基于单目视觉的水下机器人悬停定位技术与实现  

Technique and Implementation of Underwater Vehicle Station Keeping Based on Monocular Vision

作　　者：郝颖明[1] 吴清潇[1] 周船[1] 李硕[1] 朱枫[1] 

机构地区：[1]中国科学院沈阳自动化研究所

出　　处：《机器人》2006年第6期656-661,共6页Robot

基　　金：国家863计划资助项目(2002AA401001-4A)

摘　　要：以7000m载人潜水器的水下悬停为应用背景,以单目CCD摄像机为传感器,提出了基于模型的单目视觉定位和基于特征的视觉伺服两种水下机器人视觉悬停技术方案,分别适用于已知模型的观察目标和未知模型的观察目标两种情况.以自行研制的水下机器人控制系统实验平台为实验载体,利用前视摄像机和水下人工目标,建立了水下演示实验验证系统,并在室内实验水池中,分别实现了两种悬停方法的水下演示实验.演示实验表明:在两种视觉悬停方法的闭环控制下,水下机器人能够抵抗恒定水流干扰和人工位置扰动,很好地实现水下悬停作业.According to the application background of the 7000m manned underwater vehicle and with monocular CCD cameras as sensors, two visual station keeping methods for underwater vehicle are proposed, i.e. the model-based monocular vision positioning for the target when model is known, and the feature-based visual servoing for the target when model is unknown. Based on the experiment platform of underwater vehicle control system made by ourselves, an underwater demonstration experiment system is developed with forward-looking cameras and man-made targets. Underwater demonstration experiments with the two station ＂keeping methods are made in an indoor experimental pool. It is demonstrated that the vehicle can achieve station keeping under the closed loop control of the two proposed methods in the presence of man-made abrupt disturbances and current disturbances.

关 键 词：水下机器人 视觉悬停 视觉伺服 视觉定位 单目视觉 

分 类 号：TP24[自动化与计算机技术—检测技术与自动化装置]