Supervised STM Image Segmentation of Self-Assembled Molecule Layers
Presentation Type
Poster
Major
Mathematics
Abstract
Self-Assembled monolayers (SAMs) of cage molecules are useful in the construction of nanostructures and nanodevices. For the purposes of nanofabrication, it is useful to monitor the formation of defects in these SAMs and differences in molecular lattice orientation. Through autonomous and supervised image processing, this project aims to segment STM images of SAMs of cage molecules into domains of similar lattice orientation. Due to the translation invariance of the power spectra of the Fourier domain, it is possible to obtain a measure of the similarity or difference between the lattice orientations in two regions in an image and implement traditional clustering methods such as K mean and spectral clustering or more advanced methods such as Markov random fields. The results from the application of these methodologies are promising and could possibly replace manual segmentation as the industry standard.
Faculty Mentor
Timothy Lucas
Funding Source or Research Program
Summer Undergraduate Research Program, Undergraduate Research Fellowship
Location
Waves Cafeteria, Tyler Campus Center
Start Date
21-3-2014 2:00 PM
End Date
21-3-2014 3:00 PM
Supervised STM Image Segmentation of Self-Assembled Molecule Layers
Waves Cafeteria, Tyler Campus Center
Self-Assembled monolayers (SAMs) of cage molecules are useful in the construction of nanostructures and nanodevices. For the purposes of nanofabrication, it is useful to monitor the formation of defects in these SAMs and differences in molecular lattice orientation. Through autonomous and supervised image processing, this project aims to segment STM images of SAMs of cage molecules into domains of similar lattice orientation. Due to the translation invariance of the power spectra of the Fourier domain, it is possible to obtain a measure of the similarity or difference between the lattice orientations in two regions in an image and implement traditional clustering methods such as K mean and spectral clustering or more advanced methods such as Markov random fields. The results from the application of these methodologies are promising and could possibly replace manual segmentation as the industry standard.