Air bubbles detecting on ribbed smoked sheets based on fractal dimension

Air bubble impurity is one of critical defects on ribbed smoked sheets (RSSs). Only RSSs with few or without air bubbles are allowed in top industries such as wheel-manufacturing. However, air bubbles are difficult to detect because of their size, intensity and low contrast with the background. This paper proposes to apply the fractal dimension as a feature for RSS classification with regard to air bubble impurity. Images of RSS are captured in a controlled environment using a fluorescent light source. These images are resized and divided into 16 smaller areas. Then, the fractal dimension values of these areas are calculated. Finally, the standard deviation of the 16 fractal dimensions is computed for comparing with a predetermined threshold for classification. The experimentations with 500 images of RSSs of which half contains some air bubbles were conducted to verify the proposed method. The experimental results showed the classification accuracy of 98.00% using the expert grading as reference.