Exploiting the Temporal Dimension of Remotely Sensed Imagery with Deep Learning Models

Abstract

The rapid rise of artificial intelligence and the increasing availability of open Earth Observation (EO) data present new opportunities to address important global problems such as the proliferation of agricultural systems which endanger ecological sustainability. Despite the plethora of satellite images describing a given location on earth every year, very few deep learning-based solutions have harnessed the temporal and sequential dynamics of land use to map agricultural practices. This paper compares different approaches to classify agricultural land use exploiting the temporal and spectral dimensions of EO data. The results show greater efficiency of the presented deep learning-based algorithms compared to state-of-the-art approaches when mapping agricultural classes.