基于散焦图像特征的微装配机器人深度运动显微视觉伺服

Microscopic Visual Servoing for Microassembly Robot Depth Motion Based on Defocus Image Features

为了描述微操作手深度运动,采用灰度方差聚焦评价算子计算机械手散焦图像特征.散焦特征曲线理论上为单峰分布,峰值点对应显微光学焦平面深度位置.实际提取的散焦特征含有大量随机噪声,利用非线性跟踪微分器抑制噪声实现对机械手散焦图像特征及其微分信号的无颤振光滑逼近.依据散焦微分信号设计粗—精两级自寻优视觉控制器,完成微操作手对装配平面深度的精确定位.微装配深度运动实验验证了本文方法的有效性,机械手深度伺服误差为7.5μm.

To measure micromanipulator depth motion, a gray variance-based focus measure operator is utilized to compute its defocus image features. The defocus feature curve is theoretically distributed with only one peak, which represents depth position of the microseopic focal plane. The extracted defocus features usually include lots of random noises. A nonlinear tracking differentiator is developed to suppress noises and smoothly track the features and their differential values without oscillation. Based on the differential defocus signals, a coarse-to-fine self-optimizing vision controller is presented for the micromanipulator to precisely locate assembly plane along the depth way. Experimental results of microassembly depth motion demonstrate the validity of the proposed approach with a depth servoing error of 7.5 μm.