基于CPU＋GPU的大视场物镜成像畸变实时校正  

Real-time Imaging Distortion Correction of Large-field Objective Lens Based on CPU + GPU Hybrid Platform

作　　者：李明杰 胡明勇 张健 曹明坡 王位 LI Ming-jie;HU Ming-yong;ZHANG Jian;CAO Ming-po;WANG Wei(Key Laboratory of Special Display Technology of the Ministry of Education, National Engineering Laboratory of Special Display Technology, National Key Laboratory of Advanced Display Technology, Academy of Photoelectric Technology, Hefei University of Technology, Hefei 230009, Chin)

机构地区：[1]合肥工业大学光电技术研究院特种显示技术教育部重点实验室,特种显示技术国家工程实验室,合肥230009

出　　处：《光子学报》2018年第6期253-260,共8页Acta Photonica Sinica

基　　金：大视场物镜成像畸变校正合作项目资助。

摘　　要：针对工业大视场物镜畸变成像的实时校正问题,提出一种校正算法和CPU+GPU并行加速方案.根据光学畸变理论和相机标定技术,建立非球面畸变校正模型.利用棋盘样板计算光学中心和估计畸变系数,设计校正算法.在CPU+GPU并行加速方案基础上,设计内核自适应维度算法并优化运行程序,结合OPENGL驱动进行实时校正和显示.实验结果表明,本文设计的实时校正系统对高分辨率的畸变成像校正率可以达到98.2%,单帧耗时0.026 s,平均综合加速比为29.1.该系统精度高,可移植性强,简单易行,能够广泛应用于成像畸变的实时校正.To solve the real-time imaging distortion correction problem of industrial wide-field objective lens,a correction algorithm and CPU ＋ GPU parallel acceleration scheme was proposed. Firstly,a distortion correction model of aspheric surface is established according to the optical distortion theory and the camera calibration technology. Secondly,a correction algorithm is designed by using optical center and estimating distortion coefficients. Finally,based on the CPU ＋ GPU hybrid acceleration platform,the kernel adaptive dimension algorithm is designed and the optimized running program is proposed,combined with Open GL driver for realtime correction and displaying. The experimental results show that the real-time correction system can achieve a correction rate of 98. 2%,single frame time-consuming of 0. 026 s,the average overall speedup of 29. 1 for high resolution distortion imaging. The imaging system is higher precision,transplantable and more simple,and can be widely used in the real-time imaging distortion correction.

关 键 词：大视场物镜 光学畸变 高分辨率 GPU并行加速 实时校正 OPENGL显示 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]