Automatic optimization design method for free-form surface imaging system
The automatic optimization design method for free-form surface imaging systems addresses inefficiencies by using machine learning to dynamically select optimization parameters, resulting in compact and high-performance optical systems.
Patent Information
- Application Number
- CN202411211974.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-30
- Publication Date
- 2024-12-06
AI Technical Summary
The design of free-form surface imaging systems faces challenges due to the complexity of initial structure determination and optimization, increased optimization variables, slow ray tracing, and the need for different imaging performances, often resulting in inefficient and non-optimal designs that do not meet actual production and processing requirements.
An automatic optimization design method combining machine learning with traditional optimization techniques to select sampling fields, characteristic lights, and volume constraints dynamically, using gradient descent to balance performance and volume constraints, thereby optimizing free-form surface imaging systems.
This method enhances optimization efficiency, reduces computational redundancy, and achieves compact optical systems with higher performance and larger fields of view by automatically adjusting optimization strategies based on design requirements.
Smart Images
Figure CN119087666A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to an automatic optimization design method for a free-form surface imaging system, and belongs to the technical field of optical system design. Background Art
[0002] Free-form surfaces are widely used in microscopic imaging, remote sensing satellites, biomedical detection and other fields due to their excellent performance. Free-form surfaces break the rotational symmetry of traditional surface shapes, making surface design more flexible and providing designers with more design freedom. These features can not only greatly improve system performance, but also provide a feasible solution for miniaturization, lightweighting and integration of optical systems. Due to the limitations of processing and detection capabilities, free-form surfaces were initially used in the design of lighting systems. In recent years, with the continuous advancement of precision processing and detection technology, free-form surfaces have gradually been used in high-precision imaging equipment.
[0003] The design method of the imaging system is generally divided into two steps, namely the initial structure construction and the optimization design. The traditional design method can well realize the design of spherical and aspherical optical systems, but it is easy to encounter difficulties when designing free-form surface optical systems. The main reasons are as follows: First, it is difficult to solve the initial structure. Free-form surface optical systems are mostly non-rotationally symmetric optical systems. The classical aberration theory is not applicable to directly solving the initial structure of the free-form surface. It is difficult for designers to find an initial system close to the design requirements. Second, the optimization design of the optical system is difficult. The number of variables provided by the free-form surface during the optimization process is far greater than that of the traditional surface shape, resulting in a complex system optimization model. At the same time, in order to ensure the imaging quality of the entire field of view during the optimization process, the optimization process requires more sampling fields of view and characteristic rays. The ray tracing speed is slow, resulting in a significant increase in optimization time. Therefore, although the use of free-form surfaces in the design of imaging systems can greatly enhance the potential of system design, it also brings many difficulties to the design of imaging systems.
[0004] Researchers have proposed many solutions to the above-mentioned difficulty in optimizing free-form surface imaging systems. The more typical ones are: ① Fuerschbach et al. published an article titled “A new family of optical systems employing -polynomial surfaces", the article mentioned that according to the corresponding correction relationship between Zernike free-form surfaces and Zernike aberrations, the Zernike free-form surface parameters to be used are selected in the optimization process, and the imaging system design that meets the requirements is achieved through progressive optimization; ②Cheng et al. published an article "Design of an optical see-through head-mounted display with a low f-number and large field of viewusing a freeform prism", in which a free-form surface imaging system progressive approximation optimization strategy is proposed. This strategy gradually upgrades the optical surface from the traditional surface shape to a more flexible free-form surface, and in the process gradually adds more stringent aberration constraints and structural control conditions, and gradually expands the field of view and aperture size of the imaging system, so that the system performance approaches the final design requirements; ③Yang et al. published an article "Aberration analysis for freeform surfacesterms overlay on general decentered and tilted optical In the paper, it is mentioned that the aberration theory mainly produces aberrations of no more than m order (possibly including some higher order wave aberrations) based on the m-order polynomial free-form surface terms. In the design process, the variables describing the surface parameters are gradually added to the optimization process from low order to high order, so that the system aberrations are gradually corrected from low order to high order.
[0005] The above is a typical optimization strategy for free-form surface imaging systems. Although good optimization effects can be achieved for specific imaging systems, the following shortcomings still exist: ① Different imaging systems often require different imaging performances. Designers need to spend a lot of time and energy on the selection of optimization strategies. If the optimization design fails, it is necessary to re-select the optimization from the initial structure. ② The settings of the sampling field of view, characteristic light and optimization constraints in the above optimization strategy are still given by the optical designer based on design experience. Compared with the traditional rotationally symmetric structure, the optimization variables of the free-form surface imaging system have increased significantly, and redundant optimization constraints often cause a rapid decrease in the optimization rate. ③ The above optimization strategy only considers the influence of aberrations, and does not consider the influence of factors such as volume and tolerance of the imaging system on the imaging performance during actual use. The optimization design completed by the above strategy may not meet the needs of actual production and processing. Summary of the invention
[0006] The technical problem solved by the present invention is: to overcome the shortcomings of the prior art, propose an automatic optimization design method for a free-form surface imaging system, combine machine learning technology with traditional optimization design, and realize the automatic optimization design of a compact free-form surface optical system. The method has broad application prospects in the automatic optimization design evaluation of refractive optical systems, reflective optical systems and catadioptric optical systems.
[0007] The technical solution of the present invention is:
[0008] An automatic optimization design method for a free-form surface imaging system, comprising:
[0009] S1: Obtain the initial structure of the optical system, determine multiple imaging performance design targets including focal length EFL, F number, imaging MTF, etc. in the optical system according to design requirements, and set the number of cycles;
[0010] S2: According to the focal length EFL design target and F number design target, set fixed optimization constraint parameters and optimize the initial structure of the optical system;
[0011] S3: Determine whether the number of optimizations reaches the set number of cycles. If not, execute step S4. If reached, it indicates that the current system may not have the potential to achieve optimization of the existing imaging performance.
[0012] S4: Obtain a comprehensive evaluation value of the current imaging performance of the system through an imaging performance comprehensive evaluation method and perform an evaluation. If the comprehensive evaluation value does not meet the design performance index, execute step S5. If it does, determine the current imaging system as the final optimized system, and output the optical system structure and volume.
[0013] S5: Calculate the current volume of the system and the feature field number, feature ray number and optical surface number used for volume constraints in the next round of optimization by using a volume calculation method;
[0014] S6: According to the comprehensive evaluation value, the sampling field of view points and their weights used in the next round of optimization are calculated by the automatic sampling field of view selection method and the automatic sampling field of view weight calculation method;
[0015] S7: Calculate the optimization targets of volume and MTF through the automatic optimization constraint setting method, and optimize the current optical system structure based on the obtained sampling field of view and the set fixed optimization constraint parameters. After the optimization is completed, return to step S3.
[0016] Furthermore, the automatic volume calculation method is: in the imaging system, the spatial coordinates of the intersection of the edge field of view and the central field of view characteristic light rays with each optical element are traced by real light rays, and the characteristic field of view number, characteristic light ray number and optical surface number required for calculating the length, width and height of the system are recorded.
[0017] Furthermore, the automatic sampling field selection method is: according to the comprehensive evaluation value, the imaging performance of each field of view of the current system is comprehensively calculated, and the field of view point with the worst imaging performance is calculated, and the field of view point is added to the current sampling field of view.
[0018] Furthermore, the automatic calculation method of the sampling field weight is: according to the imaging performance of each field of view of the current system, an adaptive algorithm is used to calculate the weight for each current sampling field of view, and then the sampling fields whose weight values are less than a threshold and are not boundaries are deleted.
[0019] Furthermore, the principle that the adaptive method should follow is that the optimization weight of the sampling field of view with good imaging performance is reduced in the next round, and the optimization weight of the sampling field of view with poor imaging performance is increased in the next round.
[0020] Furthermore, the comprehensive evaluation method of imaging performance is as follows: according to the design requirements of the optical system, the actual design values of various imaging performances, focal length and F number in the current imaging system are calculated in turn, and the normalized weighted superposition is performed as the comprehensive evaluation value MF of the current performance of the imaging system.
[0021] Furthermore, the difference weights of different design requirements determine the optimization strategy in the optimization process. The larger the difference weight, the farther away from the design target, and in the next optimization process, the system performance needs to meet the design index first.
[0022] Furthermore, the comprehensive evaluation value MF of the current imaging performance of the system is evaluated to see whether |MF-3| is less than ε MF , where ε MF It is the allowable error in system design and its value is less than 0.03.
[0023] Furthermore, through the gradient descent machine learning method, the optimization targets of volume and MTF corresponding to the next round of optimization process are automatically calculated according to the current imaging performance.
[0024] Furthermore, the initial structure of the optical system is determined by imaging theory.
[0025] The advantages of the present invention compared with the prior art are:
[0026] (1) The present invention automatically adopts different optimization constraint strategies by adjusting the weight parameters of different imaging performances in the imaging performance evaluation method to meet the design requirements of various imaging systems.
[0027] (2) The present invention can automatically select and calculate the sampling market, characteristic light, and optimization constraints through machine learning calculation methods such as gradient descent, thereby reducing the calculation redundancy in traditional optimization methods and effectively improving the optimization efficiency.
[0028] (3) The present invention adds constraints related to processing and manufacturing, such as system volume, during the optimization process. During the optimization process, the mutual constraints between optical volume and performance can be effectively balanced, so that the optical system can fully approach its design performance limit, which is conducive to fully tapping the design potential of the optical system and obtaining a compact optical system with higher performance and larger field of view. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:
[0030] Figure 1 This is an optimization flow chart of an embodiment of the present invention using volume and system MTF as evaluation functions;
[0031] Figure 2 The initial structure of the optical imaging system according to the embodiment of the present invention with volume and system MTF as evaluation functions;
[0032] Figure 3 An optimized optical imaging system using volume and system MTF as evaluation functions according to an embodiment of the present invention;
[0033] Figure 4 A diagram showing the volume change of the optical system during the optimization process of an embodiment of the present invention. DETAILED DESCRIPTION
[0034] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.
[0035] The present invention proposes an automatic optimization design method for a free-form surface imaging system, including an automatic sampling field selection method, an automatic volume constraint setting method, and an imaging performance comprehensive evaluation method. For imaging systems with different design requirements, the present invention can adopt different optimization strategies, and use machine learning methods such as gradient descent during the optimization process to automatically select the sampling field of view, characteristic light, and reasonable volume constraints in each round of optimization, so as to better balance the constraint relationship between volume and performance.
[0036] This method is Figure 1 As shown, specifically including:
[0037] S1: The initial structure of the optical system is obtained through imaging theory calculation;
[0038] S2: Input design requirements: Design target EFL of system focal length EFL tag , Imaging MTF Design Target MTF tag , the design target of F number FNO tag , the number of cycles is M; the design constraints of the system in the optimization process are divided into two types: fixed constraints and variable constraints. The fixed optimization constraints in this case are the system focal length and F number, and the variable constraints are the system MTF constraints and volume constraints;
[0039] S3: According to the imaging MTF design goal, set fixed optimization constraints and perform preliminary system optimization to reduce the difference between the optimized system imaging performance and the system design requirements, so that the imaging properties such as system focal length (EFL) and F number are close to the design requirements;
[0040] S4: Determine whether the number of optimization times reaches the set upper limit M. If not, execute step S5. If reached, it means that the current system may not have the optimization potential to achieve the existing imaging performance, and it is necessary to re-evaluate the design conditions or re-select the initial structure.
[0041] S5: Through the comprehensive evaluation method of imaging performance, according to the design requirements of the optical system, the actual design values of MTF, focal length and F number in the current imaging system are calculated in turn, and the normalized weighted superposition is used as the comprehensive evaluation value MF of the current performance of the imaging system:
[0042]
[0043] Evaluate whether the imaging performance of the current system meets the design goal, that is, evaluate whether |MF-3| is less than ε MF , ε MF is the allowable error of system design, which is usually less than 0.03. In this case, it is 0.01. If it is not satisfied, step S6 is executed. If it is satisfied, the current imaging system is output as the final optimized system. The difference weights of different design requirements (|(EFL / EFL tag )-1|、|(FNO / FNO tag )-1| and (MTF / MTF tag )-1|) will determine the optimization strategy in the optimization process. The larger the difference weight is, the farther away from the design goal is. In the next optimization process, the system performance needs to meet the design index first.
[0044] S6: Through the volume calculation method, in the imaging system, the spatial coordinates of the intersection points of the edge field of view and the central field of view characteristic rays with each optical element are traced by real rays, where the spatial coordinates of the intersection point of the jth characteristic ray emitted from the field of view F and the reflector S are (X FjS ,Y FjS ,ZFjS ). Then the system length is the maximum distance ΔZ between the coordinates of different intersection points in the Z direction max , width is the maximum distance ΔX between different intersection coordinates in the X direction max , height is the maximum distance ΔY between different intersection coordinates in the Y direction max . And record the characteristic field number, characteristic light number and optical surface number required for the calculation of the length, width and height of the system respectively. Since the characteristic light, field of view and position of the reflector used in the system volume calculation will change with the presence and position of the intermediate image, the automatic volume calculation method needs to be used before each round of optimization to calculate the characteristic field number, characteristic light number and optical surface number used by the volume constraint in the next round of optimization.
[0045] S7: According to the comprehensive evaluation value of the current imaging system performance calculated in step S5, the automatic sampling field selection method is used to comprehensively evaluate the imaging performance of each field of view of the current system, and the field of view point with the worst imaging performance is calculated based on this, and it is added to the current sampling field of view. The weight is automatically calculated for each current sampling field of view through the automatic sampling field of view weight calculation method, and then the weight values of which are less than the threshold ε are deleted. FOV And the sampling field of view is not bounded. In this case, ε FOV The value is 0.05.
[0046] Automatic calculation method of sampling field weight: Volume adjustment will change the imaging performance of different fields of view, so it is necessary to automatically modify the weight of the sampling field of view in each round of optimization to balance the imaging performance of the entire field of view. The optimization weight calculation of each field of view is based on the current imaging performance of the system and is calculated through an adaptive algorithm. The principle that the adaptive method should follow is "the sampling field of view with good imaging performance has a lower optimization weight in the next round, and the sampling field of view with poor imaging performance has an increased optimization weight in the next round". As long as the adaptive calculation method meets the above principles, it can be used.
[0047] S8: Use the automatic setting method of optimization constraints to update the optimization constraints. In the optimization design process, constraints are divided into two types: fixed and variable. Among them, focal length and F number are fixed constraint parameters, which are given unchanged during the optimization process. Optical performance constraints such as MTF and volume are variable constraints. In each round of optimization, machine learning methods such as gradient descent are used to automatically calculate the optimization target of the next round of optimization process based on the current imaging performance, including volume and MTF, and perform optimization so that the performance of the imaging system gradually approaches the design requirements. After optimization, return to step S5.
[0048] The calculation method used in this case is the gradient descent method. The specific content of this method is as follows: The length, width and height optimization constraints of the volume in the i-th optimization process are length i 、width i、high i , in the optimization process, the system length (ΔZ max ), width(ΔX max ), high (ΔY max ) are less than this value. The calculation formula is as follows:
[0049]
[0050] step l 、step w and step h is the step length, width and height in the gradient descent method. In this case, the values are all 1. The gradient k i is the calculation gradient in the gradient descent method. Since in the imaging system, system performance and volume are mutually constrained, in this case, the gradient k i The calculation formula is k i =(MTF i -MTF tag ) / MTF tag .
[0051] In summary, this method automatically sets different optimization strategies for imaging systems with different design requirements based on the principle of progressive optimization to meet the design requirements of various imaging systems. In the optimization process, machine learning methods such as gradient descent are used to dynamically select parameters such as the sampling field of view and characteristic light used in each round of optimization to reduce the computational redundancy in existing selection methods. At the same time, machine learning methods such as gradient descent adaptively calculate the reasonable volume constraints for each round of optimization, effectively balance the mutual constraints between optical volume and performance, and make the optical system fully approach its design performance limit. This is conducive to obtaining a compact imaging system with higher performance and wider width.
[0052] The embodiment is: using the system volume and the optical system MTF value as the evaluation criteria for performance evaluation, using the free-form surface imaging system automatic optimization design method disclosed in the present invention, Figure 2 The initial structure shown is optimized. The optimization design flow chart is as follows Figure 1 As shown, the optimized imaging optical system structure is as follows Figure 3 The optimization process took 8 minutes and 28 seconds. Compared with the initial optical system optimization sampling, the final round of optimization added 3 new sampling field points and reduced 6 redundant sampling field points. The total number of sampling field points used for optimization changed from 10 to 7, which reduced the field sampling redundancy during the optimization process, reduced the optimization calculation amount, and improved the optimization efficiency. The volume change of the optical system during the optimization process is shown in Figure 2. Figure 4 As shown in the figure, the volume of the optical system after optimization by this program is reduced by 37%, the structure of the optimized optical system is more compact, and the design of the light and small optical system is better completed. The design potential of the imaging system is greatly tapped.
[0053] The above-described embodiments are only preferred specific implementations of the present invention. Common changes and substitutions made by those skilled in the art within the scope of the technical solution of the present invention should be included in the protection scope of the present invention.
Claims
1. A method for automatic optimization design of a free-form surface imaging system, characterized in that: include: S1: Obtain the initial structure of the optical system, determine multiple imaging performance design targets including focal length EFL, F number, imaging MTF, etc. in the optical system according to design requirements, and set the number of cycles; S2: According to the focal length EFL design target and the F-number design target, set the fixed optimization constraint parameters and optimize the initial structure of the optical system; S3: Determine whether the number of optimizations reaches the set number of cycles. If not, execute step S4. If reached, it indicates that the current system may not have the potential to achieve optimization of the existing imaging performance. S4: Obtain a comprehensive evaluation value of the current imaging performance of the system through an imaging performance comprehensive evaluation method and perform an evaluation. If the comprehensive evaluation value does not meet the design performance index, execute step S5. If it does, determine the current imaging system as the final optimized system, and output the optical system structure and volume. S5: Calculate the current volume of the system and the feature field number, feature ray number and optical surface number used for volume constraints in the next round of optimization by using a volume calculation method; S6: According to the comprehensive evaluation value, the sampling field of view points and their weights used in the next round of optimization are calculated by the automatic sampling field of view selection method and the automatic sampling field of view weight calculation method; S7: Calculate the optimization targets of volume and MTF through the automatic optimization constraint setting method, and optimize the current optical system structure based on the obtained sampling field of view and the set fixed optimization constraint parameters. After the optimization is completed, return to step S3.
2. The method for automatic optimization design of a free-form surface imaging system according to claim 1, characterized in that: The volume automatic calculation method is as follows: In the imaging system, the spatial coordinates of the intersections of the edge field of view and the central field of view characteristic rays with each optical element are traced by real rays, where the spatial coordinates of the intersection of the jth characteristic ray emitted from the field of view F and the reflector S are (X FjS ,Y FjS ,Z FjS ), then the system length is the maximum distance ΔZ between the coordinates of different intersection points in the Z direction max , width is the maximum distance ΔX between different intersection coordinates in the X direction max , height is the maximum distance ΔY between different intersection coordinates in the Y direction max , record the characteristic field of view number, characteristic light number and optical surface number required for calculating the length, width and height of the system.
3. The method for automatic optimization design of a free-form surface imaging system according to claim 1, characterized in that: The automatic sampling field selection method is: according to the comprehensive evaluation value, the imaging performance of each field of view of the current system is comprehensively calculated, and the field of view point with the worst imaging performance is calculated, and the field of view point is added to the current sampling field of view.
4. The method for automatic optimization design of a free-form surface imaging system according to claim 1, characterized in that: The automatic calculation method of the sampling field weight is: according to the imaging performance of each field of view of the current system, an adaptive algorithm is used to calculate the weight for each current sampling field of view, and then the sampling fields whose weight values are less than the threshold and are not borders are deleted.
5. The method for automatic optimization design of a free-form surface imaging system according to claim 4, characterized in that: The principle that the adaptive method should follow is: the sampling field of view with good imaging performance has its optimization weight reduced in the next round, and the sampling field of view with poor imaging performance has its optimization weight increased in the next round.
6. The method for automatic optimization design of a free-form surface imaging system according to claim 1, characterized in that: The comprehensive evaluation method for imaging performance is as follows: according to the design requirements of the optical system, the actual design values of various imaging performances, focal length and F number in the current imaging system are calculated in turn, and their normalized weighted superposition is used as the comprehensive evaluation value MF of the current performance of the imaging system.
7. The method for automatic optimization design of a free-form surface imaging system according to claim 6, characterized in that: The difference weights of different design requirements determine the optimization strategy during the optimization process. The larger the difference weight, the farther away from the design goal. In the next optimization, the system performance needs to meet the design indicator first.
8. The method for automatic optimization design of a free-form surface imaging system according to claim 1, characterized in that: Evaluate the comprehensive evaluation value MF of the current imaging performance of the system to see whether |MF-3| is less than ε MF , where ε MF It is the allowable error in system design and its value is less than 0.
03.
9. The method for automatic optimization design of a free-form surface imaging system according to claim 1, characterized in that: Through the gradient descent machine learning method, the optimization targets of volume and MTF corresponding to the next round of optimization process are automatically calculated according to the current imaging performance.
10. The method for automatic optimization design of a free-form surface imaging system according to claim 1, characterized in that: The initial structure of the optical system is determined through imaging theory.