Link to the University of Pittsburgh Homepage
Link to the University Library System Homepage Link to the Contact Us Form

USING MACHINE LEARNING FOR FEATURE DETECTION IN TRANSMISSION ELECTRON MICROSCOPY

CAO, CHANGJIAN (2019) USING MACHINE LEARNING FOR FEATURE DETECTION IN TRANSMISSION ELECTRON MICROSCOPY. Master's Thesis, University of Pittsburgh. (Unpublished)

[img] PDF
Restricted to University of Pittsburgh users only until 18 June 2021.

Download (11MB) | Request a Copy

Abstract

In situ testing performed in a transmission electron microscope (TEM) represents an im- portant technique for materials analysis. However, it produces a large amount of data in the form of images and video, which cannot typically be analyzed using traditional image- analysis algorithms. Therefore, a machine-learning approach is proposed to detect the shape of a body by recognizing and locating the border of the material. This supervised-learning al- gorithm is applied and a convolutional neural network is built to rapidly label all pixels. This network explores the relationship between small sub-images cropped from original images and their corresponding labels, and then it predicts the label when given new sub-images, thus generating a segmented image. In this project, the performance was assessed based on specificity, sensitivity, and accuracy of results. The overall accuracy of the present model is over 90%; however, the precision and recall rate are low due to high false-positive detection. This research suggests key factors for improving future machine-learning algorithms for TEM image analysis.


Share

Citation/Export:
Social Networking:
Share |

Details

Item Type: University of Pittsburgh ETD
Status: Unpublished
Creators/Authors:
CreatorsEmailPitt UsernameORCID
CAO, CHANGJIANchc278@pitt.edu
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Committee ChairJacobs, Tevistjacobs@pitt.edu
Committee MemberBabaee, Hessamh.babaee@pitt.edu
Committee MemberAkcakaya, Muratakcakaya@pitt.edu
Date: 18 June 2019
Date Type: Publication
Defense Date: 26 March 2019
Approval Date: 18 June 2019
Submission Date: 1 April 2019
Access Restriction: 2 year -- Restrict access to University of Pittsburgh for a period of 2 years.
Number of Pages: 42
Institution: University of Pittsburgh
Schools and Programs: Swanson School of Engineering > Mechanical Engineering
Degree: MS - Master of Science
Thesis Type: Master's Thesis
Refereed: Yes
Uncontrolled Keywords: Transmission electron microscopy, Convolutional neural networks, Supervised learning, Image processing.
Date Deposited: 18 Jun 2019 17:48
Last Modified: 18 Jun 2019 17:48
URI: http://d-scholarship.pitt.edu/id/eprint/36208

Metrics

Monthly Views for the past 3 years

Plum Analytics


Actions (login required)

View Item View Item