Web Surface Defect Inspection Based on Singularity Detection

Conventionally, defect detection of web surfaces has been performed by hardware solution for thresholding and matched filters. These techniques have made it possible to detect only the most basic defect types. In this paper, an efficient web surface defect detection method based on singularity detection was proposed. The proposed method was performed as follows: the original image was firstly smoothed by a 2-D Gaussian function and then we used the first order partial derivatives of the Gaussian function, which was known as local extrema approach, to record the values and angles of each modulus maximum at the corresponding location. The method is applied to some web surface defect images and the results are satisfactory and demonstrate good immunity to background textural noises. This method is targeted for web surface defect inspection but has the potential for broader application areas such as steel, wood and fabric defect detection.