Discrimination of grassland species and their classification in botanical families by laboratory scale NIR hyperspectral imaging: Preliminary results

The objective of this study was to discriminate by a NIR line scan hyperspectral imaging, taxonomic plant families comprised of different grassland species. Plants were collected from semi-natural meadows of the National Apuseni Park, Apuseni Mountains, Gârda area (Romania) according to botanical families. Chemometric tools such as PLS-DA were used to discriminate distinct grassland species, and assign the different species to botanical families. Species within the Poacea family and other Botanical families could be distinguished (R2=0.91 and 0.90, respectively) with greater accuracy than those species in the Fabacea family (R2=0.60). A correct classification rate of 99% was obtained in the assignment of the various species to the proper family. Moreover a complete study based on wavelength selection has been performed in order to identify the chemical compound related to each botanical family and therefore to the possible toxicity of the plant. This work could be considered as a first step for the development of a complete procedure for the detection and quantification of possible toxic species in semi-natural meadows used by grazing animals.