Carbon Nanotube Detection by Scanning Electron Microscopy

Abstract

This study describes an approach towards the detection of carbon nanotubes in scanning electron microscopic images. It is designed to work in a real-time environment for automated nanofabrication. Main difficulties are the high level of image noise and the occurrence of debris objects. A descriptor derived from the principle component analysis of the geometrical object distribution is found to be a very helpful indicator for object classification. It turns out to be largely invariant to alterations in imaging conditions. All methods described have been implemented and evaluated in largescale experiments. The procedure has also been fully integrated into a distributed control architecture. 1