Classification and Grading of Multiple Varieties of Apple Fruit

Abstract

Fruit safety plays an important role in the economy of the world for the agriculture field. Recently, it has been noticed that fruits are affected by various diseases. This creates economic failure in the agriculture field across the world. The manual investigation of multiple varieties of an apple fruit is a burdensome task that can be minimized for quality evaluation for fresh and rotten fruits using automatic computerized methods. In this work, a novel method is implemented for multiple varieties of apple fruit quality evaluation. The six different varieties of apples i.e. Fuji, York, Golden Delicious, Red Delicious, Granny Smith, and Jonagold are used for image acquisition. Firstly, images are segmented by the grab-cut method and fuzzy c-means clustering. Secondly, multiple features statistical, textural, geometrical, discrete wavelet transform, a histogram of the oriented gradient, and Laws’ texture energy are extracted and selected by principal component analysis from the feature space. Lastly, the classification of fresh and rotten apples is done by applying k-NN, LR, SRC, and SVM classifiers. The cross-validation technique with distinct values of k is used to validate the performance of the system. The proposed method achieves 92.90% (k = 5), 98.42% (k = 10), and 95.27% (k = 15) accuracy by SVM classifier. It has been seen that the extraction of proper features and selection of features results in upgraded performance. Also, the state of the art techniques are comparable with the proposed algorithm showing the probability to use for multiple fruits.