Comparison of Citrus Fruit Surface Defect Classification using Discrete Wavelet Transform, Stationary Wavelet Transform and Wavelet Packet Transform Based Features

The aim of this study is to classify the citrus fruit images based on the external defect using the features extracted in the spectral domain (transform based) and to compare the performance of each of the feature set. Automatic classification of agricultural produce by machine vision technology plays a very important role as it improves the quality of grading. Multi resolution analysis using wavelets yields better results for pattern recognition and object classification. This study details about an image processing method applied for classifying three external surface defects of citrus fruit using wavelet transforms based features and an artificial neural network. The Discrete Wavelet Transform (DWT), Stationary Wavelet Transform (SWT) and Wavelet Packet Transform (WPT) features viz. mean and standard deviation of the details and approximations were extracted from citrus fruit images and used for classifying the defects. The DWT and SWT features were extracted from 40x40 sub-windows of the fruit image. The WPT features were extracted from the full fruit image of size 640x480. The classification results pertaining to the three wavelet transforms are reported and discussed.