基于单应性矩阵的机器人视觉伺服系统研究

摘要

针对在视觉伺服过程中采用图像雅克比矩阵方法时遇到的对目标深度信息的求取等问题,并为了进一步扩大机器人的运动空间,设计了一种基于单应性矩阵的视觉控制器应用于视觉伺服控制系统.通过摄像机针孔模型求解出该单应性矩阵后,对其进行了奇异值分解,得到了机器人控制器的输入,从而实现了对机器人的控制；该控制策略运用Matlab/Simulink仿真软件,并结合在机器入学领域中被广泛采用的机器人工具箱,构建了基于图像的视觉伺服系统仿真模型,并选取一长方体表面上的4个角点作为目标物体进行了仿真抓取实验.研究结果表明,该视觉控制器使整个控制系统具有较强的稳态特性和较高的控制精度,图像特征点从当前位置运动到了期望位置,误差最终收敛于零,具有较好的控制效果.%Aiming at getting the depth information in image Jacobian matrix and enlarging robot's application area in visual servoing system, a visual controller was designed based on the homography matrix. Then the singular value decomposition was executed after getting the homography matrix according to camera pinhole model, and the answer was as the input of the robot. A visual servoing simulation model was established, combined with robotics toolbox for Matlab and Simulink. Four peaks of a cuboid were selected as a target object. The results indicate that, by using this controller, image feature points have finally reached the desired position and the error has tended to zero, the steady-state characteristics and control precision of the system are made robust, so that an ideal control effect could be achieved.