摘要
基于Eye-to-Hand单目视觉配置,研究了一种三维目标跟踪和定位算法,实现了在笛卡尔空间下三维目标的精确跟踪和定位。在目标跟踪过程中,采用基于模型的三维跟踪算法,利用虚拟视觉伺服算法使CAD模型在图像空间中的投影收敛于目标边缘,进而实现在笛卡尔空间中与三维目标精确对准,获得实时、鲁棒的三维跟踪。在目标定位过程中,在笛卡尔空间设计增量PI控制算法,使当前目标位姿收敛到期望位姿。实验结果表明,该算法能精确地跟踪和定位三维目标,位置精度控制在0.2mm以内,姿态精度控制在0.1°以内。
Based on the eye-to-hand monocular vision configuration, a 3D target tracking and positioning algorithm was developed, and the high precise 3D tracking and positioning in a Cartesian space was realized. In the tracking procedure, the model-based 3D tracking algorithm is introduced, and the virtual visual servoing (VVS) technique is employed to make the CAD model' s projection lines converge to the target edges in the image space. Correspondingly, the precise registration can be achieved between the model and the target in the Cartesian space and the realtime, robust 3D tracking can be realized. In the positioning procedure, the increment PI controller is employed, which can make the current pose converge to the desired one. The experimental results show the positioning accuracy of the algorithm can be controlled within 0.2mm along the translational directions and 0.1 ° along the rotational directions.
出处
《高技术通讯》
CAS
CSCD
北大核心
2014年第6期623-631,共9页
Chinese High Technology Letters
基金
国家自然科学基金(61227804
61105036)资助项目
关键词
CAD模型
位姿估计
三维跟踪
视觉反馈控制
视觉定位
CAD model, pose estimation, 3D tracking, visual feedback control, visual positioning