Selection of efficient wavelengths in NIR spectrum for determination of dry matter in kiwi fruit.

Abstract

The feasibility of using efficient wavelengths in the near-infrared (NIR) spectrum for the rapid determination of the dry matter (DM) in kiwi fruit was investigated. Partial least squares (PLS), synergy interval PLS (siPLS) and genetic algorithm siPLS (GA-siPLS) were comparatively performed to calibrate regression models. The number of wavelengths and the number of PLS components were optimised as per the root mean square error of cross-validation (RMSECV) in the calibration set. The performance of the final model was evaluated by the root mean square error of prediction (RMSEP) and the correlation coefficient (r) in the prediction set. Results indicate that the performance of GA-siPLS model is the best one compared to PLS and siPLS models. The optimal model was achieved with r = 0.9020 and RMSEP = 0.5315 in the prediction set. This work shows that it is feasible to determine DM in kiwi fruit using NIR spectroscopy and that GA-siPLS algorithm is most suitable in solving the problem of selection of efficient wavelengths.