Piezoelectric transducers for real-time evaluation of fruit firmness. Part II: Statistical and sorting analysis

This paper presents a statistical method for the calibration of an acoustic technique for the real-time eval-uation of fruit firmness. The technique uses an experimental setup based on two standard piezoelectrictransducers and exploits two novel stiffness indexes developed in the first part of this paper. Exten-sive experimental measurements show good correlation (r = 0.930, R2= 0.865) between the proposednon-destructive test and the traditional destructive Magness-Taylor test. An evaluation of the statisticalsignificance (t-test) of the obtained regression model parameters has been performed and validates themethod. The presented sorting analysis complements the physical detection techniques presented in thefirst part of the paper, allowing to classify individual kiwifruits with high accuracy and high predictionrate (~90%). The technology is suitable for industrial real-time and in-line applications aiming to improvewarehouse stock management and market stock uniformity.