Towards achieving online prediction of starch in postharvest sweet potato [Ipomoea batatas (L.) Lam] by NIR combined with linear algorithm

Sweet potato [Ipomoea batatas (L.) Lam] is one of the primary sources for producing high-quality starch characterized by large particles and high viscosity, and has been widely used as suitable raw materials for industrial production purposes. To quantify the starch content in postharvest sweet potatoes for industrial application, an online method based on near-infrared (NIR) data combined with chemometrics was developed using 650 samples for calibration and internal validation, and 50 samples for independent external validation. Seven informative wavelengths (910, 959, 1197, 1215, 1450, 1468 and 1699 nm) associated with the prediction of starch in 900–1700 nm range were further selected by competitive adaptive reweighted sampling (CARS) algorithm to relate to measured starch values using linear algorithms, achieving good validation performance to predict starch of sweet potato with correlation coefficients of 0.94 and error of 1.26 g/100 g. The developed NIR-based method is simple, convenient, efficient and promising. It can be applied for real-time online determination of starch content in sweet potatoes after harvest to further use in food and other industry.