JP2021060353A - Method for estimating internal error of optical system and measurement device - Google Patents

Abstract

To precisely estimate an internal error of an optical system from a measured value of transmission light even if there is a measurement error.SOLUTION: The method for estimating an internal error includes the steps of: causing a measurement system to measure transmission light that has passed through a detection target optical system 120; generating learning data in which the internal error of the detection target system and the calculated value of the transmission light form a pair; generating a discriminator from the learning data; and estimating the internal error of the detection target system by using the measured value of the transmission light and the discriminator.SELECTED DRAWING: Figure 1

Description

The present invention relates to a method of estimating an internal error of an optical system from the transmitted light of the optical system.

The MTF (Modulation Transfer Function), which indicates the performance of the optical system, indicates that the contrast or resolution at the target frequency is high because the value is large at the target frequency. A decrease in the value of MTF can be evaluated as an increase in transmitted wave surface aberration.

It is necessary to solve the inverse problem in order to identify the internal error of the optical system from the transmitted light including the resolution and the transmitted wave surface information. The internal error referred to here includes misalignment, deformation, shape error, refractive index distribution error, and the like of the optical element in the optical system.

Patent Document 1 discloses a method using deep learning as a method of solving an inverse problem. If this method is used, the inverse problem can be solved by using a discriminator generated by obtaining a large amount of training data in advance. Specifically, ray tracing calculation is performed to calculate a plurality of transmitted lights according to a plurality of internal error amounts, and a pair (set) of a plurality of internal errors and transmitted light is learned to measure the transmitted light. A classifier can be generated from the values to estimate the internal error of the optical system.

Japanese Unexamined Patent Publication No. 2018-065171

However, the measurement value of the transmitted light is superposed with the measurement error caused by the measurement system for measuring the transmitted light. As a result, the estimation of the internal error of the optical system may be erroneous due to the measurement error.

An object of the present invention is to provide a method for accurately estimating an internal error of an optical system from a measured value of transmitted light even if there is a measurement error.

The internal error estimation method as one aspect of the present invention is a learning method in which a step of causing the measurement system to measure the transmitted light transmitted through the test optical system and a pair of the internal error of the test optical system and the calculated value of the transmitted light are paired. It is characterized by having a step of generating data, a step of generating a classifier from training data, and a step of estimating an internal error of the optical system under test by using a measured value of transmitted light and a classifier.

Further, in the computer program as another aspect of the present invention, the step of causing the computer to measure the transmitted light transmitted through the test optical system to the measurement system, the internal error of the test optical system, and the calculated value of the transmitted light are obtained. Processing including a step of generating a pair of training data, a step of generating a classifier from the training data, and a step of estimating the internal error of the optical system under test using the measured value of transmitted light and the classifier. Is characterized by executing.

Further, the measuring device as another aspect of the present invention transmits the light source, the light projecting system that causes the light from the light source to enter the test optical system, and the light from the light projecting system that is transmitted through the test optical system. It has a light receiving system in which the transmitted light is incident, a light receiving sensor that receives light from the light receiving system, and an arithmetic system that performs calculations using the output from the light receiving sensor. The calculation system acquires the measured value of the transmitted light, generates training data for pairing the internal error of the optical system under test and the calculated value of the transmitted light, generates a classifier from the learning data, and generates the transmitted light. It is characterized in that the internal error of the optical system under test is estimated by using the measured value of the above and the classifier.

Further, the method for manufacturing an optical system as another aspect of the present invention includes a step of acquiring an internal error estimated by the above-mentioned internal error estimation method and a step of modifying the optical system according to the internal error. It is characterized by.

According to the present invention, even if there is a measurement error, the internal error of the optical system under test can be estimated accurately from the measured value of the transmitted light.

The flowchart which shows the internal error estimation method in Example 1 of this invention. The figure explaining the transmission wave surface calculation in Example 1. FIG. The figure which shows the example 1 of the measurement error in Example 1. FIG. The figure which shows the example 2 of the measurement error in Example 1. FIG. The flowchart which shows the classifier generation method in Example 1. FIG. The flowchart which shows the detail of the classifier generation method in Example 1. FIG. The figure which shows the structure of the transmitted wave surface measuring apparatus in Example 1. FIG. The figure which shows the modification of the internal error estimation method in Example 1. FIG. The flowchart which shows the adjustment amount calculation method in Example 1. FIG. The flowchart which shows the internal error estimation method in Example 2 of this invention. The figure which shows the component decomposition of MTF in Example 2. FIG. The figure which shows the intensity distribution measurement in Example 2. FIG. The flowchart which shows the optical system manufacturing method in Example 3 of this invention.

Hereinafter, examples of the present invention will be described with reference to the drawings.

FIG. 1 shows an internal error estimation method according to the first embodiment of the present invention. A computer can execute a process according to this internal error estimation method (internal error estimation process) according to an internal error estimation program as a computer program.

The internal error estimation method generates an internal error case of the optical system under test in step S101. Next, in step S102, in the case of internal error generated in step S101, a plurality of transmitted lights transmitted through the optical system under test are calculated and a calculated value is acquired. Next, in step S103, the measurement error is superimposed (included) on the calculated value of the number of transmitted lights obtained in step S102. Next, in step S104, a plurality of transmitted lights on which measurement errors are superimposed are learned in step S103 to generate a classifier. Next, in step S105, the transmitted light of the optical system under test is measured using an optical measuring instrument (not shown) to acquire the measured value. Further, in step S106, the internal error is estimated using the classifier generated in step S104 and the measured value of the transmitted light acquired in step S105.

In this embodiment, the transmitted light of the optical system under test is a general term for the transmitted wave surface and the intensity distribution. In the following, the above-mentioned internal error estimation method (hereinafter, simply referred to as an estimation method) will be described in more detail in the case where the transmitted light is a transmitted wave surface.

The internal error of the optical system under test referred to in step S101 is an error possessed by the internal elements constituting the optical system under test. For example, the parallel eccentricity, rotational eccentricity, and spacing error (positional deviation) of a lens or lens group (including multiple lenses) as an internal element, the error of the lens shape (including thickness), and the refractive index and refractive index distribution. Indicates the error of. An example of an internal error is a case in which these plurality of internal errors are changed within an assumed range and added together. The assumed range is preferably at least 1 times, at most 3 times, the range of the manufacturing tolerance of the internal element.

In step S102, the estimation method superimposes a plurality of internal errors in the internal error case generated in step S101 on the test optical system 120, and the on-axis and off-axis transmitted wave planes of the test optical system for each internal error ( Multiple transmitted light) is calculated.

FIG. 2 shows a method of calculating the transmitted wave surface. The on-axis light 162 in the figure is light that passes through the test optical system 120 and is focused on the optical axis of the test optical system 120. The distribution of the optical path length obtained by tracing a plurality of light rays from the object position (infinity in FIG. 2) to the condensing position of the axial light 162 is the transmitted wave plane. By performing the same calculation on the off-axis lights 161 and 163, it is possible to calculate the on-axis and off-axis transmitted wave planes of the optical system 120 to be inspected.

FIG. 2 shows a case where there are only two transmitted wave planes of off-axis light, but the present invention is not limited to this, and a transmitted wave plane of light reaching an arbitrary image height on the image plane (image sensor) can be used. it can. The plurality of transmitted wave planes include, for example, transmitted wave planes under conditions in which the orientation of the image height, the height of the image height, the zoom factor, and the like in the optical system 120 under test are different.

In step S103, the estimation method superimposes the measurement error on the transmitted wave surface calculated in step S102. The measurement error is estimated in advance as a probability distribution for each component from the components of the measurement system and their assembly accuracy. For example, it is the sum of the aberrations of the light source included in the measurement system, the aberrations of the light projecting optical system and the light receiving optical system, the error of the light receiving sensor, the arrangement error of each component, and the like. The error is added by adding the probability distributions. 3 and 4 show an example of adding errors. For example, when there are two errors of the normal distribution as shown in the following equation (1), the total value is also the normal distribution equation shown in the equation (2). For example, if the error shown on the upper side of the equation (1) is uniformly distributed for each region as shown in the following equation (3), the error shown in the equation (3) and the lower side of the equation (1) are shown. The sum with the error of the normal distribution (Equation (4)) is expressed by Eq. (5).

The probability distribution of the total error, which is the sum of a plurality of errors, has the property of approaching the normal distribution P. The σ of the normal distribution P can be calculated by the sum of squares of σ of each element as shown in the following equations (6) and (7). In the estimation method, the sum of a plurality of error elements in this way is used as the measurement error.

If the error of the measurement system is classified into the system error and the random error and added together, it may be easy to calculate the measurement error according to the measurement conditions. When expressing the transmitted wave surface with Zernike coefficients, the error of the measurement system is calculated for each coefficient. In this case, it is advisable to obtain the Zernike coefficient for the on-axis transmitted wave surface and to obtain the elliptical Zernike coefficient described later for the off-axis transmitted wave surface.

Further, the measurement error may be treated as an error whose amount does not change depending on the optical system to be measured. In this case, the measurement error amount is a simplified error amount, but it is not necessary to recalculate for each measurement target. When the measurement error is changed according to the measurement target, a more detailed measurement error amount can be calculated. The measurement error in this case corresponds to the concept of "uncertainty" with respect to the measured value.

In step S104, the pair of the plurality of internal errors obtained in steps S101 to S103 and the calculated value of the transmitted light is learned, and the discriminator is generated. In this embodiment, this pair of internal error and transmitted light is referred to as learning data.

The flowchart of FIG. 5 shows, as an example of step S104, a method of generating a classifier when the calculated value of transmitted light is a Zernike coefficient. The classifier may be a machine learning device such as a support vector machine or a random forest, but it is preferable to use a deep neural network (DNN) that can solve an inverse problem by a general-purpose method. DNN is also used in the example.

The generation method constructs a DNN in step S111. The DNN is composed of multiple layers as shown in FIG. A Zernike coefficient is input to the input layer S121. In the weight layer S122, the value of the previous layer is multiplied by the weight and added to the previous layer. This weight is adjusted in step S112, which will be described later. The activation function S123 performs a process of extracting only a portion equal to or more than a certain threshold value as meaningful information with respect to the value sent from the weight layer S122. It is preferable to use ReLU (ramp function) for this activation function. S122 and S123 are called fully connected layers, and are preferably repeated about three times in order to enhance the expressiveness of the input / output conversion of the network.

In the dropout layer S126, excessive optimization (overfitting) is prevented by performing a process of randomly blocking the route sent to the next layer. After that, the value is passed to the output layer S128 via the weight layer S127.

Next, the generation method adjusts the learning of the DNN, that is, the weight inside the DNN in step S112 of FIG. Specifically, the weight inside the DNN is adjusted by using the optimization calculation so that the difference between the output value when the Zernike coefficient is input to the DNN and the value of the internal error corresponding to the Zernike coefficient becomes small. To do. At this time, it is preferable to use the squared error for the definition of the difference and Adam (Adaptive Moment Optimization) for the optimization calculation. There are various methods for adjusting the internal structure and weight of the DNN, but any method may be used as long as the internal error can be obtained from the Zernike coefficient.

The details of the method of measuring the transmitted light in step S105 will be described. FIG. 7 shows the configuration of a transmitted wave surface measuring device as a measuring system. The measuring device includes a light source 101, 102, 103, a light projecting lens (light emitting system) 111, 112, 113, a light receiving lens (light receiving system) 131, 132, 133, a wavefront sensor (light receiving sensor) 141, 142, 143 and It is composed of a computer 150 as an arithmetic system. The optical system 120 to be inspected is arranged between the light projecting lenses 111, 112, 113 and the light receiving lenses 131, 132, 133.

The optical system 120 to be inspected is an imaging optical system such as an imaging lens for a camera or a projection lens for a projector, and includes optical elements (internal elements) 121 and 122 each of a plurality of lenses or a plurality of lens groups. Although the figure shows two optical elements 121 and 122, the number of optical elements may be three or more. The axis passing through the centers of the optical elements 121 and 122 is referred to as an optical axis, and the direction in which the optical axis extends is referred to as the optical axis direction.

The internal error (quantity) B in this embodiment indicates the amount of misalignment of the optical elements 121 and 122 in the optical system 120 to be inspected, as shown in the equation (8). In equation (8), the amount of parallel position shift in the optical axis direction of the nth optical element is ΔZ _n , the parallel position shift in the direction perpendicular to the optical axis is ΔX _n , ΔY _n , and around the axis perpendicular to the optical axis. The amount of rotational deviation of is indicated _{by Δθ Xn} and Δθ _Yn.

The plurality of lights from the plurality of light sources 101, 102, 103 such as the laser diode are incident on the projection lenses 111, 112, 113, respectively, via the fiber. A plurality of lights emitted as parallel light from the projectile lenses 111, 112, 113 are applied to the optical system 120 to be inspected. The plurality of transmitted lights transmitted through the optical system 120 to be inspected enter the light receiving lenses 131, 132, 133, respectively, pass through them, and are received by the wavefront sensors 141, 142, 143. The image signals from the wavefront sensors 141, 142, and 143 are input to the computer 150, and the computer 150 calculates the transmitted wavefront using the image signals from each wavefront sensor and calculates the elliptical Zernike coefficient of each transmitted wavefront.

A Shack-Hartmann sensor, shearing interferometer, etc. are used as the wavefront sensor. The Shack-Hartmann sensor divides and collects the wave surface (phase distribution) of incident light by a plurality of microlenses constituting the microlens array. Then, it is imaged as a plurality of focused spot images. The position of the center of gravity of the focusing spot shifts according to the inclination of the wave surface incident on the microlens array, and the wave surface inclination is calculated from the amount of deviation of the center of gravity of the focusing spot, and the two-dimensional phase distribution is obtained from the wave surface inclination. calculate.

The Talbot interferometer, which is one of the examples of shearing interferometers, diffracts the incident light by a two-dimensional diffraction grating and acts as interference fringes generated by the interference between transmitted light (0th order light) and diffracted light (for example, 1st order light). Take an image. The light and dark phases of the vertical and horizontal stripes of the interference fringes are obtained by a known method such as a phase shift method and a Fourier transform method. Since the obtained phase corresponds to the differential wave plane of the incident light, the wave plane of the incident light is calculated by integrating this.

The measuring device shown in FIG. 7 can simultaneously measure the on-axis and off-axis transmitted wavefronts, but separately, a driveable stage is used to move the wavefront sensor and light source to the image height to be measured. It may be a measuring device for making the device. In this case, a stage equipped with one light source, a light projecting lens, a light receiving lens and a wavefront sensor is driven, and the light source, the light projecting lens, the light receiving lens and the wavefront sensor are moved to an arbitrary image height position to move the transmitted wave surface. Just measure.

The intensity distribution of transmitted light coaxial with the optical axis is circular. On the other hand, the intensity distribution of the transmitted light off the axis is not circular. An elliptical Zernike polynomial may be used to decompose the wave surface aberration in such a non-circular region into a plurality of wave surface aberration components. An elliptical Zernike polynomial is a Zernike polynomial orthogonalized in an elliptical region. When the circular Zernike polynomial is CZ, the coordinate system in the pupil is x, y, the diameter in the x-axis direction is 2a, and the diameter in the y-axis direction is 2b, the elliptical Zernike polynomial EZ is expressed by the following equation (9). ..

Note that j in the equation indicates the term number of the Zernike polynomial. When finding up to 36 terms of an elliptical Zernike polynomial, the coefficient E can be expressed by the following equation (10). However, the superscript in Eq. (10) indicates the image height number, and the subscript indicates the term number of the Zernike polynomial.

In step S106, the estimation method estimates the internal error B by inputting the coefficient E of the Zernike polynomial measured in step S105 into the classifier F constructed in step S104. As shown in the following equation (11), the discriminator F functions as a function for deriving the relationship between the internal error B and the coefficient E (that is, the Zernike coefficient).

By the internal error estimation method (processing) including the above steps, the internal error of the optical system 120 to be inspected can be obtained with high accuracy.

The examples described above are only one of the examples of the present invention, and various modifications may be made to the above examples. For example, when only the positional deviation of each internal element is obtained as the internal error of the optical system under test, the maximum order of the Zernike coefficient may be suppressed to 16 terms to be strong against disturbance. If the surface shape and refractive index distribution are to be obtained in addition to the positional deviation of each internal element as an internal error, the maximum degree of the Zernike coefficient may be increased to 169 terms, and a polynomial other than the Zernike polynomial should be used. You may.

By handling the transmitted wave surface as point sequence data, it is possible to capture higher-order components. In this case, the generation method handles point sequence data according to the number of light rays calculated in step S102. In step S104, a convolutional neural network is used. In this case, a convolutional layer is added before the fully connected layer to extract the features.

As shown in FIG. 8, steps S107 and S108 may be added to adjust the internal element after measuring the transmitted wave surface of the optical system 120 to be inspected. In step S107, the estimation method calculates the adjustment amount of the internal element with respect to the transmitted wave surface measured in step S105. Here, a case where three transmitted lights pass through three image heights as shown in FIG. 7 will be described. Further, here, as shown in the following equation (12), the adjustment amount A of the internal element is the translation amount ΔX ₀ , ΔY ₀ , ΔZ _{0 of} the optical element 121 in the three-axis directions and the two axes of the optical element 122. An example in which the amount of rotation around the _{wheel is Δθ X0} and Δθ Y0 will be described.

The adjustment amount A may have a value larger than the internal error B. Therefore, it is possible to improve the estimation accuracy of the internal error B by obtaining the internal error B with the influence of the adjustment amount A removed.

Next, the details of step S107 will be described with reference to FIG. In the estimation method, the sensitivities of the optical elements 121 and 122 are obtained in step S131. The sensitivity S is the amount of change in the wave surface aberration component when each optical element moves by a unit amount in each axial direction or around each axis, and can be expressed by the following equation (13).

In step S132, the estimation method obtains a weighting coefficient w to be multiplied by each wave surface aberration component. The weighted wave surface aberration component wE is expressed by the following equation (14).

Here, the weighting coefficient w for the wave surface aberration component to be emphasized is set large. For example, when a part of the imaging screen is out of focus, the weight of the defocus component (four terms of the elliptical Zernike polynomial) is set large as shown in the following equation (15). When emphasizing the waviness of the wave surface, only the defocus component may be set as the value of the difference from a specific image height. Further, the weights of the first to third terms of the elliptical Zernike polynomial, which are easily changed depending on the measurement system, may be set small.

In step S133, the estimation method determines the adjustment amount A using the wave surface aberration component E, the weighting coefficient w, and the sensitivity S. The adjustment amount A may be determined by solving the following minimization problem. That is, the adjustment amount A may be determined so that φ becomes the smallest in the following equation (16). As for the method of determining A at this time, an existing method such as the least squares method may be used.

Further, as is clear from the equation (16), the same result can be obtained by multiplying the sensitivity S by the reciprocal w'of the weighting coefficient without weighting the aberration coefficient. In this case, A may be obtained so as to reduce φ'expressed by the following equation (17).

Next, in step S108 of FIG. 8, as shown in the following equation (18), the estimation method is the adjustment residual after adjusting the measured value E obtained in step S105 with the adjustment amount obtained in step S107. To calculate.

In the final step S106, as shown in the following equation (19), the estimation method inputs the adjustment residual calculated in step S108 into the classifier F constructed in step S104 to obtain the internal error B.

The order of each step in the flowchart described in this embodiment is merely an example, and may be in a different order as long as the internal error of the optical system under test can be finally estimated.

Next, Example 2 of the present invention will be described. In this embodiment, the transmitted light of the optical system under test is measured as its intensity distribution. The flowchart of FIG. 10 shows the internal error estimation method in this embodiment. The internal error estimation process according to this internal error estimation method can also be executed by the computer according to the internal error estimation program as a computer program.

Step S201 is the same as step S101 of FIG. 1 described in the first embodiment. In step S202, the estimation method calculates the modulation response function MTF from the intensity distribution of the transmitted light of the subject 120. Since the MTF indicates the resolution of the optical system under test according to the frequency, the MTF may be simply referred to as the resolution. The MTF is obtained, for example, by obtaining the point image distribution function PSF of the focused point image of the optical system 120 to be inspected and Fourier transforming the intensity thereof.

Further, in this step, as shown in FIG. 11, when calculating the MTF from the transmitted light, the estimation method focuses on the reference frequency to obtain the MTF at the time of defocus. In FIG. 11, the horizontal axis represents the defocus amount and the vertical axis represents the MTF. Although only two MTFs are shown in FIG. 11, in reality, there are MTFs obtained by multiplying the vertical component and the horizontal component of the MTF by the number of combinations of the measured image height and the zoom state. When learning the MTF at the time of defocus as it is, the numerical value of MTF at each defocus amount is saved as it is. The MTF may be decomposed into components such as _{peak values (MTF 1} , MTF ₂ ), defocus amount (defocus ₁ , defocus ₂ ), and full width at half maximum (width ₁ , width ₂ ) shown in FIG. 11 and stored.

The superposition of measurement errors in step S203 is basically the same as in step S103 of the first embodiment. However, only the notation method is changed to the form of the resolution saved in step S202. Further, step S204 is the same as step S104 of the first embodiment.

In step S205, the estimation method measures the intensity distribution of the transmitted light of the optical system 120 to be inspected and calculates the resolution. FIG. 12 conceptually shows the intensity distribution measurement. In the estimation method, when measuring the intensity distribution, black-and-white charts 301, 302, and 303 are arranged at the object positions of the optical system 120 to be inspected, and the images (black-and-white chart images) 311, 312, 313 are measured. The black-and-white chart image at the time of defocus is also measured by changing the arrangement of the optical system 120 to be inspected. The contrast of the black and white chart image is used as the MTF at the time of defocus. The width between the black lines in the black-and-white chart is set to the reference frequency used in step S202.

The black-and-white chart may be a simple edge chart or a patterned chart. When using an edge chart, the estimation method calculates the MTF from the blur of the edge image. When a patterned chart is used, the estimation method calculates the resolution from the chart image by obtaining the correlation between the degree of blurring of the patterned chart and the MTF in advance.

In step S206, the estimation method estimates the internal error of the optical system 120 to be inspected by using the discriminator generated in step S204 and the resolution measured in step S205.

By the internal error estimation method (processing) including the above steps, the internal error of the optical system 120 to be inspected can be obtained with high accuracy.

The flowchart of FIG. 13 shows a method of manufacturing an imaging optical system using the internal error estimation method described in the first or second embodiment.

In the inspection step of step S301, the manufacturing method acquires the measured value obtained in step S105 of FIG.

In step S302, the manufacturing method determines whether the measured value is better (OK) or worse (NG) than the reference, and if it is OK, the manufacturing is terminated. If it is NG, the internal error described in Example 1 or 2 is estimated in step S303.

In step S304, the manufacturing method acquires the internal error estimated in step S303, adjusts the imaging optical system according to the internal error, or replaces or replaces the internal element (optical element) constituting the imaging optical system. The imaging optical system is modified by processing.

According to the manufacturing method of this embodiment, an imaging optical system having high optical performance (imaging performance) can be obtained.
(Other Examples)
The present invention supplies a program that realizes one or more functions of the above-described embodiment to a system or device via a network or storage medium, and one or more processors in the computer of the system or device reads and executes the program. It is also possible to realize the processing. It can also be realized by a circuit (for example, ASIC) that realizes one or more functions.

Each of the above-described examples is only a representative example, and various modifications and changes can be made to each of the examples in carrying out the present invention.

101-103 Light source 111-113 Floodlight lens 120 Tested optical system 131-133 Light-receiving lens 141-143 Wavefront sensor 150 Computer

Claims (16)

A step to make the measurement system measure the transmitted light transmitted through the optical system under test,
A step of generating learning data for pairing the internal error of the optical system under test and the calculated value of the transmitted light, and
The step of generating a classifier from the training data and
An internal error estimation method comprising a step of estimating an internal error of the optical system to be inspected by using the measured value of the transmitted light and the discriminator. An internal error estimation method, wherein the learning data includes a measurement error of the measurement system. The internal error estimation method according to claim 2, wherein the calculated value includes the measurement error. The internal error estimation method according to any one of claims 1 to 3, wherein the measured value indicates a transmitted wave surface. The internal error estimation method according to claim 4, wherein the measurement error is an error for each wave surface aberration component of the transmitted wave surface. The internal error estimation method according to claim 4 or 5, wherein the transmitted wave surface is each transmitted wave surface of transmitted light transmitted through the optical system under test in a plurality of directions. The internal error estimation method according to any one of claims 1 to 3, wherein the measured value indicates a resolution. The internal error estimation method according to claim 7, wherein the resolution is a resolution at a plurality of image heights in the optical system under test. The internal error estimation method according to any one of claims 1 to 8, wherein the measurement error is an uncertainty of the measured value. The internal error estimation method according to any one of claims 1 to 9, wherein the measurement error includes a systematic error and a random error in the measurement system. The learning data is according to any one of claims 1 to 10, wherein the learning data is paired with the calculated value and the internal error according to the manufacturing tolerance of the internal element constituting the optical system to be tested. Internal error estimation method. The internal error estimation method according to claim 11, wherein the amount of internal error according to the manufacturing tolerance corresponds to three times the manufacturing tolerance. From claim 1, the internal error is at least one of a position error, a shape error, a refractive index error, and a refractive index distribution error of the internal element constituting the test optical system. The internal error estimation method according to any one of 12. On the computer
A step to make the measurement system measure the transmitted light transmitted through the optical system under test,
A step of generating learning data for pairing the internal error of the optical system under test and the calculated value of the transmitted light, and
The step of generating a classifier from the training data and
A computer program characterized in that a process including a step of estimating an internal error of the optical system under test is executed by using the measured value of the transmitted light and the discriminator. A floodlight system that causes light from a light source to enter the optical system under test,
A light receiving system in which the transmitted light transmitted through the test optical system among the light from the light projecting system is incident, and a sensor that receives the transmitted light from the light receiving system.
It has an arithmetic system that performs arithmetic operations using the output from the sensor.
The arithmetic system is
Obtain the measured value of the transmitted light and
Learning data for pairing the internal error of the optical system under test and the calculated value of the transmitted light is generated.
A classifier is generated from the training data,
A measuring device characterized in that an internal error of the optical system under test is estimated by using the measured value of the transmitted light and the discriminator. It is a manufacturing method of an optical system.
The step of acquiring the internal error estimated by the internal error estimation method according to any one of claims 1 to 15, and the step of acquiring the internal error.
A method for manufacturing an optical system, which comprises a step of modifying the optical system according to the internal error.

Images