Predictive Model for Optimum Fruit Maturity Grading

Abstract

Metabolic changes to the climacteric fruits like Mango during the post-harvest lifecycle has significant impact on the marketability of the fruit. One such change is in the structure of the cuticle layer on the surface of the mango and its composition changes as the fruit matures and expands. By looking at the structure of the cuticle the right maturity stage of the fruit can be determined which can decrease storage costs and increase market outcomes. The objective of this paper is to classify the mango fruit according to its maturity stage by looking at the cuticle structure of the mango fruit images as seen using light microscope under magnification. The data set consists of structural microscopic images of 2 varieties of mango viz. Banganapalli and Kili Mooku/Banglora and the classification is done using Convolutional Neural network (CNN) based on deep learning techniques. The combined results of both varieties yield a classification accuracy of 83% for all maturity stages. With this neural network model one can identify the ripening stage of the mango in a non-destructive manner that can greatly improve the mango harvesting strategies.