Detection of Ripening stage of Banganapalle Mango using KNN method on PCA-reduced EIS Data
Dibakar Roy, Avishek Adhikary
Abstract
Banganapalle is one of the most common and popu- lar subspecies of Mango in India and an geographical indication of Indian state Andhra Pradesh. Mango being a senescent type fruits spontaneously goes into the senescent stage within a few days depending upon its harvest time ripeness. In this work, two non-destructive schemes are presented to assess the ripening stage of a Banganapalle mango and to predict the number of days left for it to enter the senescent stage. The first scheme uses electrical equivalent model parameters of Banganapalle mangoes to assess the ripeness. The parameter values are estimated from electrical impedance spectroscopy (EIS) using ZView software. The second scheme directly uses the EIS data after reducing it by principle component analysis (PCA). For both schemes, two different machine learning (ML) classifiers, i.e., support vector machine (SVM) and k-nearest neighbour (KNN) are used and their performances are compared. It is found that the KNN classifier on PCA-reduced EIS data provides the highest true positive rate (TPR) and Fl score (0.86) in detecting the stage of ripening based on 582 measurements on 28 samples. Proposed techniques are validated using ethylene data for reference.