一种基于旋转矩阵分解的视觉伺服控制算法

A visual servoing algorithm based on rotation matrix decomposition

针对传统视觉伺服控制算法易使目标物品脱离摄像机视野而致伺服失败的缺点,从表征当前摄像机坐标系与期望摄像机坐标系姿态关系的旋转矩阵中分解出了等效转轴和等效转角,利用它们构造了一种可以有效控制摄像机朝向的任务函数矢量,并推导出了表征任务函数矢量变化量与摄像机运动速度间非线性映射关系的雅可比矩阵。然后构造了任务函数,并根据李雅普诺夫第二方法设计了解耦的视觉伺服控制律,最后对所设计的控制律进行了实验验证。实验结果表明,使用旋转矩阵分解方法构造的任务函数矢量来控制摄像机朝向,可使目标物体始终位于摄像机的视野之内,从而有效避免伺服失败。

To avoid the target being lost from the camera views in the application of traditional visual servoing methods,a new characteristic vector was derived from the rotation axis and a rotation angle decomposing from the rotation matrix connecting the current camera frame and the desired camera frame,as well as the Jacobian matrix representing the relationship between the time variation of the characteristic vector and the camera velocity.The proposed characteristic vector could be efficiently applied to control the camera＇s orientation so that the task function was constructed,and a decoupled control scheme was designed using the second Lyapunov method.Experimental results showed that the failures of visual servoing tasks for home environments could be avoided by using the rotation matrix decomposition method.