-tolerant fusion, the influence of external factors such as ambient light field fluctuations, minor vibrations, and temperature drift is weakened, significantly improving the anti-interference capability and field applicability of the detection method.

[0187] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the technical solution, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0188] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting the distortion of optical reflection images on a lens surface, characterized in that, include: Collect structural parameters and ambient light field data of the lens under test. The structural parameters include substrate curvature, composite film thickness and interlayer refractive index distribution. The interference effect of the structural parameters on the reflected light from the mirror is analyzed by the multi-layer curved optical detection method to obtain the measured optical data of the mirror. The optical surface composite features are obtained by feature extraction of the measured optical data of the mirror using the variational mode decomposition algorithm, and a surface morphology reconstruction model is constructed based on the optical surface composite features. Using Fresnel equations and ray tracing algorithms, the transmission and reflection coefficients of light at the interfaces of various media in the lens under test are calculated. Combined with the parasitic light field interference generated by multiple reflections inside the film, an error transfer function based on the coupling relationship between film thickness change and phase change is constructed. Based on the error transfer function, inverse compensation is performed for parasitic light field interference, the real surface phase gradient field is extracted, and the surface morphology reconstruction model is input. Combined with multi-directional reflection imaging perspective, data fusion calibration is performed to obtain the deformation evaluation result.

2. The method for detecting optical reflection image distortion on a lens surface according to claim 1, characterized in that: The steps for obtaining the measured optical data of the specular surface using the multi-layer refractive index method include: The incident angle range of light is determined based on the curvature of the substrate, and different reflective interfaces are distinguished by the thickness of the composite film layer to form multi-layer multi-path reflected beams. Effective reflected beams are selected by combining the interlayer refractive index distribution. Using the thickness of the composite film as the basis for the optical path propagation distance, and combining the interlayer refractive index distribution, the actual propagation optical path of the effective reflected beam is calculated. Combining the optical path offset of the substrate curvature, the total optical path difference between any two effective reflected beams is calculated. Based on the correspondence between the total optical path difference and the wavelength of monochromatic light, the theoretical phase difference between each effective reflected beam is calculated, and the interference fringe distribution characteristics are obtained. Based on the interference fringe distribution characteristics, the interference detection optical path of the lens under test is determined, and the reflected light interference image of the lens under test is acquired to extract the measured interference features. The cause of the fringe deviation is obtained by comparing the measured interference features with the theoretical interference features corresponding to the structural parameters, and then the measured data of the mirror optics is obtained by performing a reverse conversion.

3. The method for detecting optical reflection image distortion on a lens surface according to claim 1, characterized in that: The steps for obtaining the optical surface composite features include: Based on the multi-scale distribution characteristics of specular optical errors, the parameters of the variational mode decomposition algorithm are set, a variational mode decomposition constraint model for the measured specular optical data is constructed, and the measured specular optical data is input for iterative calculation to obtain modal components at multiple different frequency scales. Modal components are divided into corresponding modal types according to frequency scale, and specular optical detection features are extracted from the corresponding modal types; Based on the aforementioned mirror optical detection characteristics, the correlation between curved surfaces, films, and micro-surfaces is integrated, and the composite features of energy proportion, scale correlation coupling, spatial distribution, and interference response are extracted as the composite features of the optical surface.

4. The method for detecting optical reflection image distortion on a lens surface according to claim 3, characterized in that: The steps to obtain modal components at multiple different frequency scales include: Based on the components of mirror error, the modal data are divided and decomposed. A secondary penalty factor and a convergence termination condition are set. Based on the inherent fluctuation frequency range of mirror optical error, the initial center frequency range is preset to obtain the variational mode decomposition algorithm parameters. Using the measured data of the mirror optics as the original decomposition signal, the overall optical wave signal is defined as the linear superposition of multiple bandwidth-limited intrinsic mode components; The variational constraint objective is established based on minimizing the bandwidth of each modal component. The Hilbert transform is introduced to solve the one-sided spectrum of each optical wave mode. Combined with the frequency domain hybrid constraint, the frequency distribution intervals of optical errors at different scales are distinguished, and the variational mode decomposition constraint model is constructed. The intrinsic mode components of each order and the initial center frequency range are updated iteratively, and the curvature fluctuation signal, film thickness fluctuation signal and refractive index fluctuation signal are separated. After each iteration, the convergence termination condition is determined. When the signal reconstruction error is less than the set threshold, the iteration operation is terminated immediately, and multiple modal components of different frequency scales after the decomposition of the specular optical measurement data are obtained.

5. The method for detecting optical reflection image distortion on a lens surface according to claim 1, characterized in that: The steps for constructing the surface topography reconstruction model include: The optical surface composite features are integrated into a multi-dimensional feature vector as input sample data. Simultaneously, the mirror three-dimensional surface topography data of the input sample data are collected, and the spatial coordinate point height value and the regional surface contour curve are used as the model true value output label. The hierarchical logic of reconstruction is determined, and a model framework is constructed using a hierarchical fitting and global fusion modeling approach. The mathematical correlation between the optical surface composite features and surface morphology parameters is established, and a hierarchical morphology reconstruction sub-model is built. The loss function and iterative training parameters are determined based on the optical morphology evaluation index, and the morphology reconstruction sub-model is trained in combination with optical physical constraints until the preset accuracy is achieved in the morphology fitting of the corresponding level. The superposition weights are determined based on the dominant role of optical surface composite features in different regions, and the weights are fused to obtain the surface morphology reconstruction model after training.

6. The method for detecting optical reflection image distortion on a lens surface according to claim 1, characterized in that: The steps for calculating the transmission and reflection coefficients of light at the interfaces of various media in the lens under test include: Arrange all the medium layers in sequence according to the order of light incidence, and record the physical thickness of each layer, the refractive index at the corresponding working wavelength and the position of the medium boundary interface. Establish a three-dimensional rectangular coordinate system, distinguish the interface type, and determine the incident light source properties and the termination condition of ray tracing. Based on the initial propagation direction vector of the incident ray, calculate the spatial coordinates of the ray reaching the first medium interface, solve for the incident angle using the vector dot product formula, and use Snell's law to solve for the refraction angle to determine total internal reflection; Substituting the incident angle and the refraction angle into Fresnel's formula, the single-interface amplitude reflection and transmission coefficients are calculated, and the transmission and reflection coefficients are derived. The direction vector of the reflected ray is updated based on the reflection geometry, and the ray propagation path is updated by combining ray tracing. The transmission and reflection coefficients of the ray in each layer of the medium are recorded.

7. The method for detecting optical reflection image distortion on a lens surface according to claim 1, characterized in that: The steps for constructing the error transfer function include: The parasitic light field formed by multiple reflections within a single film layer is analyzed, and the phase and amplitude of each parasitic light field are calculated, thereby deriving the total reflection coefficient. The phase abrupt change is determined by the effect of film thickness variation on phase, and the parasitic interference error is obtained by analyzing the effect of the phase abrupt change on the total reflectance coefficient of the parasitic optical field. Using the parasitic interference error as input and the phase change as an intermediate variable, a linear transmission model is established to analyze the transmission relationship from film thickness deviation to phase deviation and obtain the single film deviation. The reflection coefficient deviation is obtained by combining the change in the equivalent reflection coefficient caused by a unit phase change with the total reflection coefficient. The error function of a single film layer is obtained by mapping the single film deviation to the reflection coefficient deviation, and the error function of each film layer is superimposed to obtain the error propagation function.

8. The method for detecting optical reflection image distortion on a lens surface according to claim 1, characterized in that: The steps for extracting the true surface phase gradient field include: Based on the error transfer function, the mapping relationship between the thickness deviation of each layer and the parasitic phase perturbation is determined. Combined with the interference law of multiple reflected light fields, a global total parasitic model is constructed. The global spatial phase data of the mirror under test is collected, and the continuous spatial distribution observation phase is obtained by unwrapping. The film process parameters, film thickness distribution and interlayer refractive index distribution are combined and substituted into the error transfer function to calculate the theoretical parasitic interference phase field under the current mirror state. The true phase is obtained by subtracting the theoretical parasitic interference phase field from the observed phase of the continuous spatial distribution. The true phase gradient in the x and y directions is solved by the spatial difference method and then integrated to obtain the true surface phase gradient field.

9. The method for detecting optical reflection image distortion on a lens surface according to claim 8, characterized in that: The steps to obtain the deformation assessment result include: The real surface phase gradient field is aligned with global coordinates, split into orthogonal dual components, matched with the input dimension of the surface morphology reconstruction model, and solved using the global path integral algorithm to obtain the three-dimensional surface morphology from a single viewpoint. Multi-view visual imaging data of the mirror under test from multiple spatial orientations are collected, and the equivalent surface morphology observed at the corresponding viewpoints is inferred to obtain multiple sets of morphology observation results from different viewpoints. Based on the incident angle, optical path propagation stability and imaging signal-to-noise ratio, a fusion weight is assigned to each group of different viewpoint morphology observation results. Taking the three-dimensional surface morphology under a single viewpoint as the benchmark, multiple groups of different viewpoint morphology observation results are fused and weighted to obtain the calibrated surface morphology field. Using the mirror target design morphology as a benchmark, the actual morphology deviation, peak-valley deformation deviation, and root mean square surface deformation are calculated point by point in the calibration surface morphology field. The actual reflection and refraction angle of the global light is calculated based on the calibrated surface topography field, and the comprehensive reflection image deformation is obtained by combining the peak-valley deformation deviation and the root mean square surface deformation. Based on the preset lens testing standards, the deformation of the comprehensive reflection image is divided into multiple evaluation levels, and the deformation cause and deformation area are distinguished to obtain the deformation evaluation result.

10. The method for detecting optical reflection image distortion on a lens surface according to claim 9, characterized in that: Using the three-dimensional surface morphology from a single viewpoint as a benchmark, the formula for calibrating the surface morphology field by weighted fusion correction of multiple sets of morphology observation results from different viewpoints is expressed as follows: In the formula, It is to calibrate the surface topography field. It is a three-dimensional surface morphology from a single viewpoint. It is the total number of perspectives. It is the first The fusion weight of each perspective It is the first The surface morphology obtained by inversion from a different perspective.

Images