<論文>ウェーブレット変換を用いた画像処理による欠陥検出

概要

論文の詳細を見る
In this paper, we proposes a detection of defects of industrial products with screw by using image processing with wavelet transforms. The conventional method had adopted binarization judging system, but 3-dimensional information such as surface shapes is lost by this method, therefore, defects which appeared in the surface was undetectable. Then the method using image data quantized to 256 levels is proposed. Next, the wavelet transforms can decompose a signal into low frequency and high frequency and select a required ingredient. The selected ingredient is compared with the average of normal products, and the defects are detected through the thresholding method to frequency and pixels. Experimental results show that the detectable rate of the screw defects reached 100% although a little incorrect detection occurred.
神戸市立工業高等専門学校の論文
2003-03-01