Classification of Fruit in a Box (FIB) Using Hybridization of Color and Texture Features

Abstract

The purpose of this paper is to develop an effective classification of fruit in a box by considering the color and texture features from the images. Twenty fruit types with various appearances in color and texture were selected to be analyzed in this study. Although the capability of many color or texture features were previously studied in many researches, each feature cannot be used to identify the fruit type accurately enough for practical use. In this study, we combine six features, i.e., HSV Color Histogram, Color Layout Descriptor (CLD), Color Correlogram, Gray Level Co-occurrence Matrix (GLCM), Local Binary Pattern (LBP), and Neighboring Gray Tone Difference Matrix (NGTDM) to gain high accuracy of fruit-in-a-box classification. An image preprocessing stage is applied to fruit images to prepare the images in good condition. Then, six image features are extracted from each image. Finally, the fruit classification process is adopted through the well-known classification methods such as Decision Tree, Random Forest, k-Nearest Neighbor (kNN), Support Vector Machine (SVM), Logistic Regression classifier, Linear Discriminant Analysis (LDA) classifier, Naive Bayes classifier, and Multi-Layer Perceptron (MLP). After experiments were tested and evaluated, it shows that, with the appropriate classification method, the hybridization of features yields high accuracy with independence of classification method and effectiveness in the classification of fruit in a box.