微精密鋼珠主要應用於線性滑軌、滾珠螺桿與軸承等精密零組件產品,鋼珠表面品質決定零組件壽命,常見的瑕疵有生鏽、刮傷與破裂。本研究目的為應用機器視覺於鋼珠表面作瑕疵偵測,因鋼珠體積微小且為球體構造,對於微小細緻的表面瑕疵,本研究採用高畫素鏡頭及相機,以多次取像後,降低鋼珠球面之漏檢率。本研究建立一套微精密鋼珠表面瑕疵檢測系統,其包含鋼珠驅動機構、光學檢測模組與檢測軟體,運用兩支不同運轉特性之螺桿驅動機構,推動鋼珠前進與翻轉,透過不同光學設備組合將鋼珠表面瑕疵予以突顯,最後將鋼珠影像資訊傳送至自行設計之檢測軟體作系統判斷,辨識鋼珠表面品質優劣。 為驗證檢測系統之瑕疵偵測能力,本研究運用Mean Shift與Watershed兩種影像處理方法作為對照組;其中,Mean Shift用於影像前置處理,Watershed則用於影像瑕疵偵測,最後本文做一個系統穩定性實驗,驗證檢測方法之可行性。實驗結果顯示Mean shift具有保留瑕疵邊緣,去除雜訊之優點,Watershed對於瑕疵影像只需得到部分資訊,即可達到偵測之功用,而本研究建構之機器視覺檢測系統可有效地偵測大部分鋼珠表面之細微瑕疵,達到自動化檢測之目的。
Miniature steel balls are important components to mechanical devices such as linear motion guide, ball screw and bearing. In particular, the quality of a steel ball surface determines the lifespan of itself. There are many kinds of miniature steel ball surface defects, including dent, scratch, rust and rupture, which made by different causes such as metal fatigue. The purpose of this study is to inspect and analyze the steel ball surface defects using machine vision. Steel ball, which is a sphere, can be expanded into many 2D flat planes. This study use high-resolution lens and cameras to capture the surface of a steel ball plane by plan. By means of that, the defects of the miniature steel ball can be detected. In order to verify the ability to detect of inspection system, this study proposes to two DIP methods: Mean Shift method and Watershed method. The Mean shift method is applied image pre-process and enhancement, while the Watershed method is used to detect the surface defect. Then, a design of experiment is conducted to ensure the stability of inspection system. The experimental results show that the Mean Shift method has retention the edge of defects and removing noise, and the Watershed method can detect defects with only little information. The proposed machine vision-based inspection system can detect most of surface defects of miniature steel balls successfully.