近幾年全世界觸控面板相關產業正如火如荼地發展,因其用途相當之廣泛,諸如: 提款機、汽車電子、醫療、運動、家電、電腦亦或智慧型手機等用途。由於電容式觸控面板製程繁雜,觸控感測器為觸控面板之功能核心,故其零件生產之品質將會大幅度影響觸控面板之整體品質及總成本比例。電容式觸控面板常見之表面瑕疵有裂縫、刮痕、粉塵、其他異物等,由於其表面存在之規律紋路,如以人工檢測方式進行品檢,容易因瑕疵過於微觀或檢測人員疲憊、認知不同而造成誤判,若改以自動化機器視覺進行瑕疵判斷,雖可改善此類問題,但其瑕疵檢測之演算法則顯得更為重要。 本研究發展傅立葉轉換方法及多帶通濾波器方法作為影像處理方法,多帶通濾波器可藉由帶寬調整來濾除指定之頻率元素,並使用Canny運算子進行邊緣偵測並透過二值化方法提取瑕疵主體,最後透過型態學取得瑕疵位置,並標記與分類瑕疵,經實驗驗證,以60張640x320大小的觸控面板影像進行瑕疵檢測,其正確率為96%,執行速度平均為0.15秒。
In recent years, the world of the touch panel is in full swing to develop related industries. Its application quite widely, such as: ATM, automotive electronics, medical, sports, appliances, computers or will smart phones. Due to the capacitive touch panel manufacturing process complicated, touch sensor for the function of the touch panel core, so the production quality will greatly affect the overall quality of the touch panel and the proportion of the total cost. The capacitive touch panel of common surface defects had cracks, scratches, dust and other foreign matter, etc.. Because the touch panel of the surface texture of its existence, such as the way of manual inspection, vulnerable to microscopic flaw or testing personnel fatigue, different cognitive causing mistake. If changed to carry out automated machine vision defect judgment, although these problems can be improved, but the defects detection of the algorithm is much more important. This research resent several methods for capacitive touch panel surface defect detection. First, Fourier Transform Method to convert texture image and Multi-Band-Pass Filter method filtered out regular texture. Second, based on optimal parameters for Canny method and threshold method, the defects are detected. Finally, Morphology method could be defects more obvious and marked off the Round on defects. The experimental results use sixty pieces and 640x320 image size of touch panel image, It can know it examines accuracy above 96% and carry out the speed on average for 0.15 seconds.