精密鋼珠(Steel ball)的應用普遍存在於多種日常用品中,而在精密工業中更是一個不可或缺的元件。目前鋼珠主要應用於精密軸承之製造,鋼珠的表面品質與精密度將主宰軸承運轉的磨耗情形以及使用壽命,進而影響整個機械設備的功能,因而顯示鋼珠品質檢測的重要性。鋼珠表面瑕疵包含表面凹陷、過度粗糙以及因金屬疲勞產生的碎裂等,在傳統的檢測方法中多採用人眼檢視方式,容易因疲勞及個人主觀判斷造成誤判或漏檢。 本研究之目的在運用電腦視覺對精密鋼珠表面之瑕疵做檢測與分析,主要以電腦視覺配合適當之光源取得鋼珠之單面(或部份)影像,再藉由影像處理之技術強化瑕疵之型態,並將瑕疵擷取並做分析取得其特性,最後以類神經網路分類之工具。實驗結果顯示,已經突破以往鋼珠需要的複雜前處理,在倒傳遞類神經網路分類器瑕疵辨識率也可到達九成。完成取像部分設計後,利用TRIZ的方式設計出手動檢測的機構,最後用本研究的取像架構搭配自動化控制元件,建立出一套鋼珠自動檢測機構。
Steel ball, a key component of stainless steel bearings and automotive parts, is an important element in industry. Currently, the quality of steel ball surfaces massively affects the quality of the machines, such a machine tool, which use bearing as a component for translating working platform. Therefore, the inspection of the smoothness, accuracy, diameter, precision of steel balls becomes a critical issue during production. However, due to their reflective surface, the inspection task still relies on human eyes nowadays and requires highly experienced operators to spot the tiny defect under a microscope. Building a computer vision inspection system composed with a high resolution micro-CCD camera, a progressive frame grabber, a lighting device, a designed steel ball rotation stage, and a self-developed defect detection software, we expect to achieve an automated visual inspection with high effectiveness and efficiency.