PVC卡片表面瑕疵之自動光學檢測

Automatic Optical Inspection for Surface Defects of PVC Card

吳俊賢(Chun-Shien Wu)

指導教授 ： 吳明川

國立臺北科技大學/機電學院/自動化科技研究所/碩士(2009年)

摘要

本研究目的在利用機器視覺與數位影像處理技術建立一套PVC卡片表面瑕疵之自動光學檢測系統。由於PVC卡片製作過程中，因機械定位、噴墨不良、壓合過程或其它加工過程造成卡片產生瑕疵，而主要的瑕疵包括：尺寸不良、圖案缺漏、刮痕、氣泡、異物與沾墨髒點這幾類瑕疵。現階段大多數製造業仍以傳統人工檢測方式，不僅檢測速率低、不精確、容易出錯，且無法維持穩定的品質。有鑑於此，本研究係針對這些問題，發展一套檢測技術來取代人工檢測，以提高檢測速率、改善檢測品質。
研究中針對屬高反射體之PVC卡片表面，採用暗場打光原理來解決表面反射問題，並利用背向打光方式及霍夫直線偵測原理判斷卡片尺寸大小；對於偏移或偏轉之圖案，主要利用多重金字塔與圖案的不變矩特徵配合特徵比對技術處理定位與缺漏問題；而刮痕、氣泡、異物與沾墨髒點的檢測，主要是藉由灰階共變異矩陣所求得的特徵值、主軸方向角度及相關係數值建立多重的檢測指標，將瑕疵從樣本中準確的檢測出來，最後再由倒傳遞類神經網路分類出瑕疵種類。經實驗結果顯示本方法可檢測出瑕疵，其平均瑕疵檢出率約為92%，因此，可實際應用於PVC卡片表面瑕疵的偵測，以改善人眼檢測所造成的缺失。

關鍵字

PVC卡片 ； 瑕疵檢測 ； 自動光學檢測

並列摘要

This paper develops a system of automatic optical inspection for surface defects of PVC card using machine vision and digital image processing. During the manufacturing process of PVC cards, the defects are generated because of the painting machine and artificial neglect. The types of defect are inaccurate size, scratches, spots, bubbles and pollution. Most of the manufactures still depend on human vision for inspecting. Inspection rate and result not only low, inaccuracy and error-prone, but also does not maintain a stable inspection quality. In this study, we develop automatic optical inspection techniques in order to replace human vision and increase the inspection speed, improve the inspection quality.
Because of the high reflecting surface of the PVC cards, so this paper applies dark-field imaging technique to solve the reflect problem. And inspect PVC cards size by using back lighting and Hough lines detection. In order to positioning and detecting a pattern, we using moment with resolution pyramid search method to match characteristics of the pattern. Eigenvalues, major-axis angle and Correlation coefficient of the covariance matrix of the data points in the map are used as similarity measures to evaluate the difference between two compared images. And uses Back-propagation neural network(BP) to recognize kind of the defects. Experimental results have shown that the proposed method can effectively detect defects. The average accuracy of inspection is about 92%. So using this inspection system can improve faults of human vision.

並列關鍵字

PVC card ； Defects detect ； Automatic Optical Inspection(AOI)

參考文獻

[1]許世清，機器視覺應用於PVC卡片表面瑕疵檢測，碩士論文，國立台北科技大學製造科技研究所，台北，2007。

[7]陳志忠，液晶顯示器的像素點缺陷與相對亮度均一性之自動化檢測，碩士論文，私立中原大學機械工程研究所，桃園，2001。

[9]N. G. Shankar, and Z. W. Zhong, "Defect detection on semiconductor wafer surfaces,” Microelectronic Engineering," vol. 77, 2005, pp. 337-346.

[10]鍾瑞賢，應用機器視覺與類神經網路於電機零組件品質之自動檢測-以S鐵心為例，碩士論文，元智大學工業工程研究所，桃園，1999。

[16]N. Otsu, "A threshold selection method form gray-level histogram," Man and Cybernetics, vol.9, 1979, pp.62-66.