Leather inspection through singularities detection using wavelet transforms

Abstract

The major problem in leather inspection is to separate defects from the background exhibiting a wide range of visual appearances. Leather defects, characterized by oriented structures, cannot be easily discriminated from the structures typical of the normal surface. Though gaussian filters generally represent a successfull tool to smooth out the structures on the background, a wrong choice of the resolution can preclude the detection of defective regions (singularities) in the subsequent analysis. However, wavelet transforms can be profitably used for studying the evolution of singularities across different scales. Suitable kernels for this transform does allow multiscale singularities analysis through the detection of local maxima in wavelet transform modulus.

Complexity measure of regional seasonal precipitation series based on wavelet entropy

“…Nicknamed the "Math Microscope" (Ivanov et al 1996), the WT has the function of multi-resolution analysis for nonstationary signals and good time-frequency localization characteristics. Thus, many scholars prefer to use WT, which has been applied in various fields, such as signal singularity detection (Branca et al 1997), machine fault diagnosis (El-Zonkoly and Desouki 2011), edge detection (Ducottet et al 2004), health diagnosis and quality control of architectural structures (Bayissaa et al 2008), and stock market forecasting (Hsieh et al 2011). The WT includes the continuous wavelet transform (CWT) (Kwona et al 2005) and discrete wavelet transform (DWT) (Froese et al 2006).…”