- 學位論文
鋼管鑄造及銜接瑕疵之視覺檢測技術
A Technique of Visual Inspection for Minting and Connecting of Pipes
摘要
本論文主要是利用影像視覺檢測方式來對鋼管鑄造及銜接的X光影像作自動瑕疵判讀。在X光影像方面，首先將有X光影像的底片擺至判片燈上，接著利用數位相機架定在三腳架上進行拍攝，以獲取固定大小的X光數位影像。獲取X光影像後，運用臨界值、邊緣偵測及影像外形處理等方法對X光影像做前處理，使影像的外形輪廓能達到比對分析的品質；再運用樣板比對、座標系統設定方法定義影像中有興趣的區域，即使影像中有興趣的區域與下一張影像比較，位置會上下左右地變動，系統仍能搜尋出並將其定位。最後使用影像灰階測量法、邊緣強度測定法及邊緣尋找法等，使X光影像中的鋼管瑕疵能被自動判讀顯示出來。 本文以工廠實際鑄造生產的鋼管X光檢驗影像作為樣本影像並加以實驗，結果顯示確實能達到鋼管瑕疵自動判讀的效果，所介紹的三種檢測法中平均正確率達91.34％，對於單一種鋼管特定瑕疵條件的檢測正確率甚至達100％，印證了本文所提方法的可行性。
並列摘要
This thesis utilizes visual inspection technique to find out flaws on an X-ray image for casting and connecting of pipes automatically. In acquiring X-ray image part, first we put the X-ray photograph image on a flaw-inspecting lamp, and then we use the digital camera mounted on a camera-shelf to acquire X-ray digital image of the fixed size ( the following is called X-ray image briefly). After X-ray image is acquired, at first we use the techniques including thresholding, edge detection and shape treatment to make the X-ray image good enough for being compared and analysed, and then the methods of pattern match and coorinate system setting make the region of interst (ROI) that is defined on the first image be found out on every image, although ROI on the other image appears shifted or rotated. Finally we use “measure intensity”, “caliper”, and “edge detection” techniques to find out flaws on the X-ray image of pipes automatically. After experiment with the actual X-ray images for casting and connecting of pipes, the average rate of correction of using the three techniques is 91.34%. And the rate of correction of using one of the three technique for inspecting a specific flaw in a specific kind of pipe even is 100%. These results show that the system can reach automatic flaw-inspecting, and prove the feasibility of proposed method in this thesis.