A precision viticulture UAV-based approach for early yield prediction in vineyard

Abstract

Early yield prediction is a key factor for vineyard management optimization, to obtain the desired grape production and quality. Ground observation and manual weighing are time consuming and frequently provide low representative data. The aim of the study was to develop an automated early yield estimation system (5 weeks before harvest) using high-resolution RGB images, acquired through an unmanned aerial vehicle (UAV) platform in a representative zone of vigour variability of the whole vineyard. An unsupervised recognition algorithm was applied to derive the number of clusters and size, which have been used to estimate production. This fast and accurate methodology, which operated with a low cost setup, has shown high accuracy in yield prediction, providing interesting potential to support grape production management both in vineyard and in cellar.