This paper proposes an on-line quality inspection method to detect seam, a major type of surface defect generated in rolling processes. A feature-preserving snake-projection procedure is first adopted to convert the images of suspect seams to one-dimensional sequences. Discrete Wavelet Transform is then performed on the derived sequences with features extracted from wavelet coefficients. Finally, T2 control chart is established to discriminate between real seams and false positives. Minimization of the discrimination error based on training data is also presented with illustrations. On-line implementation of the proposed method shows that it satisfies the requirements of both detection accuracy and detection speed.

This content is only available via PDF.
You do not currently have access to this content.