Use of UAV-borne multispectral data and vegetation indices for discriminating and mapping three indigenous vine varieties of the Greek Vineyard

Abstract

In line with precision viticulture, in recent years new methods of vineyard management have been introduced, so ato optimize vine cultivation and production of wine of the highest quality. Following on the methodologies developefor mapping other crop parameters, there is currently a growing research effort for the discrimination and mapping ovine varieties, as this information is useful for vineyard-scale management, local and regional inventory and planninpurposes, application of EU Directives, and support of certification and production of high quality wines. This researcfocuses on developing a methodology, based on UAV-borne multispectral data, for discriminating and mapping threvine varieties in Attica, Greece, employing three non-parametric classifiers, namely Random Forest (RF), Support VectoMachines (SVM) and Spectral Angle Mapper (SAM), and selected vegetation indices (VIs). The suggested methodologuses easy to obtain and process, cost-effective images and relies mostly on free open-source software. Study conclusionsuggest that although the multispectral images used did not result in the accurate discrimination of the vine varieties apixel level, expressed by highest overall accuracy (OA) 61.6%, they nevertheless proved useful in mapping varieties at thplot level. Therefore, it is considered effective for applications that require such level mapping.