基于单目视觉的自主足球机器人自定位研究  

The study of self-localization for autonomous soccer robot based of monocular vision

作　　者：谢云[1] 杨宜民[1] 

机构地区：[1]广东工业大学自动化学院,广东广州510090

出　　处：《电路与系统学报》2006年第1期29-34,共6页Journal of Circuits and Systems

基　　金：国家863计划资助项目(512-9935-08);广东省重点科技攻关资助项目(200027)

摘　　要：本文介绍了一种在颜色场地信息已经由视觉识别的前提下,通过单目视觉测距,实现全自主足球机器人自定位的几何方法;对摄像机CCD像面的量化、镜头的非线性畸变和摄像机定标等可能的引起误差的因素进行了详细分析,并且对应地提出了解决办法。尤其是引入了偏最小二乘回归法对摄像机内部参数进行定标,避免由变量的相关性导致的定标误差;同时也更适于实际的比赛情况。实验结果表明它是可行的。A geometrical self-localization method of autonomous soccer robot is introduced. There is a precondition that is some color landmarks in the field have been recognized by cyber-vision. A method of depth measurement by monocular vision is used in the geometrical self-localization method. Three possible factors bringing ranging error, which are quantization error of CCD image element, no-linearity aberration of its camera lens, and vidicon calibration, are analyzed. Their corresponding resolventes have been put forward. PLSR is used in vidicon calibration for avoiding the calibration error owing to the relativity among variables, especially. It＇s shown that the geometrical self-localization method is effective and robust.

关 键 词：全自主足球机器人 单目视觉 几何定位 误差分析 偏最小二乘回归 

分 类 号：TP24[自动化与计算机技术—检测技术与自动化装置]