A Classification Algorithm for Steel Bar in Coil using Wavelet Transform

There are many ways for detecting defects and classification and these methods have been applied to many areas of industry such as fabric or steel or etc. This paper proposes a method to classify defects of steel Bar In Coil (BIC) which has cylindrical shape. The wavelet transform has been used to detect or classify defects of images recently and the proposed classification algorithm uses wavelet transform to extract features of defect. The performance of classification using wavelet transform is closely related to the selection of the wavelet. A kind of standard wavelet is selected on the base of defect analysis and the selected wavelet decomposes the first order derivative of defects into wavelet function space according to the frequency. The wavelet transform is based on subband coding method and it separates frequency of input data. There are three defects which could affect the quality of BIC severely and each defect class has different frequency concentration. The proposed classification algorithm uses detail coefficients through one to three level 2D wavelet transform as a feature of distinguishing each defect. The result shows 94.8% classification accuracy. Keywords—Bar In Coil (BIC), Classification, Wavelet Transform.