Robust contact detection in micromanipulation using computer vision microscopy.

This paper presents a computer vision algorithm for visually detecting the contact between an end-effector and a target surface under an optical microscope. Without using additional sensors (e.g., proximity or force/touch sensors), this algorithm provides robustness and a sub-micrometer detection resolution. Fundamentally, after the establishment of a contact in the world frame, further vertical motion induces horizontal motion in the image plane. An analysis is presented to elaborate this fundamental mechanism. Experimental results demonstrate that this computer-vision-based method is capable of achieving contact detection between a micropipette and a glass slide surface with a resolution of 0.2 microm. Furthermore, 1300 experimental trials reveal that the presented algorithm is robust to variations in illumination intensity, microscopy magnification, and microrobot motion speed.