基于无标定显微视觉伺服的零件微装配

Micro-assembly of Parts Based on Uncalibrated Microscope Visual Servo

为完成微小零件的装配操作,获得高效的微装配性能和避免复杂的摄像机标定工作,提出了基于BROYDEN方法的图像雅可比矩阵在线辨识模型.为了实现在线辨识快速收敛的目的,应用切比雪夫多项式构成了待辨识的图像雅可比矩阵的成本函数来逼近最优值.采用辨识的图像雅可比矩阵,设计了PD控制器,该控制器满足了系统的快速性和平滑性控制要求.在显微视觉环境下,完成了不同倍率物镜下微操作机械手自动定位与夹取三维微小零件的视觉伺服任务.实验结果表明,所提方法具有较好的鲁棒性和满意的执行效果,达到了系统应用要求.

In order to assemble micro-parts and to obtain high efficiency for the microassmebly and to avoid complex camera calibration, an image Jacobian matrix online identification model based on BROYDEN is presented. We employ Chebyshev polynomial as a cost function of image Jacobian matrix identification to approximate the optimal value so as to satisfy the requirements for fast convergence of online identification. With identificated image Jacobian matrix, we design a proportional and differential （PD） controller, which meets the requirements for rapidness and smoothness of the system. In the microscope vision environment, we complete the visual servo task that micro-manipulator is positioned automatically and grips 3-D micro-parts under different magnification lens. The experimental results show that the proposed method obtains better robustness and satisfactory performance, reaching the system application requirements.