摘要
在硬件资源有限的情况下,为了支持尽可能多的子孔径进行实时波前斜率处理,提出了一种基于累加器的波前斜率处理器。该处理器的运算核心是子光斑质心计算模块,根据二维图像矩计算的可分解性以及一维矩的递推累加求解方法,用加法运算代替子孔径坐标与像素灰度的乘法运算,获得灰度重心法所需的所有二维低阶矩。该模块仅由5个累加器组成,硬件实现时避免了乘法器的使用,降低了资源消耗。仿真实验结果表明:对于22×22方形排布的哈特曼-夏克波前传感器图像,本文的结构可在FPGA内实现;在100 MHz的工作频率下,完成一帧所有子孔径斜率计算的延迟时间为0.33μs,计算误差<0.002 pixel;与传统的波前斜率处理器相比,其逻辑资源消耗减小了40%左右。所提出的结构能够在不增加额外资源的情况下,通过对原波前斜率处理器进行升级来完成,其支持的子孔径数目增加1倍左右,实现了波前斜率的高速、高精度提取。
In order to accommodate as many subapertures as possible in the high-speed wavefront slope calculation when hardware resources were limited, an accumulator based wavefront slope processor was proposed. The computational core of the processor is an array of subaperture spot centroid calculating element. According to the decomposition of 2D moment calculation and the recursive procedure of 1D moment calculation,it can substitute the multiplications between the pixel grayvalues and its coordinates with several sum operations to obtain the low order 2D geometric moments required in centroiding. The calculating element simply consists of five accumulators, and the cost has been decreased because no multipliers are needed. Experimental results indicate that when it is implemented in a Field Programmable Gate Array（FPGA） at a clock frequency of 100 MHz, the proposed architecture can obtain gradients of all subapertures in 22 × 22 Shack-Hartmann with the latency no more than 0.33 us and the error less than 0. 002 pixel. The new design has reduced the hardware resource by 40% as compared to that of the multiplier-based architecture. Furthermore, the original multiplierbased processor can be updated for Shack-Hartmann sensors and can obtain the subapertures twice as much as that of original one without additional hardware resources. It realizes the high speed measurement of wavefront with a high accuracy.
出处
《光学精密工程》
EI
CAS
CSCD
北大核心
2011年第3期501-507,共7页
Optics and Precision Engineering
基金
中国科学院知识创新工程重要方向项目(No.KGCX2-Y11-920)
国家863高技术研究发展计划资助项目(No.2008AA02Z422)
