Comparison of Various RGB Image Features for Nondestructive Prediction of Ripening Quality of “Alphonso” Mangoes for Easy Adoptability in Machine Vision Applications: A Multivariate Approach

Abstract

AbstractA study was conducted to predict ripening quality in mangoes using RGB images. Mature “Alphonso” mangoes were selected for experiments. During ripening, random samples were chosen at 24‐h interval for imaging and quality analysis. Hierarchical clustering method was employed to classify the ripening period into five stages based on quality parameters (physico‐chemical, color and textural properties). From each image, 18 features were extracted and evaluated in the prediction of ripening stages. From linear discriminant analysis, group of normalized differential index (NDI) and area features obtained from RGB channels were found better in prediction with 14.9 and 12.3% of misclassification, respectively. While employing quadratic discriminant analysis, NDI and area features predicted the ripening quality effectively with lowest misclassification of 7.9 and 3.5%, respectively. Cross and external validation of selected models had also shown effective results (96.3% correct prediction) with these features. The proposed image features are easy to extract with simple algorithms and useful to predict the ripening quality of mangoes, particularly in machine vision applications.Practical ApplicationsMango is one of the important climacteric fruits, having great demand in the world market. India is the market leader, exporting the mango fruits in fresh as well as processed form. Since, level of ripeness plays a major role in the final product quality, mango pack houses and processing industries need an easy, rapid non‐destructive quality detection mechanism for predicting the ripeness level of mangoes in order to get the desire optimum quality of final products. This study explains about the adoptability of RGB based image processing method for prediction of ripeness level as a nondestructive method with various image features. This study will be helpful to great extend in the process of predicting the ripeness level of mango using the proposed image features with suitable for machine vision systems.