An Automatic Predictive Model for Sorting of Artificially and Naturally Ripened Mangoes

Abstract

There is a fact that mangoes are being ripened using artificial ripening agents like—ethrel (ethenol) and calcium carbide and a well-known carcinogenic; consuming such mangoes might cause cancer. In this paper, a model for classification of artificially and naturally ripened mangoes based on their images using machine learning techniques is proposed. An extensive experimentation has been conducted on the dataset consists of 2440 images based on two varieties of mangoes locally termed as badami and raspuri, which were naturally and artificially ripened. Artificially ripened was further classified based on the artificial ripening agent—calcium carbide or ethrel solution. The model is based on the extraction of best color feature, best texture feature and also their fusion. To express the efficacy of this model, the best color and texture features are chosen based on the highest match of results which were obtained from five well-known classifiers like LDA, Naive Bayesian, kNN, SVM and PNN. The obtained accuracy was between 69.89 and 85.72% in all categories by choosing the best color features. Similarly, best texture features gave an accuracy between 75 and 79%. It has also been observed that accuracy decreases as the fusion of color and texture features is chosen for classification.