Determination of starch content in single kernel using near-infrared hyperspectral images from two sides of corn seeds

Abstract

Rapid, non-destructive and reliable detection of starch content in single seed is significant to facilitate the breeding of high-starch corn but difficult for a traditional method of seed composition analysis. This study investigated the possibility of using near-infrared (NIR) hyperspectral imaging technology to determine the starch content in a single kernel corn seed. The hyperspectral images including embryo-up and embryo-down orientations of a corn seed were acquired with a range of 930–2500 nm. The characteristic spectrum of each corn seed was calculated by averaging the two sides’ spectra. All spectra were preprocessed by the smoothing and derivative algorithm, and then, the characteristic wavelengths were selected by competitive adaptive reweighted sampling (CARS) method. The selected wavelengths were used as the inputs to develop partial least squares regression (PLSR) and nonlinear statistical data models with artificial neural networks (ANN) algorithm. The results indicated that the ANN prediction model based on Levenberg-Marquardt algorithm (LMA) was the optimal for starch content determination with correlation coefficient (Rp) of 0.96 and root mean square error of prediction (RMSEP) of 0.98 in prediction sets. Therefore, NIR hyperspectral imaging technology combined with appropriate chemometric analysis can be considered as a useful tool for starch content determination in corn seed at a kernel level. These results can provide a useful reference for rapid and non-destructive detection of other chemical composition in single corn seed.