Characterising and mapping vineyard canopy using high-spatial-resolution aerial multispectral images

Abstract

Airborne digital images of vineyards have potential for yielding valuable information for viticulturists and vineyard managers. This paper outlines a method of analysing high-spatial-resolution airborne images of vineyards to estimate physical variables of individual grapevines in terms of local canopy shape and size. An algorithm ("Vinecrawler") has been developed to identify individual vine rows and extract sets of reflectance values (or combinations thereof) at quasi-regular distances (approximately one pixel length) along the rows. Key vine canopy variables, including size, foliage density and shape, were calculated from the sets of reflectance values collected by Vinecrawler. The algorithm precisely identifies individual vines, allowing conversion from image coordinates (x-pixel, y-pixel) to a (row, vine) coordinate system. The (row, vine) coordinate system is a valuable tool for directing vineyard managers to particular phenomena identified from variables returned by Vinecrawler. This paper describes the computational methods used to identify vine rows in raw airborne digital imagery and the operation of the Vinecrawler algorithm used to track along vine rows and extract vine canopy size and shape descriptors and locational information.