Abstract: The influence of illumination reduces the quality of strip image, a new kind of steel strip detection method is put forward through analysis of strip defect characteristics. First of all, the logarithm of strip image gray value is decomposed into a series of sub-graph through the wavelet transform. Secondly, every sub-graph of wavelet decomposition is treated by homomorphism, and then the center-surround difference operation is used to construct difference sub-map. On this basis, difference sub-maps are selected for image fusion and exponential operation is used to get the high contrast defect image. Finally, steel strip defect is detected through the segmentation method of maximum between-cluster variance. The result of experiments shows that this method can enhance the defect image contrast with clear image details, inhibit the effect of noise, and effectively realize the rapid detection of strip defect image.

Key words: wavelet decomposition, image segmentation, defect detection