Tree Species Identification
Identifying tree species in RGB aerial images and terrestrial LiDAR using Deep Learning
Background
Global climate change is affecting ecosystem functions and the adaptive capacity of forests at an accelerating rate, therefore, it is essential to improve forest resilience. The basis for the above is an intimate knowledge of our forest ecosystems, including a timely, and effective monitoring of forest development. Hence, under a dynamic climate, monitoring species richness will be critical to designing policies that ensure the provision of ecosystem services.
Project aim
We will develop models and algorithms capable of accurately detecting tree species in the over- and understory of forests. Specifically, we aim to: 1) develop deep-learning models based on high-resolution remote sensing imagery for overstory tree species recognition; 2) assess the transferability of these models to larger scales; and 3) evaluate the capabilities of terrestrial laser scanning for understory sapling species identification.
Scientific and social context
In recent years, tree species mapping has become a hot topic among the scientific community and practitioners, driven by increasing awareness regarding the importance of maintaining forest resilience to extreme events. The outcome of this research will provide the first automatic tree species detection model that can be applied across geographic regions. The automatic species detection will serve as a basis for improved inventory planning, for the planning of forest operations, and to ensure targeted management activities for more resilient forests.
Project lead
Dr. Mirela Beloiu & Prof. Verena Griess, ETH Zurich
Collaborators:
- Dr. Ionut Sandric, University of Bucharest
- Dr. Nataliia Rehush, Swiss Federal Institute for Forest, Snow and Landscape Research WSL
- Prof. Dr. Arthur Gessler, Swiss Federal Institute for Forest, Snow and Landscape Research WSL
- Dipl. Ing. Peter Surovy, Czech University of Life Sciences Prague
- Dr. Arnadi Murtiyoso, Forest Resources Management, ETH Zurich
Students:
- Zhongyu Xia (PhD student, ongoing)
- Lucca Heinzmann, Cécile Reichmuth (Master students 2022-2023)
Funding
Swiss National Science Foundation
COST Action: CA20118 - 3DForEcoTech - Three-dimensional forest ecosystem monitoring and better understanding by terrestrial-based technologies
Contact
Dr. Mirela Beloiu (mirela.beloiu(at)usys.ethz.ch)
Publications:
Beloiu, M.; Heinzmann, L.; Rehush, N.; Gessler, A.; Griess, V.C. Individual Tree-Crown Detection and Species Identification in Heterogeneous Forests Using Aerial RGB Imagery and Deep Learning. Remote Sens. 2023, 15, 1463. external pagehttps://doi.org/10.3390/rs15051463
