An Approach for identifying Crops Types using UAV Images in the Ecuadorian Sierra

Abstract

Spectral signature analysis allows identification of the different types of terrestrial objects and characterizes behaviour of different kinds of vegetation. In Ecuador usually phenological analysis (state of vegetal growing) and crop type are based on acquired manually information. This does not allow taking agile decisions over crops management. The advantages for using UAV images propose a significant change to the current methodologies. This paper presented a correlation study of crop spectral signature using multispectral images from a UAV. Ecuadorian Sierra was the study zone to differentiate the types of crops in an agricultural field. The Inception algorithm of Tensorflow was chosen to generate a crop layer and to predict the crop type with the closest possible approximation from an image.