X線フィルム中の溶接欠陥像の特徴抽出とニューラルネットワークによるその種類判別

概要

論文の詳細を見る
The authors have already proposed a method to detect weld defect images from X-ray film and the effectiveness of the system to automatic inspection has reported. In this paper, a system to discriminate the defect type automatically using artificial neural network is discussed. In order to discriminate the defect type of the defect images abstracted by the image processing system, ten feature parameters that indicate characteristics of defects are selected. Then the dimensions of parameters of defect images are measured automatically and applied to the neural network as the input data. The neural network which was previously trained by typical defect image data outputs defect type automatically. As the result of discrimination experiments, it is confirmed that the system is effective in discrimination of defect type.
社団法人日本機械学会の論文
2000-03-25