在分析移动机器人车载摄像机位姿的特殊性质的基础上,根据摄像机的等效转轴构造辅助旋转矩阵,利用该旋转矩阵将原始待分解本质矩阵和单应矩阵转换为一类简单的、可通过初等数学运算进行分解的本质矩阵和单应矩阵。仿真实验的结果表明,该车载摄像机位姿估计算法较传统方法具有更高的精度和更快的运算速度,对摄像机等效转轴的扰动也具有很好的鲁棒性。此外,分解出的可能解的数目较传统算法减少了一半,且在除诱导单应阵的空间景物平面与地面垂直的情况下,均能直接得到移动机器人的唯一转角,为移动机器人姿态控制提供了极大的便利。
An accurate and fast pose estimation problem for on-board camera of mobile robot is investigated. Firstly the special properties of the pose for on-board camera of mobile robot are analyzed. Secondly,an auxiliary rotation matrix is constructed using the on-board camera's equivalent rotation axis,which is utilized to turn the initial essential matrix and homography matrix into a simplified kind that can be decomposed through elementary mathematical operations. Finally,some simulation experiments are designed to verify the algorithm 's rapidity,accuracy and robustness. The experimental results showthat compared to traditional algorithms,the proposed algorithm can acquire higher accuracy and faster calculating speed,together with the robustness to the disturbance of the on-board camera's equivalent rotation axis. In addition,the number of possible solutions are reduced one half,and the unique rotation angle of the mobile robot can be determined except for the condition that the 3D planar scene structure and the ground are perpendicular,which can provide great convenience for controlling the pose of the mobile robot.