Abstract:

Based on the precision of quantitative analysis models established by application of partial least squares (PLS), the effects of average and original spectra as well as 10 spectral data preprocessing methods on near infrared spectroscopy (NIRS) nondestructive determination of total sugar (TS) content in tomatoes were analyzed. The results suggested that PLS correction model based on averaged spectrum is obviously superior than that based on original spectrum, and the constant offset elimination (COE) is the best data preprocessing for NIRS nondestructive determination of tomato TS content. In 11998.9 to 7497.9 cm-1 and 4601.3 to 4256.5 cm-1 spectral region, the correlation coefficient (R) between prediction and actual values of PLS model is 0.917, the standard error of calibration (RMSEC) is 0.263%, and the standard error of prediction (RMSEP) is 0.236%. In conclusion, the average spectra and the optimized spectral data preprocessing method can effectively increase the precision of NIRS nondestructive determination of tomato TS content.

Key words: tomato, near infrared spectroscopy (NIRS), partial least squares (PLS)