QQ客服 官方微博 反馈留言

题名	提高液晶自适应系统校正效果的研究
作者	张杏云
学位类别	博士
答辩日期	2015-05
授予单位	中国科学院大学
导师	宣丽
关键词	液晶自适应光学系统 数据并行处理 响应矩阵 系统辨识 开环自适应控制
其他题名	Research on Improving the performance of Liquid Crystal Adaptive Optics Systems
学位专业	光学工程
中文摘要	虽然自适应光学技术在大口径望远镜系统中得到了广泛的应用，但由于传统的变形镜波前校正器驱动单元数少的限制，目前大都工作在红外波段。然而在可见光波段，望远镜具有更高的分辨率，大多数观测目标也具有更高的辐射能量，因此具有高驱动单元密度、能够满足可见光波段大口径望远镜需求的新型波前校正器是下一代自适应光学系统的关键。而这正是液晶波前校正器相对于传统的变形镜波前校正器的优势。 尽管前人已经解决了液晶波前校正器的响应速度慢、偏振能量损失以及宽波段校正色散等问题，然而到目前为止依然存在着阻碍液晶自适应光学系统实际应用的以下问题：1）数据处理计算量大导致计算延迟时间长。液晶校正器具有几十倍于变形镜驱动器的像素数，这使得波前重构的计算量巨大，导致毫秒级的时间延迟，限制了液晶自适应光学系统的校正带宽。2）响应矩阵测量精度低。系统中的静态像差及随机噪声与干扰会显著降低响应矩阵的测量精度，进而影响液晶自适应光学系统的校正效果。3）开环校正精度较低。为了减少能量损失，液晶自适应光学系统采用开环校正，但在开环系统中探测不到校正后的残差，无法实施控制，导致校正精度较低。 本文针对液晶自适应光学系统数据处理计算延迟时间长的问题，提出了基于多核CPU和多GPU的并行数据处理方法。液晶自适应光学系统的数据处理主要包括光斑阵列质心计算和波前重构两部分。计算量相对较小的质心计算由16核CPU并行完成，计算量巨大的波前重构算法由4片GPU并行完成，大幅提高了液晶自适应光学系统的数据处理速度：质心计算时间从214us缩短到27us，计算速度提高7.9倍； Zernike模式波前重构计算时间从690us缩短到90us，计算速度提高7.7倍。 针对液晶自适应光学系统响应矩阵测量精度低的问题，提出了一种基于最小二乘统计的响应矩阵测量方法，通过数学统计手段消除了响应矩阵测量过程中的随机噪声与干扰，使得液晶自适应光学系统响应矩阵的测量精度提高了2倍以上，校正后图像的功率谱最大提高了3倍左右。 针对液晶自适应光学系统开环校正精度较低的问题，提出用系统传递函数构造开环校正残差并设计开环自适应控制器的方法，部分消除了波前探测器与校正器的动态响应延迟误差。首先根据液晶校正器的光强调制特性，将液晶自适应光学系统的输入由Zernike系数替换为液晶校正器的驱动电压，使得35输入35输出的系统转换为了单输入单输出系统；然后通过状态空间系统辨识得到了单输入单输出系统的传递函数，并证实传递函数具有高达90%的精度和不同区域高达95%的一致性，以及两个月内小于7%的时变性，证明所获得的传递函数能够很好地描述液晶自适应光学系统，且可以作为不变函数用于控制器设计；最后根据系统的传递函数，构造出开环系统的校正残差函数，由此设计了在校正过程中参数自适应更新的一阶控制器，使得校正精度提高了32%。 本论文创新性地提出液晶自适应光学系统并行数据处理方法、最小二乘响应矩阵测量方法、开环自适应控制算法，显著地提高了液晶自适应光学系统的校正效果。论文成果对液晶自适应光学理论与应用做出了一定程度的贡献。
英文摘要	Although the adaptive optics technology has been widely applied in large aperture telescope systems, most of the adaptive optics systems (AOs) have to be working in infrared band, limited by the lack of actuators of traditional deformable mirror (DM) wavefront corrector. However, in visible wave band, not only telescope presents higher resolution, but also most of the celestial objects have higher radiating energy. Hence new type of wavefront corrector with high density of actuators that could satisfy the requirement of large aperture telescopes in visible wave band is the sticking point of next generation AOs. While this is exactly the advantage of novel liquid crystal wavefront corrector (also named as LCOS). Predecessors have solved some of the problems of LCOS, like slow response speed, energy loss due to polarization property and narrow correction wave band due to dispersion. However, up to today, there are still fellowing disadvantages that hinder the application of liquid crystal adaptive optics systems (LCAOS): 1) Large computing time delay due to vast complexity in data processing. The pixels of LCOS are dozens of times more than the actuators of DM, this brings huge computation during wavefront reconstruction, causes computing time delay in milliseconds that would evidently limits the correction bandwidth of LCAOS. 2) Low precision in the measurement of interaction matrix. Static aberration, random noise and disturbance in LCAOS would impact the precision of interaction matrix measurement, and then affect the performance of LCAOS correction. 3) Low accuracy in open loop correction. Because it’s very hard to control the residual error that is impossible to be detected in a open loop structure, which is adopted by LCAOS to reduce the energy loss due to polarization property. In this dissertation, a parallel data processing technology based on multi-CPU and multi-GPU is carried out to solve the problem of vast computational complexity for LCAOS. The two major tasks of data processing in LCAOS is centroid detection and wavefront reconstruction. The centroid detection algorithm is implemented on 16-CPUs because of its relatively low complexity, and the wavefront reconstruction algorithm is implemented on 4-GPUs due to its high complexity, thereby the speed of data processing is dramatically increased: the computing time of centroid detection is reduced from 214us to 27us, with a gain of 7.9 in speed; and the computing time of wavefront reconstruction is reduced from 690us to 90us, with a gain of 7.7 in speed. A novel interaction matrix measurement method based on least square statistics is brought out to improve the precision of interaction matrix measurement. By eliminating the influence of random noise and disturbance during measurement via statistic method, the accuracy of the interaction matrix is improved with a factor of 2, and the power spectrum of the image after correction is enhanced with a gain of nearly 3 at maximum. As to the problem of low accuracy in open loop correction, a open loop adaptive controller is presented with residual error constructed from system transfer function, to remove the error caused by the delay of dynamic response of wavefront sensor and corrector partially. At first, making use of the relationship between phase modulation and light intensity modulation of LCOS, innovatively replace the input of LCAOS from Zernike coefficients to driving voltage of LCOS, so that LCAOS could be simplified from a 35-input 35-output system to a single-input single-out (SISO) system; then the transfer function of the SISO system is identified through subspace system identification, and corroborated that the precision of the system transfer function is 90%, the uniformity of different areas of LCOS and Shack-Hartmann wavefront sensor is identical to 95%, and the time-invariance of the system is less than 7% in two months, hence the acquired system transfer function could accurately describe the liquid crystal adaptive optics system and could be used during controller design as constant function; Finally a open loop correction residual error function is constructed based on the acquired system transfer function, then a first-order controller with parameter adaptively updating during correction is designed, the correction precision of LCAOS is improved by 32%. This dissertation has developed heuristic parallel data processing technology, least square interaction matrix measurement method and open loop adaptive control algorithm, hence notably improves the performance of liquid crystal adaptive optics systems. The results of this dissertation have contributed to the theory and application for liquid crystal adaptive optics systems on a certain extent.
公开日期	2015-12-24
内容类型	学位论文
源URL	[http://ir.ciomp.ac.cn/handle/181722/48955]
专题	长春光学精密机械与物理研究所_中科院长春光机所知识产出
推荐引用方式 GB/T 7714	张杏云. 提高液晶自适应系统校正效果的研究[D]. 中国科学院大学. 2015.

个性服务

查看访问统计

相关权益政策

暂无数据

收藏/分享

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。