Crop Classification Using LSTM Neural Networks

This paper considers the problem of crop classification based on time series data obtained from a mixture of satellite data acquisition methods, namely synthetic aperture radar (SAR) and electro-optical (EO) sensors. Previous works have considered various machine learning approaches for this problem, such as support vector machine (SVM) and multilayer perceptron (MLP) models. While the existing works demonstrated the promise of these aforementioned approaches, there is still room to improve. In this paper, a long-short term memory (LSTM) approach is considered, and performance is analyzed with respect to the MLP approach. The examined approach is shown to significantly outperform existing crop classification methods.