An automated web surface inspection for hot wire rod using undecimated wavelet transform and support vector machine

This paper presents defect detection and classification method for hot wire rod in the steel industry. The detection algorithm is based on undecimated discrete wavelet transform (UDWT). The algorithm utilizes the translation invariant property of UDWT. To discriminate the real defects and pseudo defects, we use support vector machine (SVM). The total 14 feature attributes are extracted from binary and gray image. To select best model for SVM classier, we search the parameter spaces by exponentially growing sampling test. The experimental results show the proposed methods can be applied to real-world application.