Edge Based Chinese Wine Classification Using FCM and Kernel Principal Component Analysis

Abstract

Micrographs of Chinese wines show floccule, stick and granule of variant shape and size. Different wines have variant microstructure and micrographs. A semi-supervised wine classification method based on kernel principal component analysis (KPCA) method and fuzzy C-means (FCM) algorithm using edge feature from micrographs was proposed in this paper. In this work, ten Chinese wines are determined or classified in the system, such as Wuliangye, Luzhoulaojiao, Xushui, Jiannanchun, Maotai and et.al. First, the micrographs were enhanced using Rotating Kernel Transformation (RKT) Filter, and edges of these images were detected using Canny edge detector; then the edges features were mapped into parameter space through circular Hough transform, and Kernel Principal Components Analysis was performed for feature dimension reduction. We adopted the semi-unsupervised classification method improved FCM in which the cluster centers were given. The means of feature vectors generated from each wine in training set were employed as the corresponding cluster center. We compared the recognition results for different choices of parameters in Kernel Principal Components Analysis. The experimental results showed that, the classification rate was over 95%.