CN101206294A - Eccentricity regulating method for lens optical system, eccentricity regulating device and program - Google Patents

Abstract

The invention relates to a method, a device and a program for adjusting the eccentricity of an optical lens system. The CTF of a zoom lens (21) is obtained through a lens design application program (72) by eliminating the assembling error of the mounting error of a first lens set and the predicted mounting error of the first lens set from the production error of all optical elements of the zoom lens (21). The CTF is inputted into an input terminal of a nerve network (84), and the movement amount of the first lens set is inputted into the input terminal, so that the nerve network (84) learns. When adjusting the eccentricity, the first lens set is arranged at an initial position for imaging a lens evaluation drawing (14) so as to obtain the CTE. The CTF is inputted into the nerve network (84) so as to obtain a first movement amount. The first lens set is moved from the initial position with the first movement amount.

Description

The eccentric method of adjustment of lens optical system, off-center adjustment apparatus and eccentric adjustment program

Technical field

The present invention relates to eccentric method of adjustment, off-center adjustment apparatus and the eccentric adjustment program of lens optical system.

Background technology

Now, in a lot of video cameras, carry zoom lens.Zoom lens for example are made of preceding group of lens, zoom lens, back group lens, condenser lens etc.By zoom lens are moved at optical axis direction, focal length variations by mobile focusing lens, is focused.

Zoom lens, back group lens, condenser lens are installed on the lens barrel in the roughly consistent mode of its optical axis.On the other hand, preceding group of lens are installed on the lens barrel by spring-like ring etc., when off-centre is adjusted, with the face of light shaft positive cross in can move.Before after the group lens carry out off-centre adjustment, be fixed on the lens barrel with bonding agent etc.

Usually, take lens evaluation map such as exploring degree figure by zoom lens, and meanwhile overcome the spring-like ring, before making the group lens with the face of light shaft positive cross in move, thereby carry out the off-centre adjustment of lens.Then, becoming best position at the image formation state of lens evaluation map is fixed on preceding group of lens on the lens barrel.

; described method is carried out according to operator's visualization, thus in the whether best judgement of exploring degree, need masterful technique, but also there are the following problems: do not produce adaptability in batches; adjust in addition and have individual differences among the result, thereby lack reliability.

In order to solve such problem, proposed not to carry out the method that the optical axis of the off-centre adjustment of lens or optical element is adjusted by staff.For example, open the spy and described the method for using genetic algorithm equiprobability heuristic approach to adjust the optical axis of the light transmission path that constitutes by a plurality of optical elements in the 2002-122785 communique.In addition, open the spy and described the method for using neural network condenser lens to be adjusted to the focusing position of infinity in the flat 7-38798 communique.

, open in the method for adjustment of 2002-122785 communique, need carry out performance evaluation by individuality to genetic algorithm equiprobability heuristic approach by generation, so must repeatedly carry out performance measurement the spy.In addition, in probability is explored, promptly enable to find the scope that does not have eccentric corrected lens position, also be difficult in this scope, find best lens position.Therefore, can't instantaneous high precision adjust.

In addition, open in the method for adjustment of flat 7-38798 communique the spy, make the relation of the focusing position of neural network learning contrast and condenser lens, but the error during shaping error, the lens unit assembling of resin etc. for example of the error when reflection is made in study.In addition, not mentioned application to eccentric method of adjustment.

Therefore, neural network is learnt rightly, be difficult to the off-centre adjustment that high precision is carried out lens.

Summary of the invention

Eccentric method of adjustment, off-center adjustment apparatus and the eccentric adjustment program of the lens optical system that the off-centre that provides high precision at short notice to carry out lens optical system is adjusted are provided.

In order to realize described purpose, other purpose, the eccentric method of adjustment of lens optical system of the present invention comprises initial movable step, performance number calculation step, the first amount of movement calculation step, first and moves step.And, by with the lens installed surface of the light shaft positive cross of lens optical system on move the lens that are adjusted of the lens optical system that constitutes by a plurality of optical parameters, adjust and be adjusted the off-centre of lens with respect to optical axis.In described initial movable step, be adjusted lens and on the lens installed surface, move to initial position.In described performance number calculation step, use described lens optical system capture lens evaluation map, obtain the performance number of described lens optical system according to this photographic images.In the described first amount of movement calculation step,, obtain first amount of movement that is adjusted lens with described performance number neuralward network input.This neural network is considered the rigging error of the foozle of each optical parameter, each optical parameter except that being adjusted lens and describedly is adjusted the prediction installation site of lens and learns.Then, move in the step described first, be adjusted lens move to initial position add the above first amount of movement and obtain first adjust the position.

In the study of described neural network, use design application, according to a plurality of imaginary cad datas, the lens evaluation map is carried out emulation, obtain performance number.Each imaginary cad data is that the correction cad data that the cad data in the design of described lens optical system obtains is revised in the rigging error of foozle, each optical parameter except that being adjusted lens according to each optical parameter and the prediction installation site that is adjusted lens.Then, the performance number that will be obtained by described emulation and the amount of movement neuralward network input from initial position to the prediction installation site make neural network learning.

And then, preferably after described first moves step, comprise first performance number calculation step and first qualification determination step whether again.Again in the calculation step, adjust position first at first performance number, the scioptics optical system is with imaging apparatus capture lens evaluation map, obtains first performance number of adjusting the position according to this photographic images.Then, first whether in the qualification determination step,, judge whether the off-centre adjustment of lens optical system is qualified according to the performance number of obtaining in the calculation step again at first performance number.

Preferred described lens evaluation map has a plurality of lens evaluations district.And then preferably the figure and be used to of horizontal direction evaluation usefulness of each lens evaluation district with performance number of the horizontal direction on the face that is used to estimate with the light shaft positive cross of lens optical system measured the figure of vertical direction evaluation usefulness of the performance number of the vertical direction on the face with described light shaft positive cross.

In preferred implementation of the present invention, described lens optical system comprises zoom lens, and the wide-angle that lens evaluation district comprises the performance number when being used to estimate zoom lens and being positioned at wide end (ワ イ De end) is distinguished with estimating with the looking in the distance of performance number of estimating the district and being used to estimate zoom lens when being positioned at far-end (テ レ end).Wide-angle is looked in the distance with estimating four jiaos and the central authorities that the district is arranged on the central portion of lens evaluation map with estimating four jiaos and the central authorities that the district is arranged on the lens evaluation map.

Preferred described performance number is CTF.And, preferably whether also have after the qualification determination step and adjust operation again described first.This is adjusted again, and operation comprises the exploration step, calculation step, evaluation of estimate calculation procedure, the second amount of movement calculation step, second move step to second performance number again.In described exploration step, whether be judged to be in the qualification determination step when defective first, be adjusted lens and move into place in adjusting a plurality of first in the first exploration scope that the position is the center and explore points making on the lens installed surface with first.At described second performance number again in the calculation step, first explore a some capture lens evaluation map at each, obtain the performance number that each explores point according to this photographic images.In the evaluation of estimate calculation procedure, calculate an evaluation of estimate by each first exploration point.In the described second amount of movement calculation step,, obtain second amount of movement that is adjusted lens according to these a plurality of evaluations of estimate.Then, move in the step described second, make be adjusted lens move to first adjust the position add second amount of movement and obtain second adjust the position.

The preferred described operation of adjusting again also has the 3rd performance number calculation step and second qualification determination step whether again.Again in the calculation step, adjust position second at described the 3rd performance number, the scioptics optical system is with imaging apparatus capture lens evaluation map, obtains second performance number of adjusting the position according to this photographic images.Described second whether in the qualification determination step,, judge whether the off-centre adjustment of lens optical system is qualified according to the performance number of obtaining in the calculation step again at the 3rd performance number.

In addition, preferably prepare the evaluation value calculation method of the described evaluation of estimate of a plurality of calculating, with predetermined select progressively evaluation value calculation method, up to obtaining qualification determination.And the preferred described second amount of movement calculation step comprises quadric surface and forms step and coordinate Calculation step.In quadric surface forms step, with first to adjust the position be the XY axle of initial point, be to describe each in the three-dimensional coordinate of Z axle to explore point and evaluation of estimate thereof with the evaluation of estimate, according to the tracing point of evaluation of estimate, form the quadric surface of evaluation of estimate.Then, in the coordinate Calculation step, will be on this quadric surface evaluation of estimate become the maximum pairing XY coordinate of point and obtain as second amount of movement.

The preferred described operation of adjusting again also comprises following each step: when (A) described a plurality of evaluation value calculation method all is defective, in the second exploration scope than the described first exploration narrow range, determine the second exploration point that number is lacked than the described first exploration point; (B) these second are explored point and carry out described second performance number calculation step, described evaluation of estimate calculation procedure, the described second amount of movement calculation step, described second move step, described the 3rd performance number calculation step, described second qualification determination step whether more again; (C) with a plurality of evaluation value calculation methods of predetermined select progressively, carry out described step B, qualified up to obtaining.

In addition, preferred a plurality of evaluation value calculation method comprise with performance number minimum in a plurality of performance numbers be the worst-case value computing method calculated as evaluation of estimate of worst-case value, mean value calculation method that the mean value of a plurality of performance numbers is calculated as evaluation of estimate, will be used for respect to wide-angle with estimate the district and the difference value that averages out with the performance number of estimating four jiaos the district that distinguishes of looking in the distance as in the difference value computing method of described evaluation of estimate calculating at least one.The calculating wide-angle is poor with the mutual performance number in four jiaos district in the evaluation district, to the absolute value addition of this difference and the averaging of value that obtains, and calculate and to look in the distance with estimating the poor of the mutual performance number in four jiaos district in the district, to the absolute value addition of this difference and the averaging of value that obtains, with the addition reciprocal of the value of each equalization, obtain the difference value of described difference value computing method thus.And then, as evaluation value calculation method, preferably also comprise weight computation method, described weight computation method is to the weighting and calculate evaluation of estimate respectively of the evaluation of estimate that calculated by described worst-case value computing method, described mean value calculation method, described difference value computing method.

And then, preferably described first or second whether be judged to be in the qualification determination step qualified after, have first learning procedure again.First again in the learning procedure, with the input of performance number neuralward network, obtain amount of movement, according to the amount of movement of obtaining and first amount of movement or be transformed to poor apart from second amount of movement of the distance of initial position, neural network is learnt again.

The off-center adjustment apparatus that is adjusted lens of the present invention has the lens of being adjusted moving part, image pickup part, performance number operational part, neural network, the first amount of movement operational part, control part.And, just the lens optical system that constitutes by a plurality of optical parameters be adjusted lens with the lens installed surface of the light shaft positive cross of described lens optical system on move, adjust the described off-centre that is adjusted lens with respect to described optical axis.Be adjusted the lens moving part and keep being adjusted lens, it is moved on the lens installed surface.Image pickup part scioptics optical system capture lens evaluation map.The performance number operational part is obtained the performance number of lens optical system according to the photographic images of lens evaluation map.Neural network is learnt like that by described, from input layer input performance number, is adjusted the amount of movement of lens from output layer output.The first amount of movement operational part will be imported by the performance number neuralward network that the performance number operational part is obtained, and obtain first amount of movement that is adjusted lens.Then, control part control is adjusted the lens moving part, makes to be adjusted only mobile first amount of movement of lens.

And then the off-center adjustment apparatus that preferably is adjusted lens has performance number calculation mechanism and neural network learning mechanism.The performance number calculation mechanism is used design application, according to a plurality of imaginary cad datas, the lens evaluation map is carried out emulation, obtains a plurality of performance numbers.Each imaginary cad data is that the correction cad data that the cad data in the design of lens optical system obtains is revised in the rigging error of foozle, each optical parameter except that being adjusted lens according to each optical parameter, the prediction installation site that is adjusted lens.In addition, performance number that neural network learning mechanism will be obtained by emulation and the amount of movement neuralward network input from initial position to the prediction installation site make neural network learning.

The off-centre adjustment program of lens optical system of the present invention makes computing machine carry out described initial movable step, performance number calculation step, the first amount of movement calculation step, first and moves step.And, by with the lens installed surface of the light shaft positive cross of described lens optical system on move the lens that are adjusted of the lens optical system that constitutes by a plurality of optical parameters, adjust the described off-centre that is adjusted lens with respect to described optical axis.

In addition, the study of the neural network of using in the first amount of movement calculation step comprises uses design application according to a plurality of imaginary cad datas, the lens evaluation map is carried out emulation, thereby obtain the step of performance number.

In other preferred implementations of the present invention, the eccentric method of adjustment of lens optical system comprises amount of movement prediction steps and mobile step.In the amount of movement prediction steps, promptly pass by a plurality of past of performance number of amount of movement and the described lens optical system corresponding with this past amount of movement and adjust data according to comprising the amount of movement that is adjusted lens in the off-centre adjustment of carrying out in the past, prediction is adjusted the amount of movement of lens.In mobile step,, make to be adjusted lens and to move according to the described amount of movement of prediction.

Preferred described amount of movement is obtained by described a plurality of equalizations of amount of movement in the past.In addition, preferred, the described past adjust be included in in the data off-centre of carrying out adjust described in the i.e. raising rate of ratio that improves of performance number, with respect to the past amount of movement, carry out and the corresponding weighting of described raising rate, obtain described amount of movement.And then, adjusted the date that comprises the off-centre adjustment of carrying out in the past in the data in the described past, with respect to the past amount of movement, carry out the weighting corresponding with the described date, obtain described amount of movement.

Preferably, described a plurality of past were adjusted data by dividing in batches, in described amount of movement prediction steps,, predicted described amount of movement according to the past adjustment data corresponding with described batch.

In other preferred implementations of the present invention, the off-center adjustment apparatus of lens optical system comprises the amount of movement prediction section and is adjusted the lens moving part.The amount of movement prediction section is promptly pass by a plurality of past of performance number of amount of movement and the described lens optical system corresponding with this past amount of movement and is adjusted data according to comprising the amount of movement that is adjusted lens in the off-centre adjustment of carrying out in the past, predicts the described amount of movement that is adjusted lens.Being adjusted the lens moving part makes the described lens that are adjusted only move described amount of movement.

In other preferred implementations of the present invention, the off-centre adjustment program of lens optical system makes computing machine carry out amount of movement prediction steps and mobile step.In the amount of movement prediction steps, promptly pass by a plurality of past of performance number of amount of movement and the described lens optical system corresponding and adjust data according to comprising the amount of movement that is adjusted lens in the off-centre adjustment of carrying out in the past, predict the described amount of movement that is adjusted lens with this past amount of movement.In mobile step, make the described lens that are adjusted only move described amount of movement.

According to the present invention, when the setting (initial learn) of neural network, amount of movement that is adjusted lens that use obtains according to the emulation of design application and performance number at this moment so adjust the data that obtain in the operation even without the off-centre in past, also can be provided with easily.In addition, measure the performance number of the initial position that is adjusted lens,, calculate the amount of movement that is adjusted lens this performance number neuralward network input.Thus, can be than shortened the adjustment time by a larger margin in the past.

In the off-centre of having used neural network is adjusted, when performance number does not satisfy reference value, make lens move into place a plurality of exploration points in specialized range from this adjustments position, obtain each and explore the performance number of putting, according to the amount of movement of these performance number calculating optimums.Thus, can improve the adjustment precision.

And, when suitably having carried out eccentric adjustment,, obtain amount of movement with performance number neuralward network input at this moment, according to the amount of movement of obtaining and first amount of movement or be transformed to poor apart from second amount of movement of the distance of initial position, neural network is learnt again.Thus, reflected that the study again of neural network of the feature of actual lens becomes possibility, can further improve the adjustment performance.

In other preferred implementations of the present invention, past amount of movement of obtaining in the off-centre adjustment in past and performance number are at this moment stored as adjusting data in the past.Then, adjust data in the past, calculate the amount of movement that is adjusted lens according to this.Can not measure the performance number that is adjusted lens and obtain this amount of movement, so can shorten the eccentric needed time of adjusting.And then, adjust data in the past and divide by each batch of lens optical system.Usually performance number is similar to have the lens optical system of shared batch, and eccentric adjusted position becomes roughly the same possibility height.Therefore, by adjusting data in the past, can efficiently carry out off-centre adjustment with bulk management.

Description of drawings

With reference to accompanying drawing and read preferred implementation of the present invention, those skilled in the art will understand described purpose and advantage.

Fig. 1 is the skeleton diagram of the off-center adjustment apparatus of expression lens of the present invention.

Fig. 2 is the stereographic map that expression is installed in lens unit the digital camera of the state on the camera body.

Fig. 3 is the front view of expression lens evaluation map.

Fig. 4 is the block diagram of expression controller.

Fig. 5 is the key diagram of expression neural network.

Fig. 6 is that expression has 8 places to explore the curve map of the exploration scope of the first lens combination G1 when putting.

Fig. 7 is the curve map that is illustrated in the CTF of each CTF that estimates the directions X of measuring in the district and Y direction.

Fig. 8 is illustrated in the curve map of having described to adjust position AP on the three-dimensional coordinate, having explored the state of some P1～P8 and the evaluation of estimate relevant with them.

Fig. 9 is the curve map that expression generates quadric state.

Figure 10 is that expression has 4 places to explore the curve map of the exploration scope of the first lens combination G1 when putting.

Figure 11 is the process flow diagram of expression effect of the present invention.

Figure 12 is the process flow diagram of the study of expression neural network.

Figure 13 is the expression first eccentric process flow diagram of adjusting operation.

Figure 14 is the expression second eccentric process flow diagram of adjusting operation.

Figure 15 is the block diagram of the controller of expression embodiment of the present invention 2.

Figure 16 is a key diagram of representing to adjust in the past data.

Figure 17 is the process flow diagram of the effect of expression embodiment of the present invention 2.

Embodiment

As shown in Figure 1, lens off-center adjustment apparatus 10 of the present invention has lens maintenance platform 11, is adjusted lens moving part 12, camera driver portion 13, lens evaluation Figure 14, controller 15, guidance panel 16, alarm 17.Keeping platform 11 to be formed for keeping the eccentric object of adjusting at lens is the department of assembly 18 of lens unit 19.Being adjusted lens moving part 12, camera driver portion 13, controller 15 is arranged on lens and keeps on the platform 11.Also have, controller 15 has neural network 84, but also can be divided into controller and neural network.

Lens unit 19 has lens barrel 40, zoom lens (Zoom lens optical system) 21, zoom mechanism 22, AF mechanism 23, image area sensor 24, the cell controller 26 with image processing part 25, battery (omitting diagram).

As shown in Figure 2, lens unit 19 is by the anti-pawl 41a that bores of the trough of belt that is formed on lens barrel 40 rear ends, and loading and unloading are installed in the front of camera body 30 freely, constitute digital camera 31.Setting operation portion 32, release-push 33, LCD34, zoom button 35 on camera body 30 as display part.And then, also be arranged on (all not shown) such as data recording section, camera control unit and batteries of recording image data in the loading and unloading recording medium freely at camera body 30.

As shown in Figure 1, lens barrel 40 is taken in first～the 4th lens combination G1～G4 that constitutes zoom lens.The first lens combination G1 is preceding group of lens, and the second lens combination G2 is zoom lens, and the 3rd lens combination G3 is back group lens, and the 4th lens combination G4 is a condenser lens.Also have, each lens combination G1～G4 also can be a simple lens.In addition, lens optical system also can be 2 groups, 3 groups, and then focal length also can be fixed.

The first lens combination G1 and the 3rd lens combination G3 are kept by fixed lens support 41,43 respectively.On the other hand, the second lens combination G2 and the 4th lens combination G4 are kept by mobile lens support 42,44 respectively.Mobile lens support 42 is installed on the lens barrel 40 movably by leading screw 46a and guide rod 47.Stepper motor by zoom mechanism 22 makes this leading screw 46a rotation, and thus, the second lens combination G2 moves, and carries out zoom.Mobile lens support 44 is installed on the lens barrel 40 movably by leading screw 46b and guide rod 47.Stepper motor by AF mechanism 23 makes this leading screw 46b rotation, and thus, the 4th lens combination G4 moves, and focuses.

In the installation exercise of zoom lens 21, at first, second～the 4th lens combination G2～G4 is installed successively on lens barrel 40, the first lens combination G1 is installed then.The fixed lens support 41 of the first lens combination G1, can be moved on lens installed surface 40a to the lens installed surface 40a application of force by force application components such as spring-like ring (omitting diagram).After the off-centre adjustment of the first lens combination G1 finished, fixed lens support 41 was fixed on the lens installed surface 40a by bonding agent etc.

Be adjusted lens moving part 12 and have lens holder 50, directions X moving part 51, Y direction moving part 52, Z direction moving part 53, base portion 54, when off-centre is adjusted, the first lens combination G1 is moved.Lens holder 50 can clamped controlling the position and remove between the release position of clamping of fixed lens support 41 and being installed on the Y direction moving part 52 to displacement of fixed lens support 41 both sides.Y direction moving part 52 is supported by directions X moving part 51, and in addition, directions X moving part 51 is supported by Z direction moving part 53.Driver 55 is set in base portion 54.Controller 15 drives directions X moving part 51, Y direction moving part 52, Z direction moving part 53 by this driver 55.

Directions X moving part 51 moves Y direction moving part 52, makes above-below direction (with the paper direction of the figure vertical direction of the first lens combination G1 at lens installed surface 40a by fixed lens support 41 thus.Below be called " directions X ") move.Y direction moving part 52 moves lens holder 50, makes left and right directions (with the paper direction of the figure parallel direction of the first lens combination G1 at lens installed surface 40a thus.Below be called " Y direction ") move.Z direction moving part 53 moves directions X moving part 51, thus the first lens combination G1 is pressed on the lens installed surface 40a.

The cell controller 26 control zoom mechanism 22 of camera driver portion 13 scioptics unit 19, AF mechanism 23, image processing part 25.Image area sensor 24 for example is CCD, is configured in the image planes side of the 4th lens combination G4.When off-centre is adjusted the eccentric state that detects the first lens combination G1 in the operation, use image area sensor 24, when the shooting that lens are estimated Figure 14 (figure image) is carried out the second lens combination G2 (zoom lens) when being positioned at wide end (below be called when wide " (ワ イ De Time) ") and is positioned at far-end (below be called " far time (テ レ Time) ") totally 2 times.Image processing part 25 generates the view data (figure view data) of figure image according to the signal from image area sensor 24.When wide and the figure view data when far away send to controller 15 by camera driver portion 13.Controller 15 calculates CTF according to 2 figure view data.CTF (Contrast Transfer Function: contrast transfer function) be one of the performance number of expression lens performance, the state of the contrast of presentation video.The value of CTF is high more, and the exploring degree is high more.In addition, eccentric adjustment amount and CTF have correlationship, so can know the amount of movement of the off-centre of lens from CTF.Also have, the performance number of lens is not limited to CTF.

Lens are estimated the place ahead that Figure 14 is configured in lens unit 19, by roughly illumination equably of lighting device 56.As shown in Figure 3, lens are estimated looking in the distance that the wide-angle that uses when Figure 14 is provided with the CTF that estimates when wide uses when estimating district 61～65 and estimating CTF when far away and are distinguished 66～70 with estimating.Wide-angle is with estimating district 61～65th, lays respectively at the evaluation district of upper right, the bottom right of lens evaluation Figure 14, upper left, lower-left and central authorities.On the other hand, look in the distance, lay respectively at the evaluation district of upper right, the bottom right of the central portion 14a of lens evaluation Figure 14, upper left, lower-left and central authorities with estimating district 66～70th.Though omit in the drawings, estimate the figure (below be called " directions X evaluation figure ") and being used to of CTF that record in the district 61～70 is used to estimate the directions X of zoom lens 21 at each and estimate the figure of the CTF of Y direction (below be called " Y direction evaluation figure ").It for example is the white and black vertical stripe decorative pattern that forms that alternately vertically disposes that directions X is estimated figure.On the other hand, Y direction evaluation figure for example is the horizontal stripe decorative pattern that white and black alternately landscape configuration forms.In addition, the figure of distinguishing and looking in the distance and also can write down the figure that is used to estimate performance number radially in the district and be used to estimate tangential performance number with evaluation at wide-angle with estimating.

Figure image data extraction wide-angle when wide is with the view data of estimating district 61～65, and the figure image data extraction when far away is looked in the distance with the view data of estimating district 66～70.These 10 evaluation districts comprise directions X respectively and estimate figure and Y direction evaluation figure, so pass through the shooting of image area sensor 24, finally obtain 10 of directions Xs, 20 CTF of 10 totals of Y direction.Therefore, the eccentric state of the first lens combination G1 installation site is represented by 20 CTF as 1 group.

As shown in Figure 4, controller 15 has CAD database 71, lens design application program 72, learns with database 73, the first eccentric adjustment part 75, picture appraisal portion 76, lens exploration amount particular portion 77, the second eccentric adjustment part 78.

The cad data of the optical texture of record a plurality of expressions first～the 4th lens combination G1～G4 in CAD database 71 (radius-of-curvature of each lens combination, lens thickness, refractive index, lens interval etc.).These cad datas have the basic cad data in the design and consider manufacturing, assembling or alignment error and revised basic cad data and a plurality of imaginary cad data that obtains.

Following the obtaining of a plurality of imaginary cad datas.At first, in basic cad data, keep under the state of installation site (initial position) in design in installation site the first lens combination G1, rigging error or the alignment error of foozle, second～the 4th lens combination G2～G4 of first～the 4th lens combination G1～G4 that produces when thinking to make, assemble or install are revised the optical texture of first～the 4th lens combination G1～G4.Below, should be called the first correction cad data by revised basic cad data.

Then, revise in the cad data, the position of the first lens combination G1 is changed to the first prediction installation site from initial position first.Thus, obtain (1-1) imaginary cad data.This (1-1) imaginary cad data will comprise from the distance (directions X and Y direction) of initial position to the first prediction installation site amount of movement as the first lens combination G1.This amount of movement is equivalent to the offset of initial position as benchmark.Equally, revise in the cad data, the position of the first lens combination G1 is changed to N prediction installation site (N is the natural number more than 2) from initial position, obtain (1-N) imaginary cad data first.This (1-N) imaginary cad data comprises the amount of movement of the first lens combination G1 of prediction installation site from initial position to N.

Then, in basic cad data, remain in installation site under the state of initial position the first lens combination G1, according to first rigging error or the alignment error of revising foozle, second～the 4th lens combination G2～G4 of first～the 4th different lens combination G1～G4 of the error of cad data, revise the optical texture of first～the 4th lens combination G1～G4.Below, this revised basic cad data is called M correction cad data (M is the natural number more than 2).

Then, in M revises cad data, the position of the first lens combination G1 is changed to the first prediction installation site from initial position.Thus, obtain (M-1) imaginary cad data.This (M-1) imaginary cad data comprises from the amount of movement of the first lens combination G1 of initial position to the first prediction installation site.Equally, in M revises cad data, the position of the first lens combination G1 is changed to N prediction installation site (N is the natural number more than 2) from initial position, obtain (M-N) imaginary cad data.This (M-N) imaginary cad data comprises the amount of movement of the first lens combination G1 of prediction installation site from initial position to N.

Controller 15 is estimated the figure view data of Figure 14 and is carried out emulation according to a cad data that takes out from CAD database 71 with 72 pairs of lens of lens design application program, calculate 10 of directions Xs, 10 (adding up to 20) CTF of Y direction.Then, the amount of movement of the first lens combination G1 that comprises in 20 CTF of total (1 group of CTF) that will be calculated by lens design application program 72 and the cad data use data as study, records in the study usefulness database 73.

As lens design application program 72, optical design and the evaluation software " ZEMAX (trade name) " that can use the リ of Co., Ltd. one デ イ Application テ ッ Network ス to make.ZEMAX can reproduce the optical element shape that design finishes from CAD application program input object, and the error that can obtain variety of way simply comprises performance number.Also have, the lens design application program then can be utilized various application software if can carry out the application software of optical design and evaluation.

The first eccentric adjustment part 75 has neural network 84 and Error Calculation portion 85.As shown in Figure 5, neural network 84 is used first eccentric the adjustment in the operation, is made of input layer 84a, middle layer 84b, output layer 84c.Input layer 84a is by 20 unit O1[i] (i=1～20) formation.To unit O1[1]～O1[10] input estimates 10 CTF of the corresponding directions X in district 61～70 with each, to unit O1[11]～O1[20] input and each evaluation distinguish 10 CTF of 61～70 corresponding Y directions.Output layer 84c is by 2 unit O3[k] (k=1,2) formation.From O3[1] amount of movement of directions X of the output first lens combination G1, from O3[2] amount of movement of the Y direction of the output first lens combination G1.Offset when the amount of movement of these directions Xs and Y direction is equivalent to initial position as benchmark.Middle layer 84b is by the unit O2[j of stated number] (j be natural number) arbitrarily constitute.Also have, neural network 84 can be made of software or hardware, and this external enwergy is used the product of selling on the various markets.

Each unit O1[i of input layer 84a] use coupling coefficient w21[j] each unit O2[j of [i] and middle layer 84b] coupling.In addition, each unit O2[j of middle layer 84b] use coupling coefficient w32[k] each unit O3[k of [j] and output layer 84c] coupling.Each unit O2[j of middle layer 84b] and each unit O1[i of input layer 84a] relation by following [mathematical expression 1] expression.In addition, each unit O3[k of output layer 84c] and each unit O2[j of middle layer 84b] relation by following [mathematical expression 2] expression.

[mathematical expression 1]

$O2\left[j\right]=\underset{i}{\mathrm{\Σ}}f\left(w21\right[j\left]\right[i]\×O1[i]-\mathrm{\θ}2[j\left]\right)$

[mathematical expression 2]

$O3\left[k\right]=\underset{j}{\mathrm{\Σ}}f\left(w32\right[k\left]\right[j]\×O2[j]-\mathrm{\θ}3[k\left]\right)$

In described mathematical expression, f represents the S function, θ 2[j] each unit O2[j of expression middle layer 84b] threshold value, θ 3[k] each unit O3[k of expression output layer 84c] threshold value.Also have coupling coefficient w21[j] [i], w32[k] [j], θ 2[j], θ 3[k] set at random in advance.

When the setting (initial learn) of neural network 84, read a plurality of study data simultaneously with database 73 from study.Each study group is made of 1 group of CTF and amount of movement.The CTF of 10 directions Xs among this 1 group of CTF is to O1[1]～O1[10] input, the CTF of 10 Y directions is to O1[11]～O1[20] input.Then, from O3[1] amount of movement of the directions X of the first lens combination G1 that obtains by [mathematical expression 1] and [mathematical expression 2] of output, from O3[2] export the amount of movement of the Y direction of the first lens combination G1.

Error Calculation portion 85 is according to the amount of movement O3[k of the first lens combination G1 that will learn to obtain with 1 group of CTF neuralward network 84 input of data] with identical study amount of movement T[k with the first lens combination G1 that comprises in the data], calculate coupling factor w32[k] the error delta w32[k of [j]] [j] and threshold value θ 3[k] and error delta θ 3[k].Error delta w32[k] [j] and Δ θ 3[k] by following [mathematical expression 3] expression.

[mathematical expression 3]

d3[k]＝(T[k]-O3[k]×(1.0-O3[k])/μ)，

Δw32[k][j]＝ε×d3[k]×O2[j]，

Δθ3[k]＝ε×d3[k]

Here, ε, μ are constants arbitrarily.The error delta w32[k that calculates] [j] and coupling coefficient w32[k] [j] addition, upgrade w32[k] [j].In addition, the error delta θ 3[k that calculates] and threshold value θ 3[k] addition, upgrade threshold value θ 3[k].

And then, Error Calculation portion 85 is according to input layer O1[i] input, from middle layer O2[j] output, the coupling coefficient w32[k of renewal] [j], and described d3[k], calculate coupling factor w21[j] the error delta w21[j of [i]] [i] and threshold value θ 2[j] and error delta θ 2[j].Error delta w21[j] [i] and Δ θ 2[j] by following [mathematical expression 4] expression.

[mathematical expression 4]

$d2\left[j\right]=O2\left[j\right]\×(1.0-O2[j\left]\right)\left(\underset{k}{\overset{2}{\mathrm{\Σ}}}d3\right[k\left]w32\right[k\left]\right[j\left]\right)(2.0/\mathrm{\μ}),$

Δw21[j][i]＝ε×d2[j]×O1[i]，

Δθ2[j]＝ε×d2[j]

The error delta w21[j that calculates] [i] and coupling coefficient w21[j] [i] addition, upgrade w21[j] [i].In addition, the error delta θ 2[j that calculates] and threshold value θ 2[j] addition, upgrade θ 2[j].

As mentioned above, use data by each study, carry out coupling coefficient w32[k repeatedly] [j], w21[j] [i] and threshold value θ 3[k], θ 2[j] renewal, be used for the study end of the setting of neural network 84.

First eccentric the adjustment in the operation, if each unit O1[i at input layer 84a] CTF that calculates by CTF determination part 76b of input, then from each unit O3[k of output layer 84c] output with initial position as the amount of movement of the first lens combination G1 of benchmark (below be called " first amount of movement ") X1, Y1.Installation site (initial position) in the design of the mobile controller 87 specific first lens combination G1 and the first amount of movement X1, Y1 addition and the adjustment position AP that obtains.Then, drive directions X moving part 51 or Y direction moving part 52, the first lens combination G1 is moved to adjusting position AP.

Picture appraisal portion 76 has contrast calculating part 76a, CTF determination part 76b, detection unit 76c, storer 76d.At contrast calculating part 76a, by camera driver portion 13 from image processing part 25 input figure view data when wide and the figure view data when far away.The figure image data extraction wide-angle of contrast calculating part 76a leniently the time be with the view data of estimating district 61～65, and then, the figure image data extraction when far away is looked in the distance with the view data of estimating district 66～70.In each evaluation district 61～70, record directions X and estimate figure and Y direction evaluation figure, so contrast calculating part 76a calculates 10 contrasts of directions X, 10 contrasts of Y direction (adding up to 20) according to the view data in 10 evaluation districts that extract.The contrast C of directions X and Y direction is by following [mathematical expression 5] expression.

[mathematical expression 5]

$C=\frac{L_b-L_d}{L_b+L_d}$

Here, L _bThe high-high brightness of a plurality of white lines of expression light, L _dThe high-high brightness of a plurality of black lines of expression dark-part.The contrast of being obtained by contrast calculating part 76a is the contrast C of the output image of lens evaluation map image _oOn the other hand, the contrast C of the input picture of writing lens evaluation map image in storer 76d _i CTF determination part 76b is according to the contrast C of input picture and output image _o, C _i,, obtain CTF by following [mathematical expression 6].

[mathematical expression 6]

$\mathrm{CTF}=\frac{C_o}{C_i}$

Here, C _iBe the contrast of input picture, C _oIt is the contrast of output image.Thus, obtain the CTF of each directions X of estimating district 61～70 and the CTF of Y direction.

Whether the CTF in the detection unit 76c carry out being obtained by CTF determination part 76b 1 group has all surpassed the qualification determination of the certain reference value that writes down among the storer 76d.Be judged to be when qualified, promptly the first amount of movement X1, Y1 and 1 group of CTF at this moment record study again as study again and use the database 90 with data from initial position to the distance of adjusting position AP.When carrying out for the second time later off-centre adjustment, study is used with database 73 with database 90 vicarious learnings again, the coupling coefficient or the threshold value of renewal neural network 84.In addition, also can use study to use data and learn again after the second time and use the data both sides, determine the coupling coefficient or the threshold value of neural network 84 with the study again of database 90 with the study of database 73.

Be judged to be when defective, the first amount of movement X1, Y1 and the 1 group CTF corresponding with this amount of movement recorded the storer 78d of the second eccentric adjustment part 78.Simultaneously, the regulation exploration amount that writes down among the lens exploration amount particular portion 77 readout memory 77a (by initial position and the specific amount of movement of first amount of movement) from adjustment position AP.

The exploration amount of the first lens combination G1 as shown in Figure 6, to adjust in the scope of round C1 that position AP is the center.Here, the radius of justifying C1 is as the adjustment working range of the first lens combination G1 and below 2/3 of radius of predetermined round C.Be provided with 8 in the inside of circle C1 and explore some P1～P8.Explore some P1～P8 for these 8 and set arbitrarily, but in the present embodiment, forming to adjust position AP is the foursquare angle at center and the central authorities of Ge Bian.Distance from adjustment position AP to each limit is about 0.7 times (1/ ) of the radius of circle C1.On one side the first lens combination G1 is moved, each explore point and carry out the shooting of lens evaluation Figure 14 and based on the CTF of CTF determination part 76b measure on one side.Thus, explore point by each and calculate 1 group of CTF.At storer 78d,, write down its exploration amount and 1 group of CTF at this moment by moving of the first lens combination G1.It is the amount of movement (offset) of the first lens combination G1 at center that this exploration amount is equivalent to adjust position AP.

As shown in Figure 4, the second eccentric adjustment part 78 has evaluation of estimate calculating part 78a, quadric surface generating unit 78b, amount of movement particular portion 78c, storer 78d.

Evaluation of estimate calculating part 78a calculates an evaluation of estimate according to exploring 1 a group of CTF that is recorded among the storer 78d with respect to one.In the embodiment of Fig. 6, obtain 8 evaluations of estimate.Method as calculating evaluation of estimate has 4 kinds of methods shown below.

First kind is to be the worst-case value computing method that worst-case value calculates as evaluation of estimate with CTF minimum among 1 group of CTF.Second kind is the mean value calculation method that the mean value of 1 group of CTF is calculated as evaluation of estimate.The third is to be used for the difference value computing method that difference value that evaluation district 61～64 with respect to four jiaos and 66～69 CTF average out calculates as evaluation of estimate.The 4th kind is that the evaluation of estimate that is calculated by worst-case value computing method, mean value calculation method, difference value computing method is weighted and addition, with the weight computation method of this additive value as evaluation of estimate.The evaluation of estimate of calculating is corresponding with the exploration amount of the first lens combination G1, and records among the storer 78d.

Here, the worst-case value computing method are described.For example, Fig. 7 represents to estimate the CTF in district 61～70 with respect to each.Among the figure, " " expression is distinguished 61～65 CTF with respect to wide-angle with estimating, and " △ " expression is with respect to looking in the distance with the CTF that estimates district 66～70.Evaluation of estimate calculating part 78a will be with respect to wide-angle with the CTF that estimates the Y direction that is positioned at upper right evaluation district 61 in the district 61～65 as worst-case value BP, with this worst-case value BP as evaluation of estimate.

Then, the difference value computing method are described.Is CTF_1_X_W, CTF_2_X_W, CTF_3_X_W, CTF_4_X_W with respect to wide-angle with the CTF of the directions X in four jiaos the district 61～64 that estimates the district.And, according to following expression formula, calculate the average CTF_X_W of whole combinations of each poor absolute value.

CTF_X_W＝

(|CTF_1_X_W-CTF_2_X_W|+|CTF_2_X_W-CTF_3_X_W|+

|CTF_3_X_W-CTF_4_X_W|+|CTF_4_X_W-CTF_1_X_W|+

|CTF_1_X_W-CTF_3_X_W|+|CTF_2_Y_T-CTF_4_X_W|)/6

Equally, about with respect to wide-angle with the CTF of the Y direction in four jiaos the district 61～64 that estimates the district, with respect to looking in the distance with the directions X in four jiaos the district 66～69 that estimates the district and the CTF of Y direction, calculate average CTF_Y_W, CTF_X_T, the CTF_Y_T of whole combinations of the absolute value of each difference.Then, shown in following expression formula, the aggregate value of the inverse that each is average is as difference value.

Difference value=1/CTF_X_W+1/CTF_Y_W+1/CTF_X_T+1/CTF_Y_T.

Also have, about mean value calculation method and weight computation method, because understand easily, so omit detailed explanation.

Quadric surface generating unit 78b uses XYZ coordinate shown in Figure 8, makes adjustment position AP and exploration point P1～P8 (with reference to Fig. 6) corresponding with the XY coordinate, makes its evaluation of estimate corresponding with the Z coordinate, and describes.Then, as shown in Figure 9, generate by near the quadric surface 95 the tracing point.Adjust position AP and become XY origin (0,0).

Amount of movement particular portion 78c explores evaluation of estimate and becomes maximum point in quadric surface 95.In this exploration, use quadratic programming.In quadric surface shown in Figure 9 95, become maximum in a SP evaluation of estimate.Therefore, amount of movement particular portion 88c will put the X coordinate of SP and Y to sit target value specific as the second amount of movement Y2 of the second amount of movement X2 of the directions X of the first lens combination G1 and Y direction respectively.Mobile controller 87 drives directions X moving part 51 or Y direction moving part 52, makes the first lens combination G1 from adjusting the only mobile X2 of position AP, Y2 on lens installed surface 40a.

4 kinds of evaluation value calculation methods adopt according to predetermined order.Then, use the evaluation value calculation method that adopts, obtain 8 evaluations of estimate according to 8 groups of CTF that read by storer 76d.Then, by quadratic programming, calculate the second amount of movement X2, Y2.Behind the first lens combination G1 only mobile this second amount of movement X2, Y2, at 1 group of CTF of this position finding.Then, carry out whole CTF in 1 group and whether become the above whether qualification determination of reference value.Change evaluation value calculation method on one side, Yi Bian carry out this operation, qualified up to obtaining.Owing to change like this evaluation value calculation method, so, to each other during characteristic changing, also can tackle rightly at the zoom lens 21 of same batch even because foozle etc. for example.

When whole 4 kinds of evaluation value calculation methods are all defective, as shown in figure 10, reduce the exploration amount of the first lens combination G1 after, explore once more.At this moment, the exploration amount of the first lens combination G1 is in the scope of circle C2.The radius of circle C2 is below 1/2 of radius of circle C1.Be provided with 4 in the inside of circle C2 and explore some P9～P12.In the present embodiment, get 4 and explore some P9～P12 on square, the distance from adjustment position AP to each limit is about 0.7 times (1/ ) of the radius of circle C2.

The first lens combination G1 is moved from adjusting position AP, each explore group that some P9～P12 respectively measure CTF on one side.In storer 78d,, also be recorded in the current exploration amount and the 1 group of CTF of this moment that explore in the round C2 except the exploration amount that is recorded in the last time of exploring in the round C1 and 1 group of CTF at that time.Then, evaluation of estimate calculating part 78a uses and the exploration point P1～P8 identical order of circle in the C1, explores the CTF mensuration of a P9～P12.When whole 4 kinds of evaluation value calculation methods are all defective, can not carry out off-centre adjustment, with alarm 17 notices.

Below, with reference to the process flow diagram of Figure 11～Figure 14, eccentric method of adjustment of the present invention is described.As shown in figure 11, in the present invention, at first under off-line state, carry out the study of neural network 84 in advance, enter the first eccentric operation of adjusting then.The first eccentric result who adjusts is judged to be when qualified, and the first amount of movement X1, the Y1 of the first lens combination G1 and CTF at this moment use data as study again, record study again and use in the database 90.On the other hand, the first eccentric result who adjusts is judged to be when defective, carries out the second eccentric operation of adjusting.Also have, also can with first eccentric adjust after the operation whether the qualification determination group is gone into first eccentric the adjustment in the operation.

Here, with reference to Figure 12, the study of neural network 84 is described.If the study start button of pressing operation panel 16, controller 15 takes out cad data one by one from CAD database 71.

Controller 15 calculates 1 group of CTF according to each cad data with lens design application program 72.With the amount of movement of 1 group of CTF calculating and the first lens combination G1 corresponding with it as study with data recording to study with database 73.

Read 1 group of CTF and amount of movement from study with database 73, with 84 inputs of 1 group of CTF neuralward network.Then, by Error Calculation portion 85, as mentioned above, calculate the coupling coefficient of neural network 84 and the error of threshold value.Error that neural network 84 will be calculated by Error Calculation portion 85 and coupling coefficient and threshold value addition are before this upgraded coupling coefficient and threshold value.Like this, read study one by one with data (group of amount of movement and CTF) with database 73, carry out the renewal of coupling coefficient and threshold value from study.When using global learning to use data, the study of neural network 84 finishes, and can carry out the first eccentric operation of adjusting.

Then, with reference to Figure 13, the first eccentric operation of adjusting is described.If keep platform 11 that lens unit 19 is installed at lens, then each circuit of camera driver portion 13 and lens unit 19 is connected by department of assembly 18.If the adjustment start button of pressing operation panel 16 then is adjusted lens moving part 12 and drives moving part 51～53, after lens holder 50 enters in the lens barrel 40, the both sides of clamping fixed lens support 41.Under this state, directions X moving part 51 and 52 work of Y direction moving part move to initial position with the first lens combination G1.At this initial position, the optical axis of the first lens combination G1 that is clamped by lens holder 50 is consistent with the center line of lens barrel 40 in design.

If the first lens combination G1 is arranged on initial position, then begins CTF and measure.At first, Figure 14 is estimated at far-end and wide end capture lens respectively in camera driver portion 13 scioptics unit 19.The figure view data of taking sends to the picture appraisal portion 76 of controller 15 as output image.The contrast calculating part 76a of picture appraisal portion 76 when wide and 2 figure view data when far away calculate 10 of directions Xs, 10 contrasts of Y direction respectively.CTF determination part 76b is according to the contrast C as lens evaluation Figure 14 of input picture that writes down among the storer 76d _iContrast C with each output image that calculates by contrast calculating part 76a _o, each is estimated district 61～70, obtain 10 CTF respectively at directions X and Y direction.

The input layer 84a input of 20 CTF neuralward networks 84 obtaining by CTF mensuration.Thus, export the first amount of movement X1, the Y1 of the first lens combination G1 from the output layer 84c of neural network.Mobile controller 87 drives moving part 51,52, only moves the first amount of movement X1, Y1.Thus, the first lens combination G1 moves to from initial position and adjusts position AP.After the first lens combination G1 is moved, carry out the mensuration of 20 CTF once more.

Then, according to the CTF that measures, whether carry out qualification determination by detection unit 76c.20 CTF all are reference value when above, are qualified.When qualified, picture appraisal portion 76 records study again with in the database 90 with the first amount of movement X1, Y1 and its 20 CTF.

First eccentric the adjustment in the operation when defective, enter the second eccentric operation of adjusting shown in Figure 14.This second eccentric adjustment in the operation, at first scioptics exploration amount particular portion 77, specific 8 exploration amounts.Mobile controller 87 moves to 8 exploration point P1～P8 from adjusting position AP according to each exploration amount successively with the first lens combination G1.Then, explore some P1～P8, measure 20 CTF respectively at each.Each CTF that explores point records the storer 78d of the second eccentric adjustment part 78 with the exploration amount.

The evaluation of estimate calculating part 78a of the second eccentric adjustment part 78 adopts first kind of evaluation value calculation method in 4 kinds of evaluation value calculation methods, calculates 1 evaluation of estimate according to 20 CTF that write down among the storer 78d.The evaluation of estimate that calculates is corresponding with the exploration amount of the first lens combination G1 and record among the storer 78d.The calculating of evaluation of estimate and undertaken by exploring point to the record of storer 78d.After calculating 8 evaluations of estimate, quadric surface generating unit 78b generates quadric surface 95 shown in Figure 9 according to exploration amount and the evaluation of estimate of this moment.Amount of movement particular portion 78c specific evaluation of estimate on the quadric surface 95 that generates becomes maximum some SP.Then, the X of a SP is sat target value as being that the second amount of movement X2 of directions X of benchmark is specific to adjust position AP, it is specific as the second amount of movement Y2 that with adjustment position AP is benchmark that the Y of a SP is sat target value.Mobile controller 87 drives moving part 51,52, and the first lens combination G1 is only moved the second amount of movement X2, Y2 from adjusting position AP on lens installed surface 40a.

After making the first lens combination G1 only move the second amount of movement X2, Y2, carry out the mensuration of 20 CTF.Then, by detection unit 76c, carry out the whether qualification determination of 20 CTF.Being judged to be when qualified, the second amount of movement X2, Y2 are transformed to behind the amount of movement of initial position, with 1 group of CTF as learning again with data recording to study again with in the database 90.

On the other hand, be judged to be when defective, the first lens combination G1 returned adjust position AP, change evaluation value calculation method then.Evaluation of estimate calculating part 78a is according to second kind of evaluation value calculation method, 20 CTF that use is read from storer 78d with respect to an exploration amount, calculate an evaluation of estimate, make this evaluation of estimate corresponding with the exploration amount of the first lens combination G1 and record among the storer 78d.After calculating 8 evaluations of estimate, quadric surface generating unit 78b and amount of movement particular portion 78c carry out and first kind of processing that evaluation value calculation method is same, and specific is the second amount of movement X2, the Y2 of benchmark to adjust position AP.According to this second amount of movement X2, Y2, from adjusting position AP the first lens combination G1 is moved, carry out based on the CTF mensuration of CTF determination part 76b with based on the whether qualification determination of detection unit 76c with described order.

Its result being judged to be when qualified, will be transformed to behind the amount of movement of initial position with the second amount of movement X2, the Y2 that second kind of evaluation value calculation method obtained, with 1 group of CTF as learning again with data recording to study again with in the database 90.On the other hand, be judged to be when defective, same with described order, by the specific second amount of movement X2 of the third evaluation value calculation method, Y2, carry out the moving of the first lens combination G1, CTF measures, and qualification determination whether.

So repeat above-mentioned processing, up to qualification determination occurring.When if all evaluation value calculation method is all defective, lens exploration amount particular portion 77 is just set the exploration amount of further having dwindled the exploration amount of the first lens combination G1 and having obtained.Mobile controller 87 moves to 4 exploration point P9～P12 from adjusting position AP according to this exploration amount successively with the first lens combination G1, explores some mensuration CTF by each.

Evaluation of estimate calculating part 78a uses first kind of evaluation value calculation method, calculates an evaluation of estimate according to 20 CTF that record by the exploration amount among the storer 78d.Then, by quadric surface generating unit 78b and amount of movement particular portion 78c, according to the quadric surface that generates from 4 evaluations of estimate, the 3rd amount of movement of the specific first lens combination G1 (from adjusting the displacement of position AP) X3, Y3.Mobile controller 87 moves the first lens combination G1 according to the 3rd amount of movement X3, Y3 from adjusting position AP on lens installed surface 40a.

After with the 3rd amount of movement X3, Y3 the first lens combination G1 being moved,, carry out the mensuration of 20 CTF by CTF determination part 76b.Then, carry out whether qualification determination based on detection unit 76c.Be judged to be when qualified, the 3rd amount of movement X3, Y3 be transformed to behind the amount of movement of initial position, with 1 group of CTF as learning again with data recording to study again with in the database 90.

On the other hand, be judged to be when defective, evaluation of estimate calculating part 78a uses second kind of evaluation value calculation method to calculate evaluation of estimate, carries out and described same processing.Then, whether carry out qualification determination by detection unit 76c.The change evaluation value calculation method repeats the calculating of evaluation of estimate and qualification determination whether, up to qualification determination occurring.When whole four kinds of evaluation value calculation methods are all defective, alarm 17 is given the alarm, notify the operator.

As mentioned above, first eccentric the adjustment in the operation when qualified, the first amount of movement X1, Y1 and 20 CTF record study again with in the database 90, in addition, second eccentric the adjustment in the operation when qualified, will be equivalent to when qualified the second amount of movement X2, Y2 from the amount of movement of initial position and this 20 CTF as learning again with data recording to study again with the database 90.This is learnt with data when the off-centre of carrying out next time is adjusted again, and with one of data, 84 inputs of neuralward network utilize in the renewal of coupling coefficient and threshold value as a plurality of study again.

Becoming when qualified, lens holder 50 is backed out from the first lens combination G1, keep platform 11 to take off lens unit 19 from lens.Then, with bonding agent or Screw etc. that fixed lens support 41 is fixing, the first lens combination G1 is fixed on eccentric adjusted position.Also have, the periphery of the front of the first lens combination G1 is pushed by spring-like ring (not shown), so when lens kept platform 11 to take off lens unit 19, the first lens combination G1 did not move.In addition, also can be under the state that lens unit 19 is installed on the lens maintenance platform 11, fixed first lens group G1.In addition, will keep platform 11 to take off as defective products from lens by the lens unit of alarm 17 notices.

Then, keep installing on the platform 11 new lens unit at lens.Utilize study data again, behind the coupling coefficient and threshold value of renewal neural network 84, carry out first eccentric operation, the second eccentric adjustment operation adjusted, qualified up to obtaining.

Also have, in the present embodiment, Application of Neural Network use the back transfer model of S function, but also can use the Radial Basis Function Network that uses Gaussian function.

Also have, in the present embodiment, generate quadric processing, and come by quadratic programming specific on quadric surface evaluation of estimate be " MATLAB " that has used サ イ バ ネ Star ト company to make in the processing of peaked point, but need not be confined to this.

Below, with reference to Figure 15～Figure 17 embodiments of the present invention 2 are described.In this embodiment 2, replace controller 15 shown in Figure 1, use controller 100 shown in Figure 15.Controller 100 has adjusts database 101, batch input part 102, the first eccentric adjustment part 103, picture appraisal portion 76, lens exploration amount particular portion 77, the second eccentric adjustment part 78, mobile controller 87 in the past.Also have, picture appraisal portion 76, lens exploration amount particular portion 77, the second eccentric adjustment part 78, mobile controller 87 are identical with embodiment 1, so mark identical symbol.

Adjust in the past in the database 101, as shown in figure 16, the data 101a that record is adjusted about the off-centre of the first lens combination G1 that carries out in the past (below be called " adjusting data in the past ").Adjust in the past and write down among the data 101a the first lens combination G1 at the adjusted amount of movement XP in the past of directions X with at the adjusted CTF of amount of movement YP, adjusted position in the past of Y direction.CTF and embodiment 1 are same, are 10 of directions Xs, 10 in Y direction (adding up to 20).In addition, adjust the data of the raising rate that comprises eccentric adjusted CTF among the data 101a in the past.This raising rate be with the adjusted CTF of off-centre before divided by the off-centre adjustment CTF and the value that obtains.And then, adjust the data that also comprise the date of having carried out eccentric adjustment among the data 101a in the past.

Adjusting data 101a in the past divides by the batch of zoom lens 21.Here, " LOT1 ", " LOT2 ", " LOT3 " expression shown in Figure 15 are made zoom lens 21 by the production line that differs from one another.If lens unit 19 is set in the lens off-center adjustment apparatus, then the batch of zoom lens 21 is imported to the first eccentric adjustment part 103 from batch input part 102.

The first eccentric adjustment part 103 has 103a of data extract portion and prediction section 103b, uses first eccentric the adjustment in the operation.The 103a of data extract portion adjusted the corresponding past of batch of taking out the database 101 and being imported by batch input part 102 from the past and adjusts data.

Prediction section 103b adjusts data according to the past of taking out, and prediction is from amount of movement X1, the Y1 of initial position (below, be called " first amount of movement ").As the method for calculating the first amount of movement X1, Y1, use following any means: will pass by the value of amount of movement XP, YP equalization as the equalization method of the first amount of movement X1, Y1; Will be proportional and to the value of past amount of movement XP, YP weighting raising rate method of weighting as the first amount of movement X1, Y1 with the raising rate of CTF; Make over time of having carried out eccentric adjustment and in the past amount of movement XP, YP be associated, nearest increase weighting more, thereby calculate the temporal association method of amount of movement; Made up the method for raising rate method of weighting and temporal association method.

The equalization method is carried out according to following [mathematical expression 7].Here, XP[i] be the past amount of movement of directions X of adjusting the first lens combination G1 of data 101a in the past, YP[i] be the past amount of movement of Y direction, " N " is the number of each past in batches adjusting data 101a, X1 is first amount of movement of the directions X of the first lens combination G1, and Y is first amount of movement of Y direction.Also have, " i " is " 1 "～" N ".

[mathematical expression 7]

$\mathrm{<figure-callout\; id="1"\; label="X"\; filenames="S2007101988032E00041.png"\; state="\{\{state\}\}">X</figure-callout>}1=\frac{\underset{i}{\overset{N}{\mathrm{\&Sigma;}}}\mathrm{XP}\left[i\right]}{N},$

$\mathrm{<figure-callout\; id="1"\; label="Y"\; filenames="S2007101988032E00041.png"\; state="\{\{state\}\}">Y</figure-callout>}1=\frac{\underset{i}{\overset{N}{\mathrm{\&Sigma;}}}\mathrm{YP}\left[i\right]}{N}$

In addition, raising rate method of weighting is carried out according to following [mathematical expression 8].Here, u[i] be the raising rate of CTF.

[mathematical expression 8]

$\mathrm{<figure-callout\; id="1"\; label="X"\; filenames="S2007101988032E00041.png"\; state="\{\{state\}\}">X</figure-callout>}1=\underset{i}{\overset{N}{\mathrm{\&Sigma;}}}u\left[i\right]\mathrm{XP}\left[i\right],$

$\mathrm{<figure-callout\; id="1"\; label="Y"\; filenames="S2007101988032E00041.png"\; state="\{\{state\}\}">Y</figure-callout>}1=\underset{i}{\overset{N}{\mathrm{\&Sigma;}}}u\left[i\right]\mathrm{YP}\left[i\right]$

By being weighted according to the raising rate like this, can carry out off-centre adjustment with high precision more.

In addition, temporal association method carries out according to following [mathematical expression 9].Here, TN[i] be now, TP[i] be the time of having carried out eccentric adjustment in the past, β is a constant arbitrarily.

[mathematical expression 9]

$\mathrm{<figure-callout\; id="1"\; label="X"\; filenames="S2007101988032E00041.png"\; state="\{\{state\}\}">X</figure-callout>}1=\underset{i}{\overset{N}{\mathrm{\&Sigma;}}}\frac{\mathrm{\&beta;}}{\left(\mathrm{TN}\right[i]-\mathrm{TP}[i\left]\right)}\mathrm{XP}\left[i\right],$

$\mathrm{<figure-callout\; id="1"\; label="Y"\; filenames="S2007101988032E00041.png"\; state="\{\{state\}\}">Y</figure-callout>}1=\underset{i}{\overset{N}{\mathrm{\&Sigma;}}}\frac{\mathrm{\&beta;}}{\left(\mathrm{TN}\right[i]-\mathrm{TP}[i\left]\right)}\mathrm{YP}\left[i\right]$

Also have, about TN[i] and TP[i], but be replaced into the numerical value of calculation date in advance.In [mathematical expression 9], old past amount of movement XP, YP of date reduces the influence of the first amount of movement X1, Y1.On the other hand, new past amount of movement XP, YP of date increases the influence of the first amount of movement X1, Y1.Thus, when switching in batches,, adjusted data according to the past of after this a plurality of lens units 19 and also can predict suitable amount of movement automatically, so do not need specially to specify the data that are used for prediction and calculation even when the character of lens unit 19 changes.

In addition, the method that has made up raising rate method of weighting and temporal association method is carried out according to following [mathematical expression 10].

[mathematical expression 10]

$\mathrm{<figure-callout\; id="1"\; label="X"\; filenames="S2007101988032E00041.png"\; state="\{\{state\}\}">X</figure-callout>}1=\underset{i}{\overset{N}{\mathrm{\&Sigma;}}}\frac{\mathrm{\&beta;}}{\left(\mathrm{TN}\right[i]-\mathrm{TP}[i\left]\right)}u\left[i\right]\mathrm{XP}\left[i\right],$

$\mathrm{<figure-callout\; id="1"\; label="Y"\; filenames="S2007101988032E00041.png"\; state="\{\{state\}\}">Y</figure-callout>}1=\underset{i}{\overset{N}{\mathrm{\&Sigma;}}}\frac{\mathrm{\&beta;}}{\left(\mathrm{TN}\right[i]-\mathrm{TP}[i\left]\right)}u\left[i\right]\mathrm{YP}\left[i\right]$

Mobile controller 87 is according to the first amount of movement X1, Y1 by prediction section 103b prediction, and control directions X moving part 51 or Y direction moving part 52 make the first lens combination G1 move from initial position on lens installed surface 40a.CTF determination part 76b measures the CTF of the first lens combination G1 of the position after moving.

Then, with reference to process flow diagram shown in Figure 17, the eccentric method of adjustment of embodiment 2 is described.If keep platform 11 that lens unit 19 is installed at lens, then camera driver portion 13 is connected by department of assembly 18 with lens unit 19.By moving part 51～53, the first lens combination G1 is arranged on initial position.In addition, by batch input part 102, the batch of zoom lens 21 is to first eccentric adjustment part 103 inputs.

The first eccentric adjustment part 103 was adjusted from the past and is taken out the past adjustment data 101a corresponding with the batch of input the database 101.Prediction section 103b adjusted data 101a according to the past, by equalization method, raising rate method of weighting, temporal association method, made up any one method in the method for raising rate method of weighting and temporal association method, calculating with the initial position is the first amount of movement X1, the Y1 of benchmark.

Mobile controller 87 is according to the first amount of movement X1, the Y1 that are calculated by prediction section 103b, and control directions X moving part 51 or Y direction moving part 52 make the first lens combination G1 move to from initial position on lens installed surface 40a and adjust position AP.CTF determination part 76b is adjusting position AP mensuration CTF.Then, as mentioned above, whether carry out qualification determination by detection unit 76c.

Be judged to be when qualified, with the first amount of movement X1, Y1 and CTF at this moment as adjusting data recording in the past to adjusting database 101 in the past.On the other hand, be judged to be when defective, the first amount of movement X1, Y1 and CTF at this moment recorded the storer 78d of the second eccentric adjustment part 78.

First eccentric the adjustment in the operation when defective, enter second eccentric the adjust operation identical with embodiment 1.Then, carry out, qualified up to obtaining, perhaps defective up to whole evaluation value calculation methods.

Also have, in embodiment 2, distinguishing what adjust data in the past is the batch of zoom lens 21 in batches, but for example also can adjust data in the past with the numbering of the mould that respectively constitutes lens of zoom lens or the differentiations such as numbering in chamber.At this moment, put aside over by the combination that constitutes the numbering of lens about each and adjust data, the prediction amount of movement.Thus, the precision of prediction of amount of movement improves.In addition, data are adjusted in differentiation such as determination data in the past in the time of also can be with the numbering of the automatic assembling machine of lens unit or assembling.

Do not have under the state of measured data as the beginning of lens production the time, according to described embodiment 1, input contains the performance number of error and amount of movement at this moment, carries out the study of neural network.Produce just begun after, the mensuration by CTF and used the calculating of the amount of movement of neural network is carried out the off-centre adjustment of lens, uses adjusted amount of movement and performance number at this moment, and neural network is learnt again.The measured data that the off-centre of savings lens is adjusted by the stage that can be observed certain tendency in batches, is transferred to the off-centre adjustment of the lens of embodiment 2.So according to the situation that obtains of measured data, the method for the amount of movement of lens is obtained in suitable change, thus, can keep high precision, can carry out off-centre adjustment simultaneously at short notice.

The present invention is only with the preceding group of object of adjusting as off-centre, but also can use the condenser lens that is positioned at the rear end, and if then in assembling, then also can use the lens combination of centre.

In addition, in described embodiment, the off-centre adjustment of the zoom lens that are built-in with imageing sensor has been described, but also can have used the present invention the common zoom lens that do not have imageing sensor, the zoom lens of fixed focal length.In addition, be not limited to video camera, in the off-centre adjustment of the employed lens combinations of all optical instruments such as telescope or binoculars, also can use the present invention.

The present invention can carry out various distortion, change in the scope that does not break away from invention spirit, at this moment also should be interpreted as being included in protection scope of the present invention.

Claims (25)

1. the eccentric method of adjustment of a lens optical system, it is by the eccentric method of adjustment that comprises the lens optical system that a plurality of optical parameters of being adjusted lens constitute, it is characterized in that,

By with the lens installed surface of the light shaft positive cross of described lens optical system on move the described lens that are adjusted, adjust the described off-centre that is adjusted lens with respect to described optical axis, the eccentric method of adjustment of described lens optical system comprises:

The initial movable step, it moves the described lens that are adjusted to initial position on described lens installed surface;

The performance number calculation step, it by described lens optical system, with imaging apparatus capture lens evaluation map, obtains the performance number of described lens optical system at described initial position according to this photographic images;

The first amount of movement calculation step, it imports described performance number neuralward network, obtain described first amount of movement that is adjusted lens, wherein, described neural network consider the foozle of described each optical parameter, except that the described rigging error that is adjusted described each optical parameter the lens with describedly be adjusted the prediction installation site of lens and learn;

First moves step, its with described be adjusted lens move to described initial position add the above first amount of movement and obtain first adjust the position.

2. the eccentric method of adjustment of lens optical system according to claim 1 is characterized in that,

The study of described neural network comprises following steps:

Use design application, according to a plurality of imaginary cad datas, described lens evaluation map is carried out emulation, thereby obtain a plurality of described performance numbers, wherein, described each imaginary cad data is the foozle according to described each optical parameter, the correction cad data that obtains except that the described cad data that is adjusted in the rigging error of described each optical parameter the lens and the design that described lens optical system is revised in the described prediction installation site that is adjusted lens;

The performance number that to be obtained by described emulation and import to described neural network to the amount of movement of described prediction installation site from described initial position is learnt described neural network.

3. the eccentric method of adjustment of lens optical system according to claim 1 and 2 is characterized in that, also comprises:

First performance number is calculation step again, and it adjusts the position described first, by described lens optical system, takes described lens evaluation map with imaging apparatus, obtains the described performance number of the described first adjustment position according to this photographic images;

The first qualification determination step whether, it judges according to the described performance number of obtaining in the calculation step again at described first performance number whether the off-centre adjustment of described lens optical system is qualified.

4. according to the eccentric method of adjustment of any described lens optical system in the claim 1～3, it is characterized in that,

Described lens evaluation map has a plurality of lens evaluations district, and the figure and being used to of horizontal direction evaluation usefulness that described each lens evaluation district has a performance number of the horizontal direction on the face that is used to estimate with described light shaft positive cross is measured the figure of vertical direction evaluation usefulness of the performance number of the vertical direction on the face with described light shaft positive cross.

5. the eccentric method of adjustment of lens optical system according to claim 4 is characterized in that,

Described lens optical system comprises zoom lens, looking in the distance of the described performance number of the wide-angle that described a plurality of lens evaluations district comprises the described performance number when being used to estimate described zoom lens and being positioned at wide end when estimating district and described zoom lens and be positioned at far-end with estimating the district, described wide-angle is with estimating four jiaos and the central authorities that the district is arranged on described lens evaluation map, described looking in the distance with estimating four jiaos and the central authorities that the district is arranged on the central portion of described lens evaluation map.

6. according to the eccentric method of adjustment of any described lens optical system in the claim 1～5, it is characterized in that,

Described performance number is CTF.

7. according to the eccentric method of adjustment of any described lens optical system in the claim 3～6, it is characterized in that,

Also comprise and adjust operation again, the described operation of adjusting again has following steps:

Explore step, whether it is judged to be in qualification determination step when defective described first, makes the described lens that are adjusted move into place in adjusting a plurality of first in the first exploration scope that the position is the center with described first and explore points on the described lens installed surface;

Second performance number is calculation step again, and it respectively first explores the described lens evaluation map of some shooting described, obtains the described performance number of described each first exploration point according to this photographic images;

The evaluation of estimate calculation procedure, it calculates an evaluation of estimate according to described each performance number by described each first exploration point;

The second amount of movement calculation step, it obtains described second amount of movement that is adjusted lens according to described a plurality of evaluations of estimate;

Second moves step, its make described be adjusted lens move to described first adjust the position add the above second amount of movement and obtain second adjust the position.

8. the eccentric method of adjustment of lens optical system according to claim 7 is characterized in that,

The described operation of adjusting again also has following step:

The 3rd performance number is calculation step again, and it adjusts the position described second, by described lens optical system, takes described lens evaluation map with described imaging apparatus, obtains the performance number of the described second adjustment position according to this photographic images;

The second qualification determination step whether, it judges according to the described performance number of obtaining in the calculation step again at described the 3rd performance number whether the off-centre adjustment of described lens optical system is qualified.

9. according to the eccentric method of adjustment of claim 7 or 8 described lens optical systems, it is characterized in that,

The described second amount of movement calculation step comprises following step:

Quadric surface forms step, and it uses with first and adjusts the position as the XY axle of initial point, with the three-dimensional coordinate of evaluation of estimate as the Z axle, in this three-dimensional coordinate, describe described each explore point and evaluation of estimate thereof, according to the tracing point of evaluation of estimate, the quadric surface of formation evaluation of estimate;

The coordinate Calculation step, it will be obtained as described second amount of movement by the pairing XY coordinate of the point of evaluation of estimate maximum on described quadric surface.

10. according to the eccentric method of adjustment of any described lens optical system in the claim 7～9, it is characterized in that,

The evaluation value calculation method that calculates described evaluation of estimate has a plurality of, and is with predetermined select progressively, described qualified up to obtaining.

11. the eccentric method of adjustment of lens optical system according to claim 10 is characterized in that,

The described operation of adjusting again also comprises:

When (A) described a plurality of evaluation value calculation method all is defective, in the second exploration scope, determine the second exploration point that number is lacked than the described first exploration point than the described first exploration narrow range;

(B) these second are explored point and carry out described second performance number calculation step, described evaluation of estimate calculation procedure, the described second amount of movement calculation step, described second move step, described the 3rd performance number calculation step, described second qualification determination step whether more again;

(C) with a plurality of evaluation value calculation methods of predetermined select progressively, carry out described step B, qualified up to obtaining.

12. the eccentric method of adjustment according to claim 10 or 11 described lens optical systems is characterized in that,

Described evaluation value calculation method comprise with performance number minimum in described a plurality of performance numbers be the worst-case value computing method calculated as described evaluation of estimate of worst-case value, will described a plurality of performance numbers the mean value calculation method calculated as described evaluation of estimate of mean value, will be used for respect to described wide-angle with estimate the district and difference value that the described performance number in four jiaos the district distinguishing with evaluation of looking in the distance averages out as in the difference value computing method of described evaluation of estimate calculating at least one;

Calculate described wide-angle with estimating the poor of the mutual described performance number in four jiaos district in the district, to the absolute value addition of this difference and the averaging of value that obtains, and calculate described looking in the distance with estimating the poor of the mutual described performance number in four jiaos district in the district, to the absolute value addition of this difference and the averaging of value that obtains, with the addition reciprocal of the value of each equalization, obtain described difference value thus.

13. the eccentric method of adjustment of lens optical system according to claim 12 is characterized in that,

Described evaluation value calculation method also comprises weight computation method, and described weight computation method is to the weighting and calculate evaluation of estimate respectively of the evaluation of estimate that calculated by described worst-case value computing method, described mean value calculation method, described difference value computing method.

14. eccentric method of adjustment according to claim 3 or 8 or 11 described lens optical systems, it is characterized in that, also comprise first learning procedure again, described first again learning procedure whether be judged to be in the qualification determination step when qualified described first or second, described performance number is obtained amount of movement to described neural network input, according to amount of movement of obtaining and described first amount of movement or be transformed to poor apart from second amount of movement of the distance of initial position, described neural network is learnt again.

15. the off-center adjustment apparatus of a lens optical system, it is by the off-center adjustment apparatus that comprises the lens optical system that a plurality of optical parameters of being adjusted lens constitute, it is characterized in that,

By with the lens installed surface of the light shaft positive cross of described lens optical system on move the described lens that are adjusted, adjust the described off-centre that is adjusted lens with respect to described optical axis, the off-center adjustment apparatus of described lens optical system comprises:

Be adjusted the lens moving part, it keeps the described lens that are adjusted, and the described lens that are adjusted are moved;

Image pickup part, it is by described lens optical system capture lens evaluation map;

The performance number operational part, its photographic images according to described lens evaluation map is obtained the performance number of described lens optical system;

Neural network, it imports described performance number from input layer, from the described amount of movement that is adjusted lens of output layer output, wherein, described neural network consider the foozle of described each optical parameter, except that the described rigging error that is adjusted described each optical parameter the lens, describedly be adjusted the prediction installation site of lens and learn;

The first amount of movement operational part, it will be imported to described neural network by the described performance number that described performance number operational part is obtained, and obtain described first amount of movement that is adjusted lens;

Control part, it controls the described lens moving part that is adjusted, and makes the described lens that are adjusted only move described first amount of movement.

16. the off-center adjustment apparatus of lens optical system according to claim 15 is characterized in that, also comprises:

The performance number calculation mechanism, it uses design application, according to a plurality of imaginary cad datas, described lens evaluation map is carried out emulation, obtain a plurality of described performance numbers, wherein, described each imaginary cad data be according to described each optical parameter foozle, revise the correction cad data that the cad data in the design of described lens optical system obtains except that the described rigging error of described each optical parameter the lens, the described prediction installation site that is adjusted lens of being adjusted;

Neural network learning mechanism, the performance number that it will be obtained by described emulation and import to described neural network to the amount of movement of described prediction installation site from described initial position is learnt described neural network.

17. the off-centre of a lens optical system is adjusted program, it is to adjust program by the off-centre that comprises the lens optical system that a plurality of optical parameters of being adjusted lens constitute, it is characterized in that,

By with the lens installed surface of the light shaft positive cross of described lens optical system on move the described lens that are adjusted, adjust the described off-centre that is adjusted lens with respect to described optical axis, the off-centre adjustment program of described lens optical system makes computing machine carry out following steps:

The initial movable step, it makes the described lens that are adjusted move to initial position on described lens installed surface;

The performance number calculation step, it uses described lens optical system capture lens evaluation map at described initial position, obtains the performance number of described lens optical system according to this photographic images;

The first amount of movement calculation step, it imports described performance number neuralward network, obtain described first amount of movement that is adjusted lens, wherein, described neural network consider described each optical parameter foozle, learn except that the described rigging error that is adjusted described each optical parameter the lens and alignment error;

First moves step, and it makes the described lens that are adjusted move at described initial position and add the above first amount of movement and the adjustment position that obtains.

18. the off-centre of lens optical system according to claim 17 is adjusted program, it is characterized in that,

The study of described neural network comprises following steps:

Use design application, according to a plurality of imaginary cad datas, described lens evaluation map is carried out emulation, thereby obtain a plurality of described performance numbers, wherein, described each imaginary cad data is the foozle according to described each optical parameter, the correction cad data that obtains except that the described cad data that is adjusted in the rigging error of described each optical parameter the lens and the design that described lens optical system is revised in the described prediction installation site that is adjusted lens;

The performance number that to be obtained by described emulation and import to described neural network to the amount of movement of described prediction installation site from described initial position is learnt described neural network.

19. the eccentric method of adjustment of a lens optical system, it is by the eccentric method of adjustment that comprises the lens optical system that a plurality of optical parameters of being adjusted lens constitute, it is characterized in that,

By with the lens installed surface of the light shaft positive cross of described lens optical system on move the described lens that are adjusted, adjust the described off-centre that is adjusted lens with respect to described optical axis, the eccentric method of adjustment of described lens optical system comprises:

The amount of movement prediction steps, it is promptly pass by a plurality of past of performance number of amount of movement and the described lens optical system corresponding with this past amount of movement and adjusts data according to comprising the described amount of movement that is adjusted lens in the off-centre adjustment of carrying out in the past, predicts the described amount of movement that is adjusted lens;

Mobile step, it moves the described lens that are adjusted according to the described amount of movement of prediction.

20. the eccentric method of adjustment of lens optical system according to claim 19 is characterized in that,

By described a plurality of equalizations of amount of movement in the past, obtain described amount of movement.

21. the eccentric method of adjustment according to claim 19 or 20 described lens optical systems is characterized in that,

The described past is adjusted packet and contains the i.e. raising rate of ratio that performance number improves described in the off-centre adjustment of carrying out in the past, carries out and the corresponding weighting of described raising rate with respect to described amount of movement in the past, obtains described amount of movement.

22. the eccentric method of adjustment according to any described lens optical system in the claim 19～21 is characterized in that,

The described past is adjusted the date that data comprise the off-centre adjustment of carrying out in the past, carries out the weighting corresponding with the described date with respect to described amount of movement in the past, obtains described amount of movement.

23. the eccentric method of adjustment according to any described lens optical system in the claim 19～22 is characterized in that,

Described a plurality of past is adjusted data by dividing in batches, in described amount of movement prediction steps, adjusts data according to the described past corresponding with described batch, predicts described amount of movement.

24. the off-center adjustment apparatus of a lens optical system, it is by the off-center adjustment apparatus that comprises the lens optical system that a plurality of optical parameters of being adjusted lens constitute, it is characterized in that,

By with the lens installed surface of the light shaft positive cross of described lens optical system on move the described lens that are adjusted, adjust the described off-centre that is adjusted lens with respect to described optical axis, the off-center adjustment apparatus of described lens optical system comprises:

The amount of movement prediction section, it is promptly pass by a plurality of past of performance number of amount of movement and the described lens optical system corresponding with this past amount of movement and adjusts data according to comprising the described amount of movement that is adjusted lens in the off-centre adjustment of carrying out in the past, predicts the described amount of movement that is adjusted lens;

Be adjusted the lens moving part, it makes the described lens that are adjusted only move described amount of movement.

25. the off-centre of a lens optical system is adjusted program, it is to adjust program by the off-centre that comprises the lens optical system that a plurality of optical parameters of being adjusted lens constitute, it is characterized in that,

By with the lens installed surface of the light shaft positive cross of described lens optical system on move the described lens that are adjusted, adjust the described off-centre that is adjusted lens with respect to described optical axis, the off-centre adjustment program of described lens optical system makes computing machine carry out following steps:

The amount of movement prediction steps, it is promptly pass by a plurality of past of performance number of amount of movement and the described lens optical system corresponding with this past amount of movement and adjusts data according to comprising the described amount of movement that is adjusted lens in the off-centre adjustment of carrying out in the past, predicts the described amount of movement that is adjusted lens;

Mobile step, it makes the described lens that are adjusted only move described amount of movement.

Images