TanDEM-X Forest Mapping Using Convolutional Neural Networks

Antonio Mazza,Francescopaolo Sica,Giuseppe Scarpa,Paola Rizzoli

doi:10.3390/rs11242980

Abstract

In this work, we face the problem of forest mapping from TanDEM-X data by means of Convolutional Neural Networks (CNNs). Our study aims to highlight the relevance of domain-related features for the extraction of the information of interest thanks to their joint nonlinear processing through CNN. In particular, we focus on the main InSAR features as the backscatter, coherence, and volume decorrelation, as well as the acquisition geometry through the local incidence angle. By using different state-of-the-art CNN architectures, our experiments consistently demonstrate the great potential of deep learning in data fusion for information extraction in the context of synthetic aperture radar signal processing and specifically for the task of forest mapping from TanDEM-X images. We compare three state-of-the-art CNN architectures, such as ResNet, DenseNet, and U-Net, obtaining a large performance gain over the baseline approach for all of them, with the U-Net solution being the most effective one.

Highlights

Forests are of paramount importance for the Earth’s ecosystem, since they play a fundamental role in reducing the concentration of carbon dioxide in the atmosphere and in regulating global warming
The dataset of bistatic TanDEM-X images used for the current work was acquired over the state of Pennsylvania, USA during the first year of the mission and belongs to the global dataset of nominal acquisitions used for the generation of the TanDEM-X digital elevation model (DEM)
We have explored the use of convolutional neural networks for the purpose of forest mapping from TanDEM-X products

Summary

Introduction

Forests are of paramount importance for the Earth’s ecosystem, since they play a fundamental role in reducing the concentration of carbon dioxide in the atmosphere and in regulating global warming. The study of deforestation and development of global forest coverage and biomass is necessary to assess how forests impact the ecosystem In this framework, remote sensing represents a powerful tool for a regular monitoring at a global scale of vegetated areas. As well known, passive imaging systems are useless under cloudy conditions, whereas synthetic aperture radar (SAR) systems, providing a continuous large-scale coverage ranging from mid- to very high-resolution, can operate effectively regardless of weather and daylight conditions. This feature is important for tropical zones which are characterized by heavy rain seasons. The derived product has been made available in May 2019 and can be downloaded free of charge for scientific use [12]

Methods

Results

Conclusion