Multi Features based Fruit Classification Using different Classifiers

Abstract

The fruits available naturally will be having different colors and shape in appearance. Humans can identify the type of fruit by seeing their shape and color without any difficulty. Here a practical approach has been offered in this paper to classify the fruit images based on the Color and Shape of the fruit. Five thousand images were taken from the standard Fruit-360 dataset for the experiment; the dataset contains Apple, Banana, Cherry, Grapes, and Mango. The color_moment and shape of the fruits were considered to extract the features from different fruit images. In this proposed work, three feature vectors are constructed. In the color_moment feature extraction, here statistical features such as mean and standard deviation of three-color channels (RGB) are computed. The binarized images of fruits were used to extract shape-based features, and a multifeatured vector consisting of color_moment and shape features were used. The SVM, MLP, and RF classifiers are used for the classification process. The recognition accuracy of 99.98% has been achieved using the combined feature vector (multifeatured vector) and RF classifier. This paper’s contribution is that the color_moments feature extraction is carried out directly on fruit images without using any pre-processing techniques such as gray-scale and binary conversion on fruit images.