关键词: 平整度, 客观评级, 织物褶皱, 傅里叶变换, 支持向量机

Abstract:

In order to solve the problems that the manual evaluation of fabric smoothness is subjective and the existing objective evaluation has low accuracy, a fabric smoothness evaluation method based on Fourier transform, spectral feature extraction and support vector machine was proposed. Firstly, the images of the standard replicas and fabric samples were acquired; the obtained images were preprocessed and transformed to the Fourier frequency domain; and a set of low-pass filters were constructed in the frequency domain, and the frequency interval of the wrinkle information in the spectrogram was determined by frequency domain filtering and inverse Fourier transformation, which is called the wrinkle contribution interval. The wrinkle contribution interval was divided into several feature subintervals, and the spectral amplitude in each interval was integrated, which is constructed into the feature vector. The feature vector set was constructed by the feature vectors of all training samples, and utilized to train the support vector machine, which can objectively evaluate the fabric smoothness. The total number of the adopted fabric samples is 132, among which 24 and standard template images are taken as the training set, and the rest are taken as the test set. The results show that the algorithm performs well on fabric smoothness evaluation, and the evaluation accuracy rate is up to 96.30%.

Key words: smoothness, objective evaluation, fabric wrinkle, Fourier transform, support vector machine