Vine Identification and Characterization in Goblet-Trained Vineyards Using Remotely Sensed Images

This paper proposes a novel approach for living and missing vine identification characterization in goblet-trained plots using aerial images. Given the periodic structure of goblet vineyards, RGB color coded parcel image is analyzed proper processing techniques order to determine locations vines. Vine achieved by implementing marker-controlled watershed transform where centers vines serve as object markers. As result, precise mortality rate calculated each parcel. Moreover, all vines, even overlapping ones, are fully recognized providing information about their size, shape, green intensity. The presented automated yields accuracy values exceeding 95% when obtained results assessed with ground-truth data. unsupervised can be applied any type presenting similar spatial patterns requiring only input.