Linear and non-linear pattern recognition models for classification of fruit from visible-near infrared spectra

Environment and genotype affect the composition, quality, storability and s ensory properties of plant-based products. Visible-near infrared (NIR) spec tral measurements are used increasingly to monitor fruit properties such as maturity, sensory properties and storability non-destructively both prior to harvest and during storage. To explore this problem, at harvest and afte r storage, visible-NIR spectra containing 1024 individual data points were measured on kiwifruit berries sourced from six pre-harvest fruit management treatments. These raw spectra were processed by principal component analys is (PCA), or by Fourier, Hartley, Haar, Hurst, range renormalisation or pol ar coordinate transforms (PCT) in order to extract a smaller set of feature s selected independently of treatment. In order to reduce their dimensional ity further, the extracted features were processed by canonical variate ana lysis. The ability of various connectionist and linear discrimination patte rn recognition models to predict the treatment source of unknown fruit on t he basis of these features was evaluated. Thus far, this work has establish ed that the performance of the non-linear model was shown to be significant ly better in comparison to the linear model. From these results, it has als o been shown that both the feature extraction and selection techniques have a marked effect on the ability to classify fruit by treatment source and s torage date. In general, the best classifications were based on features ex tracted using the Fast Fourier Transform (FFT) method, but the best perform ance in any single classification was given by the Haar transform (HT) in c onjunction with the scaled conjugated gradient learning method, (C) 2000 El sevier Science B.V. All rights reserved.