摘要
研究折光鱼眼全景相机的标定方法,采用统一球面映射原理构建相机成像模型;采用FAST算法提取棋盘格标定图像的畸变角点特征.以由粗到精的方式标定全景相机,具体分为两步:用线性矩阵的奇异值分解方法初始估计全景相机的内、外参数;并在此基础上采用非线性盒约束优化方法实现对内、外参数的精细化估计.实验结果验证了方法的有效性.
A calibration method was proposed for fish-eye omni-directional camera,and the unifying spherical projection model was used to model the camera.FAST(features from accelerated segment test)algorithm was used to detect the distorted image corners of calibration grid pattern.A two step coarse-to-fine calibration method was proposed.SVD was used(singular value decomposition)to get a coarse estimate of the intrinsic and extrinsic parameters of fish-eye camera.Then,based on the former results,the nonlinear optimization technique with box constraints was used to get a fine estimation of camera parameters.The experimental results validate the performance of our proposed methods.
出处
《华中科技大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2013年第S1期99-102,共4页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
中央高校基本科研业务费专项资金资助项目(HIT.NSRIF201172)
国家自然科学基金资助项目(61075081
61273339)
高等学校博士学科点专项科研基金资助项目(20122302120039)
关键词
机器人视觉
相机标定
统一球面投影模型
折光鱼眼全景视觉
角点提取
盒约束优化
robot vision
camera calibration
unifying spherical projection model
dioptric fisheye omni-directional camera
corner detection
box-constrained nonlinear optimization