TW200831972A - Method, device and program for adjusting decentration of lens optical system - Google Patents

Abstract

CTF of a zoom lens (21) is obtained by applying a lens design application (72) according to a manufacture error of each optical element of the zoom lens (21), an assembly error except an installation error of a first lens group, and a predicted installation error of the first lens group. A neural network (84) learns in a manner of that the obtained CTF is inputted to an input port of a neural network (84), and a movement amount the first lens group at that time is outputted from the output port. In a decentration adjustment, the first lens group is set at an initial position and takes picture of a lens evaluation chart (14), to obtain the CTF. The obtained CTF is inputted to the neural network (84), to obtain a first movement amount. The first lens group is moved from the initial position with only the first movement amount.

Description

200831972 IX. Description of the Invention: [Technical Field of the Invention] The present invention relates to an eccentric § circumference method and an eccentric adjustment device and an eccentric adjustment program for a lens optical system. [Prior Art] Many cameras now load a zoom lens. The zoom lens is composed of, for example, a front group lens, a zoom lens, a rear group lens, and a focus lens. By moving the zoom lens toward the optical axis direction, the focal length changes, and the focus adjustment is performed by moving the focus lens. The zoom lens, the rear group lens, and the focus lens are attached to the mirror body such that their optical axes are substantially uniform. On the other hand, the front group lens is attached to the mirror body via a spring-like ring or the like, and is movable in a plane orthogonal to the optical axis during eccentric adjustment. After the eccentric adjustment of the front group lens, it is fixed to the mirror body with an adhesive or the like. Generally, a lens evaluation chart such as a resolution chart is imaged by a zoom lens, and the eccentric adjustment of the lens is performed while moving against the spring-like ring and moving the front group lens in a direction orthogonal to the optical axis. Then, the front group lens is fixed to the mirror body at a position where the imaging state of the lens evaluation chart becomes optimum. However, since the above method is adjusted by visual observation by the operator, in addition to the need for proficiency in determining whether the resolution is optimal or not, there is also a mass production suitability, and there is a personal difference in the adjustment result and lack of reliability. Sexual problem. In order to solve such a problem, it is proposed to perform an eccentric adjustment or an optical axis adjustment of an optical element without manual manipulation. For example, Japanese Laid-Open Patent Publication No. Hei. No. Hei. No. Hei. No. Hei. Japanese Laid-Open Patent Publication No. Hei 7-38798 discloses a method of adjusting a focus lens to an infinity in-focus position using a neural network. However, in the adjustment method of the Japanese Patent Publication No. 2002-1 227 85, since the probability exploration of the genetic algorithm requires evaluation of the performance of each generation of each body, it is necessary to perform the sexual energy measurement again and again. Moreover, in the exploratory exploration, even if a suitable range of lens positions without eccentricity can be found, it is difficult to find the optimum lens position within this range. Therefore, it is impossible to adjust in an accurate manner in an instant. Further, in the adjustment method of Japanese Laid-Open Patent Publication No. Hei 7-3879-8, although the relationship between the neural network learning contrast and the focus position of the focus lens is made, the learning does not reflect the manufacturing error, such as the molding error of the resin product, the lens. The error of the unit group immediately. Also, the application of the eccentricity adjustment method is not mentioned at all. Therefore, the neural network cannot be properly learned, and it is difficult to perform eccentric adjustment of the lens with high precision. SUMMARY OF THE INVENTION An object of the present invention is to provide an eccentricity adjustment method for a lens optical system, an eccentricity adjusting device, and an eccentricity adjustment program, which can perform eccentric adjustment of a lens optical system with high precision in a short time. In order to achieve the object and other objects, the eccentricity adjustment method of the lens optical system of the present invention includes an initial movement step, a performance calculation step, a first movement amount calculation step, and a first movement step. Further, by adjusting the lens of the lens optical system composed of a plurality of optical elements to move on the lens mounting surface orthogonal to the optical axis of the lens optical system, the eccentricity of the adjusted lens of the 200831972 lens is adjusted. . In the initial moving step, the adjusted lens is moved to the starting position on the lens mounting surface. In this performance 値 calculation step, the lens evaluation chart is taken using the lens optical system, and the performance of the lens optical system is determined from the photographic image. In the first movement amount calculation step, the performance 値 is input to the neural network, and the first movement amount of the lens is adjusted. This neural network is based on the manufacturing error of each optical component, the assembly error of each optical component except the lens being adjusted, and the predicted mounting position of the adjusted lens. On the other hand, in the first moving step, the adjusted lens is moved to the first adjustment position, which is the position after the first movement amount is added to the home position. In the learning of the neural network, a design application is used to simulate a lens evaluation chart from a plurality of imaginary CAD data to obtain performance defects. Each imaginary CAD data has been corrected for the CAD of the design of the optical system of the lens optical system based on the manufacturing error of each optical component, the assembly error of each optical component except the lens being adjusted, and the predicted mounting position of the lens. data. Further, the performance 値 obtained by the simulation and the amount of movement from the starting position to the predicted mounting position are input to the neural network, and the neural network is learned. Further, after the first moving step, it is preferable to include the first performance 値 recalculation step and the first pass/fail determination step. In the first performance 値 recalculation step, at the first adjustment position, the lens evaluation chart is taken by the photographic element via the lens optical system, and the performance 値 at the first adjustment position is obtained from the photographic image. On the other hand, in the first pass/fail determination step, it is judged whether or not the eccentricity adjustment of the lens optical system is acceptable based on the performance 値' obtained in the first performance 値 recalculation step. 200831972 The lens evaluation chart preferably has a plurality of lens evaluation areas. Further, each lens evaluation region preferably has a horizontal direction evaluation pattern 'which is used to evaluate the horizontal direction of the surface orthogonal to the optical axis of the lens optical system 値; and a vertical direction evaluation pattern for measuring The performance in the vertical direction on the plane orthogonal to the optical axis. In a preferred embodiment of the present invention, the lens optical system includes a zoom lens, and the lens evaluation area includes a wide-angle evaluation area for evaluating the performance of the zoom lens at the wide-angle end; and a telephoto evaluation area, Used to evaluate the performance of the zoom lens at the telephoto end. The wide-angle evaluation area is set at the four corners and the center of the lens evaluation chart, and the telephoto evaluation area is set at the four corners and the center of the central part of the lens evaluation chart. This property is preferably a CTF. Further, after the first pass/fail determination step, it is preferable to have a readjustment step. The re-adjustment step includes a search step, a second performance 値 recalculation step, an evaluation 値 calculation step, a second movement amount calculation step, and a second movement step. In the search step, when the first pass/fail determination step is determined to be unsatisfactory, the adjusted lens is moved to the plurality of first search points located in the first search range centering on the first adjustment position on the lens mounting surface. . In the second performance 値 recalculation step, a lens evaluation chart is taken at each first search point, and the performance 値 at each search point is obtained from the captured image. In the evaluation 値 calculation step, an evaluation 算 is calculated for each of the first exploration points. In the second movement amount calculation step, the second movement amount of the lens to be adjusted is obtained based on the plurality of evaluations. In the second moving step, the adjusted lens is moved to the second adjustment position, which is the position after the second movement amount is added to the first adjustment position. Preferably, the re-adjusting step further includes a third performance 値 recalculation step 200831972 and a second pass/fail determination step. In the third performance 値 recalculation step ‘ at the second adjustment position, the lens evaluation chart is taken by the photographic element via the lens optical system, and the performance 値 at the second adjustment position is obtained from the photographic image. In the second pass/fail determination step, it is determined whether or not the eccentricity adjustment of the lens optical system is acceptable based on the performance 求 obtained in the third performance 値 recalculation step. Further, it is preferable to prepare a plurality of evaluation 値 calculation methods for calculating the evaluation ’, and select an evaluation 値 calculation method in a predetermined order as soon as they are qualified. Further, the second movement amount calculation step preferably includes a secondary curved surface forming step and a coordinate calculating step. In the quadric surface forming step, each of the exploration points and the evaluation points thereof are drawn on the XY axis whose origin is the first adjustment position, and the three-dimensional coordinates which are evaluated as the Z axis, and are formed according to the points of the evaluation flaws drawn. Evaluate the quadratic surface of 値On the other hand, in the coordinate calculation step, the XY coordinates corresponding to the point at which the evaluation 値 becomes maximum on the quadric surface are obtained as the second movement amount. The re-adjusting step further includes the following steps: (A), in the case where the plurality of evaluation methods are all failed, the second evaluation range is narrower than the first search range. a second search point having a smaller number than the first search point; and (B) a step of performing the second performance 値 recalculation step, the evaluation 値 calculation step, the second movement amount calculation step, and the second search point 2 moving step, third performance 値 recalculation step, and second pass/fail determination step; and (C) step, selecting a plurality of evaluation 値 calculation methods in a predetermined order, and executing the step as soon as it is qualified (B). Further, as a plurality of evaluation evaluation methods, it is preferable to include at least one of the following: the worst calculation method is to calculate the worst performance of the plurality of performances 値 200831972 and as an evaluation 値; The average calculation method is to calculate the average 値 of a plurality of performance 値 as an evaluation 値; and the difference 値 calculation method is to calculate the performance of the area in the four corners of the wide-angle evaluation area and the telephoto evaluation area. The difference between the balances is used as an assessment. The calculation of the difference 値 of the difference 値 calculation method is to calculate the difference in performance 区域 between the four corner regions in the wide-angle evaluation region, and to average the sum of the absolute 値 of the difference, and calculate the evaluation region for the telephoto. The difference between the performances of the four corners is averaged, and the sum of the absolute 値 is averaged, and then the reciprocal of each of the average 値 is added and found. Further, as the evaluation 値 calculation method, it is preferable to include a weight calculation method for assigning weights to the evaluation 算出 calculated by the worst 値 calculation method, the average 値 calculation method, and the difference 値 calculation method, and calculating the evaluation value. Further, it is preferable that the first re-learning step is provided after the first or second pass/fail determination step is judged as pass. In the first re-learning step, the performance 値 is input to the neural network, and the amount of movement is calculated, and the difference between the obtained movement amount and the first movement amount or the second movement amount that is converted into the distance from the home position is The neural network is re-learning. The eccentricity adjusting device for an adjusted lens according to the present invention includes: a tuned lens moving unit, an imaging unit, a performance 値 calculating unit, a neural network, a first movement amount calculating unit, and a control unit. Further, the lens of the lens optical system composed of a plurality of optical elements is moved on a lens mounting surface orthogonal to the optical axis of the lens optical system, and the eccentricity of the adjusted lens with respect to the optical axis is adjusted. The adjusted lens moving portion holds the adjusted lens and moves it on the lens mounting surface. The Department of Photography photographed through the lens optical system -10- 200831972 Mirror evaluation table. The performance 値 calculation unit evaluates the performance of the lens optical system from the photographic image of the lens evaluation chart. The neural network learns as described above, inputting performance from the input layer, and outputting the amount of movement of the adjusted lens from the output layer. The first movement amount calculation unit inputs the performance 求 obtained by the performance 値 calculation unit into the neural network, and obtains the first movement amount of the adjusted lens. On the other hand, the control unit controls the adjusted lens shifting force unit so that the adjusted lens moves only by the first movement amount. Further, the eccentricity adjusting device of the lens to be adjusted is preferably provided with a performance calculation means and a neural network learning means. The performance calculation method uses a design application to simulate a lens evaluation chart from a plurality of hypothetical CAD data, and to obtain a plurality of performance defects. Each of the imaginary CAD data has corrected the CAD data of the CAD data on the design of the lens optical system based on the manufacturing error of each optical element, the assembly error of each optical element except the lens to be adjusted, and the predicted mounting position of the lens to be adjusted. In addition, the neural network learning method inputs the neural network and learns the neural network by using the performance obtained by the simulation and the amount of movement from the starting position to the predicted installation position. The eccentricity adjustment program of the lens optical system of the present invention causes the computer to execute the initial movement step, the performance 値 calculation step, the first movement amount calculation step, and the first movement step. Further, the eccentricity of the adjusted lens on the optical axis is adjusted by moving the adjusted lens of the lens optical system composed of the plurality of optical elements on the lens mounting surface orthogonal to the optical axis of the optical system of the lens. Further, the learning of the neural network used in the first movement amount calculation step includes a performance seeking step of simulating the lens evaluation chart from a plurality of hypothetical CAD data using a design application program to obtain performance 値. -11- 200831972 In another preferred embodiment of the present invention, the eccentricity adjustment method of the lens optical system includes a movement amount prediction step and a movement step. In the movement amount prediction step, a plurality of past adjustment data including the past movement amount of the movement amount of the adjusted lens which has been eccentrically adjusted in the past, and the performance of the lens optical system corresponding to the past movement amount are included. , predicts the amount of movement of the lens being adjusted. In the moving step, the adjusted lens is moved in accordance with the predicted amount of movement. Preferably, the amount of movement is preferably obtained by zp homogenizing i of the plurality of past movement amounts. Further, in the past adjustment data, it is preferable to include an improvement rate of the ratio in which the performance of the eccentricity is improved in the past, and the amount of movement is determined by weighting the past movement amount in accordance with the improvement rate. Further, in the past adjustment data, the date and time of the eccentricity adjustment performed in the past may be included, and the amount of movement of the past movement amount may be obtained in accordance with the weighting of the date and time. Preferably, the plurality of past adjustment data are distinguished according to each batch number, and in the movement amount prediction step, the movement amount is preferably predicted based on past adjustment data corresponding to the batch number. In still another preferred embodiment of the present invention, the lens optical system eccentricity adjusting device includes a movement amount predicting portion and an adjusted lens moving portion. The movement amount prediction unit predicts based on a plurality of past adjustment data including the past movement amount of the movement amount of the adjusted lens that has been eccentrically adjusted in the past and the performance of the lens optical system corresponding to the past movement amount. The amount of movement of the lens is adjusted. The adjusted lens moving portion moves the adjusted lens only by the amount of movement. In other preferred embodiments of the present invention, the eccentricity of the lens optical system -12-200831972 adjusts the program to cause the computer to perform the motion amount prediction step and the moving step. In the movement amount prediction step, a plurality of past adjustment data including the past movement amount of the movement amount of the adjusted lens which has been eccentrically adjusted in the past, and the performance 値 of the lens optical system corresponding to the past movement amount are included. The amount of movement of the adjusted lens is predicted. In the moving step, the adjusted lens is moved only by the amount of movement. According to the present invention, since the amount of movement of the adjusted lens and the performance at that time based on the simulation of the design application are used in the setting of the neural network (initial learning), even if there is no eccentric adjustment work in the past The information obtained can also be set simply. Further, the performance of the adjusted lens at the initial position is measured, and this performance is input to the neural network to calculate the amount of movement of the adjusted lens. Therefore, the adjustment time can be shortened more greatly than before. In the case of using the eccentricity adjustment of the neural network, the performance 値 does not satisfy the reference ,, the lens is moved from its adjustment position to a plurality of exploration points located within a predetermined exploration range, and the performance at each exploration point is obtained, These performances determine the optimal amount of movement. Therefore, the adjustment accuracy can be improved. However, when the eccentricity adjustment is properly performed, the performance 那时 at that time is input to the neural network, and the amount of movement is calculated, and the amount of movement from the first movement amount or the distance from the first movement amount or the conversion to the start position is obtained. 2 The difference in the amount of movement allows the neural network to re-learn. Thus, re-learning of the neural network reflecting the characteristics of the actual lens can be realized, and the adjustment performance can be further improved. In other preferred embodiments of the present invention, the amount of past movements obtained in the past eccentricity adjustment and the performance at that time are stored as past adjustments. Then, based on this past adjustment data, the amount of movement of the adjusted lens is calculated. Since it is not necessary to measure the performance of the tuned lens, the amount of movement can be ascertained, so the time required for eccentric adjustment is shortened. In addition, the past adjustment data is distinguished according to the batch numbers of the lens optics. The performance of the lens optical system having a common batch number is similar to that of a similar number, and the position after the eccentric adjustment becomes substantially the same is highly likely. Therefore, the eccentricity adjustment can be performed efficiently by managing the past adjustment data according to the batch number. The above objects and advantages will be apparent to those skilled in the art from a <RTIgt; [Embodiment] As shown in Fig. 1, a lens eccentricity adjusting device 1 of the present invention includes a lens holder 1 1 , an adjusted lens moving unit 1 2 , a camera driving unit 13 , a lens evaluation chart 14 , and a controller. 1 5. Operation panel 1 6 and alarm 1 7. The lens holder 1 1 is formed with a seat portion 1 8 for holding the lens unit 19 which is an eccentric adjustment object. The adjusted lens moving unit 1, the camera driving unit 13 and the controller 15 are provided in the lens holder 11. Further, although the controller 丨5 has the neural network 804, the controller and the neural network can be separated. The lens unit 19 includes a lens barrel 40, a zoom lens (zoom lens optical system) 21, a zoom mechanism 22, an AF mechanism 23, an image area sensor 24, a unit controller 26 having the image processing unit 25, and a battery ( Omit the illustration). As shown in Fig. 2, the lens unit 19 is detachably attached to the front surface of the camera body 30 via a shutter switch 41a formed at the rear end of the lens barrel 40 to constitute a digital camera. . In the camera body 3, an operation unit 3, a shutter button 33, an LCD 3 4 as a display unit, and a zoom button 35 are provided. Further, the camera body 30 is also provided with a data recording unit, a camera controller, a battery, and the like (none of which are shown) for recording video data on a recording medium that is freely detachable. -14- 200831972 As shown in Fig. 1, the lens barrel 40 accommodates the first to fourth lens groups G1 to G4 constituting the zoom lens 21. The first lens group G 1 is a front group lens ‘the second lens group G 2 is a zoom lens, the third lens group G 3 is a rear group lens, and the fourth lens group G4 is a focus lens. Further, each of the lens groups G1 to G4 may be a single lens. Further, the lens optical system may be two groups, three groups or the like. Further, the focal length may be fixed. The first lens group G1 and the third lens group G3 are each held by the fixed lens holders 41 and 43. On the other hand, the second lens group G 2 and the fourth lens group ^ G 4 are each held by the moving lens holders 4 2, 4 4 . The movable lens holder 4 2 is movably attached to the lens barrel 40 by means of a screw rod 46a and a guide rod 47. The screw rod 46a is rotated by the stepping motor of the zoom mechanism 22, whereby the second lens group G2 is moved to zoom. The moving lens holder 44 is attached to the lens barrel 40 so as to be freely movable by a screw bar 46b and a guide bar 47. The screw rod 46b is rotated by the stepping motor of the AF mechanism 23, whereby the fourth lens group G4 is moved to adjust the focus point. In the assembly operation of the zoom lens 21, first, the second to fourth lens groups G2 to G4 are sequentially attached to the lens barrel 40, and then the first lens group G1 is mounted. The biasing member (not shown) such as a spring-like ring biases the fixed lens holder 41 of the first lens group G1 toward the lens mounting surface 40a, and is movable on the lens mounting surface 40a. After the eccentricity adjustment of the first lens group G1 is completed, the fixed lens holder 41 is fixed to the lens attachment surface 40a with an adhesive or the like. The adjusted lens moving unit 1 2 includes a lens holder 50, an X-direction shifting unit 51, a Y-direction shifting unit 52, a Z-direction shifting unit 53, and a bottom portion 54, and moves the first lens group G1 during eccentricity adjustment. The lens holder 50 is attached to the Y-direction shifting portion 52 so that it can hold the position of the grip -15-200831972 on both sides of the fixed lens holder 4! and release the clamping position of the fixed lens holder 41. Displacement between. The Y-direction shifting portion 52 is supported by the X-direction shifting portion 51, and the X-direction shifting portion 51 is supported by the Z-direction shifting portion 53. The driver 55 is placed in the bottom portion 5 4 . The controller 15 drives the X-direction shifting unit 51, the Y-direction shifting unit 5 2, and the Z-direction shifting unit 53 via the driver 55. The X-direction shifting portion 51 moves the Y-direction shifting portion 52 to move the first lens group G1 toward the upper direction of the lens mounting surface 40a via the fixed lens holder 41 (the direction perpendicular to the paper surface of the drawing surface. "X direction") "Moving. The Y-direction shifting portion 5 2 moves the lens holder 50 to the left-right direction of the lens mounting surface 40a (the direction parallel to the plane of the drawing surface.) Called "Y direction"). The Z-direction shifting portion 5 3 presses the X-direction shifting portion 51 to press the first lens group G1 against the lens mounting surface 40a. The camera drive unit 13 controls the zoom mechanism 22, the AF mechanism 23, and the image processing unit 25 via the unit controller 26 of the lens unit 19. The video area sensor 24 is, for example, a CCD, and is disposed on the image plane side of the fourth lens group G4. f When the eccentricity state of the first lens group G1 is detected in the eccentric adjustment work, the image area sensor 24 is used, and when the second lens group G2 (zoom lens) is at the wide-angle end (hereinafter referred to as "wide angle" ) and the lens evaluation chart 14 (chart image) is taken twice at the telephoto end (hereinafter referred to as "the telephoto"). The image processing unit 25 generates image data (chart image data) of the chart image based on the signal from the image area sensor 24. The chart image data at the wide angle and the telephoto position is transmitted to the controller 15 via the camera drive unit 13. The controller 1 5 calculates ctF from two chart image data. The CTF (Contrast Transfer Function) is one of the performance characteristics of the lens performance. -16 - 200831972 indicates the state of contrast of images. The higher the CTF recovery, the higher the resolution. Further, since the lens eccentricity has a correlation with the CTF, the amount of movement of the eccentricity of the lens can be known from the CTF. Further, the performance of the lens is not limited to CTF. The lens evaluation chart 14 is disposed in front of the lens unit 19 and is substantially uniformly illuminated by the illumination device 56. In the lens evaluation chart 14, as shown in FIG. 3, the wide-angle evaluation areas 61 to 65 used for evaluating the CTF at the wide angle, and the telephoto evaluation areas 66 to 70 used for evaluating the CTF at the telephoto end are set. . The wide-angle evaluation areas 61 to 65 are each located in the evaluation areas of the upper right, lower right, upper left, lower left, and center of the lens evaluation chart 14. On the other hand, the telescope evaluation areas 66 to 70 are each located in the evaluation areas of the upper right, lower right, upper left, lower left, and center of the central portion 14a of the lens evaluation chart 14. Although omitted on the drawing, in each of the evaluation areas 61 to 70, a pattern for evaluating the CTF of the zoom lens 21X direction (hereinafter referred to as "X direction evaluation pattern") and a figure for evaluating the CTF in the Y direction are recorded. (hereinafter referred to as "Y-direction evaluation graph"). The X-direction evaluation pattern is, for example, configured to alternately configure white and black vertical stripe patterns in the longitudinal direction. On the other hand, the Y-direction evaluation pattern is, for example, configured to alternately configure white and black horizontal stripe patterns in the lateral direction. In addition, in the wide-angle evaluation area and the telescope evaluation area, a graph for evaluating the radial performance , or a graph for evaluating the performance 切 of the tangential direction may be recorded. The image data of the wide-angle evaluation area 6 1 to 6 5 is extracted from the chart image data at the wide angle, and the image data of the evaluation area 66 to 70 for the telephoto is extracted from the chart image data at the telephoto. Since the one-dimensional evaluation areas respectively include the X-direction evaluation pattern and the Y-direction evaluation pattern, the image area sensor 24 captures the image, and finally finds one in the X direction and 10 in the Y direction. A total of 20 CTFs of 200831972 are used to indicate that the first lens group G1 is in an eccentric state at one mounting position by using 20 CTFs as one group. As shown in FIG. 4, the controller 15 includes a CAD database 7 1 , a lens design application 72 , a learning database 73 , a first eccentricity adjustment unit 75 , an image evaluation unit 76 , and a lens search amount specifying unit 7 . 7. The second eccentricity adjustment unit 78. In the CAD database 71, a plurality of CAD data indicating the optical structures of the first to fourth lens groups G1 to G4 (the radius of curvature of each lens group, the lens thickness, the refractive index, the lens interval, and the like) are recorded. These CAD materials have basic CAD data on the design, and a number of hypothetical CAD data that are manufactured, assembled, or considered for installation errors and have been corrected for basic CAD data. A plurality of hypothetical CAD data are obtained in the following manner. First, in the basic CAD data, the mounting position (starting position) of the first lens group G1 is maintained at the design mounting position (starting position), depending on the first to fourth positions that may occur at the time of manufacture, assembly, or installation. The manufacturing errors of the lens groups G1 to G4 and the assembly errors or mounting errors of the second to fourth lens groups G2 to G4 correct the optical structures of the first to fourth lens groups G1 to G4. Hereinafter, the corrected basic CAD data is referred to as the first corrected CAD data. Next, the position of the second lens group G1 is changed from the initial position to the first predicted women's position in the first corrected CAD data. Thereby, the (1 - 1) hypothetical C A D data can be obtained. The first (1 - 1) virtual C A D data includes the distance (X direction and γ direction) from the start position to the first predicted mounting position, which is used as the amount of movement of the first lens group G1. This amount of movement corresponds to the amount of eccentricity based on the starting position. In the same manner, in the first correction cad data, the position of the first lens group G1 is changed from the initial position to the Nth prediction -18-200831972 installation position (N-series 2 or more natural numbers), and the first (N-N) is obtained. ) Imaginary CAD data. The first (1 - N) virtual CAD data includes the amount of movement of the first lens group G1 from the start position to the Nth predicted mounting position. Then, in the state in which the mounting position of the first lens group G1 is held at the home position, the manufacturing error of the first to fourth lens groups G1 to G4 differs from the error of the first corrected CAD data in the basic CAD data. The optical errors of the first to fourth lens groups G1 to G4 are corrected by the combination error or mounting error of the second to fourth lens groups G2 to G4. Hereinafter, the corrected basic CAD data is referred to as a third-order corrected CAD data (a natural number of M system 2 or more). Next, the CAD data is corrected in the third step, and the position of the first lens group G1 is changed from the initial position to the first predicted mounting position. In this way, the (Μ-1) hypothetical CAD data can be obtained. The first (Μ-1) virtual CAD data includes the amount of movement of the first lens group G1 from the start position to the first predicted mounting position. Similarly, in the third, the CAD data is corrected, and the position of the first lens group G1 is changed from the initial position to the third predicted mounting position (the natural number of the Ν 2 or more), and the imaginary CAD data of the first (Μ - Ν) is obtained. . This (Μ - Ν) imaginary CAD data includes the amount of movement of the first lens group G 1 from the start position to the second predicted mounting position. The controller 15 simulates the chart image data of the lens evaluation chart 14 by the lens design application 72 based on one CAD data taken out from the CAD database 71, and calculates 10 CTFs in the X direction and 1 C direction in the X direction. CTF (20 in total). Then, a total of 20 CTFs (one set of CTFs) calculated by the lens design application 72 and the amount of movement of the first lens group G1 included in the CAD data are used as learning materials, and are recorded in the learning database - 19- 73 ° 200831972 As the lens design application 72, the optical design and evaluation software "ZEMAX (product name)" manufactured by LEA DIN GTEX Co., Ltd. can be used. ZEMAX inputs objects from the CAD application software, revisiting the shape of the finished optical component, and easily finding the performance of various forms with errors. In addition, the lens design application can be used in a variety of applications as long as it is an application for optical design and evaluation. The first eccentricity adjustment unit 75 includes a neural network 84 and an error calculation unit 85. As shown in Fig. 5, the neural network 84 is used in the first eccentricity adjustment step, and is composed of an input layer 84a, an intermediate layer 84b, and an output layer 84c. The input layer 84a is composed of 20 units 〇l[i] (i = 1 to 20). In the unit 01[1] to 〇1[1〇], 10 CTFs in the X direction corresponding to the respective evaluation areas 61 to 70 are input, and the units 〇i[ii] to 〇ι[2〇] are input and the evaluation areas 61 are provided. ~70 corresponds to 10 CTFs in the Y direction. The output layer 84c is composed of two units 〇3[k] (k=l, 2). The amount of movement in the X direction of the first lens group G1 is output from the unit 〇3 [1], and the amount of movement of the first lens group G1 in the Y direction is output from the unit 0 3 [2]. The amount of movement in the X direction and the Y direction corresponds to the amount of eccentricity based on the starting position. The intermediate layer 84b is composed of a predetermined number of units 〇2U] (j is an arbitrary self-number). In addition, the neural network 84 can be constructed in a software or hardware manner, and various commercially available software can be used. Each unit 〇1 of the input layer 84a is combined with a combination coefficient W21[j][n and each unit 〇2U of the intermediate layer 84b]. Further, each unit 〇2[j] of the intermediate layer 84b is combined with each unit 〇3 [k] of the output layer 84c by the coupling coefficient w32[k][j]. The relationship between each unit 02 [j] of the intermediate layer 84b and each unit 01 [1] of the input layer 84a is represented by the following [Formula 1]. Further, the relationship between each unit 〇3[k] of the output layer 84c and each of the -20-200831972 units 〇2 [ j ] of the intermediate layer 84b is expressed by the following [Formula 2]. [Formula 1] 02[j] = Tf(w2l[j]\i] x OIU] - Θ2[β) i [Formula 2] 〇m = Σ/(^32[^][7·] X 02 [j] - 03[k]) j (In the above equation, f denotes a sigmoid function, 02[j] denotes a threshold 各2[j] of the intermediate layer 84b, 0 3 [k] indicates the threshold 各 of each unit 03 [k] of the output layer 84c. Further, the combination coefficients w21[j][i], w32[k][j], Θ 2[j], And 0 3 [k]. When the neural network 84 is set (initial learning), a plurality of learning materials are simultaneously read from the learning database 7 3. Each learning group is composed of one set of CTFs and moving amounts. Among the 1 set of CTFs, 10 X-direction CTF inputs 〇1[1]~〇1[10], 10 Y-direction CTF inputs 〇1[11]~(〇1 [20]. Then, from The unit 03 [1] outputs the movement amount in the X direction of the first lens group G1 obtained by the [first formula] and the [second equation], and outputs the movement of the first lens group G1 in the Y direction from the unit 〇3 [2]. The error calculation unit 85 calculates the movement amount 03[k] of the first lens group G1 and the first lens group G1 included in the same learning data based on the input of one set of CTFs for learning data to the neural network 84. Shift The quantity T[k] is calculated as the error Δ w32[k]U] of the combination coefficient w3 2[k][j] and the error Δ Θ 3[k] of the threshold 値0 3 [k]. Equation] represents the error Δ w32[k][j] and the error Δ 0 3[k]. -21- 200831972 [3rd formula] d3[k] = (T[k] - 03[k] x (1.0 - 03 [k]) / μ\ Aw32[A:] [/] = sx J3[A:] x 〇2[several A03[k] = sxd3[k] Here, ε and / / are arbitrary constants. The error Δ w32[k][j] is added to the combination coefficient w32[kn"], and then w32[k]U] is updated. Further, the calculated error Δ Θ 3[k] and the threshold 値0 3 [k] is added, and 0 3 [k] is updated again. Further, the error calculation unit 85 is based on the input layer 01 [input, the output from the intermediate layer 〇 2 [j], and the updated coupling coefficient W32 [ k][j], and the above d 3 [ k ], calculate the error of the error Δ w 2 1 [ j ][丨] and the threshold 値Θ 2[j] of the combination coefficient w 2 1 [ j ] [ i ] Δ 0 2[j]. The error Δ w2im[i] and the error Δ 0 2[j] are expressed by the following [Form 4]. [Form 4] = 〇2 [1.〇-〇2[/ ])[|^3[咖32|&gt;][^(2.0///),

Aw2l[j][i] = εχ d2[j] χ 0\\ί\ ^^[j] - ^xd2[j] The calculated error Δ w21U][i] and the combination coefficient w2l[]][i ] Add 'and then w2im[n update. Further, the calculated error Δ Θ 2 ["] is added to the threshold 値 0 2 [j], and 0 2 [j] is updated. According to the above method, the combination coefficients w32[k][j], w2 1[j][i], and threshold 値0 3[k], 0 2U] are repeatedly updated for each learning data, and the neural network-22 - 200831972 8 4 The end of the required learning is set. In the first eccentricity adjustment step, when the CTF calculated by the CTF measuring unit 76b is input to each unit 0 1 [1 ] of the input layer 8 4 a, the starting position is output from each unit 03 [k] of the output layer 84c. The amount of movement of the first lens group G1 based on the reference (hereinafter referred to as "first movement amount") X1, Y1. The shift controller 87 adds the adjustment position AP of the first movement amount X1, Y1 to the design mounting position (starting position) of the first lens group G1. Then, the X-direction shifting portion 51 or the Y-direction shifting portion 52 is driven to move the first lens group G1 to the adjustment position AP on the lens mounting surface 40a. The image evaluation unit 76 includes a comparison calculation unit 76a, a CTF measurement unit 76b, a determination unit 76c, and a memory 76d. The comparison calculation unit 76a receives the chart image data at the wide angle and the chart image data at the telephoto position from the image processing unit 25 via the camera drive unit 13. The comparison calculation unit 76a extracts the image data of the wide-angle evaluation areas 6 1 to 6 5 from the chart image data at the wide angle, and extracts the image data of the telephoto evaluation areas 6 6 to 7 0 from the chart image data at the telephoto position. Since the X-direction evaluation pattern and the Y-direction evaluation pattern are recorded in each of the evaluation areas 61 to 70, the comparison calculation unit 76a calculates 10 comparisons in the X direction and 10 in the γ direction from the image data of the extracted 10' evaluation areas. Comparison (20 total). The contrast C in the X direction and the γ direction is expressed by the following [5th formula]. -23- 200831972 [Formula 5] η — Lh — ~v^ The maximum brightness of the complex white line in the unlit portion of the Lb, and the maximum brightness of the complex black line in the dark portion. The contrast c of the output image of the comparison lens is evaluated by the comparison unit 7 6 a. . In the face, the lens is evaluated as a comparison of the input image of the chart image with the C1 record body 76d. The CTF measuring unit 76b is based on the input image and the output image C. , Ci, use the following [6th formula] to find CTF. [Form 6] CTF = ^

Ct Here, Ci is the contrast of the input image, C. By outputting the video image, it is determined whether the CTF in the X direction of each of the evaluation areas 61 to 70 and the CTF ° determination unit 76c determine whether or not all of the 1 CTFs obtained by the CTF measuring unit 76b exceed the memory 76d. The determination of whether the record is fixed or not. When the judgment is passed, a set of CTFs of the first movement amount Xb Y1 from the distance to the adjustment position AP is used as the material for re-learning, and is recorded in the re-learning 90. The re-learning database 90 is used to replace the learning database 7 3 and update the neural network 8 4 when the second and subsequent partial deviations are performed.

Ld represents the comparison of the other side of the memory. The reference start position in the Y direction group and the data center adjustment at that time are combined with the number -24- 200831972 number and threshold. In addition, the combination of the learning data of the learning database 7 3 and the re-learning data of the re-learning database 90 may be used for the second time or later to determine the coupling coefficient and threshold of the neural network 84. When the determination is unsuccessful, the first movement amount X1, Y1 and one set of CTFs corresponding to the movement amount are recorded in the memory 78d of the second eccentricity adjustment unit 78. At the same time, the lens search amount specifying unit 77 reads out a predetermined amount of search (the amount of movement from the adjusted position and the adjustment position AP specified by the first movement amount) recorded in the memory 77a. As shown in Fig. 6, the amount of search for the first lens group G1 is within the range of the circle C1 centering on the adjustment position A P . Here, the radius of the circle C1 is set to be 2/3 or less of the radius of the circle C of the adjustment operation range of the first lens group G1. Inside the circle C1, eight search points P1 to P8 are provided. Although these points P 1 to P 8 are arbitrarily determined, in the present embodiment, the corners of the square centered on the adjustment position A P and the center of each side are set. The distance from the adjustment position AP to each side is about 0.7 times the radius of the circle C 1 (1 / / &quot; 2). While the first lens group G1 is being moved, the lens evaluation chart 14 is photographed at each search point, and the CTF measurement by the CTF measuring unit 76b is performed. Thereby, one set of CTFs is calculated for each search point. Each time the first lens group G1 moves, the amount of exploration and the set of C TFs at that time are recorded in the memory 78 d. This amount of exploration corresponds to the amount of movement (eccentricity) of the first lens group G 1 centering on the adjustment position AP. As shown in Fig. 4, the second eccentricity adjusting unit 78 includes an evaluation 値 calculating unit 78a, a quadric surface generating unit 78b, a movement amount specifying unit 78c, and a memory 780d. The evaluation calculation unit 7 8 a calculates an evaluation 値 from one set of CTFs recorded in the memory -25-200831972 7 8d for one search point. In the embodiment of Fig. 6, eight evaluations are obtained. As a method of calculating the evaluation flaw, there are four methods shown below. The first type is the worst C of the smallest CTF among the 1 set of CTFs and is used as the worst 値 calculation method for evaluating 値. The second type calculates the average enthalpy of one set of CTFs and uses it as an average 値 calculation method for evaluating 値. The third type is a method for calculating the difference 値 which is used to evaluate the difference between the CTFs of the evaluation areas 61 to 64 and 66 to 69 in the four corners. The fourth type ^ adds and adds the endowment calculated by the worst 値 calculation method, the average 値 calculation method, and the difference 値 calculation method, and adds this as the weighting calculation method of the evaluation 値. The calculated evaluation 値 and the search amount of the first lens group G 1 are assigned and recorded in the memory 78 d. Here, the worst-case calculation method will be described. For example, assume that the CTF for each of the evaluation areas 61 to 70 is as shown in Fig. 7. In Fig. 7, "□" indicates the CTF for the wide-angle evaluation areas 61 to 65, and "△" indicates the CTF for the telephoto evaluation areas 66 to 70. The evaluation 78 calculation unit 78a sets the CTF in the Y direction of the evaluation area 61 located in the upper right among the wide-angle evaluation areas 61 to 65 as the worst 値 BP, and uses this worst 値 BP as the evaluation 値. Next, a method of calculating the difference 値 will be described. The CTFs in the X direction of the evaluation areas 61 to 64 of the four corners of the wide-angle evaluation area are CTF_1_X_W, CTF_2-X-W, CTF_3_X_W, CTF_4_X-W. Then, using the following equation to calculate the average of all combinations of the absolute 各自 of the respective differences -26- 200831972 CTF - X - W = - (I CTF 丄 X - W - CTF - 2 - X - WI + I CTF - 2 An X-W-CTF-3-X-WI+I CTF-3X-W-CTF-4-X-WI+I CTF-4-X-W-CTF-1 1 X-WI + I CTF丄X-W-CTF-3-1 X-WI + I CTF-2-1-Y-T-CTF-4-1 X-WI) /6 Similarly, for the wide-angle evaluation area of the four corners of the evaluation area 61~64 The CTF in the Y direction and the CTF' in the X direction of the evaluation areas 66 to 69 in the four corners of the evaluation area for the telephoto and the CTF in the Y direction are calculated as the average CTF_Y_W, CTF_X_T, and CTF_Y_T of all combinations of the absolute differences of the differences. ^ Then, as shown in the following equation, the sum of the reciprocals of the averages is taken as the difference 値.

Difference 値 = 1 / CTF - X - W + 1 / CTF - Y - W + 1 / CTF - X - T + 1 / CTF - Y - T In addition, the method for calculating the average 及 and the method for calculating the weight are easy to understand, so Detailed description is omitted. The quadric surface generating unit 78b uses the XYZ coordinates shown in Fig. 8 so that the adjustment position AP and the search points P1 to P8 (see Fig. 6) correspond to the XY 'coordinates, and the evaluation 値 corresponds to the Z coordinate Way to draw. Then, as shown in Fig. 9, a quadric surface 95 is produced as it passes through the vicinity of the plotted point. The adjustment position AP becomes the origin (〇, 〇) of the XY coordinates. The movement amount specifying portion 78c is a point at which the quadric surface 95 is evaluated and becomes the largest. Explore here, using a secondary plan. In the quadric surface 9 5 shown in Fig. 9, it is evaluated that the 値 becomes maximum at the point S P . Therefore, the movement amount specifying unit 7 8 c specifies the X coordinate of the point SP and the γ of the γ coordinate as the second movement amount X2 in the X direction of the first lens group G1 and the second movement -27-200831972 in the Y direction. . The shift controller 87 drives the X-direction shifting portion 51 or the Y-direction shifting portion 52, and moves the first lens group G1 on the lens mounting surface 40a from the adjustment position AP by only X2 and Y2. Four evaluation methods are used in the order determined in advance. Then, using the evaluation 値 calculation method used, eight evaluation 値 were obtained from the eight sets of CTFs sold from the memory 7 6d. Next, the second movement amount X2, Y2 is calculated by the secondary plan method. After the first lens group G1 is moved only by the second movement amounts X2 and Y2, one set of CTFs is measured at this position. Then, a determination is made as to whether or not all of the CTFs in one group become qualified or not. This step is performed while changing the evaluation and calculation method until it is qualified. Since the evaluation 値 calculation method is changed in this way, even if the characteristics change between the zoom lenses 2 1 of the same lot number due to manufacturing errors or the like, for example, it can be appropriately handled. In the case where all of the four types of evaluation 値 calculation methods are unsatisfactory, as shown in Fig. 10, the amount of exploration of the first lens group G1 is made smaller, and then the search is performed. At this time, the amount of exploration of the first lens group G1 is within the range of the circle C 2 . The radius of the circle C2 is 1/2 or less of the radius of the circle C1. Four search points P 9 to P 1 2 are provided inside the circle C2. In the present embodiment, four exploration points P9 to P12 are taken in a square, and the distance from the adjustment position AP to each side is about 0.7 times (1 /, 2) of the radius of the circle C2. The group of CTFs is measured at each of the search points P9 to P12 while moving the first lens group G1 from the adjustment position AP. In the memory 78d', in addition to the amount of exploration above the circle C1 and the group of CTFs at that time, the amount of exploration explored in the circle C2 and the set of CTFs at that time are also recorded. Then, the evaluation 値 calculation unit 78a measures the CTFs of the search points P9 to P12 in the same manner as the search points -28 - 200831972 P1 to P8 in the circle C1. In the case where all of the four evaluation methods are unqualified, the eccentricity cannot be adjusted, and the alarm is notified by the alarm 17. Next, the eccentricity adjustment method of the present invention will be described with reference to the flowcharts of Figs. 1 to 14. As shown in Fig. 1, in the present invention, first, the learning of the neural network 84 is performed in advance, and then the first eccentricity adjusting step is entered. As a result of the first eccentricity adjustment, when the determination is successful, the first movement amount X1, γ1 of the first lens group G1 and the CTF at that time are used as re-learning materials, and are recorded in the re-learning database 90. On the other hand, if the result of the first eccentricity adjustment is unsatisfactory, the second eccentricity adjustment step is performed. Further, the determination of the pass or fail after the first eccentricity adjustment step may be incorporated into the first eccentricity adjustment step. Here, the learning of the neural network 84 will be described with reference to FIG. When the learning start button of the operation panel 16 is pressed, the controller 15 extracts the CAD data one by one from the CAD library 7 1 . The controller 15 calculates a set of CTFs using the lens design application 72 based on each CAD data. The amount of movement of the calculated one set of CTFs and the corresponding first lens group G1 is used as learning material, and is recorded in the learning database 73.

One set of CTFs and movement amounts are read from the learning database 73, and one set of CTFs is input to the neural network 84. Then, the error calculating unit 85 calculates the error of the coupling coefficient and the threshold 神经 of the neural network 84 as described above. The neural network 84 adds the error calculated by the error calculating unit 85 and the combination coefficient and the threshold 至 up to that time, and updates the coupling coefficient and the threshold 値. In this way, the learning materials (the amount of movement and the group of CTF -29-200831972) are read one by one from the learning database 73, and the binding coefficient and the threshold are updated. When all the learning materials have been used, the learning of the neural network 84 is completed, and the first eccentricity adjusting step can be performed. Next, the first eccentricity adjustment step will be described with reference to Fig. 13. When the lens unit 19 is attached to the lens holder 1 1 , the respective circuits of the camera driving unit 13 and the lens unit 19 are connected via the seat portion 18 . When the adjustment start button of the operation panel 16 is pressed, the adjusted lens moving portion 12 drives the shifting portions 51 to 53. Therefore, after the lens holder 50 enters the lens barrel 40, the lens holders are fixed to the both sides of the lens holder. In this state, the X-direction shifting portion 51 and the Y-direction shifting portion 52 operate to move the first lens group G1 to the home position. At this starting position, the optical axis of the first lens group G1 clamped by the lens holder 50 and the center line of the lens 40 are designed to match. After the first lens group G 1 is set to the initial position, the CTF is measured. First, the camera driving section 13 photographs the lens evaluation chart 14 at the telephoto end and the wide-angle end via the lens unit 19, respectively. The captured chart image data is transmitted as an output image to the image evaluation unit 76 of the controller 15. The comparison calculation unit 76a of the image evaluation unit 76 calculates the comparison between 10 in the X direction and 10 in the Y direction from the two kinds of chart image data at the wide angle and the telephoto. The CTF measuring unit 76b compares the contrast Ci of the lens evaluation chart 14 as the input image recorded by the memory 76d with the contrast C of each of the output images calculated by the comparison calculating unit 76a. For each of the evaluation areas 61 to 70, 10 CTFs are obtained in the X direction and the γ direction, respectively. The input layer 8 4 a of the 20 C T F input neural network 8 4 is measured by C T F . Therefore, the first movement amounts X1, Y1 of the first lens group G1 are output from the output layer 8 4 c of the neural network. The shift controller 87 drives the shifting unit -30-200831972 51, 52 to move only the first movement amount X1, Y1. Therefore, the first lens group G 1 is moved from the start position to the adjustment position ΑΡ. After the first lens group G 1 is moved, 20 CTFs are measured. Then, based on the measured CTF, the determination unit 76c determines whether or not the pass is successful. It is acceptable when all 20 CTFs are above the benchmark. When it is qualified, the image evaluation unit 76 records the first movement amount X1, Y1 and the 20 CTFs in the re-learning database 90. When the first eccentricity adjustment step becomes unsatisfactory, the process proceeds to the second eccentricity adjustment step shown in Fig. 14 . In the second eccentricity adjustment step, first, the lens exploration amount specifying unit 77 specifies eight exploration amounts. The shift controller 87 sequentially shifts the first lens group G1 from the adjustment position AP to the search points P1 to P8 at eight points in accordance with each search amount. Then, 20 CTFs are measured at each of the search points P1 to P8. The CTF of each of the search points is recorded together with the amount of exploration in the memory 78d of the second eccentricity adjustment unit 78. The evaluation 値 calculation unit 78a of the second eccentricity adjustment unit 78 uses the first one of the four types of evaluation 値 calculation methods, and calculates one evaluation 之中 from among the 20 CTFs recorded in the memory 78 d. The calculated evaluation 记录 is recorded in the memory 78 d in correspondence with the amount of search of the first lens group G 1 . The evaluation of each exploration point and the calculation of the memory for 7 8 d. After the eight evaluations are calculated, the quadric surface generating unit 78b generates the quadric surface 95 shown in Fig. 9 based on the amount of exploration and the evaluation 那时 at that time. The movement amount specifying portion 78c specifically evaluates the point SP which becomes the largest on the generated quadric surface 95. Then, 'the X of the X coordinate of the point SP. is specified as the second movement amount X2 in the X direction based on the adjustment position ap, and the Y of the Y coordinate of the point SP is specified as the second movement based on the adjustment position AP. The amount Y 2 . The shift controller 87 drives the shifting units 51 and 52 to move the first lens group G1 on the lens mounting surface 40a from the adjustment position AP by only the second movement amounts X2 and Y2. After the first lens group G1 is moved by only the second movement amount X 2 and Y 2 , 20 CTFs are measured. Next, the determination unit 76c determines whether or not the 20 CTFs are qualified. When the determination is made, the second movement amount X 2, Y 2 is converted into the movement amount from the home position, and is used as the re-learning material together with the group CTF, and is recorded in the re-learning database 90. On the other hand, when the determination is unsatisfactory, the first lens group G1 is returned to the adjustment position AP, and the evaluation 値 calculation method is changed. The evaluation/calculation unit 78a calculates one evaluation 値 using 20 CTFs for one search amount read from the memory 78d by using the second evaluation 値 calculation method, and then evaluates the 透镜 and the first lens group. The manner in which the amount of exploration of G 1 corresponds is recorded in the memory 78d. After the eight evaluations are calculated, the quadrangular surface generating unit 78b and the movement amount specifying unit 78c perform the same processing as the first type of evaluation and calculation method, and specify the second movement amount X2 based on the adjustment position AP. Y2. Based on the second movement amounts X2 and Y2, the first lens group G1 is moved from the adjustment position AP, and the CTF measurement by the CTF measuring unit 76b and the passing of the determination unit 7 6 c are performed in the above-described order. No decision. As a result, when the determination is passed, the second movement amount X2 and Y2 obtained by the second evaluation 値 calculation method are converted into the movement amount from the start position, and the data is used as the re-learning data from the first group CTF. And recorded in the re-learning database 90. On the other hand, in the case where the determination is unsatisfactory, the third evaluation amount X calculation method specific second movement amount X2, Y2 is performed in the same manner as the above-described procedure, and the movement of the second lens group G1 and the measurement of the CTF are performed and passed. Whether or not to decide. -32- 200831972 In this way, the above processing is repeated until the qualification is judged. In the case where all of the evaluations and the calculation method are unsatisfactory, the lens search amount specifying unit 7 sets the amount of exploration of the first lens group G1 to a smaller amount of search. The shift controller 87 sequentially shifts the first lens group G1 from the adjustment position A P to the search points P 9 to P 1 2 at four points based on the search amount, and measures the CTF for each search point. The evaluation/calculation unit 718 calculates a single evaluation 从 from the 20 CTFs recorded in the memory 78d using the first evaluation 値 calculation method. Then, the second curved surface generating unit 78b and the movement amount specifying unit 78c specify the third movement amount of the first lens group G1 (moving distance from the adjustment position AP) based on the quadric surface created from the four evaluation points. X 3, Y 3 . The shift controller 87 moves the first lens group G1 from the adjustment position AP on the lens mounting surface 40a based on the third movement amounts X3 and Y3. After the first lens group G1 is moved by the third movement amounts X3 and Y3, 20 CTFs are measured by the CTF measuring unit 76b. Then, the pass/fail determination of the judgment unit 76c is performed. When the determination is made, the third movement amount X3, Y3 is converted into the movement amount from the home position, and is used as the material for re-learning together with the group CTF, and is recorded in the re-learning database 90. On the other hand, when the determination is unsatisfactory, the evaluation 値 calculation unit 78a calculates the evaluation 使用 using the second evaluation 値 calculation method, and performs the same processing as described above. Then, the determination unit 76c performs the pass/fail determination. The evaluation method is calculated until the qualification is judged, and the evaluation and the pass/fail determination are repeated. In the case where all of the four evaluation methods are unqualified, the alarm 7 is called and the operator is notified. -33-200831972 As described above, when the first eccentricity adjustment step is passed, the first movement amount XI, Y1, and 20 CTFs are recorded in the re-learning database 90, and the second eccentricity adjustment step becomes In the case of the qualifier, the amount of movement 'from the start position of the second movement amount X 2 and Y 2 at the time of passing the qualification and the 20 CTFs are used as the material for re-learning and recorded in the re-learning database 90. This re-learning data is used as one of a plurality of re-learning materials in the case of the next eccentricity adjustment, and is input to the neural network 84 for use in the update of the combination coefficient and the threshold. When P becomes acceptable, the lens holder 50 is separated from the first lens group G1, and the lens unit 19 is removed from the lens holder 1 1 . Then, the lens holder 41 is fixedly fixed by an adhesive or a small screw, and the first lens group G1 is fixed to the position where the eccentricity is adjusted. Further, since the outer circumference of the front surface of the first lens group G1 is pressed by a spring-like ring (not shown), the first lens group G1 does not move when the lens unit 19 is removed from the lens holder 11. Further, the first lens group G 1 may be fixed in a state where the lens unit 19 is attached to the lens holder 1 1 . Further, the lens unit notified by the alarm 17 is taken out as a defective product and detached from the lens holder 1 1 .

The U connector "mounts a new lens unit 兀 to the lens holder 1 1 . After the re-learning data is used, the coupling coefficient and the threshold of the neural network 8 are updated, and the first eccentricity adjustment step and the second eccentricity adjustment step are executed as soon as they are qualified. Further, in the present embodiment, although the Back Propagation model using the sigmoid function is applied to the neural network, a Radial Basis function network using a Gaussian function can also be applied. In addition, in the present embodiment, the process of generating a quadric surface and the point of using the secondary plan method to specifically evaluate the point on the quadric surface to become the largest flaw in the quadratic surface are used by the CYBERNET company. MATLAB", but not necessarily limited to this. Next, a second embodiment of the present invention will be described with reference to Figs. 15 to 17 . In the second embodiment, instead of the controller 15 shown in Fig. 1, the controller 100 shown in Fig. 15 is used. The controller 1A includes a past adjustment database 1〇1, a batch number input unit 102, a first eccentricity adjustment unit 103, an image evaluation unit 76, a lens search amount specifying unit 77, and a second eccentric adjustment unit 7 8 ^ And shifting the controller 87. In addition, since the image evaluation unit 76, the lens search amount specifying unit 77, the second eccentricity adjusting unit 78, and the shift controller 87 are the same as in the first embodiment, the same reference numerals are attached. In the past, the database 1 0 1 is adjusted, and as shown in Fig. 16, the information on the eccentricity adjustment of the first lens group G 1 that has been performed in the past is recorded as 1 0 1 a (hereinafter referred to as "past adjustment data"). In the past, the data 1 1 1 a is adjusted, and the past adjustment amount XP in which the first lens group G1 has been adjusted toward the X direction, the past adjustment amount YP adjusted in the Y direction, and the CTF at the adjusted position i are recorded. Similarly to the first embodiment, the CTF is one in the X direction and ten in the Y direction (a total of 20). In addition, in the past, the data was adjusted 1 〇 1 a, including information on the improvement rate of the eccentrically adjusted CTF. This improvement rate is obtained by dividing the eccentrically adjusted CTF by the CTF before the eccentric adjustment. In addition, in the past, the adjustment of the data 1 0 1 a also included information on the date and time of the eccentric adjustment. The past adjustment data 101a is distinguished according to each lot number of the zoom lens 21. Here, "LOTI", "L〇T2", and "LOT3" indicate that the zoom lens 21 is manufactured in a different production line. When the lens unit 19 is set to the eccentricity adjusting device, the batch number of the zoom lens 2 1 is input from the lot number input unit 102 to the first eccentricity adjusting unit 103 of -35-200831972. The first eccentricity adjusting unit 1 0 3 includes a data extracting unit 1 〇 3 a and a predicting unit 1 0 3 b for the first eccentricity adjusting step. The data extracting unit 1 〇 3 a takes out the past adjustment data corresponding to the batch number entered by the batch number input unit 102 from the past adjustment database 1 0 1 . The prediction unit 1 〇 3 b predicts the movement amount X 1 and Y 1 from the home position (hereinafter referred to as "first movement amount") based on the extracted past adjustment data. As a method of calculating the first movement amounts X 1 and Y 1 , any one of the following methods is used: an average method of averaging the first movement amounts XI and Y1 and a CTF improvement in which the past movement amounts XP and YP are averaged. The improvement rate weighting method for weighting the past movement amounts XP and YP as the first movement amounts XI and Y1, and the time when the eccentricity adjustment has been performed in the past and the past movement amounts XP, YP are associated with each other. The method of calculating the amount of movement and the method of combining the improvement rate weighting method and the time correlation method are performed in such a manner that the weighting is larger. The averaging method was carried out according to the following [Formula 7]. Here, 'ΧΡΠ' is the past movement amount in the X direction of the first lens group G1 of the past adjustment data 101a, YP[i] is the past movement amount in the Y direction, and "N" is the adjustment data 1 in the past of each batch number. The number of 1 a, X 1 is the first movement amount in the x direction of the first lens group G 1 , and Y1 is the first movement amount in the Y direction. In addition, '[" is "1" to "N". -36 - 5 200831972 [Type 7] ΣΧΡ\ί]

Xl = ^- Ν ΣΥΡ[ί] Υ1 = ^- Ν Further, the improvement rate weighting method is performed according to [8th formula] below. Here, u[i] is the improvement rate of CTF.

[Equation 8] x\ = iu\i\xp[q, n = iu[i\YP\i] Thus, by performing weighting in accordance with the improvement rate, eccentricity adjustment can be performed with higher precision. Further, the time correlation method is performed according to the following [Form 9]. Here, TN[i] is the current time, and TP[i] is the time when the eccentricity adjustment is performed in the past, and an arbitrary constant is used. [9th formula] NΧΙ = Σ β Ν Π-Σ , · (ΤΝ\ι]-ΤΡ[ι])β XP\il (ΤΝ\ί]-ΤΡ[ί]) -37- 200831972 In addition, regarding TN[ i], ΤΡ [Π pre-replaced into a number of words. In [9th formula], the influence of the shift of the old date and time on the first movement amount X1 and Υ1 becomes small. The past movement amount ΧΡ, ΥΡ of the other β period is larger for the first movement amount XI. Therefore, even in the case of the lens unit 19 at the time of batch number switching, since the appropriate amount of movement can be automatically predicted from the past 5 of the plurality of lens units 19, it is not necessary to specifically refer to another desired material. Further, the improvement rate weighting method and the time correlation method group &lt; are performed according to the following [Form 10]. Here, u[i], ΤΝ[Π, system and improvement rate weighting method or time-off method are the same. [Formula 10] • The momentum of the date XP, i, the effect of the new day Y 1: the qualitative change: the adjustment data: the method of predictive calculation - and TP[1] ΝΧΙ = Σ β Ν 71-Σ ί ( TN[q-TP[Q)β i (TN[i] - TP\i]) u\i]YP\i] The shift controller 87 controls the X direction based on the body X 1 and Y1 predicted by the prediction unit 103b The shifting unit 51 or the Y-direction shifting unit 1 lens group G1 measures the first lens group G 1 at the moved position from the starting position shifting portion 76b on the lens mounting surface 40a. Next, referring to FIG. The flow chart is shown to illustrate the eccentricity adjustment method in the form. When the lens unit 19 is attached to the lens, the camera driving unit 13 and the lens unit 19 set the first lens group G1 to the initial ? 1 movement amount 52 by the shifting portions 51 to 53 via the seat portion 1, and the first step. The CTF of the CTF amount. : Second Embodiment The holders 1 1 8 are connected. position. Further, -38- 200831972, the batch number of the zoom lens 2 1 is input to the first eccentricity adjusting unit 1 〇 3 by the lot number input unit 1 0 2 '. The first eccentricity adjustment unit 1 0 3 extracts the past adjustment data l〇la corresponding to the entered lot number from the past adjustment database 1 〇 i . The prediction unit l3b calculates the data 1 0 1 a based on the past, and uses the averaging method, the improvement rate weighting method, the time correlation method, and the method of combining the improvement rate weighting method and the time correlation method to calculate the start. The position is the first movement amount XI, Y1 of the reference. The shift controller 870 controls the X-direction shifting unit 51 or the Y-direction shifting unit 5 2 based on the first movement amounts X1 and Y1 calculated by the predicting unit 1 〇3 b, and the first lens group G1 on the lens mounting surface 40a. Moves from the starting position to the adjustment position AP. The CTF measuring unit 76b measures the CTF at the adjustment position AP. Then, as described above, the determination unit 76c determines whether or not the pass is successful. When the judgment is passed, the first movement amount X 1 , Y1 and the current CTF are used as past adjustment data, and are recorded in the past adjustment database 1 0 1 . On the other hand, when the determination is unsatisfactory, the first movement amount X 1 , Y 1 and the current C TF are recorded in the memory 78 d of the second eccentricity adjustment unit 78. &amp; When the first eccentricity adjustment step becomes unsatisfactory, the process proceeds to the second eccentricity adjustment step as in the first embodiment. Then, the execution is completed until the qualification is passed, or the evaluation is completed and the calculation method becomes unqualified. Further, in the second embodiment, the batch number for distinguishing the past adjustment data is used as the batch number of the zoom lens 2, but for example, the past adjustment may be made based on the number of the mold constituting the lens or the number of the cavity of the zoom lens. data. In this case, the past adjustment data is stored in advance for each combination of the numbers of the respective constituent lenses, and the amount of movement is predicted. In this way, the prediction accuracy of the movement amount is increased from -39 to 200831972. Further, the past adjustment data may be distinguished based on the number of the automatic assembly machine of the lens unit or the measurement data at the time of assembly. In the state where there is no actual measurement data at the start of lens production, according to the first embodiment, the performance 値 containing the error and the amount of movement at that time are input, and the neural network is learned. After the initial production, the measurement of the CTF is performed, and the eccentricity adjustment of the lens is performed by using the calculation of the movement amount of the neural network, and the neural network is performed using the adjusted movement amount and the performance 那时 at that time. study again. The eccentricity adjustment of the lens of the second embodiment is shifted to the stage where the eccentricity of the storage lens is adjusted and the tendency of each lot number is known to be fixed. In this way, the method of obtaining the amount of movement of the lens can be appropriately changed in response to the acquisition of the measured data, whereby the eccentricity adjustment can be performed in a short time with high precision while maintaining high precision. Although the present invention only uses the front lens group as an object of eccentric adjustment, it can also be applied to a focus lens located at the rear end, and can be applied to an intermediate lens group as long as it is incorporated. Further, in the above-described embodiment, although the eccentric adjustment of the zoom lens of the built-in image sensor is described, the present invention can also be applied to a general zoom lens or a fixed focal length zoom that does not have the image sensor. lens. Further, the present invention is not limited to a camera, and the present invention can be applied to eccentric adjustment of a lens system for an optical device as a whole by a telescope or double glasses. The present invention can be variously modified and modified without departing from the spirit of the invention, and should be construed as being included in the scope of protection of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a view showing the eccentricity adjusting device of the lens of the present invention - 40 - 200831972. Fig. 2 is a perspective view showing a digital camera in a state in which a lens unit is attached to a camera body. Figure 3 is a front elevational view of the lens evaluation chart. Figure 4 is a block diagram of the controller. Figure 5 is an explanatory diagram showing a neural network. Fig. 6 is a view showing a search range of the first lens group G1 in the case where there are eight points of the search point.

Fig. 7 is a graph showing the CTF in the X direction and the CTF in the Y direction measured in each evaluation area. Fig. 8 is a view showing a state in which the adjustment position AP, the search points pi to P8, and their evaluation points are drawn on the three-dimensional coordinates. Fig. 9 is a view showing a state in which a quadratic surface is generated. Fig. 10 is a diagram showing the search range of the first lens group G 1 in the case where there are four points of the search point. Fig. 1 is a flow chart showing the action of the present invention. Figure 12 is a flow chart showing the learning of a neural network. Fig. 13 is a flow chart showing the first eccentricity adjustment step. Fig. 14 is a flow chart showing the second eccentricity adjustment step. Fig. 15 is a block diagram showing a controller according to a second embodiment of the present invention. Figure 16 shows an explanatory diagram of the adjustment data. Fig. 17 is a flow chart showing the operation of the second embodiment of the present invention. [Description of main component symbols] -41 - 200831972 10 11 12 13 14 15 16 17 18 19 21 22 23 24 55 84 Lens eccentric adjustment device Lens holder Adjusted lens moving part Camera drive part Lens evaluation chart controller Operation panel alarm Seat lens unit zoom lens zoom mechanism AF mechanism image area sensor driver neural network-42-

Claims (1)

200831972 X. Patent Application Range: 1. An eccentricity adjustment method for a lens optical system, the lens optical system being composed of a plurality of optical elements including an adjusted lens, wherein the optical lens and the lens are optically The eccentricity adjustment method of the lens optical system includes the following steps: the eccentric adjustment method of the lens optical system includes: the initial moving step is performed on the lens The adjusted lens is moved to the initial position on the surface; the performance 値 calculation step is performed at the initial position, and the lens evaluation chart is taken by the photographic element via the lens optical system, and the performance of the lens optical system is obtained from the photographic image. The first movement amount calculation step is to input the performance 値 to the neural network, and to obtain the first movement amount of the adjusted lens, and the learning of the neural network considers the manufacturing error of the optical elements, and the adjusted a grouping error of the optical elements excluding the lens, and a predicted mounting position of the adjusted lens; and a first moving step Adjusting lens moves first adjustment position, the position location system coupled with the first movement amount to the starting position. 2. The eccentricity adjustment method of the lens optical system according to the first application of the patent scope, wherein the learning of the neural network comprises the following steps: the performance seeking step is to simulate from a plurality of hypothetical CAD data by using a design application program. The lens evaluation chart is used to obtain a plurality of the performance parameters, and the imaginary CAD data has been based on manufacturing errors of the optical elements, the grouping errors of the optical elements except the adjusted lens, and the adjusted lens The corrected CAD data of the CAD data of the design of the lens optics is corrected by predicting the installation position; and the learning step is to obtain the performance 値 obtained by the simulation, and from the starting position to the The amount of movement until the installation position is predicted, the neural network is input, and the neural network is learned. 3. The eccentricity adjustment method of the lens optical system according to claim 1 or 2, further comprising the steps of: the first performance 値 recalculation step is performed at the first adjustment position via the lens optical system The photographic element captures a lens evaluation chart, and the performance 値 at the first adjustment position is obtained from the photographic image; and the first pass/fail determination step is based on the performance obtained in the first performance 値 recalculation step 値It is determined whether the eccentricity adjustment of the lens optical system is acceptable. 4. The eccentricity adjustment method of the lens optical system according to any one of claims 1 to 3, wherein the lens evaluation chart is provided with a plurality of lens evaluation regions, the lens evaluation regions having a horizontal direction evaluation pattern, A performance for evaluating the horizontal direction on the plane orthogonal to the optical axis; and a pattern for the vertical direction evaluation for measuring the vertical direction of the surface orthogonal to the optical axis. 5. The method of eccentricity adjustment of a lens optical system according to claim 4, wherein the lens optical system comprises a zoom lens, and the plurality of lens evaluation regions include a wide-angle evaluation region for evaluating the zoom lens at a wide angle The performance 値; and the telescope evaluation area for evaluating the performance 値 of the zoom lens at the telephoto end, and the wide-angle evaluation area is disposed at four corners and the center of the lens evaluation chart, the telephoto The evaluation area is set at the center of the lens evaluation chart at the four corners -44-200831972 and falls to the center. The @心调整方法 of the lens optical system according to any one of claims 1 to 5, wherein the performance is CTF. 7. The method of adjusting the center of the lens optical system according to any one of claims 3 to 6, further comprising the step of re-adjusting, comprising the step of: τ: the step of exploring, the first pass When the determination step determines that the vacancies are not matched, the adjusted lens is moved to the plurality of first explorations in the first search range centered on the first if adjustment position on the lens mounting surface; the second performance In the 値recalculation step, the imaginary lens evaluation chart is taken at each of the first search points, and the performance 値 is obtained from the photographic images at the first search points; the evaluation 値 calculation step is based on the performance 値, Calculating the evaluation 値 for each of the first search points; the second movement amount calculation step is: determining the second movement amount of the adjusted lens based on the plurality of evaluation ;; and the second movement step of adjusting the lens The process moves to the second adjustment position, which is the position after the second movement amount is added to the first adjustment position. 8. The eccentricity adjustment method of the lens optical system according to claim 7, wherein the re-adjusting step further comprises the following steps: a third performance 値 recalculation step, in the second adjustment position, via the lens optics Taking the photographic element photographing lens evaluation chart 'the performance of the photographic image from the second adjusted position 値; and the second pass/fail determination step based on the performance obtained in the third performance 値 recalculation step値, determine whether the eccentricity of the optical system of the lens is -45-200831972. 9. The eccentricity adjustment method of the lens optical system according to claim 7 or 8, wherein the second movement amount calculation step includes the following steps: a quadric surface formation step using the first adjustment position as an origin The XY axis, and the three-dimensional coordinate with the evaluation 値 as the Z axis, the respective exploration points and their evaluation points are drawn on the three-dimensional coordinates, and the quadratic surface for evaluating the 値 is formed according to the points of the evaluated 値; and the coordinates are calculated. In the step, the XY coordinates corresponding to the point at which the 値 becomes the largest on the quadric surface are obtained as the second movement amount. The eccentricity adjustment method of the lens optical system according to any one of claims 7 to 9, wherein the evaluation method for calculating the evaluation 有 has a plurality of calculation methods, and the predetermined order is obtained as long as the qualification is obtained. select. 1 1 · The method for adjusting the eccentricity of the lens optical system according to claim 10, wherein the re-adjusting step further comprises the following steps: (A) Step, in which the plurality of evaluation methods are all failed. When the second search range is narrower than the first search range, the second search point is determined to be smaller than the first search point; (B) the step is to execute the second search point 2 performance 値 recalculation step, the evaluation 値 calculation step, the second movement amount calculation step, the second movement step, the third performance 値 recalculation step, and the second pass/fail determination step; and (C) In the step, a plurality of evaluation methods are selected in accordance with a predetermined order, and the step (B) is executed as soon as it is qualified. 1 2 · The eccentricity adjustment method of the lens optical system according to the first or the eleventh patent application scope, wherein -46- 200831972 The evaluation calculation method includes at least one of the following: Calculating is the worst performance of the plurality of performance defects 値 and as the evaluation 値; the average 値 calculation method is to calculate the average 値 of the plurality of performance 値 and as the evaluation 値; and the difference 値 calculation method, Calculating the difference 値 used to obtain the balance 该 in the four corners of the wide-angle evaluation area and the evaluation area for the telephoto is used as the evaluation 値; the calculation of the difference , is to calculate the wide-angle evaluation area The difference between the performances of the four corners of the area is averaged, and the sum of the absolute enthalpies of the difference is averaged, and the difference between the performances of the four corners of the evaluation area is calculated. And the average of the absolute 値 of the difference is averaged, and each of them is obtained by adding the reciprocal of the mean 値. The eccentricity adjustment method of the lens optical system according to Item 12 of the patent application, wherein the evaluation 値 calculation method further includes a weight calculation method, the method for calculating the worst 値, the method for calculating the average 値, and The evaluation 算出 calculated by the difference 値 calculation method is assigned a weighting to calculate an evaluation 値. 1 . The eccentricity adjustment method of the lens optical system according to claim 3, 8, or 11 of the patent application, further comprising a first re-learning step, which is determined in the third or second pass or fail determination step When the performance 値 is input to the neural network, and the amount of movement is determined, the difference between the obtained movement amount and the first movement amount or the second movement amount converted to the distance from the home position is The neural network is re-learning. 15. An eccentricity adjusting device for a lens optical system, the lens optical system comprising a plurality of optical elements including an adjusted lens of the package -47-200831972, wherein the device is in the optical system of the lens The eccentricity adjusting device of the lens optical system includes the following members: the eccentricity adjusting device of the lens optical system is moved by moving the lens mounting surface orthogonal to the optical axis, and the eccentricity adjusting device of the lens optical system includes: the lens moving portion is adjusted, and the adjusted lens is held And moving the lens mounting surface; the photographing unit photographs the lens evaluation chart via the lens optical system; and the performance/calculation unit estimates the performance of the lens optical system from the photographic image of the lens evaluation chart. The neural network inputs the performance 从 from the input layer and outputs the amount of movement of the tuned lens from the output layer. The learning of the neural network considers the manufacturing errors of the optical components, and the The assembly error of the optical element and the predicted mounting position of the adjusted lens; the first movement amount calculation unit obtains the performance 値 calculation unit The performance of the neural network input Zhi, and the lens is adjusted to obtain the first movement amount; and a control unit, the lens system is adjusted to the first movement moves only the amount of a manner of controlling the lens moving portion is adjusted. 1 6. The eccentricity adjustment device of the lens optical system according to the fifteenth aspect of the patent application, which further comprises the following components: a performance calculation means, using a design application, simulating the lens evaluation from a plurality of hypothetical CAD data a chart to obtain a plurality of performance 値, and each of the imaginary CAD data has been based on a manufacturing error of the optical elements, a grouping error of the optical elements except the tuned lens, and a predicted installation of the tuned lens Corrected CAD data of the CAD data of the design of the optical system of the lens; and -48-200831972 neural network learning means, the performance 求 obtained by the simulation, and from the starting position to the Predict the amount of movement up to the installation location, enter the neural network, and let the neural network learn. An eccentricity adjustment program for a lens optical system, the lens optical system being composed of a plurality of optical elements including an adjusted lens, the program being oriented by the optical axis of the lens optical system Moving on the lens mounting surface and adjusting the eccentricity of the adjusted lens to the optical axis, the eccentricity adjustment program of the lens optical system causes the computer to perform the following steps: the initial moving step is performed on the lens mounting surface The adjusted lens is moved to the starting position; the performance 値 calculating step is to use the lens optical system to take a lens evaluation chart at the starting position, and obtain the performance of the lens optical system from the photographic image; the first movement amount calculation step Entering the performance 向 to the neural network and determining the first amount of movement of the tuned lens, and the learning of the neural network considers manufacturing errors of the optical elements, and the optical elements except the tuned lens Assembly error and installation error; and a first moving step of moving the adjusted lens to an adjustment position at which the position is added The position after the first movement amount. 1 8. The eccentricity adjustment program of the lens optical system of claim 17 of the patent application, wherein the learning of the neural network comprises the following steps: the performance seeking step is a design application, from a plurality of hypothetical CAD The data is simulated by the lens evaluation chart to obtain a plurality of the performance parameters, and the imaginary CAD data has been based on manufacturing errors of the optical elements, the grouping errors of the optical elements except the adjusted lens, and the Correcting the predicted mounting position of the lens and correcting the corrected CAD data of the CAD data of the design of the lens optics; and the learning steps, the performance obtained by using the simulation, and starting from Yanhai The amount of movement up to the predicted installation location is entered into the neural network and the neural network is learned. 1 9. An eccentricity adjustment method for a lens optical system, the lens optical system being composed of a plurality of optical elements including an adjusted lens, the method being orthogonal to an optical axis of the lens optical system The eccentricity adjustment method of the lens optical system includes the following steps: the movement amount prediction step is based on the inclusion of the eccentric adjustment in the past. Predicting the amount of movement of the adjusted amount of movement of the lens and the plurality of past adjustment data of the performance of the lens optical system corresponding to the past movement amount, predicting the amount of movement of the adjusted lens; and moving the step based on The amount of movement predicted is moved by the adjusted lens. 2 0. The eccentricity adjustment method of the lens optical system according to claim 19, wherein the movement amount is obtained by averaging the plurality of past movement amounts. 2 1. The eccentricity adjustment method of the lens optical system according to claim 19 or 20, wherein the past adjustment data includes an improvement rate of a ratio of the eccentricity adjustment performed in the past that has been improved, the movement amount The past movement amount is obtained by weighting the improvement rate. The eccentricity adjustment method of the lens optical system according to any one of the above-mentioned claims, wherein the past adjustment data includes a date and time of an eccentric adjustment performed in the past, the movement amount being In the past, the amount of movement has been obtained by weighting the date and time. The eccentricity adjustment method of the lens optical system according to any one of claims 19 to 22, wherein the plurality of past adjustment data are classified according to each batch number, and in the movement amount prediction step, according to the correspondence The past adjustment data of the batch number predicts the amount of movement. 24. An eccentricity adjusting device for a lens optical system, the lens optical system comprising a plurality of optical elements including an adjusted lens, the device being orthogonal to an optical axis of the lens optical system The eccentricity adjusting device of the lens optical system includes the following components: the movement amount predicting unit is based on the eccentricity adjustment performed in the past Predicting the amount of movement of the adjusted amount of movement of the lens and the plurality of past adjustment data of the performance of the lens optical system corresponding to the past movement amount, predicting the amount of movement of the adjusted lens; and adjusting the lens moving portion , the adjusted lens moves only the amount of movement. 2 5. An eccentricity adjustment program for a lens optical system, the lens optical system being composed of a plurality of optical elements including an adjusted lens, the program being aligned with the optical axis of the lens optical system Moving on the lens mounting surface, and adjusting the eccentricity of the adjusted lens to the optical axis, the eccentricity adjustment program of the lens optical system causes the computer to perform the following steps: The movement amount prediction step is based on the inclusion that has been performed in the past Deviating the amount of past movement of the amount of movement of the adjusted lens and the plurality of past adjustment data of the performance of the lens optical system corresponding to the past movement amount, predicting the amount of movement of the adjusted lens; and moving the step, The adjusted lens is caused to move only the amount of movement. -51 -