A method for generating an initial structure of an RC telescope of an infrared Fourier spectrometer

Abstract

This invention discloses a method for generating the initial structure of an RC telescope for an infrared Fourier spectrometer, comprising: establishing a parameter configuration object; generating candidate parameter combinations in a five-dimensional input parameter space and performing adaptive hybrid sampling based on historical valid sample features; for each candidate parameter combination, performing a first-order analytical closure solution of the RC spectrum to eliminate invalid samples that do not satisfy the RC closure relationship; performing Fourier spectrometer constraint screening on samples that pass the first-order analytical closure of the RC spectrum; performing analytical three-dimensional geometric tracing on samples that pass the constraint screening, and dynamically solving the secondary mirror size and the primary mirror center aperture size during the tracing process; performing multi-objective sorting on valid samples that pass the analytical three-dimensional geometric tracing and dynamic size solution; and performing hierarchical organization and structured export of the sample results. This invention can solve the problems of reliance on manual experience, insufficient parameter space coverage, and unrealistic three-dimensional light transmission evaluation in existing technologies.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of infrared optical system design and computer-aided optical design, and in particular relates to a method for generating the initial structure of an RC telescope for an infrared Fourier spectrometer. Background Technology

[0002] Infrared Fourier transform spectrometers typically consist of a front-end telescope system, a subsequent collimation or relay system, an interferometry module, a focusing module, and a detection module. The front-end telescope system collects scene radiation, defines the field of view, and provides the subsequent optical link with an incident beam that meets interface requirements. Therefore, the structural quality of the front-end telescope system directly affects the subsequent design and performance verification of the entire system. For this type of system, a reflective front-end telescope offers advantages such as chromatic aberration-free operation, suitability for larger apertures, and a larger back-end workspace, making the RC telescope a commonly used candidate configuration.

[0003] For example, Chinese patent document CN113218506A discloses an infrared dual-band Fourier transform imaging spectrometer. The target being measured enters a front-view telescope system via a scanning mirror, and the front-view telescope system images the target's radiation information onto an interferometric system. Chinese patent document CN115265780A discloses a Fourier spectrometer device with optimized laser sampling wavelength. This device includes an ultra-stable laser, a small field-of-view telescope, an interferometric module, a laser detector, a detector module, and a control acquisition and processing computer, arranged sequentially according to the optical path transmission.

[0004] In current engineering practice, the initial structure of front-end RC telescopes typically relies on designers pre-determining key parameters such as primary mirror focal length, primary-secondary mirror spacing, radius of curvature, and cone constant based on experience, and then conducting trial-and-error optimization using commercial optical software. This approach is highly dependent on human experience, has low parameter scanning efficiency, and often fails to converge when the initial parameter values ​​deviate from the physically achievable range. Furthermore, using only a small number of single-variable scans or paraxial criteria makes it difficult to simultaneously cover the coupling relationships between aperture, full field of view, system F-number, back focal length ratio, and primary mirror F-number, and also makes it difficult to accurately reflect the actual blurring conditions at multiple points on the image plane under different fields of view, including secondary mirror obstruction, primary mirror center aperture backlight obstruction, and blurring at multiple points on the image plane.

[0005] Furthermore, existing methods often rely on numerical iteration when determining the intersection of light rays and quadric surfaces, which results in high computational cost, sensitivity to initial values, and a tendency to converge to non-physical solutions. In data-driven optical design, in addition to a small number of optimizable initial structures, it is also necessary to obtain a batch of high-quality initial structure datasets that are physically consistent, have realistic light transmission, are rankable in performance, and are relatively evenly distributed. However, existing methods still have shortcomings in sample generation efficiency, physical interpretability, hierarchical derivation capability, and dataset balance, making it difficult to simultaneously meet the dual requirements of subsequent optical software verification and intelligent optical design training.

[0006] Therefore, there is an urgent need to provide a method for generating the initial structure of the RC telescope of an infrared Fourier spectrometer, so as to automatically generate high-quality initial front-end structures in batches under system-level input conditions, while also taking into account the needs of subsequent optical software verification and the construction of training datasets for artificial intelligence optical design. Summary of the Invention

[0007] This invention provides a method for generating the initial structure of the RC telescope of an infrared Fourier spectrometer, which can solve the problems of existing technologies such as reliance on manual experience, insufficient parameter space coverage, unrealistic three-dimensional light transmission evaluation, low yield of effective sample generation, and difficulty in constructing high-quality balanced datasets.

[0008] A method for generating the initial structure of an RC telescope for an infrared Fourier spectrometer includes the following steps: (1) Establish a parameter configuration object and determine the input range, front-end interface constraints and operation control parameters required for the subsequent automatic generation of the initial structure; (2) Based on the input range, candidate parameter combinations are generated in the five-dimensional input parameter space; and adaptive hybrid sampling is performed by combining the features of historical effective samples; (3) For each set of candidate parameter combinations, perform RC first-order analytical closure solution and remove invalid samples that do not satisfy the RC closure relationship; (4) Based on the front-end interface constraints, Fourier spectrometer constraint screening is performed on the samples that pass the first-order analytical closure of RC; (5) Perform analytical three-dimensional geometric tracing on the samples that have passed the constraint screening, and dynamically solve the secondary mirror size and the primary mirror center hole size during the tracing process; (6) Perform multi-objective sorting on the effective samples solved by analytical three-dimensional geometric tracing and dynamic size calculation; (7) After completing the multi-objective sorting, the sample results are hierarchically organized and structured for export.

[0009] In step (1), the input range includes: primary mirror aperture range, full field of view range, system F number range, back focal length ratio range, and primary mirror F number range, which together constitute the five-dimensional input parameter space of the candidate parameter combination; The constraints include: maximum wavenumber, spectral resolution, and Jacquinot quantity threshold, which characterize the interface constraints of the mid-infrared ground-based passive Fourier spectrometer front-end system. The operational control parameters include: target effective sample count, number of candidate samples generated per batch, maximum original sample budget, number of pupil sampling rings, magnification constraint range, effective central occlusion ratio constraint range, upper limit of primary mirror center aperture ratio, lower limit of effective light ratio, secondary mirror shading margin coefficient, primary mirror aperture margin coefficient, assembly and adjustment allowance, maximum number of iterations for dynamic dimension solving, dynamic dimension convergence tolerance, multi-objective ranking weight, sample cleaning threshold, and balanced export of target sample count.

[0010] In step (2), the candidate parameter combination is generated by mapping the five-dimensional unit hypercube joint sampling to the range of real physical parameters, and the maximum and minimum distance Latin hypercube sampling is used. When the number of historical valid samples reaches the preset preheating threshold, the low quantile and high quantile values ​​of each sampling dimension are counted, and an expansion ratio is superimposed on this to form a key sampling interval. Global sampling is performed within the original parameter range and local sampling is performed within the key sampling interval according to the preset mixing ratio. Then, the two parts of samples are merged and their order is shuffled to form a new candidate parameter combination.

[0011] In step (3), the first-order analytic closure of RC includes the following relationship: ; ; ; ; ; ; in, The total focal length of the system. Back focal length, The focal length of the primary mirror. Main mirror aperture, For the system F number, Back focal length ratio, The F-number of the primary mirror, The distance between primary and secondary mirrors, The distance from the secondary mirror to the focal plane. The magnification of the secondary lens; if the candidate sample meets any of the following conditions: , , , , If the sample is not within the preset magnification range, it is considered an invalid sample and is removed.

[0012] Furthermore, the first-order analytic closure of RC also includes the following relationship: ; ; ; ; in, The radius of curvature of the primary mirror. The radius of curvature of the secondary mirror. The principal mirror cone constant, The secondary mirror cone constant; based on the primary mirror cone constant... and secondary mirror cone constant Samples from non-true RC family samples were screened based on hyperboloid conditions; and aspherical severity indices were calculated. ,in, This is the severity weighting coefficient for aspherical surfaces of the secondary mirror.

[0013] In step (4), Fourier spectrometer constraint screening is performed on the samples that pass the first-order analytical closure of RC, specifically as follows: Calculate half field of view and maximum optical path difference : ; ; in, For full field of view, For spectral resolution; further, the Fourier spectrometer constraint is expressed as: ,in The highest wave number; like If the value is greater than the preset threshold, the corresponding candidate parameter combination is determined to be invalid; otherwise, it is retained for subsequent 3D geometric tracing steps.

[0014] In step (5), analytical three-dimensional geometric tracing is performed on the samples that have passed the constraint screening, specifically as follows: A hexagonal annular pupil sampling point is generated within the effective aperture of the primary mirror; oblique incident rays are constructed for the on-axis field, intermediate field, and edge field; the ray parametric equation is substituted into the implicit equation of the quadratic surface corresponding to the primary or secondary mirror to form a quadratic equation and the smallest positive root is analytically obtained as the physical intersection point; the implicit equation of the quadratic surface is analytically differentiated to obtain the normal vector, and the propagation direction of the reflected light is determined based on the normal vector; the reflected light is propagated to the secondary mirror plane, the central aperture plane of the primary mirror, and the image plane to obtain the light footprint of each plane and the effective ray landing point of the image plane.

[0015] In step (5), the dimensions of the secondary mirror and the central aperture of the primary mirror are dynamically solved during the tracing process, specifically as follows: Initialize the secondary mirror radius and the primary mirror center aperture radius to empty; perform three-dimensional geometric tracing for multiple field angles, and calculate the maximum radius of the secondary mirror planar light footprint and the maximum radius of the primary mirror center aperture planar light footprint respectively; multiply the maximum radius of the secondary mirror planar light footprint by the secondary mirror shading margin coefficient to obtain a new secondary mirror radius, and multiply the maximum radius of the primary mirror center aperture planar light footprint by the primary mirror aperture margin coefficient and add the adjustment allowance to obtain a new primary mirror center aperture radius; when the difference between the new secondary mirror radius and the previous iteration's secondary mirror radius and the difference between the new primary mirror center aperture radius and the previous iteration's primary mirror center aperture radius are both less than the preset tolerance, the iteration ends; otherwise, continue updating the solution; if the obtained effective central occlusion ratio exceeds the preset range, or the primary mirror center aperture ratio is greater than the preset threshold, or the effective light ratio is lower than the preset threshold, the corresponding candidate parameter combination is determined to be invalid.

[0016] In step (6), multi-objective sorting is performed, specifically as follows: The root mean square (RMS) speckle radius and comprehensive RMS index are calculated based on the effective ray landing points on the image plane for the on-axis, intermediate, and edge fields. And calculate the effective central occlusion ratio. Inter-field uniformity proxy index Aspherical severity index characterizing the difficulty of aspherical surface processing ;by As a multi-objective evaluation vector, Pareto non-dominated sorting is performed on the effective samples, and crowding distance is calculated within the same Pareto layer for fine sorting.

[0017] Inter-field uniformity proxy index The solution is as follows: Let the root mean square radii of the speckle patterns of the on-axis field, intermediate field, and edge field be respectively... , and Its average value is The standard deviation is ; Calculate the coefficient of variation ; Calculate the range coefficient ; Calculate the center offset coefficient ; Calculate the edge decay coefficient ; in To synthesize the root mean square diffuse speckle, a weighted summation of the coefficient of variation, range coefficient, center bias coefficient, and edge decay coefficient is performed to obtain the inter-field homogeneity surrogate index. ; , , , For the corresponding weights.

[0018] Compared with the prior art, the present invention has the following beneficial effects: This invention enables batch automatic exploration within a five-dimensional system parameter space; it rapidly ensures that candidate samples belong to the true RC family through first-order analytical closure of the RC model; it accurately reflects the secondary mirror occlusion, central aperture occlusion, and effective light transmission through analytical three-dimensional geometric tracing and dynamic size solving; it achieves effective sample selection under multi-objective conditions through Pareto non-dominated sorting, crowding distance, and unified evaluation fine sorting; and it can simultaneously output effective sample sets, cleaned sample sets, balanced sample sets, and seed sample sets for optical software verification, facilitating subsequent data-driven optical design and software verification. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating the overall process of automatically generating the initial structure in an embodiment of the present invention.

[0021] Figure 2 This is a schematic diagram of the first-order analytical closure relationship of the RC telescope in an embodiment of the present invention.

[0022] Figure 3 This is a sampling diagram of the hexagonal annular pupil in an embodiment of the present invention.

[0023] Figure 4 This is a schematic diagram of three-dimensional geometric tracing in an embodiment of the present invention.

[0024] Figure 5 This is a diagram showing the footprint envelope and dynamic dimension solution in an embodiment of the present invention.

[0025] Figure 6 This is an example of the occlusion ratio – RMS distribution diagram in an embodiment of the present invention.

[0026] Figure 7 This is a uniformity-RMS distribution diagram in an embodiment of the present invention.

[0027] Figure 8 This is a Pareto color distribution map in an embodiment of the present invention.

[0028] Figure 9 This is a representative sample point diagram of the final seed sample in an embodiment of the present invention. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] It should be noted that, unless otherwise specified, the features in the following embodiments and implementation methods can be combined with each other.

[0031] like Figure 1 As shown, a method for generating the initial structure of an RC telescope for an infrared Fourier spectrometer includes the following steps: S1. Establish the parameter configuration object for the target system and determine the input range, constraints, and operational control parameters required for the subsequent automatic generation of the initial structure. This step corresponds to... Figure 1 The "Parameter Configuration and Input" module.

[0032] S11. Create a parameter configuration object.

[0033] First, establish a parameter configuration object, which should include at least the primary lens aperture range, full field of view range, system F-number range, back focal length ratio range, primary lens F-number range, maximum wavenumber, spectral resolution, and Jacquinot threshold.

[0034] Among them, the primary mirror aperture range, the full field of view range, the system F-number range, the back focal length ratio range, and the primary mirror F-number range together constitute the five-dimensional input parameter space of the candidate structure; the highest wavenumber, spectral resolution, and Jacquinot quantity threshold are used to characterize the interface constraints of the mid-infrared ground-based passive Fourier spectrometer front-end system.

[0035] S12. Set the task profile.

[0036] To adapt to different mid-infrared ground-based passive Fourier spectrometer front-end tasks, multiple profiles can be preset as parameter envelope templates. Different profiles correspond to different five-dimensional parameter boundaries and Fourier spectrometer interface constraints.

[0037] In the current implementation, a wide-range exploration profile, a medium-range exploration profile, and a narrow-range focusing profile are set, with the medium-range exploration profile being the currently enabled configuration preferred.

[0038] S13. Set the running control parameters.

[0039] Further settings are made for global operation control parameters, including: target effective sample count, number of candidate samples generated per batch, maximum original sample budget, number of pupil sampling rings, magnification constraint range, effective central occlusion ratio constraint range, upper limit of primary mirror center aperture ratio, lower limit of effective light ratio, secondary mirror shading margin coefficient, primary mirror aperture margin coefficient, assembly and adjustment margin, maximum number of iterations for dynamic dimension solving, dynamic dimension convergence tolerance, multi-objective sorting weight, sample cleaning threshold, and balanced export of target sample count.

[0040] S2. Generate candidate parameter combinations in the five-dimensional parameter space, and implement adaptive hybrid sampling based on the historical effective sample distribution when the conditions are met. This step corresponds to... Figure 1 The "Joint Sampling and Adaptive Hybrid Sampling" module in the document.

[0041] S21. Establish a five-dimensional parameter sampling space.

[0042] primary mirror aperture Full field of view System F number Back focal length ratio and the F-number of the primary mirror Five parameters are used as joint sampling dimensions, and the upper and lower limits of each parameter are determined by the parameter configuration object selected in step S1.

[0043] S22, Joint sampling by the implementing unit exceeding cubic meters.

[0044] First, sampling points are generated in a five-dimensional unit hypercube space. The maximum-minimum distance Latin hypercube sampling method is adopted. This method generates multiple sets of Latin hypercube samples, compares the minimum pairwise distances of each set of samples, and finally selects the sample set with the largest minimum pairwise distance as the sampling result of the current batch, so as to improve the uniform distribution of five-dimensional sampling points in the parameter space.

[0045] S23. Implement mapping to the real physical parameter range.

[0046] Let the sampling value of a certain dimension in the unit hypercubic meter be... The corresponding physical parameter range is Then the physical parameter values ​​after this dimension mapping are expressed as: ; S24. Perform adaptive hybrid sampling.

[0047] When the number of historical valid samples reaches the preset preheating threshold, the distribution of historical valid samples in each sampling dimension is statistically analyzed, the low quantile and high quantile values ​​of each dimension are calculated respectively, and the expansion ratio is superimposed on this basis to form the key sampling interval of each dimension.

[0048] Let the set of valid historical samples in a certain dimension be . Its lower quantile and higher quantile are denoted as , respectively. and The original global range is The expansion ratio is Then the key sampling interval of this dimension can be represented as: ; ; According to a preset mixing ratio, global sampling samples are generated within the original parameter range, and local sampling samples are generated within the dynamically shrunk key sampling interval. Finally, the two sets of samples are merged and their order is shuffled to form a new combination of candidate parameters. This method improves the efficiency of generating high-quality and effective samples while maintaining global exploration capabilities.

[0049] S3. Convert the candidate parameter combinations into first-order geometric skeleton parameters of the RC (Ritchie-Kreutz) system, and remove invalid samples that do not satisfy the RC closure relation. This step corresponds to... Figure 1 The "RC first-order analytic closure" module, and together with Figure 2 The corresponding diagram is the first-order analytic closure relation of the RC.

[0050] S31. Calculate the total focal length, back focal length, and primary lens focal length of the system.

[0051] For each combination of candidate parameters, first calculate the total focal length of the system. Back focal length and primary lens focal length : ; ; ; in, Main mirror aperture, For the system F number, Back focal length ratio, The main mirror's F-number.

[0052] S32. Calculate the distance between the primary and secondary mirrors, the distance from the secondary mirror to the focal plane, and the magnification of the secondary mirror.

[0053] Further calculate the distance between the primary and secondary mirrors Distance from secondary lens to focal plane and secondary mirror magnification : ; ; ; S33. Perform a basic closure legality determination.

[0054] If a candidate sample meets any of the following conditions, it is determined to be an invalid sample and is removed: ; ; ; ; Not within the preset magnification range; S34. Calculate the radius of curvature and cone constant.

[0055] For samples that pass the basic closure validity test, the radius of curvature of the primary mirror is further calculated. Secondary mirror curvature radius Primary mirror cone constant and secondary mirror cone constant : ; ; ; ; S35. Screen true RC family samples and calculate aspherical severity.

[0056] By determining whether the primary and secondary mirror cone constants satisfy the following hyperboloid conditions, samples that do not belong to the true RC family are eliminated.

[0057] ; ; Furthermore, for subsequent multi-objective ranking, an aspherical severity index can also be calculated. : ; in, This is the severity weighting coefficient for aspherical surfaces of the secondary mirror.

[0058] S4. Based on the front-end interface requirements of the mid-infrared ground-based passive Fourier transform spectrometer, perform field-view-resolution coupling constraint screening on samples that pass the first-order analytical closure of the RC spectrum. This step corresponds to... Figure 1 The "Fourier Spectrometer Constraint Screening" module.

[0059] S41. Calculate the half field of view.

[0060] From full field of view Calculate half field of view : ; S42. Calculate the maximum optical path difference.

[0061] According to spectral resolution Calculate the maximum optical path difference : ; S43. Calculate the Jacquinot threshold. And determine whether the threshold constraint is met.

[0062] Based on the highest wave number Maximum optical path difference and half field of view ,calculate: ; like If the candidate sample does not meet the front-end interface constraints, it is determined and removed; otherwise, it is retained and proceeds to the subsequent 3D geometric tracing steps.

[0063] S5. Perform real ray tracing on the candidate samples that have passed the above screening in three-dimensional space, and dynamically solve for the secondary mirror size and the primary mirror center aperture size during the tracing process. This step corresponds to... Figure 1 The "Analytical 3D Geometric Tracing and Dynamic Dimension Solving" module in the text, and together with Figures 3-5 The content shown corresponds to this. Among them, Figure 3 Corresponding to pupil sampling, Figure 4 Corresponding to three-dimensional geometric tracing, Figure 5 The corresponding dynamic solution for the secondary mirror size and the primary mirror center hole is based on the footprint envelope.

[0064] S51. Construct pupil sampling points within the effective aperture of the primary mirror.

[0065] A hexagonal annular pupil sampling point array is constructed within the effective aperture of the primary mirror, namely: one sampling point is set at the center, and the second sampling point is set at the center. Six sampling points are set on each sampling ring.

[0066] When the number of sampling rings is At that time, the total number of sampling points for: ; This sampling method can ensure light coverage in the central, intermediate, and edge areas of a circular primary mirror aperture.

[0067] S52. Construct multiple incident rays representing the field of view.

[0068] For the same candidate structure, construct three representative fields of view: the on-axis field, the intermediate field, and the edge field. The intermediate field is taken as 0.85 times the half field of view commonly used in engineering. Combine the pupil sampling points obtained in step S51 with the corresponding field of view directions to construct a parallel ray array with oblique incidence.

[0069] S53. Analyze the intersection points of the light rays with the primary and secondary mirrors.

[0070] Each ray can be expressed as a parametric equation: ; in, The initial point of the light ray. This is the propagation direction vector.

[0071] Substituting the parametric equations of the light rays into the implicit equations of the quadratic surfaces of the primary and secondary mirrors respectively, we obtain a quadratic equation in one variable concerning the parameter t. After solving this quadratic equation, we take the intersection point corresponding to the smallest positive root as the physical intersection point of the light ray with the corresponding optical surface.

[0072] S54. Calculate the intersection point normal vector and update the reflection direction.

[0073] After finding the intersection point of the ray with the primary or secondary mirror, the analytical gradient of the implicit equation of the corresponding quadratic surface is calculated to obtain the normal vector at the intersection point. According to the law of specular reflection, the direction of propagation after reflection is... Represented as: ; in, Let be the incident direction vector. It is the unit normal vector.

[0074] S55, propagated to the secondary mirror plane, the primary mirror center hole plane, and the image plane.

[0075] After reflection from the primary mirror and the secondary mirror, the light continues to propagate to the secondary mirror plane, the central aperture plane of the primary mirror, and the final image plane. This yields the light footprint of the secondary mirror plane, the return light footprint of the central aperture plane of the primary mirror, and the effective landing point coordinates on the image plane, respectively.

[0076] S56. Dynamically solve for the dimensions of the secondary mirror and the central hole of the primary mirror.

[0077] The radius of the secondary mirror and the radius of the central aperture of the primary mirror are not preset at the beginning, but are solved dynamically through iteration.

[0078] Specifically, steps S51 to S55 are performed for each representative field of view.

[0079] The maximum envelope radius of the secondary mirror plane optical footprint in multiple representative fields of view. The maximum envelope radius of the return optical footprint of the primary mirror center aperture plane After taking the maximum value, update the secondary mirror radius. and the radius of the central aperture of the primary mirror : ; ; in, This is the secondary mirror shading margin coefficient. The main mirror aperture allowance coefficient. This is the allowance for adjustment.

[0080] S57. Determine whether the dynamic dimension solution has converged.

[0081] After the (k+1)th update, compare the changes in the radius of the old and new secondary mirrors and the radius of the central aperture of the primary mirror: ; ; in, This is the preset convergence tolerance.

[0082] When both of the above conditions are met, the dynamic dimension solution is determined to be converged; otherwise, the updated secondary mirror radius and primary mirror center hole radius are used as new boundary conditions to continue the tracing iteration.

[0083] S58. Determine the physical validity.

[0084] After the dynamic dimensions are solved, the true diameter of the secondary mirror, the diameter of the central aperture of the primary mirror, the effective central occlusion ratio, and the effective light ratio are further calculated.

[0085] If any of the following conditions are met, the candidate sample is deemed invalid and removed: the effective central occlusion ratio is not within the preset range; the proportion of the central aperture of the primary mirror is greater than the preset threshold; or the effective light ratio is lower than the preset threshold.

[0086] This step ensures that the preserved sample has true light transmission capability and is feasible for engineering.

[0087] S6. For valid samples obtained through analytical 3D geometric tracing and dynamic dimensional solving, perform image quality evaluation, light transmission evaluation, and multi-object ranking. This step corresponds to... Figure 1 The "Performance Evaluation and Multi-Objective Ranking" module in the document.

[0088] S61. Calculate the root mean square radius of the speckle and the overall root mean square index for each representative field of view. For each representative field of view, extract the coordinates of the effective ray points on the image plane, calculate the position of its centroid, and further calculate the root mean square radius of the speckle relative to the centroid.

[0089] The comprehensive root mean square index is calculated based on the root mean square radius (RMS) values ​​of multiple representative fields of view. Used to characterize the overall tracking image quality level of the system: ; S62. Construct a proxy index for inter-field uniformity.

[0090] Let the root mean square radii of the three representative fields of view be respectively , , Its average value is The standard deviation is .

[0091] First, calculate the coefficient of variation: ; Next, calculate the range coefficient: ; Then calculate the center offset coefficient: ; Further calculate the edge decay coefficient: ; Finally, by summing the weighted values ​​of the above terms, we obtain the inter-field uniformity proxy index. It is used to evaluate whether the image quality distribution is uniform across the entire field of view. ; S63. Solve for the effective central occlusion ratio. ; in, Main mirror aperture, This is the true diameter of the secondary mirror. Main mirror center hole diameter S64. Constructing a multi-objective evaluation vector For each valid sample, construct a multi-objective evaluation vector: ; in, To integrate the root mean square index, For an effective central occlusion ratio, As a proxy index for inter-field uniformity, This is an aspherical severity index.

[0092] S65. Perform Pareto non-dominated sorting.

[0093] Using the multi-objective evaluation vector from step S63 as the optimization objective, Pareto non-dominated ranking is performed on all valid samples. Through Pareto non-dominated ranking, all samples can be divided into multiple Pareto layers, and the Pareto level of each sample can be obtained.

[0094] S66. Perform fine-grained ranking and unified evaluation at the same level.

[0095] For samples located within the same Pareto layer, crowding distance is further calculated to maintain the discreteness of sample distribution in the target space. In the unified evaluation fine-ranking stage, a light-permeability penalty term constructed from the occlusion ratio is used instead of the occlusion ratio: ; The light transmission penalty term, the comprehensive root mean square index, the inter-field uniformity proxy index, and the aspherical severity index are normalized and then weighted and summed according to preset weights to obtain a unified evaluation index. To achieve fine sorting of samples within the same Pareto layer: ; in, , , , The normalized evaluation terms , , , For the corresponding weights.

[0096] S67. Generate sample level identifiers.

[0097] Based on the Pareto ranking results, crowding distance, and unified evaluation metrics, valid samples are further assigned global rankings and grade labels. These are then used for subsequent training set construction, balanced set extraction, and seed sample selection for optical software validation.

[0098] S7. After completing the multi-objective sorting, the sample results are hierarchically organized, structured, and visualized. This step corresponds to... Figure 1 The "Data Layered Export" module in the middle, and with Figures 6 to 9 The results are shown in the corresponding visualization figures.

[0099] S71. Generate a complete set of valid samples.

[0100] All valid samples obtained through steps S1 to S6 are summarized to form a complete set of valid samples. The complete set of valid samples includes at least the input parameters, RC first-order geometric skeleton parameters, dynamic size solution results, effective central occlusion ratio, RMS of each representative field of view, comprehensive root mean square speckle index, inter-field uniformity surrogate index, aspherical severity index, Pareto level, crowding distance, and unified evaluation index for each valid sample.

[0101] S72. Generate a performance evaluation table and a performance distribution chart of all valid samples.

[0102] Based on the performance evaluation data of the entire valid sample set, a performance evaluation table is generated, and a visualization chart representing the performance distribution of the entire valid sample set is generated simultaneously.

[0103] Among them, the effective central occlusion of all valid samples Compared with the comprehensive root mean square diffusion index Perform the corresponding drawing to obtain, as follows Figure 6 The effective central occlusion ratio – tracking RMS distribution is shown in the figure. Figure 6 This is used to demonstrate the distribution relationship between light occlusion and image quality diffusion among all valid samples, thereby assisting in the identification of candidate structural regions with low occlusion and low diffusion spots.

[0104] Furthermore, the inter-field uniformity index of all valid samples is used as a proxy. Compared with the comprehensive root mean square diffusion index Perform the corresponding drawing to obtain, as follows Figure 7 The diagram shows the inter-field homogeneity index – the tracking RMS distribution. Figure 7 This is used to demonstrate the distribution relationship between the image quality consistency and overall diffusion level of all valid samples across multiple fields of view, thereby avoiding the selection of the initial structure based solely on a single field of view or a single RMS index.

[0105] S73. Generate Pareto hierarchical distribution map.

[0106] Based on the Pareto non-dominated ranking results obtained in step S6, all valid samples are labeled according to the Pareto hierarchy, and a ranking is generated as follows: Figure 8 The Pareto hierarchy diagram is shown. Figure 8 This is used to display the stratified results of all valid samples in the multi-objective evaluation space, where samples at lower Pareto levels have better overall trade-off performance in terms of comprehensive root mean square speckle, effective central occlusion ratio, inter-field uniformity, and aspherical severity.

[0107] S74. Generate a clean sample set, a balanced sample set, and a seed sample set.

[0108] Based on the Pareto hierarchy, crowding distance, and unified evaluation index obtained in step S6, the entire valid sample set is further screened to generate a clean sample set, a balanced sample set, and a seed sample set for subsequent optical software verification. The clean sample set is used to remove samples with abnormal performance or evaluation indices that do not meet preset thresholds; the balanced sample set is used to maintain a balanced distribution of samples across different parameter and performance ranges; and the seed sample set is used to provide an initial structure for further optimization of the subsequent optical design software.

[0109] S75. Generate the final seed sample point map.

[0110] While generating the seed sample set, the final seed sample point map is generated based on the sample number, Pareto level, and unified evaluation index of the seed samples. Figure 9 This diagram illustrates the final distribution of the exported seed samples, facilitating designers in selecting initial values ​​for subsequent optical optimization from the seed sample set. Due to the large number of seed samples, a representative seed sample point array is selected for display to ensure clarity.

[0111] The embodiments described above provide a detailed explanation of the technical solutions and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, additions, and equivalent substitutions made within the scope of the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for generating the initial structure of an RC telescope for an infrared Fourier spectrometer, characterized in that, Includes the following steps: (1) Establish a parameter configuration object and determine the input range, front-end interface constraints and operation control parameters required for the subsequent automatic generation of the initial structure; (2) Based on the input range, candidate parameter combinations are generated in the five-dimensional input parameter space; and adaptive hybrid sampling is performed by combining the features of historical effective samples; (3) For each set of candidate parameter combinations, perform RC first-order analytical closure solution and remove invalid samples that do not satisfy the RC closure relationship; (4) Based on the front-end interface constraints, Fourier spectrometer constraint screening is performed on the samples that pass the first-order analytical closure of RC; (5) Perform analytical three-dimensional geometric tracing on the samples that have passed the constraint screening, and dynamically solve the secondary mirror size and the primary mirror center hole size during the tracing process; (6) Perform multi-objective sorting on the effective samples solved by analytical three-dimensional geometric tracing and dynamic size calculation; (7) After completing the multi-objective sorting, the sample results are hierarchically organized and structured for export.

2. The method for generating the initial structure of the RC telescope for an infrared Fourier spectrometer according to claim 1, characterized in that, In step (1), the input range includes: primary mirror aperture range, full field of view range, system F-number range, back focal length ratio range, and primary mirror F-number range; the front-end interface constraints include: maximum wavenumber, spectral resolution, and Jacquinot quantity threshold.

3. The method for generating the initial structure of the RC telescope for an infrared Fourier spectrometer according to claim 1, characterized in that, In step (2), the candidate parameter combination is generated by mapping the five-dimensional unit hypercube joint sampling to the range of real physical parameters, and the maximum and minimum distance Latin hypercube sampling is used. When the number of historical valid samples reaches the preset preheating threshold, the low quantile and high quantile values ​​of each sampling dimension are counted, and an expansion ratio is superimposed on this to form a key sampling interval. Global sampling is performed within the original parameter range and local sampling is performed within the key sampling interval according to the preset mixing ratio. Then, the two parts of samples are merged and their order is shuffled to form a new candidate parameter combination.

4. The method for generating the initial structure of the RC telescope for an infrared Fourier spectrometer according to claim 1, characterized in that, In step (3), the first-order analytic closure of RC includes the following relationship: ； ； ； ； ； ； in, The total focal length of the system. Back focal length, The focal length of the primary mirror. Main mirror aperture, For the system F number, Back focal length ratio, The primary mirror's F-number, The distance between primary and secondary mirrors, The distance from the secondary mirror to the focal plane. The magnification of the secondary lens; if the candidate sample meets any of the following conditions: , , , , If the sample is not within the preset magnification range, it is considered an invalid sample and is removed.

5. The method for generating the initial structure of the RC telescope for an infrared Fourier spectrometer according to claim 4, characterized in that, The first-order analytic closure of the RC framework also includes the following relations: ； ； ； ； in, The radius of curvature of the primary mirror. Let be the radius of curvature of the secondary mirror. The principal mirror cone constant, The secondary mirror cone constant; based on the primary mirror cone constant... and secondary mirror cone constant Samples from non-true RC family samples were screened based on hyperboloid conditions; and aspherical severity indices were calculated. ,in, This is the severity weighting coefficient for aspherical surfaces of the secondary mirror.

6. The method for generating the initial structure of the RC telescope for an infrared Fourier spectrometer according to claim 1, characterized in that, In step (4), Fourier spectrometer constraint screening is performed on the samples that pass the first-order analytical closure of RC, specifically as follows: Calculate half field of view and maximum optical path difference : ； ； in, For full field of view, For spectral resolution; further, the Fourier spectrometer constraint is expressed as: ,in The highest wave number; like If the value is greater than the preset threshold, the corresponding candidate parameter combination is determined to be invalid; otherwise, it is retained for subsequent 3D geometric tracing steps.

7. The method for generating the initial structure of the RC telescope for an infrared Fourier spectrometer according to claim 1, characterized in that, In step (5), analytical three-dimensional geometric tracing is performed on the samples that have passed the constraint screening, specifically as follows: Generate a hexagonal annular pupil sampling point within the effective aperture of the primary mirror; construct oblique incident rays for the on-axis field, intermediate field, and edge field; substitute the ray parameter equation into the implicit equation of the quadratic surface corresponding to the primary mirror or secondary mirror to form a quadratic equation in one variable and analytically obtain the smallest positive root as the physical intersection point. The implicit equation of the quadric surface is analytically differentiated to obtain the normal vector, and the propagation direction of the reflected light is determined based on the normal vector. The reflected light is propagated to the secondary mirror plane, the central aperture plane of the primary mirror, and the image plane to obtain the light footprint of each plane and the effective light landing point of the image plane.

8. The method for generating the initial structure of the RC telescope for an infrared Fourier spectrometer according to claim 1, characterized in that, In step (5), the dimensions of the secondary mirror and the central aperture of the primary mirror are dynamically solved during the tracing process, specifically as follows: Initialize the secondary mirror radius and the primary mirror center aperture radius to empty; perform three-dimensional geometric tracing for multiple field angles, and calculate the maximum radius of the secondary mirror planar light footprint and the maximum radius of the primary mirror center aperture planar light footprint respectively; multiply the maximum radius of the secondary mirror planar light footprint by the secondary mirror shading margin coefficient to obtain the new secondary mirror radius, and multiply the maximum radius of the primary mirror center aperture planar light footprint by the primary mirror aperture margin coefficient and add the adjustment allowance to obtain the new primary mirror center aperture radius; when the difference between the new secondary mirror radius and the previous iteration's secondary mirror radius and the difference between the new primary mirror center aperture radius and the previous iteration's primary mirror center aperture radius are both less than the preset tolerance, the iteration ends; otherwise, continue to update the solution; if the obtained effective central occlusion ratio exceeds the preset range, or the primary mirror center aperture ratio is greater than the preset threshold, or the effective light ratio is lower than the preset threshold, the corresponding candidate parameter combination is determined to be invalid.

9. The method for generating the initial structure of the RC telescope for an infrared Fourier spectrometer according to claim 1, characterized in that, In step (6), multi-objective sorting is performed, specifically as follows: The root mean square (RMS) speckle radius and comprehensive RMS index are calculated based on the effective ray landing points on the image plane for the on-axis, intermediate, and edge fields. And calculate the effective central occlusion ratio. Inter-field uniformity proxy index Aspherical severity index characterizing the difficulty of aspherical surface processing ;by As a multi-objective evaluation vector, Pareto non-dominated sorting is performed on the effective samples, and crowding distance is calculated within the same Pareto layer for fine sorting.

10. The method for generating the initial structure of the RC telescope for an infrared Fourier spectrometer according to claim 9, characterized in that, Inter-field uniformity proxy index The solution is as follows: Let the root mean square radii of the speckle patterns of the on-axis field, intermediate field, and edge field be respectively... , and Its average value is The standard deviation is ; Calculate the coefficient of variation ; Calculate the range coefficient ; Calculate the center offset coefficient ; Calculate the edge decay coefficient ; in To synthesize the root mean square diffuse speckle, a weighted summation of the coefficient of variation, range coefficient, center bias coefficient, and edge decay coefficient is performed to obtain the inter-field homogeneity surrogate index. ; , , , For the corresponding weights.

Images