基于角点灰度分布特征的棋盘格角点检测  

Checkerboard Corner Detection Based on Corner Gray Distribution Feature

作　　者：伍明 伍俊龙 马帅 杨康建[1,2] 杨平[1,2] Wu Ming;Wu Junlong;Ma Shuai;Yang Kangjian;Yang Ping(Key Laboratory of Adaptive Optics,Chinese Academy of Sciences,Chengdu,Sichuan 610209,China;Institute of Optics and Electronics,Chinese Academy of Sciences,Chengdu,Sichuan 610209,China;University of Chinese Academy of Sciences,Beijing 100049,China)

机构地区：[1]中国科学院自适应光学重点实验室,四川成都610209 [2]中国科学院光电技术研究所,四川成都610209 [3]中国科学院大学,北京100049

出　　处：《激光与光电子学进展》2020年第1期101-108,共8页Laser & Optoelectronics Progress

基　　金：中国科学院西部之光人才培养引进计划。

摘　　要：在摄像机标定过程中,较差的光照条件和镜头畸变会造成棋盘格角点漏检和角点冗余。分析了角点的灰度分布特性,提出一种基于角点灰度分布特征的棋盘格角点检测算法。为确保在较差光照条件和镜头畸变下,棋盘格图像角点不漏检,算法首先利用角点的灰度分布特性提取候选角点;然后通过迭代的方式提高候选角点的精度并再次结合棋盘格角点的灰度分布特性剔除候选角点中非角点处的伪角点,避免棋盘格角点冗余;最后通过角点处的邻近点合并获得最终的棋盘格角点坐标。实验结果证明,在较差的光照条件和镜头畸变条件下,本算法角点无漏检和冗余。将该算法提取的棋盘格角点应用于摄像机标定,结果显示重投影误差的均方差在0.1pixel范围以内,优于其他算法。During the camera calibration process,it is necessary to deal with the missed detection and redundancy of checkerboard corners caused by poor illumination conditions and lens distortion.In this study,we analyze the gray distribution properties of the corners and propose an algorithm for corner detection based on the checkerboard corners′gray distribution features.To ensure that the checkerboard corners are not missed under poor illumination conditions and lens distortion,the proposed algorithm firstly extracts candidate corners using the gray distribution characteristics of the corners.Then,we improve the accuracy of the candidate corners through iteration,and eliminate the fake corners based on the gray distribution characteristics of the checkerboard corners,avoiding the redundant corners.Finally,we extract the corner coordinates of the checkerboard by combining the nearest neighbor points.Experimental results show that there is no omission and redundancy in the corners under poor illumination conditions and lens distortion.By applying the checkerboard corners extracted by the proposed algorithm in camera calibration,a mean square error of the re-projection error less than 0.1 pixel has been achieved,which is better than those provided by existing algorithms.

关 键 词：测量 棋盘格 角点检测 灰度分布 摄像机标定 

分 类 号：TN206[电子电信—物理电子学]