2019 ASHS Annual Conference

The potential of near-infrared (NIR) hyperspectral imaging for rapid evaluation of starch concentration (SC) in potato and sweet potato was investigated. The hyperspectral images of both samples were obtained, then the resulting reflectance spectra (R_S) were corrected and transformed into absorbance spectra (A_S), and exponent spectra (E_S). Full wavelength partial least squares regression (PLSR) models were established based on spectral profiles with measured reference values. Six groups of feature wavelengths were chosen from R_S, A_S and E_S based on two feature selection methods including regression coefficient (RC) of PLSR and the first derivative and mean centering iteration algorithm (FMCIA), and were successively used to build simplified models. The optimal models were obtained using FMCIA on the basis of the E_S. After further reducing the number of feature wavelengths, only six wavelengths (1028, 1068, 1135, 1208, 1262 and 1460 nm) were selected and utilized to develop the simplest FMCIA-Es-PLSR model for predicting SC, yielding a high accuracy with R²_P of 0.963 and RMSEP of 0.023. In addition, the SC on potato and sweet potato were visualized based on an equation to apply the simplest models to spectral images.