Crop Rotation Modeling for Deep Learning-Based Parcel Classification from Satellite Time Series

TL;DR

This paper introduces a deep learning model that captures annual crop rotation patterns for improved satellite-based parcel classification, and releases a large annotated dataset for agricultural monitoring.

Contribution

It presents the first deep learning approach modeling crop rotations and a large-scale multi-year dataset for parcel classification.

Findings

Over 6.3 mIoU points improvement over state-of-the-art

First large-scale multi-year agricultural dataset released

Model effectively captures inter- and intra-annual crop dynamics

Abstract

While annual crop rotations play a crucial role for agricultural optimization, they have been largely ignored for automated crop type mapping. In this paper, we take advantage of the increasing quantity of annotated satellite data to propose the first deep learning approach modeling simultaneously the inter- and intra-annual agricultural dynamics of parcel classification. Along with simple training adjustments, our model provides an improvement of over 6.3 mIoU points over the current state-of-the-art of crop classification. Furthermore, we release the first large-scale multi-year agricultural dataset with over 300,000 annotated parcels.