一种使用多核CPU的相关哈特曼-夏克波前处理机  

A Correlating Hartmann-Shack Wavefront Processor Using Multi-Core CPU

作　　者：陈冰倩[1,2] 何杰铃 王彩霞[1] 魏凌[1] 

机构地区：[1]中国科学院光电技术研究所自适应光学重点实验室,四川成都610209 [2]中国科学院大学,北京100049

出　　处：《中国激光》2016年第3期221-229,共9页Chinese Journal of Lasers

基　　金：国家自然科学基金(61108082);国家重大科学仪器设备开发专项(2012YQ120080)

摘　　要：斜率算法对哈特曼-夏克传感器非常重要,而基于归一化互相关斜率算法的哈特曼-夏克传感器可满足强噪声情况下点光源和扩展目标的斜率探测需求。为解决互相关斜率处理器可移植性差的问题,提出用多核中央处理器(CPU)来实现斜率探测。在优化归一化互相关算法的基础上,编程实现斜率探测,并从多核并行和运算向量化两方面优化程序,提升斜率提取速度。当模板分辨率为9 pixel×9 pixel时,优化程序在Intel(R)Core(TM)i7-3770k四核计算机上运行,完成400个15 pixel×15 pixel的子孔径斜率提取,用时约为340μs。搭建了自适应光学系统,对优化后的斜率算法进行系统闭环测试。结果表明,使用多核CPU的相关哈特曼-夏克自适应光学系统能在强噪声环境下进行有效的波前校正。Gradient algorithm is very important for the Hartmann-Shack sensor. The Hartmann-Shack sensor based on the normalized cross correlation （NCC） gradient algorithm shows strong noise immunity. In addition, it can detect both point source and extended target. A multi- core CPU （central processing unit） is used to implement NCC gradient algorithm to get better portability compared to the existing processor. The NCC algorithm is optimized and programmed. The program is improved by multi-core parallelization and advanced vector extension. When the template resolution is 9 pixel×9 pixel, the optimized program running on a 4-core computer, Intel （TM） Core （R） i7-3770k, spends about 340 μs to calculate the slope of 400 sub-apertures with 15 pixel×15 pixel. An adaptive optics system using optimized gradient program is built and experiences closed-loop test. The experimental results show that the CPU- based Hartmann- Shack adaptive optics system using NCC gradient algorithm can correct wavefront distortions effectively, even in strong noise environment.

关 键 词：自适应光学 波前传感 多核中央处理器 目标跟踪 归一化互相关 波前斜率探测 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]