回上一畫面 ----市民農園研發服務團隊
產學合作計畫－鏡片鍍膜品質之自動光學檢測系統研發
The designs and applications of an automatic optical
inspection (AOI) system for the quality analysis of the coating layer of
mirror componentsMOST 109-2622-E-035-016 –
2020/11/01 ~ 2021/10/31
計畫中文關鍵詞鏡片元件、瑕疵、邊緣齊整度，低成本自動監控系統
計畫英文關鍵詞Mirror components, Defects, Edge regularity, Low-cost automatic
control system
計畫中文摘要
本研究計畫主要是針對鏡片元件之鍍膜品質分析，開發一套自動光學檢測系統研發，探討汽車照後鏡等產品其中元件之鏡面部份，在一般製程中的研磨拋光之瑕疵問
題，以及鍍膜製程之汙點、刮痕、裂邊、孔洞與邊緣齊整度，反射光場分佈均勻度等問題。
本研究計畫為期一年，透過開放源始碼的版本以編撰監控程式，建立低成本自動監控系統，針對攝影機所拍攝到的鏡面元件之反射光場影像，以全自動的方式計算鏡
面元件反射光場分佈均勻度，再利用邊緣檢測和樣本比對，使用包含彈性二值化、影像分割、影像差值分析等影像處理的方法，再提出相關的模組化偵測技術，進行
鏡面元件外形瑕疵偵測判斷依據及分類，進而改善檢測系統除錯的能力。並對鏡面元件鍍膜製程之汙點、刮痕、裂邊、孔洞與邊緣齊整度以快速區塊比對方式進行
，讓影像在邊緣分佈不合格處之紋理能完全解析，透過類神經與基因演算法迅速找出最佳目標樣板，以利後續瑕疵之整合判斷。除了系統更有彈性，而且每張影像判
斷時間希望能降低到0.2秒以內，誤判率降低到0.25%以內。
計畫英文摘要
The proposal ‘The designs and applications of an automatic optical
inspection (AOI) system for the quality analysis of the coating layer of mirror
components’ is focusing on the measurement and alignment in assembly
manufacturing process of mirror components. A low-cost automatic control system
with Qt platform can save money and increase quality.
It can find out the edge regularity information automatically. We will also use
different methods to determine the non-uniform factors and the light field area
for each type of mirror components. A special algorithm is presented in the
inter-hole’s position searching and statistic process of uniformity. This
algorithm can combine the procedures of segmentation process and nonlinear grey
scale mapping. On the process of searching, analyzing, and recognizing of LCD,
IC and mirror components images from reflected light and thermal imaging
system, includes the quality verification and defects position calculation. The
original image is converted into a binary image with noise suppression filter.
Next, edge detection method is used to be compared with standard sample. The
system becomes more flexible and the processing speed 0.2 s becomes faster in
comparison to the previous system. The proposed method for chip defect
detection, using unsophisticated and economical equipment, is also verified.
The proposed method provides highly accurate results with a 0.15% error rate. In
addition, we can design a novel algorithm and the relative field with the help
of Taiwan I.BRAVE company. The project will promote
the research capabilities and construct the foundation of the industry for the
AOI technology. We believe this work can significantly help researchers working
in advanced Multi-layer component fabrication.