自動化觸控面板瑕疵位置偵測

Automated Defect Detection and Location Determination on Touch Panels

邱子庭(ZI-TING QIU)

指導教授 ： 林宏達

朝陽科技大學/理工學院/工業工程與管理系/碩士(2019年)

摘要

透明基板元件具有好的穿透性與不易刮傷之特性，已被廣泛地運用於光學和電子產品相關製造的材料，隨著智慧型手機與平板電腦的普及，其需求量也日益增加。現今常使用的電容式觸控面板是由多層膜或兩層以上的貼合透明基板玻璃所組成，其表面雖為透明玻璃，但內部具有規則性的結構紋路。因透明工件具有較佳的視覺穿透性，當瑕疵發生在透明工件表面時，無法得知瑕疵的位置位於正面或反面。且產品在進行生產時，通常會在正面進行加工，對瑕疵位於正面或反面有不同處理方式，正面通常被使用較嚴格的標準看待。瑕疵所發生位置可回溯製程找出發生異常原因，因此能正確地分辨瑕疵發生於正面或反面對透明工件之品質確保是相當重要的。
本研究針對觸控面板的瑕疵偵測與瑕疵正反面位置判斷提出自動化檢測的方法。利用希爾伯特-黃轉換(Hilbert-Huang Transform, HHT)增強瑕疵與背景之對比度，再以區間估計進行二值化，將背景與瑕疵分離，達到瑕疵偵測之目的。接續使用標籤化方法將瑕疵進行標記並提取特徵值例如：灰階值平均值、灰階標準差之差值、面積與周長，並以隨機森林法(Random Forest, RF)判斷瑕疵位於透明基板的正面或反面。本研究使用462張影像進行實驗，其結果在瑕疵偵測之瑕疵檢出率為85.75%瑕疵誤判率為0.33%，瑕疵正反面位置判斷之正確類別率為98.35%。

關鍵字

觸控面板 ； 方向性背景紋路 ； 瑕疵偵測 ； 瑕疵正反面判斷 ； 希爾伯特-黃轉換 ； 隨機森林法

並列摘要

Touch panels commonly used in many electronic devices today are composed of a multilayer film and two or more transparent substrate glass. The surface of a touch panel is transparent glass with regular structural textures inside. Since transparent material has better visual transparency, it is difficult for inspectors to know where the defect is located on the front or back side when defects occur on the surfaces of transparent materials. And when products are manufactured in production line, it is usually processed on the front sides of the products. There are different quality requirements on the front and back sides, and the front is usually treated with stricter standards. The locations of the defects can be traced back to the process to find out the causes of the production abnormality. Therefore, it is important to correctly distinguish the defect locations occurring on the front or back side to ensure the quality of product using transparent materials.
This research proposes an automated defect detection and location determination of touch panels by machine vision system. First, the Hilbert-Huang Transform (HHT) is used to enhance the contrast between the defects and the background. Second, statistical interval estimation is used to segment background and defects to achieve the purpose of defect detection. Third, we combine and label the detected defects on both sides of the transparent substrates and extract the labelled defect features. Fourth, based on the extracted features, the random forest (RF) method is applied to determine where the defects are located on the front or back side of the transparent substrates. Experimental results show that the defect detection rate achieves up 85.75%, and the false alarm rates lower to 0.33%. The accurate determination rate of defect location is 98.24% on the front and back sides of the touch panels.

並列關鍵字

Touch panels ； Directional textures ； Defect detection ； Location determination ； Hilbert-Huang Transform ； Random Forest.

參考文獻

[1] Abdulmunem, M. E., Badr, A. A., “Hilbert Transform and its Applications: A survey,” International Journal of Scientific & Engineering Research, Vol.8, No.2, pp. 699-704 (2017).

[2] Alexander, F. J., “Machine learning,” Computing in Science and Engineering, Vol.15, No.5, pp. 9-11 (2010).

[3] Alickovic, E., and Subasi, A., “Detection of Alzheimer Disease Based on Histogram and Random Forest,” International Conference on Medical and Biological Engineering, Vol. 73, pp. 91–96 (2019).

[4] Banerjee, S., and Chowdhury, A. Y., “Case Based Reasoning in the Detection of Retinal Abnormalities Using Decision Trees,” Procedia Computer Science, Vol.46, pp. 402-408 (2015).

[5] Belgiu, M., Dragut, L., “Random forest in remote sensing: A review of applications and future directions,” ISPRS Journal of Photogrammetry and Remote Sensing, Vol.114, pp. 24-31 (2016).