透過您的圖書館登入
IP:3.15.6.77
  • 學位論文

應用機器視覺於熱熔保險絲檢測

Applying Computer Vision in Fuses Inspection

指導教授 : 田方治
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


本論文研究對象以熱融保險絲(Fuse)為研究對象,由於熱融保險絲在鑄造研磨時會有鏟絲殘留、剝落或是撞針不足等現象,而熱融保險絲的管壁相當微薄,造成瑕疵用肉眼不易察覺。目前工廠主要依賴以人工目視檢測方式找出瑕疵,但人容易因長時間工作使眼睛疲勞而導致人為檢測疏失,降低檢測之可靠度,使出貨的品質降低,進而影響公司信譽及不符合成本效益。 本研究目的在於提出一個自動光學檢測方式,取代現行人工目視檢測,大大的提升檢測的效率,時間成本大大的縮短。其中最困難被發現的小鏟絲,也可以被正確的找尋出來。本論文自己提出一個利用統計學找離異值的檢測方式,並且與類神經網路中較著名的倒傳遞類神經網路與學習向量化類神經網路的結果相互比較。結果發現檢測結果,不論在速度上或可靠度上,都遠遠高於類神經網路檢測與人工檢測方式。 本系統檢測一顆熱融保險絲的平均時間為0.24秒,系統檢測正確率達95.94%。

並列摘要


In the highly competitive environment, an industry should provide better performance, lower cost, and lower price products to survive and win more share of the market. The automatic manufacturing process is essential to achieve this goal. In this study, we use the fuses as the samples to inspect all the defects produced in the casting process. We proposed an approach by using statistical method to find the defects, and compared to the BP and LVQ neural networks. The results indicate that the speed and the reliability of the method we proposed are better than the other two methods. It only takes 0.24 seconds to inspect a fuse in this system. The correct rate reaches about 95.94%. This system is much faster and preciser than human and neural network inspections.

參考文獻


[27] 葉郡維,料盤中之液晶顯示器驅動IC瑕疵檢測,碩士論文,國立台北科技大學自動化科技研究所,臺北,2007。
[28] 張翔昇,應用機器視覺方法於鋼珠表面瑕疵之檢測,碩士論文,私立元智大學工業工程與管理研究所,桃園,2004。
[1] Amet, A.L. and A. Ercil, “Texture Defect Detection Using Domain Co-occurrence matrices,” IEEE, Image Analysis and Interpretation, 1998, pp.205-210.
[3] Burrus, C. S., R. A. Gopinath and H. Guo, “Introduction to Wavelets and Wavelet,” Transforms A Primer, Prentice Hall, 1998.
[6] Dmitry, Chetverikov, “Texture Imperfections ” Pattern Recognition Letters, Vol. 6, 1987, pp. 45-50.

延伸閱讀