影像辨視技術應用於3D面板貼合檢測

Image recognition technology used in the 3D display panel bonding process check

林長緯(Chang-Wei Lin)

指導教授 ： 劉益宏

中原大學/工學院/機械工程研究所/碩士(2012年)

https://doi.org/10.6840/cycu201200983

摘要

影像辨識技術在目前工業界的應用己相當普遍，因為它的檢測速度快，檢查錯誤率低，同時能即時對檢查數據進行分析與異常回饋，己逐漸取代以往的人工檢測作業方式。而隨著顯示器技術的提升，具有3D影像功能的顯示器將會是未來的基本需求，其中柱狀透鏡陣列(Lenticular Lens Array)的立體顯示技術，在觀看時不必配戴眼鏡即能觀賞3D立體影像，且擁有售價便宜的優點。但目前柱狀透鏡陣列技術的顯示器仍是依賴人員進行檢測，確認貼合紋路的傾斜角度是否在規格之內，此作法相當費時、且不準確。
因此本研究提出一套適用於「3D面板貼合檢測之自動化光學檢測系統」，利用CCD鏡頭擷取影像，再經由檢測系統檢測柱狀透鏡陣列板的貼合狀況，取代人員檢測，減少檢測的誤判率與差異性，亦可提升生產速度、降低生產成本，達到製程監控的目的。經實驗結果得知，本研究所開發的檢測系統，對於檢測環境有光源強弱之變化、檢測影像中殘有背景雜訊或者在進行黑紋傾角偵測時，黑紋不完整且有破損的狀況下，仍然可有效檢測出面板貼合狀況，整體檢出率可達96%。

關鍵字

影像辨識 ； 柱狀透鏡陣列 ； 顯示器

並列摘要

Image recognition technology has been quite common in industrial applications.
Compared with the traditional manual inspection method , it is superior for its speedy testing , high accuracy and reliability. With the rise of display technology, TV with 3D imaging capabilities will be the basic needs of the future. The Lenticular Lens Array with no aided glasses and has the advantages of lower prices among stereoscopic display technology. Now, flat display panel of Lenticular Lens Array still relying on manual inspection methods. It is very time consuming and not accurate.
In this thesis, a Image recognition technology is proposed to deal with the critical problem mentioned above. This technology will replace the traditional manual inspection method because it has high speed detection and low detection error rate. It is able to flat display panel bonding process check from different lighting environment and Image noise.
Experimental results show this technology is able to flat display panel bonding process check from different lighting environment and Image noise, and also achieves average recognition rate 96%.

並列關鍵字

Image recognition technology ； Lenticular Lens Array ； Display panel

參考文獻

[18]松橋工作室，"Visual Basic 6 完美的演繹"，知城數位科技股份有限公司，2003。

[1] Bellas, N., S. M. Chai, M. Dwyer, and D. Linzmeier, "FPGA implementation of a license plate recognition SoC using automatically generated streaming accelerators," in Proc. 20th IPDPS, Rhodes Island, Apr.25-29, 2006, pp.8-15.

[4] Draghici, S., "A neural network based artificial vision system for license plate recognition," Int. Journal of Neural Systems, vol.8, no.1, pp.113-126, Feb. 1997.

[5] Hsieh, C.-T., Y.-S. Juan, and K.-M. Hung, "Multile license plate detection for complex background," in Proc. 19th Int. Conf. on Advance Iinformation Networking and Applications, Taipei, Taiwan, Mar28-30, 2005, pp.389-392.

[6] Kamat, V. and S. Ganesan, "An efficient implementation of the Hough transform for detecing vehicle license plates using DSP'S," in Proc. Conf.Real-Time Technology and Application., Chicago, IL, May 15-17, 1995,pp.58-59.