Multivariate data analysis for classification of pineapple maturity

Abstract

A range of physical and chemical parameters were investigated for their efficacy in classifying pineapple according to maturity or the presence of marbling defects. The tests were conducted on a sample of 115 “Pattavia” cultivar pineapples which were later sorted into three sets according to their differing maturity levels based on translucent yellow area of the longitudinal cut-open surface, and a further set comprising marbled fruit. The destructive parameters that were measured included soluble solids content, titratable acidity and flesh firmness. Nondestructive measurement techniques including acoustic impulse response tests and specific gravity determination by means of water replacement were carried out on each sample. Discriminant analysis using a canonical discriminant function with leave-one-out cross validation was performed to segregate the fruits into each class. The classification into maturity and marbling classes was achieved using multiple parameters with overall performance of 75.7%. The specific gravity, stiffness coefficient ( f 3 2 m 2/3 where f 3 is the third resonant frequency at the peaks of the acoustic spectrum, and m is fruit weight) and the soluble solids content appeared to be the most important parameters for distinguishing pineapples into three main groups: (1) class A (more mature fruit), (2) class B (mature fruit), and (3) class C (less mature fruit) and class M (marbled fruit). In addition, class M could be separated from class C by the titratable acidity, the flesh firmness and the stiffness coefficients based on f 1. For classification of mere maturity class, the performance was improved to 77.0%. The maturity was showed to be characterized by the specific gravity, the acidity and the soluble solids content. In differentiation of maturity using only nondestructive parameters, the specific gravity again gave the highest correlation for discriminant score of the discriminant function 1. The total classification performance was 82.8%. The stiffness coefficients complemented the classification through the discriminant function 2.