隱形眼鏡邊緣瑕疵檢測演算法之設計

Design of Edge Defect Detection Algorithm

洪竟堯(Jing-Yao Hung)

指導教授 ： 彭德保

國立交通大學/管理學院/工業工程與管理系所/碩士(2010年)

https://doi.org/10.6842/NCTU.2010.00969

摘要

近年來使用隱形眼鏡矯正視力的消費者眾多，由於隱形眼鏡使用時會與眼球角膜直接接觸，因此隱形眼鏡的品質與消費者眼睛的健康有直接的影響。早期隱形眼鏡製造商大多採用人工目視檢查鏡片瑕疵，然而人工方式可能因個人主觀判斷不同以及個人疲勞程度不同，而對瑕疵之判別標準不一，因此鏡片品質難以作統一標準的把關。
近年來，隨著自動光學檢測技術蓬勃發展，隱形眼鏡之自動光學檢測機台已有些研究，但是隱形眼鏡邊緣瑕疵之檢測演算法部分，則尚有改善空間。本研究將改善過去萃取邊緣資訊的方法，並結合影像處理中邊緣偵測的概念，使用突波濾波器以及LOG濾波器，對隱形眼鏡邊緣進行兩階段式的檢測。針對具有大小邊情況之正常片隱形眼鏡所造成的偽陽性(False Positive)，以及邊緣較細之邊緣瑕疵隱形眼鏡所造成的偽陰性(False Negative)，應用本研究所提出的演算法都能有效的檢測出來，對於檢測正確率的提升，以及檢測誤判率的降低，有非常顯著的改善效果。

關鍵字

隱形眼鏡 ； 電腦視覺 ； 瑕疵檢測 ； 邊緣偵測 ； 濾波器

並列摘要

There are plenty of consumers using contact lenses to correct vision in recent years. Due to the contact lens will contact with users eyeball directly, the quality of contact lens has to be assured not to influence users eye health. Contact lens was generally inspected by quality inspectors in the past. However, owing to the diversity of individual subjective judgment and fatigue situation, there is no consistent inspection standard. The lens quality cannot always be guaranteed.
Recently, with the development of auto-optical inspection (AOI) technology, several researches has been done on contact lens auto inspection. However, the available algorithms for edge defect detection still left room for improvement. This thesis proposed a robust edge defect detection algorithm based on surplus filter and LOG filter. The proposed algorithm performed extreme well for false positive detection of uneven yet qualified lenses and false negative detection of defective lenses with thinner edge. Experimentations showed that the proposed algorithm had a great improvement in the accuracy rate and reduction in the error rate of edge defect detection.

並列關鍵字

contact lens ； computer vision ； defect detection ； edge detection ； filter

參考文獻

【24】 王文政，“隱形眼鏡乾片檢測演算法之設計與開發”，國立交通大學工業工程與管理學系碩士論文，(2009).

【29】 沈威廷，“導線架多層式打線端點自動校正之光學檢測系統設計與開發”，國立交通大學工業工程與管理學系碩士論文，(2009).

【1】 M. B. Ahmad and T. S. Choi, “Local Threshold And Boolean Function Based Edge Detection,” IEEE Transactions on Consumer Electronics, vol. 45, no. 3, pp. 674-679, 1999.

【2】 J. Ai and X. Zhu, “Analysis and Detection of Ceramic-Glass Surface Defects Based on Computer Vision,” Proceedings of the 4th World Congress on Intelligent Control and Automation, vol. 4, pp. 3014-3018, June 2002.

【4】 D. Casasent and R. Krishnapuram, “Curved Object Location by Hough Transformation and Inversions,” Pattern Recogintion, vol. 20, no. 2, pp.181-188, 1987.