Sprouting detection at early stages in individual CWAD and CWRS wheat kernels using SWIR spectroscopy

Reflectance spectra (1,250–2,390 nm) of two classes of western Canadian wheat (CWAD and CWRS) were studied to detect sprouting damage in individual kernels at the early stages of germination. Alpha-amylase activity levels were used as an indicator for the sprouting stage. Partial least squared discriminant analysis (PLSDA) and Logistic regression methods were used to build classification models. The optimal threshold α-amylase activity value for the separation of sprout damage classes was determined according to the area under the ROC curves. The results show that both PLSDA and logistic regression could distinguish kernels with an α-amylase activity larger than 1 SKB unit of activity from those with less enzyme activity. A total classification accuracy of over 91 and 86% was obtained for CWRS and CWAD, respectively.