Prediction of Lycopene Content in Late Mature Tomato Based on NIR Spectroscopy and siPLS

Abstract

In order to detect the lycopene content in late mature tomatoes accurately, the synergy interval partial least squares (siPLS) was used to establish the prediction model based on NIR spectra. The sampling set included ten cultivars which were selected by the content of lycopene in advance. Four abnormal samples were eliminated from calibration set using “Spectrum Outlier diagnostic” and “Leverage diagnostic”. The best processing method was determined by comparing the influence to the model parameters in a variety of spectral preprocessing method. The whole wavelength range was divided into 70 intervals by the synergy interval method, the best combination was 12, 32, 47, 59, 13, 46. The model’s correlation coefficient (R) was 0.927, the root mean square error of calibration (RMSEC) was 8.64μg/g, the root mean square error of prediction(RMSEP) was 9.19μg/g, the prediction accuracy was 92.4%. Results indicate that it is feasible to make non-destructive and rapid diagnostics of lycopene content using the prediction model.