Fast and Accurate Discrimination of Brachiaria brizantha (A.Rich.) Stapf Seeds by Molecular Spectroscopy and Machine Learning

Brachiaria brizantha is the most common forage plant used in cattle pastures. The field plant population is the main parameter for cattle nutrition; it is mainly determined by the genotype and physiological quality of the seed, that is, seed vigor. Seed vigor standard tests are considered time-consuming and laborious. Fourier transformed infrared (FTIR) spectroscopy was used to classify the seed cultivar and vigor from two different genotypes of B. brizantha, namely, Marandu and Paiaguás cultivars. Two batches from each group were used for classification by applying FTIR and machine learning algorithms. The algorithms with a higher overall accuracy in the leave-one-out cross-validation were also validated by an external validation using a dedicated set of samples exclusively separated for this purpose. The results indicate that molecular spectroscopy combined with machine learning analysis presents great potential for the classification of B. brizantha seeds.