Classification of Sun‐Dried Bulk Raisins by Using Statistical Texture Features

ABSTRACT

Classification of bulk sun‐dried raisins as one of the most important export products of Iran poses a significant challenge for producers and buyers due to their visual similarity to impurities and damaged grains. This research provides an expert system to evaluate the classification of sun‐dried dark‐color bulk raisins with dark color‐based impurities by analyzing images of bulk raisins. For this purpose, a machine vision setup was employed to capture 525 images of raisins across various mixture ranges. Subsequently, textural features were extracted using three methods of Gray Level Co‐occurrence Matrix (GLCM), Gray‐Level Run‐length Matrix (GLRM), and Local Binary Pattern (LBP) to comprehensively analyze the textural properties. Various classification models including Decision Tree (DT), Discriminate Analysis (DA), Support Vector Machine (SVM), K‐Nearest Neighborhood (KNN), and Artificial Neural Network (ANN) models were applied. DA model achieved the accuracy of 98.61% and 97.92% on the test datasets for all and GLCM features, respectively. In order to optimize feature importance, Maximum Relevance Minimum Redundancy (MRMR) and Chi‐Square Test (CST) algorithms were employed, and achieved accuracies of 95.83% and 77.38% for 6‐class and 7‐class datasets, respectively. Therefore, the results of the proposed approach can be utilized in designing a system for measuring the purity and quality of raisins.