Method and system for adjusting optical lens

Abstract

The present invention relates to the field of optical lens technology, and specifically to an optical lens calibration method and system. First, an original optical lens parameter vector is obtained, and an image quality evaluation value of the original parameter vector is calculated by a ray tracing method, so as to effectively distinguish the optical lens that needs to be calibrated and determine the first optical lens parameter vector; then, the first optical lens parameter vector is input into a conditional generative adversarial network to generate a second optical lens parameter vector, and the conditional generative adversarial network introduces image quality evaluation loss and manufacturing constraint loss, which can guide the generator to generate an optical lens parameter vector that meets imaging requirements and complies with physical manufacturing limitations; then, the image quality evaluation value of the second optical lens parameter vector is calculated by a ray tracing method to ensure that the second optical lens parameter vector can meet the imaging requirements, and then the optical lens is calibrated according to the second optical lens parameter vector, thereby improving the accuracy of optical lens calibration and reducing the calibration cost.

Description

Technical Field

[0001] The present invention relates to the technical field of optical lenses, and in particular to an adjustment method and system for an optical lens. Background Art

[0002] Optical lenses are widely used in various fields, including photography, astronomical observation, medical imaging and optical sensing. In these applications, the adjustment of optical lenses is a key step to ensure that the performance of the optical system meets the design requirements. Imprecise adjustment may cause blurred or distorted images or uneven distribution of light energy, thus affecting the performance of the entire system. Therefore, precise adjustment of optical lenses is an indispensable part of achieving high-quality imaging and meeting the needs of professional applications.

[0003] There are some common problems with traditional optical lens calibration technology. For example, the calibration process requires complex mechanical equipment, takes a long time to calibrate, and requires high skills of operators. In addition, traditional calibration methods often cannot meet the requirements of high-end applications in terms of accuracy and repeatability. These limitations not only increase costs, but also affect the performance and reliability of the final product.

[0004] The prior art has attempted to improve the calibration process through intelligent methods. For example, a Chinese patent with publication number CN117850032A proposes a calibration method and system for an optical lens. First, a target image is collected based on the target object, and several target areas are determined according to the target image; then, an optical lens calibration simulation is performed to obtain an actual imaging image and an ideal imaging image; an actual matrix and an ideal matrix are determined according to each target area, a difference matrix is ​​calculated according to the actual matrix and the ideal matrix, and the optical lens is initially adjusted according to each difference matrix; then, the actual imaging image, the ideal imaging image, the actual matrix and the ideal matrix after the initial adjustment are determined; the corresponding actual resolution, ideal resolution, actual contrast and ideal contrast are determined according to the actual imaging image and the ideal imaging image; then, the difference matrix is ​​determined, and the difference matrix, the actual resolution, the ideal resolution, the actual contrast and the ideal contrast are integrated into calibration reference data; finally, the optical lens is calibrated according to the calibration reference data.

[0005] This method solves the problems existing in traditional adjustment methods. However, most of the existing technologies adjust the installation position of the optical lens in the optical system, or directly correct the image, but do not take into account that when the manufacturing parameters of the optical lens itself need to be adjusted, inaccurate adjustment will lead to serious cost increases.

[0006] To this end, a method and system for adjusting an optical lens are proposed. Summary of the invention

[0007] The purpose of the present invention is to provide an optical lens calibration method and system. First, an original optical lens parameter vector is obtained, and based on the original optical lens parameter vector, a multi-scale point array diagram is calculated by a ray tracing method, and a first image quality evaluation value is calculated according to the multi-scale point array diagram, so as to effectively distinguish the optical lens that needs to be calibrated, and the optical lens vector that needs to be calibrated is determined as the first optical lens parameter vector; then, the first optical lens parameter vector is input into a conditional generative adversarial network to generate a second optical lens parameter vector, and the conditional generative adversarial network introduces image quality evaluation loss and manufacturing constraint loss, which can guide the generator to generate an optical lens parameter vector that meets the imaging requirements and complies with physical manufacturing constraints; then, based on the second optical lens parameter vector, a multi-scale point array diagram is calculated by a ray tracing method, and a second image quality evaluation value is calculated according to the multi-scale point array diagram, so as to ensure that the second optical lens parameter vector can meet the imaging requirements, and then the optical lens is calibrated according to the second optical lens parameter vector, so as to improve the accuracy of optical lens calibration, thereby reducing the calibration cost.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A method for adjusting an optical lens, comprising:

[0010] S1. Obtaining an original optical lens parameter vector; the original optical lens parameter vector includes adjustable optical lens parameters;

[0011] S2. Calculate a first image quality evaluation value using an imaging quality detection method and perform analysis; the first image quality evaluation value is calculated based on the original optical lens parameter vector; when the first image quality evaluation value is greater than a standard image quality evaluation value, determine the original optical lens parameter vector as the first optical lens parameter vector;

[0012] S3. Inputting the first optical lens parameter vector as a conditional vector into the trained conditional generative adversarial network to generate a second optical lens parameter vector;

[0013] S4. Calculate a second image quality evaluation value using the imaging quality detection method and perform analysis; the second image quality evaluation value is calculated based on the second optical lens parameter vector; when the second image quality evaluation value is greater than the standard image quality evaluation value, return to step S3; when the second image quality evaluation value is not greater than the standard image quality evaluation value, adjust the optical lens according to the second optical lens parameter;

[0014] The imaging quality detection method calculates a multi-scale spot diagram by a ray tracing method, and calculates an image quality evaluation value according to the multi-scale spot diagram; the calculation formula of the image quality evaluation value is: ;

[0015] in, Indicates the number of scales of the spot diagram; Indicates The weight of the scale point diagram evaluation function; Indicates The number of points in a scaled dot plot; Indicates In the scale point diagram The coordinates of the points, Indicates The geometric centroid coordinates of the scale point diagram.

[0016] Preferably, the conditional generative adversarial network performs adversarial training by alternately updating generator parameters and discriminator parameters according to a loss function; the loss function includes a generator loss function and a discriminator loss function.

[0017] Preferably, image quality evaluation loss and manufacturing constraint loss are introduced into the generator loss function to guide the generator to generate an optical lens parameter vector that meets the imaging requirements and actual manufacturing constraints; the generator loss function is as follows:

[0018] ;

[0019] in, represents the noise vector; represents the noise distribution; represents the conditional vector; represents the conditional vector distribution; represents the generated optical lens parameter vector output by the generator; represents the probability that the discriminator judges the generated optical lens parameter vector to be true; represents the image quality evaluation value calculated based on the generated optical lens parameter vector; represents the standard image quality evaluation value; represents the manufacturing constraint loss; and Represents the weight factor, which is used to balance the impact of different loss terms.

[0020] Preferably, the discriminator loss function is used to maximize the probability that the real optical lens parameter vector is judged to be true, and minimize the probability that the generated optical lens parameter vector is judged to be true; the discriminator loss function calculation formula is:

[0021] ;

[0022] in, represents the real optical lens parameter vector; represents the joint distribution of the real optical lens parameter vector and the condition vector; Represents the probability that the true optical lens parameter vector is judged to be true.

[0023] Furthermore, the training data includes an optical lens parameter vector to be adjusted and an optical lens parameter vector that has been adjusted; the optical lens parameter vector that has been adjusted is obtained by adjusting the optical lens parameter vector to be adjusted using a traditional method; in the training process of the conditional generative adversarial network, the conditional vector is the optical lens parameter vector to be adjusted, and the real optical lens parameter vector is the optical lens parameter vector that has been adjusted.

[0024] Preferably, the ray tracing method specifically comprises the following steps:

[0025] Initializing optical system structure information; the optical system structure information includes object plane, mirror plane and image plane structure parameters;

[0026] Perform multi-scale sampling on the initial incident light to obtain a first incident light, and construct a multi-scale sampling matrix structure; the initial incident light is a light emitted from a point on the object plane;

[0027] Analyzing the multi-scale sampling matrix structure, calculating the intersection information of the first incident light and the mirror surface, and obtaining a multi-scale intersection matrix;

[0028] Analyzing the multi-scale intersection matrix, calculating normal vector information according to the intersection information, and obtaining a multi-scale normal vector matrix;

[0029] Analyzing the multi-scale intersection matrix and the multi-scale normal vector matrix, calculating outgoing light information according to the intersection information and the normal vector information, and obtaining a multi-scale refraction matrix;

[0030] The multi-scale refraction matrix is ​​analyzed, and the intersection information of the second incident light and the image plane is calculated to obtain a multi-scale point array matrix; the second incident light is the light after the first incident light is refracted by the mirror surface.

[0031] An optical lens calibration system comprises a first parameter acquisition module, a second parameter acquisition module, an imaging quality detection module and an optical lens calibration module;

[0032] The first parameter acquisition module is used to acquire a first optical lens parameter vector; specifically, it includes: receiving original optical lens parameters, calling an imaging quality detection module to calculate a first image quality evaluation value according to the original optical lens parameter vector, and when the first image quality evaluation value is greater than a standard image quality evaluation value, determining the original optical lens parameter vector as the first optical lens parameter vector;

[0033] The second parameter acquisition module is used to receive the first optical lens parameter vector and acquire a second optical lens parameter vector through a conditional generative adversarial network;

[0034] The imaging quality detection module is used to calculate the image quality evaluation value according to the input optical lens parameter vector; specifically includes: calculating the multi-scale point array diagram by the ray tracing method; calculating the image quality evaluation value according to the multi-scale point array diagram; the calculation formula of the image quality evaluation value is:

[0035] ;

[0036] in, Indicates the number of scales of the spot diagram; Indicates The weight of the scale point diagram evaluation function; Indicates The number of points in a scaled dot plot; Indicates In the scale point diagram The coordinates of the points, Indicates The geometric centroid coordinates of the scale point diagram;

[0037] The optical lens calibration module is used to calibrate the optical lens; specifically includes: receiving the second optical lens parameter vector; calling the imaging quality detection module to calculate the second image quality evaluation value according to the second optical lens parameter vector; when the second image quality evaluation value is greater than the standard image quality evaluation value, re-executing the second parameter acquisition module; when the second image quality evaluation value is not greater than the standard image quality evaluation value, calibrating the optical lens according to the second optical lens parameter vector.

[0038] Preferably, the conditional generative adversarial network performs adversarial training by alternately updating generator parameters and discriminator parameters according to a loss function; the loss function includes a generator loss function and a discriminator loss function.

[0039] Preferably, image quality evaluation loss and manufacturing constraint loss are introduced into the generator loss function to guide the generator to generate an optical lens parameter vector that meets the imaging requirements and actual manufacturing constraints; the generator loss function is as follows:

[0040] ;

[0041] in, represents the noise vector; represents the noise distribution; represents the conditional vector; represents the conditional vector distribution; represents the generated optical lens parameter vector output by the generator; represents the probability that the discriminator judges the generated optical lens parameter vector to be true; represents the image quality evaluation value calculated based on the generated optical lens parameter vector; represents the standard image quality evaluation value; represents the manufacturing constraint loss; and Represents the weight factor, which is used to balance the impact of different loss terms.

[0042] Preferably, the discriminator loss function is used to maximize the probability that the real optical lens parameter vector is judged to be true, and minimize the probability that the generated optical lens parameter vector is judged to be true; the discriminator loss function calculation formula is:

[0043] ;

[0044] in, represents the real optical lens parameter vector; represents the joint distribution of the real optical lens parameter vector and the condition vector; Represents the probability that the true optical lens parameter vector is judged to be true.

[0045] Preferably, the ray tracing method specifically comprises the following steps:

[0046] Initializing optical system structure information; the optical system structure information includes object plane, mirror plane and image plane structure parameters;

[0047] Perform multi-scale sampling on the initial incident light to obtain a first incident light, and construct a multi-scale sampling matrix structure; the initial incident light is a light emitted from a point on the object plane;

[0048] Analyzing the multi-scale sampling matrix structure, calculating the intersection information of the first incident light and the mirror surface, and obtaining a multi-scale intersection matrix;

[0049] Analyzing the multi-scale intersection matrix, calculating normal vector information according to the intersection information, and obtaining a multi-scale normal vector matrix;

[0050] Analyzing the multi-scale intersection matrix and the multi-scale normal vector matrix, calculating outgoing light information according to the intersection information and the normal vector information, and obtaining a multi-scale refraction matrix;

[0051] The multi-scale refraction matrix is ​​analyzed, and the intersection information of the second incident light and the image plane is calculated to obtain a multi-scale point array matrix; the second incident light is the light after the first incident light is refracted by the mirror surface.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] 1. The present invention adopts a conditional generative adversarial network to adjust the parameters of the optical lens, and inputs the optical lens parameter vector to be adjusted as a conditional vector into the generator, which can automatically generate a parameter vector for adjusting the optical lens. Especially in complex optical systems, the conditional generative adversarial network can explore a wider design space and find more optical lens adjustment schemes in a short time, which greatly improves the efficiency of lens adjustment. In addition, the present invention performs imaging quality detection on the generated optical lens parameter vector through a ray tracing method, and performs adjustment when the imaging quality meets the standard, which can avoid rework and reduce adjustment costs.

[0054] 2. The present invention introduces image quality evaluation loss and manufacturing constraint loss into the generator loss function of the conditional generative adversarial network, guiding the generator to generate optical lens parameters that can simultaneously meet the imaging quality standards and physical manufacturing limitations, avoiding the generation of optical lens parameters that have excellent performance in theory but cannot be manufactured in practice, reducing time consumption and material waste in the adjustment process, and reducing the rework rate and related costs.

[0055] 3. The present invention calculates a multi-scale point diagram through a ray tracing method, detects imaging quality based on the multi-scale point diagram, and performs multi-scale sampling of light to simulate the interaction between light and lens at different scales in detail, more comprehensively evaluate the imaging performance of the optical lens, accurately identify the optical lens that needs to be adjusted, and more accurately determine whether the generated optical lens meets the standard, thereby improving the accuracy of optical lens adjustment and reducing costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 A schematic diagram of a flow chart of a method for adjusting an optical lens according to the present invention;

[0057] Figure 2 This is a schematic diagram of the structure of an optical lens adjustment system of the present invention;

[0058] Figure 3 It is a schematic diagram of the flow chart of the multi-scale point diagram calculation method of the present invention. DETAILED DESCRIPTION

[0059] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] The adjustment of optical lenses is a key step to ensure that the performance of the optical system meets the design requirements. Since each lens in the optical system has a direct impact on the final image quality, and the control of aberrations is related to the geometric accuracy and visual effect of the lens imaging, it is particularly important to accurately adjust the lens parameters to reduce aberrations, which includes the careful adjustment of parameters such as the radius of curvature and aperture diameter of the lens; for example, the focal length directly affects the clarity and magnification of the image, while the radius of curvature determines the quality of light focusing and the distribution of the light beam, and the aperture diameter is directly related to the luminous flux and resolution of the lens; inaccurate adjustment of lens parameters not only affects the performance of the entire system, but also causes a waste of time and money. Therefore, accurate adjustment of optical lens parameters is an indispensable link to achieve high-quality imaging and meet the needs of professional applications.

[0061] The present invention provides an optical lens adjustment method and system, which can accurately adjust the optical lens and reduce costs. The effectiveness of the present invention will be described below using two embodiments.

[0062] Embodiment 1

[0063] The specific implementation process of the embodiment of the present application is to convert the method of the present invention into Figure 2 For specific system module implementation, see Figure 1 The specific steps of the method proposed in the present invention are described in the following. The specific steps include:

[0064] Acquire an original optical lens parameter vector; the original optical lens parameter vector includes adjustable optical lens parameters;

[0065] The first image quality evaluation value is calculated and analyzed by using an imaging quality detection method; the first image quality evaluation value is calculated according to the original optical lens parameter vector; when the first image quality evaluation value is greater than a standard image quality evaluation value, the original optical lens parameter vector is determined as the first optical lens parameter vector;

[0066] Inputting the first optical lens parameter vector as a conditional vector into a trained conditional generative adversarial network to generate a second optical lens parameter vector;

[0067] The imaging quality detection method is used to calculate a second image quality evaluation value and analyze it; the second image quality evaluation value is calculated according to the second optical lens parameter vector; when the second image quality evaluation value is greater than the standard image quality evaluation value, return to the previous step; when the second image quality evaluation value is not greater than the standard image quality evaluation value, adjust the optical lens according to the second optical lens parameter;

[0068] The imaging quality detection method calculates a multi-scale spot diagram by a ray tracing method, and calculates an image quality evaluation value according to the multi-scale spot diagram; the calculation formula of the image quality evaluation value is:

[0069] ;

[0070] in, Indicates the number of scales of the spot diagram; Indicates The weight of the scale point diagram evaluation function; Indicates The number of points in a scaled dot plot; Indicates In the scale point diagram The coordinates of the points, Indicates The geometric centroid coordinates of the scale point diagram.

[0071] Furthermore, the embodiment of the present application obtains an original optical lens parameter vector for an industrial camera, including the front surface curvature radius, the rear surface curvature radius, the thickness and the aperture diameter, and calculates the image quality evaluation values ​​respectively. The specific data is shown in Table 1; in different embodiments, there are different parameter variables for different application scenarios. For example, for the structural parameters of the lens in the optical system, the original optical lens parameters include the number of mirrors, the mirror distance, the mirror type and the tilt angle, etc.; in this embodiment, the standard image quality evaluation value is set to 0.01, therefore, the first and second groups of optical lens parameters in Table 1 meet the imaging requirements, and the third and fourth groups of optical lens parameters need to be adjusted; in different embodiments, the standard image quality evaluation value is set based on specific application requirements, industry standards and research in related fields.

[0072] Table 1 Original optical lens parameters and image quality evaluation values ​​of Example 1

[0073]

[0074] Furthermore, the conditional generative adversarial network performs adversarial training by alternately updating generator parameters and discriminator parameters according to a loss function; the loss function includes a generator loss function and a discriminator loss function.

[0075] Through adversarial training between the generator and the discriminator, it is ensured that the generated optical lens parameter vector can not only meet the imaging requirements but also pass the authenticity judgment of the discriminator, making the generated optical lens parameter vector more reliable and feasible.

[0076] Furthermore, image quality evaluation loss and manufacturing constraint loss are introduced into the generator loss function to guide the generator to generate an optical lens parameter vector that meets the imaging requirements and actual manufacturing constraints; the generator loss function is as follows:

[0077] ;

[0078] in, represents the noise vector; represents the noise distribution; represents the conditional vector; represents the conditional vector distribution; represents the generated optical lens parameter vector output by the generator; represents the probability that the discriminator judges the generated optical lens parameter vector to be true; represents the image quality evaluation value calculated based on the generated optical lens parameter vector; represents the standard image quality evaluation value; represents the manufacturing constraint loss; and Represents the weight factor, which is used to balance the impact of different loss terms.

[0079] By adding image quality evaluation loss and manufacturing constraint loss to the generator loss function, the output of the generator is further refined and optimized, which can ensure that the generated optical lens parameters not only meet the imaging quality requirements, but also meet the actual manufacturing constraints, and can reduce the adjustment cost while maintaining high-quality imaging.

[0080] Furthermore, the discriminator loss function is used to maximize the probability that the real optical lens parameter vector is judged to be true, and minimize the probability that the generated optical lens parameter vector is judged to be true; the discriminator loss function calculation formula is:

[0081] ;

[0082] in, represents the real optical lens parameter vector; represents the joint distribution of the real optical lens parameter vector and the condition vector; Represents the probability that the true optical lens parameter vector is judged to be true.

[0083] Through the training of the discriminator, it is possible to more effectively distinguish between real parameters and generated parameters, further improving the authenticity and application value of the generated optical lens parameters.

[0084] Further, see Figure 3 The ray tracing method specifically comprises the following steps:

[0085] Initializing optical system structure information; the optical system structure information includes object plane, mirror plane and image plane structure parameters;

[0086] Perform multi-scale sampling on the initial incident light to obtain a first incident light, and construct a multi-scale sampling matrix structure; the initial incident light is a light emitted from a point on the object plane;

[0087] Analyzing the multi-scale sampling matrix structure, calculating the intersection information of the first incident light and the mirror surface, and obtaining a multi-scale intersection matrix;

[0088] Analyzing the multi-scale intersection matrix, calculating normal vector information according to the intersection information, and obtaining a multi-scale normal vector matrix;

[0089] Analyzing the multi-scale intersection matrix and the multi-scale normal vector matrix, calculating outgoing light information according to the intersection information and the normal vector information, and obtaining a multi-scale refraction matrix;

[0090] The multi-scale refraction matrix is ​​analyzed, and the intersection information of the second incident light and the image plane is calculated to obtain a multi-scale point array matrix; the second incident light is the light after the first incident light is refracted by the mirror surface.

[0091] Furthermore, the object plane structure parameters include object distance, object plane size and object plane position, etc.; the mirror surface structure parameters include optical lens parameters that need to be adjusted, mirror surface position and mirror surface material, etc.; wherein the mirror surface material is used to determine the optical properties such as the refractive index and reflectivity of the optical lens; the image plane structure parameters include image distance, image plane size and image plane position, etc.; in the embodiment of the present application, ray tracing is performed based on the existing professional optical simulation software, and for each ray tracing in the optical lens adjustment process, it is ensured that other optical system structure information except the optical lens parameters that need to be adjusted in the optical system remains consistent.

[0092] Furthermore, light sampling can be performed from three aspects: wavelength, field of view, and pupil; the embodiment of the present application uses the polar coordinate method to sample the pupil; after obtaining the multi-scale point diagram matrix through the ray tracing method, the multi-scale point diagram matrix is ​​converted into a discrete point diagram in a plane coordinate system, and the image quality evaluation value is calculated based on the discrete point diagram.

[0093] The propagation path of light and the corresponding point diagram are accurately calculated through steps such as multi-scale light sampling and matrix analysis, which improves the accuracy of image quality evaluation, provides a reliable theoretical basis for the adjustment process, helps to reduce the number of trial and error in actual adjustment, speeds up adjustment and reduces costs.

[0094] The embodiment of the present application applies a conditional generative adversarial network to a method for adjusting an optical lens, generates an optical lens parameter vector according to input conditions, and introduces image quality evaluation loss and manufacturing constraint loss into the generator loss function, so that the generated optical lens parameter vector meets the imaging requirements as much as possible and does not violate manufacturing constraints. Then, the optical lens is adjusted according to the generated optical lens parameter vector, which can greatly improve the adjustment efficiency and accuracy, thereby reducing errors and reducing costs. In addition, the embodiment of the present application proposes an image quality evaluation calculation method based on a multi-scale point array diagram for evaluating imaging quality. The evaluation function comprehensively considers the aberrations of the point array diagram at different scales, and provides more detailed feedback for the adjustment process.

[0095] Embodiment 2

[0096] In the first embodiment, the method of the present invention is combined with the system to achieve accurate adjustment of an optical lens used in an industrial camera. The method proposed by the present invention will be described again in the embodiment of the present application.

[0097] In the embodiment of the present application, an optical lens applied to a microscope is calibrated, and the specific steps are as follows:

[0098] S1. Obtaining an original optical lens parameter vector; the original optical lens parameter vector includes adjustable optical lens parameters;

[0099] S2. Calculate a first image quality evaluation value using an imaging quality detection method and perform analysis; the first image quality evaluation value is calculated based on the original optical lens parameter vector; when the first image quality evaluation value is greater than a standard image quality evaluation value, determine the original optical lens parameter vector as the first optical lens parameter vector;

[0100] S3. Inputting the first optical lens parameter vector as a conditional vector into the trained conditional generative adversarial network to generate a second optical lens parameter vector;

[0101] S4. Calculate a second image quality evaluation value using the imaging quality detection method and perform analysis; the second image quality evaluation value is calculated based on the second optical lens parameter vector; when the second image quality evaluation value is greater than the standard image quality evaluation value, return to step S3; when the second image quality evaluation value is not greater than the standard image quality evaluation value, adjust the optical lens according to the second optical lens parameter;

[0102] The imaging quality detection method calculates a multi-scale spot diagram by a ray tracing method, and calculates an image quality evaluation value according to the multi-scale spot diagram; the calculation formula of the image quality evaluation value is:

[0103] ;

[0104] in, Indicates the number of scales of the spot diagram; Indicates The weight of the scale point diagram evaluation function; Indicates The number of points in a scaled dot plot; Indicates In the scale point diagram The coordinates of the points, Indicates The geometric centroid coordinates of the scale point diagram.

[0105] Furthermore, the embodiment of the present application obtains an original optical lens parameter vector for a microscope, including the front surface curvature radius, the rear surface curvature radius, the thickness and the aperture diameter, and calculates the image quality evaluation values ​​respectively. The specific data is shown in Table 2; in different embodiments, there are different parameter variables for different application scenarios. For example, for the structural parameters of the lens in the optical system, the original optical lens parameters include the number of mirrors, the mirror distance, the mirror type and the tilt angle, etc.; in this embodiment, the standard image quality evaluation value is set to 0.001, therefore, only the optical lens parameters of the third group in Table 2 need to be adjusted; in different embodiments, the standard image quality evaluation value is set based on specific application requirements, industry standards and research in related fields.

[0106] Table 2 Original optical lens parameters and image quality evaluation values ​​of Example 2

[0107]

[0108] Furthermore, the conditional generative adversarial network performs adversarial training by alternately updating generator parameters and discriminator parameters according to a loss function; the loss function includes a generator loss function and a discriminator loss function.

[0109] Furthermore, image quality evaluation loss and manufacturing constraint loss are introduced into the generator loss function to guide the generator to generate an optical lens parameter vector that meets the imaging requirements and actual manufacturing constraints; the generator loss function is as follows:

[0110] ;

[0111] in, represents the noise vector; represents the noise distribution; represents the conditional vector; represents the conditional vector distribution; represents the generated optical lens parameter vector output by the generator; represents the probability that the discriminator judges the generated optical lens parameter vector to be true; represents the image quality evaluation value calculated based on the generated optical lens parameter vector; represents the standard image quality evaluation value; represents the manufacturing constraint loss; and Represents the weight factor, which is used to balance the impact of different loss terms.

[0112] Furthermore, the discriminator loss function is used to maximize the probability that the real optical lens parameter vector is judged to be true, and minimize the probability that the generated optical lens parameter vector is judged to be true; the discriminator loss function calculation formula is:

[0113] ;

[0114] in, represents the real optical lens parameter vector; represents the joint distribution of the real optical lens parameter vector and the condition vector; Represents the probability that the true optical lens parameter vector is judged to be true.

[0115] Furthermore, the training data includes an optical lens parameter vector to be adjusted and an optical lens parameter vector that has been adjusted; the optical lens parameter vector that has been adjusted is obtained by adjusting the optical lens parameter vector to be adjusted using a traditional method; in the training process of the conditional generative adversarial network, the conditional vector is the optical lens parameter vector to be adjusted, and the real optical lens parameter vector is the optical lens parameter vector that has been adjusted.

[0116] Furthermore, the ray tracing method specifically comprises the following steps:

[0117] Initializing optical system structure information; the optical system structure information includes object plane, mirror plane and image plane structure parameters;

[0118] Perform multi-scale sampling on the initial incident light to obtain a first incident light, and construct a multi-scale sampling matrix structure; the initial incident light is a light emitted from a point on the object plane;

[0119] Analyzing the multi-scale sampling matrix structure, calculating the intersection information of the first incident light and the mirror surface, and obtaining a multi-scale intersection matrix;

[0120] Analyzing the multi-scale intersection matrix, calculating normal vector information according to the intersection information, and obtaining a multi-scale normal vector matrix;

[0121] Analyzing the multi-scale intersection matrix and the multi-scale normal vector matrix, calculating outgoing light information according to the intersection information and the normal vector information, and obtaining a multi-scale refraction matrix;

[0122] The multi-scale refraction matrix is ​​analyzed, and the intersection information of the second incident light and the image plane is calculated to obtain a multi-scale point array matrix; the second incident light is the light after the first incident light is refracted by the mirror surface.

[0123] Furthermore, the object plane structure parameters include object distance, object plane size and object plane position, etc.; the mirror surface structure parameters include optical lens parameters that need to be adjusted, mirror surface position and mirror surface material, etc.; wherein the mirror surface material is used to determine the optical properties such as the refractive index and reflectivity of the optical lens; the image plane structure parameters include image distance, image plane size and image plane position, etc.; in the embodiment of the present application, ray tracing is performed based on the existing professional optical simulation software, and for each ray tracing in the optical lens adjustment process, it is ensured that other optical system structure information except the optical lens parameters that need to be adjusted in the optical system remains consistent.

[0124] Furthermore, light sampling can be performed from three aspects: wavelength, field of view, and pupil; the embodiment of the present application uses the polar coordinate method to sample the pupil; after obtaining the multi-scale point diagram matrix through the ray tracing method, the multi-scale point diagram matrix is ​​converted into a discrete point diagram in a plane coordinate system, and the image quality evaluation value is calculated based on the discrete point diagram.

[0125] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for adjusting an optical lens, characterized in that: include: S1. Obtaining an original optical lens parameter vector; the original optical lens parameter vector includes adjustable optical lens parameters; S2. Calculate a first image quality evaluation value using an imaging quality detection method and perform analysis; the first image quality evaluation value is calculated based on the original optical lens parameter vector; when the first image quality evaluation value is greater than a standard image quality evaluation value, determine the original optical lens parameter vector as the first optical lens parameter vector; S3. Inputting the first optical lens parameter vector as a conditional vector into the trained conditional generative adversarial network to generate a second optical lens parameter vector; S4. Calculate a second image quality evaluation value using the imaging quality detection method and perform analysis; the second image quality evaluation value is calculated based on the second optical lens parameter vector; When the second image quality evaluation value is greater than the standard image quality evaluation value, return to step S3; When the second image quality evaluation value is not greater than the standard image quality evaluation value, adjusting the optical lens according to the second optical lens parameter; The imaging quality detection method calculates a multi-scale spot diagram by a ray tracing method, and calculates an image quality evaluation value according to the multi-scale spot diagram; the calculation formula of the image quality evaluation value is: ； in, Indicates the number of scales of the spot diagram; Indicates The weight of the scale point diagram evaluation function; Indicates The number of points in a scaled dot plot; Indicates In the scale point diagram The coordinates of the points, Indicates The geometric centroid coordinates of the scale point diagram.

2. The method for adjusting an optical lens according to claim 1, characterized in that: The conditional generative adversarial network performs adversarial training by alternately updating the generator parameters and the discriminator parameters according to the loss function; the loss function includes the generator loss function and the discriminator loss function.

3. The method for adjusting an optical lens according to claim 2, characterized in that: Image quality evaluation loss and manufacturing constraint loss are introduced into the generator loss function to guide the generator to generate an optical lens parameter vector that meets the imaging requirements and actual manufacturing constraints; the generator loss function is as follows: ； in, represents the noise vector; represents the noise distribution; represents the conditional vector; represents the conditional vector distribution; represents the generated optical lens parameter vector output by the generator; represents the probability that the discriminator judges the generated optical lens parameter vector to be true; represents an image quality evaluation value calculated according to the generated optical lens parameter vector; represents the standard image quality evaluation value; represents the manufacturing constraint loss; and Represents the weight factor, which is used to balance the impact of different loss terms.

4. The method for adjusting an optical lens according to claim 2, characterized in that: The discriminator loss function is used to maximize the probability that the real optical lens parameter vector is judged to be true, and minimize the probability that the generated optical lens parameter vector is judged to be true; the discriminator loss function calculation formula is: ； in, represents the real optical lens parameter vector; represents the joint distribution of the real optical lens parameter vector and the condition vector; Represents the probability that the true optical lens parameter vector is judged to be true.

5. The method for adjusting an optical lens according to claim 1, characterized in that: The ray tracing method specifically comprises the following steps: Initializing optical system structure information; the optical system structure information includes object plane, mirror plane and image plane structure parameters; Perform multi-scale sampling on the initial incident light to obtain a first incident light, and construct a multi-scale sampling matrix structure; the initial incident light is a light emitted from a point on the object plane; Analyzing the multi-scale sampling matrix structure, calculating the intersection information of the first incident light and the mirror surface, and obtaining a multi-scale intersection matrix; Analyzing the multi-scale intersection matrix, calculating normal vector information according to the intersection information, and obtaining a multi-scale normal vector matrix; Analyzing the multi-scale intersection matrix and the multi-scale normal vector matrix, calculating outgoing light information according to the intersection information and the normal vector information, and obtaining a multi-scale refraction matrix; The multi-scale refraction matrix is ​​analyzed, and the intersection information of the second incident light and the image plane is calculated to obtain a multi-scale point array matrix; the second incident light is the light after the first incident light is refracted by the mirror surface.

6. An optical lens adjustment system, characterized in that: It includes a first parameter acquisition module, a second parameter acquisition module, an imaging quality detection module and an optical lens adjustment module; The first parameter acquisition module is used to acquire a first optical lens parameter vector; specifically, it includes: receiving original optical lens parameters, calling an imaging quality detection module to calculate a first image quality evaluation value according to the original optical lens parameter vector, and when the first image quality evaluation value is greater than a standard image quality evaluation value, determining the original optical lens parameter vector as the first optical lens parameter vector; The second parameter acquisition module is used to receive the first optical lens parameter vector and acquire a second optical lens parameter vector through a conditional generative adversarial network; The imaging quality detection module is used to calculate the image quality evaluation value according to the input optical lens parameter vector; specifically includes: calculating the multi-scale point array diagram by the ray tracing method; calculating the image quality evaluation value according to the multi-scale point array diagram; the calculation formula of the image quality evaluation value is: ； in, Indicates the number of scales of the spot diagram; Indicates The weight of the scale point diagram evaluation function; Indicates The number of points in a scaled dot plot; Indicates In the scale point diagram The coordinates of the points, Indicates The geometric centroid coordinates of the scale point diagram; The optical lens calibration module is used to calibrate the optical lens; specifically includes: receiving the second optical lens parameter vector; calling the imaging quality detection module to calculate the second image quality evaluation value according to the second optical lens parameter vector; when the second image quality evaluation value is greater than the standard image quality evaluation value, re-executing the second parameter acquisition module; when the second image quality evaluation value is not greater than the standard image quality evaluation value, calibrating the optical lens according to the second optical lens parameter vector.

7. The optical lens calibration system according to claim 6, characterized in that: The conditional generative adversarial network performs adversarial training by alternately updating the generator parameters and the discriminator parameters according to the loss function; the loss function includes the generator loss function and the discriminator loss function.

8. The optical lens calibration system according to claim 7, characterized in that: Image quality evaluation loss and manufacturing constraint loss are introduced into the generator loss function to guide the generator to generate an optical lens parameter vector that meets the imaging requirements and actual manufacturing constraints; the generator loss function is as follows: ； in, represents the noise vector; represents the noise distribution; represents the conditional vector; represents the conditional vector distribution; represents the generated optical lens parameter vector output by the generator; represents the probability that the discriminator judges the generated optical lens parameter vector to be true; represents the image quality evaluation value calculated based on the generated optical lens parameter vector; represents the standard image quality evaluation value; represents the manufacturing constraint loss; and Represents the weight factor, which is used to balance the impact of different loss terms.

9. The optical lens calibration system according to claim 7, characterized in that: The discriminator loss function is used to maximize the probability that the real optical lens parameter vector is judged to be true, and minimize the probability that the generated optical lens parameter vector is judged to be true; the discriminator loss function calculation formula is: ； in, represents the real optical lens parameter vector; represents the joint distribution of the real optical lens parameter vector and the condition vector; Represents the probability that the true optical lens parameter vector is judged to be true.

10. The optical lens calibration system according to claim 6, characterized in that: The ray tracing method specifically comprises the following steps: Initializing optical system structure information; the optical system structure information includes object plane, mirror plane and image plane structure parameters; Perform multi-scale sampling on the initial incident light to obtain a first incident light, and construct a multi-scale sampling matrix structure; the initial incident light is a light emitted from a point on the object plane; Analyzing the multi-scale sampling matrix structure, calculating the intersection information of the first incident light and the mirror surface, and obtaining a multi-scale intersection matrix; Analyzing the multi-scale intersection matrix, calculating normal vector information according to the intersection information, and obtaining a multi-scale normal vector matrix; Analyzing the multi-scale intersection matrix and the multi-scale normal vector matrix, calculating outgoing light information according to the intersection information and the normal vector information, and obtaining a multi-scale refraction matrix; The multi-scale refraction matrix is ​​analyzed, and the intersection information of the second incident light and the image plane is calculated to obtain a multi-scale point array matrix; the second incident light is the light after the first incident light is refracted by the mirror surface.

Images