rsdtlib: Remote sensing with deep-temporal data library

For over a decade, satellite based remote sensing data have been intensively used for Deep Learning (DL) to help to identify Land Cover (LC) and Land Use (LU), and to detect urban and vegetation changes. Usually, these tasks are carried out with few samples or short and low-dimensional time series....

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:SoftwareX 2023-05, Vol.22, p.101369, Article 101369
Hauptverfasser: Zitzlsberger, Georg, Podhoranyi, Michal, Martinovič, Jan
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:For over a decade, satellite based remote sensing data have been intensively used for Deep Learning (DL) to help to identify Land Cover (LC) and Land Use (LU), and to detect urban and vegetation changes. Usually, these tasks are carried out with few samples or short and low-dimensional time series. In a recent study demonstrating urban change detection and monitoring, a windowed high-dimensional large time series (deep-temporal) was leveraged that not only considered a large amount of observations but also combined multiple modes for a higher temporal resolution. The software used in this approach for pre-processing, called rsdtlib, is described in the underlying work. It is made available to help others in the field of remote sensing to use this approach for Deep and Machine Learning (ML) solutions. The software is scalable to support a wide range of demands, including providing single observation samples, observation pairs, multiple modes, and the construction of windowed deep-temporal time series. Its output data is in a DL/ML training ready format and the software solution integrates well with existing remote sensing tools and services. The rsdtlib software is hosted on Github as an open source project to invite other researchers and practitioners in the remote sensing domain to utilize it.
ISSN:2352-7110
2352-7110
DOI:10.1016/j.softx.2023.101369