Optical Properties of Multilayered Tissues of Different Varieties of Apples and Inspection Models of Internal Quality

Abstract

Apples being multilayered fruits possess optical properties which respond to spectroscopy-based quality inspection models. The variation in optical properties of apple tissues is crucial for efficient spectroscopic detection. Absorption coefficients (µa) and the reduced scattering coefficients (µ's) of three apple varieties had wavelength range from 475 nm to 1600 nm using a double integrating sphere technique. The quantitative prediction models for soluble solid content (SSC), firmness index (FI), and pH for synergistic Interval (SI), competitive adaptive reweighted sampling (CARS), and genetic algorithm (GA) with partial least square (PLS) was accurate. Interestingly, CARS-PLS model using µa provided the best quantitative predictions for SSC (Rp = 0.9833, RMSEP = 0.2630) and pH (Rp = 0.8429, RMSEP = 0.1229). Additionally, the GA-PLS model based on µ's yield accurate prediction for FI (Rp = 0.9372, RMSEP = 0.1478). On the other hand, differences in apple peel color were also observed among the varieties. The qualitative discrimination model random forest (RF) and backpropagation (BP) for apple varieties color detection achieved highest accuracy (100%). These findings confirmed the feasibility of combining optical properties with color detection to identify apple variety.

Keywords: Optical properties, Double integrating sphere technique, Quality, Machine Learning, Apple