LAUSR

Fusarium-damaged kernels (FDK) contain a wide spectrum of mycotoxins, affecting the quality and safety of wheat used as food and feed. At present, traditional methods to detect FDK are time-consuming and laborious. Therefore, we propose herein a new spectral classification index (NSCI) method that can provide simple and low-cost FDK detection by analysing spectra in the wavelength range 350–2500 nm. The proposed index was based on the spectral reflectance and its first derivative. Frequency histograms were plotted for each class of index value, and Gaussian curve fitting was carried out for each histogram. Wheat kernels were then classified by using the intersection of the Gaussian curves as a threshold. The classification of NSCI for spectral data obtained the detection accuracy of 0.97, with a specificity of 0.99, a sensibility of 0.93 and a training time of 15.07 s. Compared with other spectral indexes and machine learning methods, the NSCI was more equilibrated in terms of efficiency and accuracy. Meanwhile, the threshold could be tuned to adjust accuracy, sensitivity or specificity to satisfy different practical needs. We also applied the NSCI for kernel hyperspectral data in another year, and the classification results is promising. The proposed method has the potential for the rapid and simple detection of FDK in wheat.