Quantitative stereo vision for the stereo light microscope: an attempt to provide control feedback for a nanorobot system

摘要

Quantitative microscopy is aprerequisite to fulfill the sensing tasks that arise with the demand to not only visually observe microscopic specimen but to measure and manipulate structures at the microscopic scale.With the term quantitative microscope, sensor systems are addressed which consist of a microscope equipped with devices for capturing digital images and with tools for processing the image data in terms of extracting a geometrie representation of the observed scene.This thesis deals with the implementation of a quantitative stereo light microscope dedicated for the survey of generic and dynamic scenes.The considered microscope system is used predominantly, but not exclusively, for the visual control of a robot aimed at the manipulation of j.lm-sized objects with sub-micron precision.Although the research is focused on this application, much of the material is generic and can be adapted for any type of measurement task at this scale.Particular emphasis is put on the discussion of algorithms for visual inference which are robust with respect to the often difficult contrast situation occurring in light microscopy.The first part of the thesis reports on investigations of the resolution and accuracy performance of the stereo light microscope.The discussion on resolution shows