Erfassung von Stadtbäumen unter Einsatz von künstlichen neuronalen Netzen und Fernerkundungsdaten
Research output: Contribution to journal › Research article › Contributed › peer-review
Contributors
Abstract
Information on urban trees is essential, e.g. for planning purposes. While on-site data acquisition is costly and time consuming, an increasing amount of remote sensing data, including digital orthophotos or terrain models, is readily available. To make efficient use of these, the development of suitable analysis methods is crucial. In this context, various machine learning techniques, in particular artificial neural networks (ANN), are currently the subject of intensive research. This contribution summarizes the findings of a pilot study conducted in the greater Leipzig area. It investigates to what extent the potential of convolutional neural networks (CNNs) can be used to derive information on urban trees from remote sensing data. The results show that individual trees can be localized and different tree genera can be classified with an overall accuracy of up to 72 %. Furthermore, the tree age, height and crown diameter can be determined with mean error values (RMSE) of up to 9 years, 1.8 m, and 1.0 m, respectively.
Details
Original language | German |
---|---|
Pages (from-to) | 31-40 |
Journal | gis.Science - Die Zeitschrift für Geoinformatik |
Volume | 1 |
Publication status | Published - 2020 |
Peer-reviewed | Yes |
External IDs
Scopus | 85081261734 |
---|---|
ORCID | /0000-0002-3085-7457/work/154192802 |
Keywords
Keywords
- Künstliche Neuronale Netze, Fernerkundungsdaten, Stadtbäume, CNNs, CNNs