Spatial and Spectral features for Horticulture mapping

Abstract

Remote sensing data allows the continuous mapping of horticultural crops in periurban areas. These are very important for their functions in the provision of local food and for other ecosystem services they provide. This work presents a methodological development and the results of a hierarchical classification of the horticultural periurban area of Cordoba city, based on the spectral and spatial properties of satellite imagery. The methodology present is automatable, making it suitable for continuous monitoring. The classification obtained with the RF algorithm yields a global kappa of 0.77 and in particular for the horticultural class a precision of 0.82. With a hierarchical classification only of the horticultural area result in an amount of 1860 ha. With spectral information taken in radiometer fields campaigns evaluated by spectral angle mapper, we can observe as using Sentinel 2 spectra and parrot camera produce better separability of horticultural crops that the hyperspectral one.