상세 보기
Physics-Based Vision System for Handling and Inspection of Ultra-Thin Glass in Hydrodynamic Processes
- Kim, Kyung Hoon;
- Jeon, Jae Wook
WEB OF SCIENCE
0SCOPUS
0초록
The mass production of Ultra-Thin Glass (UTG) faces significant hurdles in automated handling and inspection due to its transparency, residual moisture, and flexibility. While deep learning (DL) approaches offer high performance, they are often limited by computational latency and high hardware costs. This paper proposes a cost-effective, physics-based vision solution to address these bottlenecks, marking the industry’s first feasibility verification for mass production. The system integrates a meniscus-induced imaging technique to detect transparent UTG with 100% reliability and a moisture-resilient crack detection module. By optimizing stage surface roughness (Ra ≈ 5.0) and applying morphological filtering, the system distinguishes water droplets from cracks, reducing the Missed detection rate to 7.7% without heavy computation. Additionally, a Zone-based Alignment algorithm compensates for optical distortion and cassette tolerances to ensure precise robotic loading. Validation over 281 continuous cycles demonstrated a 100% detection rate, 0% breakage, and a rapid processing time of 377 ms. Notably, this rule-based approach reduces Capital Expenditure (CAPEX) by $34,560 per unit by eliminating the need for GPUs. The proposed solution provides a robust, economically viable standard for automating advanced display material manufacturing.
키워드
- 제목
- Physics-Based Vision System for Handling and Inspection of Ultra-Thin Glass in Hydrodynamic Processes
- 저자
- Kim, Kyung Hoon; Jeon, Jae Wook
- 발행일
- 2026
- 유형
- Article
- 저널명
- IEEE Access
- 권
- 14
- 페이지
- 76396 ~ 76426
- 언어
- ENG
- 출판사
- Institute of Electrical and Electronics Engineers Inc.
- 발행국가
- 미국
- 분량
- 31 페이지
- ISSN
- E 2169-3536
P 2169-3536