Optical imaging with machine learning for the automated characterization of micro- and nanoscale devices
Purandare, Sanyogita
Loading…
Permalink
https://hdl.handle.net/2142/105225
Description
Title
Optical imaging with machine learning for the automated characterization of micro- and nanoscale devices
Author(s)
Purandare, Sanyogita
Issue Date
2019-04-18
Director of Research (if dissertation) or Advisor (if thesis)
Goddard, Lynford
Doctoral Committee Chair(s)
Goddard, Lynford
Committee Member(s)
Choquette, Kent
Koyejo, Oluwasanmi
Popescu, Gabriel
Schwing, Alexander
Department of Study
Electrical & Computer Eng
Discipline
Electrical & Computer Engr
Degree Granting Institution
University of Illinois at Urbana-Champaign
Degree Name
Ph.D.
Degree Level
Dissertation
Keyword(s)
Machine Learning
Optical Imaging
Optical Image Processing
Abstract
This dissertation presents techniques and applications associated with enhancing the sensitivity of the optical images in nanotechnology.
First, the dissertation presents a novel interpretable machine learning technique for defect detection and classification in noisy optical images of semiconductor wafer die. This solution is designed to solve the imbalanced data-set classification problem for noisy images with some feature similarities in the different classes. A baseline comparison is performed by using a standard technique for defect detection and classification. Secondly, a standard loss function is modified and combined with some portions of the new technique to implement defect detection and classification in optical images of semiconductor wafer die. The performances of both the techniques are compared. An optimization technique is presented for a specific category of functions encountered in this dissertation and the algorithm's performance is compared with standard optimization algorithm(s).
Two phase retrieval techniques are compared: diffraction phase microscopy (DPM) and transport of intensity (TIE). TIE is modified for low intensity and zero intensity images. Modified TIE is optimized for theoretical analysis of Gaussian mode phase retrieval. Experimental trials are conducted and optimized hyperparameters are used for phase retrieval.
Use this login method if you
don't
have an
@illinois.edu
email address.
(Oops, I do have one)
IDEALS migrated to a new platform on June 23, 2022. If you created
your account prior to this date, you will have to reset your password
using the forgot-password link below.