CN101266334A - Wavefront-free detector self-adapting optical system based on random paralleling gradient descent algorithm - Google Patents

Abstract

The invention relates to an adaptive optics system without wavefront detector based on the random paralleling gradient descent algorithm, comprising an imaging sensor, a wavefront corrector and a driving circuit thereof and a random paralleling gradient descent algorithm. The imaging sensor detects the blur image spot or image generated by the distortion wavefront at real time and the control pressure corresponding with the wavefront corrector for making the image spot or image clear is obtained by the random paralleling gradient descent algorithm and applied on the wavefront corrector by the driving circuit, therefore the wavefront corrector generates compensation dosage for removing the wavefront distortion, thus the above operation is repeatedly performed to correct the wavefront error, so as to obtain idea imaging effect. The adaptive optics system based on the random paralleling gradient descent algorithm does need the wavefront detector and influence function of the wavefront corrector, with features of simple structure, compactness, light weight and high efficiency for light energy utilization.

Description

Calm preceding detector self-adapting optical system based on the random paralleling gradient descent algorithm

Technical field

The invention belongs to optical technical field, relate to a kind of ADAPTIVE OPTICS SYSTEMS of novel no Wavefront sensor, relate in particular to a kind of based on detector self-adapting optical system before random paralleling gradient descent algorithm calm.

Background technology

Adaptive optics is the new optical technique that grows up over year surplus in the of past in the world 20.It utilizes the combination of electronic technology and optical technology, by real-time measurement-control-correction, the dynamic optical wavefront error is proofreaied and correct in real time, makes optical system have automatic adaptation change of external conditions, keeps the ability of optimum Working.Adaptive optical technique is a powerful measure of improving the beam quality of the resolution characteristic of imaging system and laser system, it has solved the dynamic disturbance problem that perplexs optics circle centuries, at aspects such as high resolution observation and the transmission of high concentration degree laser energy important application is arranged.

As shown in Figure 1, ADAPTIVE OPTICS SYSTEMS is made up of three parts such as wave front detector, wavefront controller and wave-front corrector usually.The objective fuzzy image that is subjected to dynamic disturbance such as atmospheric turbulence and forms obtains target image clearly after ADAPTIVE OPTICS SYSTEMS is proofreaied and correct.Wave-front corrector adopts distorting lens usually.Distorting lens is the thin mirror surface of the regular some motor drivers of arranging in the back side, and the Beam Wave-Front phase place that reflects from distorting lens is by the face shape decision of distorting lens, and the minute surface face shape of distorting lens is by the control voltage decision of each driver.Therefore just can control the Wave-front phase of incident beam by the control voltage of each driver of controlled deformation mirror.A committed step in the work of ADAPTIVE OPTICS SYSTEMS closed loop is to determine the actuator voltage value of distorting lens according to wavefront measurements, and the process that obtains distorting lens driver control voltage from wavefront sensor measurements is called wave front restoration.The type of wave front restoration method and Wavefront sensor and distorting lens is closely related.

Residual wave front signal after the Wavefront sensor measurement update also obtains wave-front corrector driver control voltage corresponding before the residual wave according to certain recovery algorithm, controls the controlled voltage of computing in control computer.The control voltage signal is applied on each driver of wave-front corrector through overdrive circuit, makes the rectifying plane shape that wave-front corrector produces to be needed, and the incident distorted wavefront is carried out phase correction.Carry out closed loop work so again and again, make the residual wave front signal, on imaging system, obtain picture rich in detail near perfect condition.

Satisfying real-time detection and restoring the wave front detector that requires is one of important component part of ADAPTIVE OPTICS SYSTEMS.The ADAPTIVE OPTICS SYSTEMS of external early stage development adopts dynamic shearing to interfere (Shear Intefermeter) type wavefront sensing technology, in the modern ADAPTIVE OPTICS SYSTEMS based on Shack-Hartmann (Shark-Hartmann) wavefront sensing technology.In the modern ADAPTIVE OPTICS SYSTEMS based on Shack-Hartmann wave front sensor, wave front restoration calculation is finished by the linear matrix computing, can finish real-time height, technology comparative maturity fast with digital signal processor (DSP) technology.But there is a deadly defect in Hartmann sensor, and promptly sub-aperture segmentation causes the efficiency of light energy utilization lower.In addition, Wavefront sensor and imaging system are two separate parts in the ADAPTIVE OPTICS SYSTEMS that is imaged as purpose, the system architecture complexity.These factors will limit in the ADAPTIVE OPTICS SYSTEMS of sub-aperture segmentation type Wavefront sensor under faint beacon condition such as Shack-Hartmann and use.

No Wavefront sensor ADAPTIVE OPTICS SYSTEMS does not need to use independently wave front detector, therefore have simple in structure, volume is little, in light weight, cost is low, efficiency of light energy utilization advantages of higher.

No Wavefront sensor ADAPTIVE OPTICS SYSTEMS based on the random paralleling gradient descent algorithm is directly carried out wavefront correction according to the image that imaging sensor obtains, need not reference picture, need not to demarcate the influence function of wave-front corrector, algorithm is simple and calculated amount is little, and speed is fast.

Summary of the invention

The purpose of this invention is to provide a kind of calm preceding detector self-adapting optical system based on the random paralleling gradient descent algorithm, this system not only can work in the point target situation but also can work in the expansion target conditions, need not wave front detector, do not need to measure the influence function of wave-front corrector yet, have simple in structure, advantages such as volume is little, and is in light weight.

The objective of the invention is to realize by following technical scheme:

A kind of calm preceding detector self-adapting optical system based on the random paralleling gradient descent algorithm of the present invention is made of imaging sensor, wave-front corrector, wave-front corrector driving circuit, random paralleling gradient descent algorithm and computing machine, and wherein imaging sensor is made up of imaging len and CCD; When ADAPTIVE OPTICS SYSTEMS is worked, at first gather a frame figure and pass to computing machine by CCD, by COMPUTER CALCULATION image patch or picture appraisal function, then by random perturbation signal of SPGD algorithm controls computing machine output, disturbing signal is amplified after-applied to wave-front corrector through the wave-front corrector driving circuit, because the control voltage of wave-front corrector changes, its shape changes, become image patch or image on CCD to change after causing incident light wave by imaging len, gathered a frame (data) image again and passed to computing machine by CCD this moment, calculate image patch or picture appraisal function once more by computing machine, the SPGD algorithm is according to the gradient direction of the variation Estimation Optimization of the image patch of twice calculating of computing machine or picture appraisal function, and control computer is with the control voltage of suitable gain along the gradient direction output wave-front corrector of estimating, this voltage amplifies after-applied to wave-front corrector through the wave-front corrector driving circuit, finish iteration one time; Carry out so repeatedly being proofreaied and correct fully, obtain desirable imaging effect until wavefront distortion.

Described system need not wave front detector, does not also need to measure the influence function of wave-front corrector; Not only the point target situation can be worked in but also the expansion target conditions can be worked in.

The random paralleling gradient descent algorithm calculates the evaluation function J of this image patch or image, J=∑ I ², I is the luminous energy of each unit of imaging sensor.J is wave-front corrector control voltage u=[u ₁..., u _j..., u _N] function.

By the random paralleling gradient descent algorithm all N wave-front corrector control voltages are applied random perturbation Δ u then _n=δ _nπ _n, δ wherein _nBe the amplitude of disturbance, π _nBe the direction of disturbance, the changes delta J of image patch that causes thus or picture appraisal function can be calculated by following formula:

ΔJ＝J(u ₁+Δu ₁，u ₂+Δu ₂，...u _N+Δu _N)-J(u ₁，u ₂，...u _N) (1)

According to the Taylor formula:

$\frac{\mathrm{\ΔJ}}{{\mathrm{\Δu}}_n}\≈\frac{\∂J}{{\∂u}_n}+\underset{l\≠n}{\overset{N}{\mathrm{\Σ}}}\frac{\∂J}{{\∂u}_l}\·\frac{{\mathrm{\Δu}}_l}{{\mathrm{\Δu}}_n}---\left(2\right)$

As disturbance Δ u _nAt random and separate, second mathematical expectation in the following formula is 0.When N is very big, g=Δ J/ Δ u _nOne of being real gradient is well approximate.In the SPGD algorithm, because Bei Nuli distribution B (N, 0.5) (N is a wave-front corrector control channel number) is obeyed in disturbance, so above-mentioned condition satisfies.

The amplitude of supposing disturbance is δ, and the direction of disturbance is π, ${\mathrm{\π}}_n^{\left(m\right)}=\±1$ And obey Bei Nuli distribution B (N, 0.5).According to the gradient direction Δ J/ Δ u that estimates, then the voltage of wavefront correction should for

$u_n^{(m+1)}=u_n^{\left(m\right)}-\mathrm{\γ}\frac{\mathrm{\ΔJ}}{\mathrm{\Δ}{u}_n}---\left(3\right)$

If the maximal value that will search for image patch or picture appraisal function, γ should be less than zero, otherwise γ should be greater than zero.Consider the amplitude of disturbance | Δ u _n| the arbitrary control voltage for wave-front corrector all is identical, so formula (3) can be reduced to:

$u_n^{(m+1)}=u_n^{\left(m\right)}-{\mathrm{\γ}}^{\′}{\mathrm{\π}}_n^{\left(m\right)}\mathrm{\ΔJ}---\left(4\right)$

γ '=gamma/delta wherein.For all control channel, write as vector form, then have:

u ^(m+1)＝u ^(m)-γ′π ^(m)ΔJ (5)

A little voltages are loaded into wave-front corrector, then can partly proofread and correct the ripple distortion.Followingly carry out repeatedly, proofreaied and correct fully up to wavefront distortion.

The present invention compared with prior art advantage is:

(1) the present invention just can finish the closed-loop corrected of ADAPTIVE OPTICS SYSTEMS according to the measurement result of imaging sensor, does not need independently Wavefront sensor, has simplified the structure of ADAPTIVE OPTICS SYSTEMS, has reduced the volume and weight of ADAPTIVE OPTICS SYSTEMS.Avoid the shortcoming of Wavefront sensor and the separate beam split of imaging system in the common ADAPTIVE OPTICS SYSTEMS, waste incident light energy simultaneously, improved the efficiency of light energy utilization of ADAPTIVE OPTICS SYSTEMS.

(2) the calm preceding detector self-adapting optical system based on the random paralleling gradient descent algorithm of the present invention not only can work in the point target situation but also can work in the expansion target conditions, and is applied widely, can be widely used in the wavefront correction of multiple optical system.

(3) the present invention changes the Estimation Optimization gradient direction according to the measurement image evaluation function and carries out work, need not beacon or reference picture, need not to measure the influence function of wave-front corrector, uses simple.

(4) the present invention carries out work according to measuring image patch or picture appraisal function Estimation Optimization gradient direction, promptly obtains poor to image patch or picture appraisal function before and after the control voltage disturbance of wave-front corrector.This metering system is the aberration of correcting optical system self simultaneously, and sensor optics system self aberration is inevitably under practical application, so the present invention has advantages of high practicability.

(5) random paralleling gradient descent algorithm computation process of the present invention is subtracting each other of vector and multiplies each other, and algorithm is simple, is fit to very much realize fast with technology such as DSP, FPGA.

Description of drawings

Fig. 1 is the basic structure synoptic diagram of traditional extraterrestrial target imaging self-adaptive optical system;

Fig. 2 is the structural representation based on detector self-adapting optical system before random paralleling gradient descent algorithm calm of the present invention;

Fig. 3 is of the present invention based on the picture appraisal of detector self-adapting optical system before random paralleling gradient descent algorithm calm and the principle of work synoptic diagram of random paralleling gradient descent algorithm.

Embodiment

The invention will be further described below in conjunction with drawings and Examples.

Embodiment

As shown in Figure 2, ADAPTIVE OPTICS SYSTEMS based on the random paralleling gradient descent algorithm is made of imaging sensor 1, wave-front corrector 2, wave-front corrector driving circuit 3, random paralleling gradient descent algorithm 4 and computing machine 5, wherein imaging sensor 1 is made up of imaging len 6 and CCD7, and computing machine 5 is responsible for collection and processing, the realization of SPGD algorithm and the output of distorting lens control signal of image patch or image.

During ADAPTIVE OPTICS SYSTEMS work, at first gather a frame figure and pass to computing machine 5 by CCD7, calculate image patch or picture appraisal function by computing machine 5, then by random perturbation signal of SPGD algorithm controls computing machine 5 outputs, disturbing signal is amplified after-applied to wave-front corrector 2 through wave-front corrector driving circuit 3, because the control voltage of wave-front corrector 2 changes, its shape changes, cause incident light wave to become image patch or image on CCD7 to change by imaging len 6 backs, gathered a frame (data) image again and passed to computing machine 5 by CCD7 this moment, calculate image patch or picture appraisal function once more by computing machine 5, the SPGD algorithm is according to the gradient direction of the variation Estimation Optimization of the image patch of computing machine 5 twice calculating or picture appraisal function, and control computer 5 is with the control voltage of suitable gain along the gradient direction output wave-front corrector of estimating 2, this voltage amplifies after-applied to wave-front corrector 2 through wave-front corrector driving circuit 3, finish iteration one time.Carry out so repeatedly being proofreaied and correct fully until wavefront distortion.

Claims (2)

1, a kind of calm preceding detector self-adapting optical system based on the random paralleling gradient descent algorithm, it is characterized in that: this system is made of imaging sensor (1), wave-front corrector (2), wave-front corrector driving circuit (3), random paralleling gradient descent algorithm (4) and computing machine (5), and wherein imaging sensor (1) is made up of imaging len (6) and CCD (7); When ADAPTIVE OPTICS SYSTEMS is worked, at first gather a frame figure and pass to computing machine (5) by CCD (7), calculate image patch or picture appraisal function by computing machine (5), then by random perturbation signal of SPGD algorithm controls computing machine (5) output, disturbing signal is amplified after-applied to wave-front corrector (2) through wave-front corrector driving circuit (3), because the control voltage of wave-front corrector (2) changes, its shape changes, causing incident light wave to pass through imaging len (6) back becomes image patch or image on CCD (7) to change, gathered a frame (data) image again and passed to computing machine (5) by CCD (7) this moment, calculate image patch or picture appraisal function once more by computing machine (5), the SPGD algorithm is according to the gradient direction of the variation Estimation Optimization of the image patch of computing machine (5) twice calculating or picture appraisal function, and control computer (5) is with the control voltage of suitable gain along the gradient direction output wave-front corrector of estimating (2), this voltage amplifies after-applied to wave-front corrector (2) through wave-front corrector driving circuit (3), finish iteration one time; Carry out so repeatedly being proofreaied and correct fully, obtain desirable imaging effect until wavefront distortion.

2, the calm preceding detector self-adapting optical system based on the random paralleling gradient descent algorithm according to claim 1, it is characterized in that: this system need not wave front detector, does not also need to measure the influence function of wave-front corrector; Not only the point target situation can be worked in but also the expansion target conditions can be worked in.

Images