Abstract: It proposes a method that mixture feature aberrance fabric defects can be recognized by constructing self-adaptive orthogonal wavelet. Firstly, optimize object of designing wavelet should be assured. Then, two-passage method can be adapted to reconstruct the architecture of quadrature mirror filter (QMF) in accuracy. The object function is exported, and set up the function relationship between optimize object and QMF coefficient through object function. Finally, the constructed self-adaptive orthogonal wavelet can be applied to recognize three kinds of mixture feature aberrance defect, which can verify the feasibility of this method through two-layer disperse wavelet decomposition.

Key words: self-adaptive orthogonal wavelet, mixture feature aberrance defect, feature extraction, edge enhancement, pattern recognition