Browse

Improvement of Labeling Performance using K-Means Clustering in TFT-LCD Defect Classification Process
TFT-LCD결함 분류 공정에서의 K-Means Clustering을 이용한 라벨링 자동화 성능 향상에 대한 연구

Cited 0 time in Web of Science Cited 0 time in Scopus
Authors
김성욱
Advisor
박희재
Major
공과대학 기계항공공학부
Issue Date
2018-02
Publisher
서울대학교 대학원
Keywords
TFT-LCDDefectClassificationLabelingK-Means ClusteringFeaturesKernel
Description
학위논문 (석사)-- 서울대학교 대학원 : 공과대학 기계항공공학부, 2018. 2. 박희재.
Abstract
The paper focuses on the improvement of defect classification for
a TFT-LCD by enhancing labeling performance. Defects occur in the
manufacturing process of TFT-LCD, which have to be repaired or
disposed depending on the size and type, thereby lowering the
productivity. Defects are detected through optical systems and image processing algorithms. Machine learning algorithms are used to
classify the detected defects, and information gathered from the
process provides feedback. The process requires an initial input of
labeled defects that is crucial to the later learning process. Any faulty
inputs given at the initial stage are consequential to impede a proper
learning process.
I propose a method for effectively labeling defects using a k
means clustering algorithm to solve this problem. Previous research
only used features that can be visually confirmed. I argue that adding
the values obtained by passing kernel over the defect data in addition
to visually confirmed features. Using this feature, we could better
classify defects that were not previously classified. Experimental
defect data occur during the TFT-LCD process.
Language
English
URI
http://hdl.handle.net/10371/141387
Files in This Item:
Appears in Collections:
College of Engineering/Engineering Practice School (공과대학/대학원)Dept. of Mechanical Aerospace Engineering (기계항공공학부)Theses (Master's Degree_기계항공공학부)
  • mendeley

Items in S-Space are protected by copyright, with all rights reserved, unless otherwise indicated.

Browse