External Defect classification of Citrus Fruit Images using Linear Discriminant Analysis Clustering and ANN classifiers

摘要

Linear Discriminant Analysis (LDA) is one technique for transforming raw data into a new feature space in which classification can be carried out more robustly. It is useful where the within-class frequencies are unequal. This method maximizes the ratio of between-class variance to the within-class variance in any particular data set and the maximal separability is guaranteed. LDA clustering models are used to classify object into different category. This study makes use of LDA for clustering the features obtained for the citrus fruit . images taken in five different domains. Sub-windows of size 40x40 are cropped from the citrus fruit images having defects such as pitting, splitting and stem end rot. Features are extracted in four domains such as statistical features, fourier transform based features, discrete wavelet transform based features and stationary wavelet transform based features. The results of clustering and classification using LDA and ANN classifiers are reported