Near infrared reflectance spectrometry classification of lettuce using linear discriminant analysis

Abstract

This study proposes a methodology for lettuce classification employing near infrared reflectance spectrometry and variable selection.

Abstract

This study proposes a methodology for lettuce classification employing near infrared reflectance spectrometry and variable selection. For this purpose, genetic algorithm (GA), successive projections algorithm (SPA) and stepwise (SW) formulation were employed to choose reduced subset of variables (wavenumbers) for linear discriminant analysis (LDA) models. The proposed method was applied to a set of 104 lettuce samples of three different cultivation types (organic, hydroponic and conventional). The classification results of LDA/GA, LDA/SPA and LDA/SW models were assessed in terms of the correct classification rate (CCR) obtained for the test samples. The best results were found with LDA/GA models, achieving a CCR of 95.4% in the test set, whereas the LDA/SW and LDA/SPA models correctly classified at 77.3% and 68.2%, respectively. The results obtained in this investigation suggest that the proposed method is a promising alternative for the assessment of the authenticity of lettuce cultivation type.