Citrus fruit external defect classification using wavelet packet transform features and ANN

Automatic grading and sorting of agricultural products gain importance with the advent of machine vision technology. Features extracted from the images in either spatial or frequency domain can be used for defect classification. The research work reported in this paper describes about the development and implementation of wavelet packet transform (WPT) based image-processing algorithm applied to classify the citrus fruit external defects viz. pitting, splitting and stem-end rot. ANN was used as a classifier. The mean and standard deviation calculated for the detail as well as the approximation sub-windows of the wavelet packet transformed images of the citrus fruits were used as features. The classification results of the algorithm are reported and its limitation is discussed.