Refine
Document Type
- Article (1)
- Bachelor Thesis (1)
Language
- English (2)
Has Fulltext
- yes (2)
Keywords
- UAV (2)
- Remote Sensing (1)
- U-Net (1)
- canopy analysis (1)
- deep learning (1)
- forest damage assessment (1)
- forest sucession (1)
- natural disaster analysis (1)
- rainforest (1)
- semantic segmentation (1)
Institute
The increasing number of severe storm events is threatening European forests. Besides the primary damages directly caused by storms, there are secondary damages such as bark beetle outbreaks and tertiary damages due to negative effects on the market. These subsequent damages can be minimized if a detailed overview of the affected area and the amount of damaged wood can be obtained quickly and included in the planning of clearance measures. The present work utilizes UAV-orthophotos and an adaptation of the U-Net architecture for the semantic segmentation and localization of windthrown stems. The network was pre-trained with generic datasets, randomly combining stems and background samples in a copy–paste augmentation, and afterwards trained with a specific dataset of a particular windthrow. The models pre-trained with generic datasets containing 10, 50 and 100 augmentations per annotated windthrown stems achieved F1-scores of 73.9% (S1Mod10), 74.3% (S1Mod50) and 75.6% (S1Mod100), outperforming the baseline model (F1-score 72.6%), which was not pre-trained. These results emphasize the applicability of the method to correctly identify windthrown trees and suggest the collection of training samples from other tree species and windthrow areas to improve the ability to generalize. Further enhancements of the network architecture are considered to improve the classification performance and to minimize the calculative costs.
With the development of new imaging and near field remote sensing UAV technologies, the field of forest canopy and tree crown investigations in the context of rainforest ecology gained new perspectives in recent years. This study demonstrates the feasibility of structural analysis of canopy for the assessment of succession states of tropical lowland rainforests using a consumer UAV and open access ground reference data. Therefore, the canopy of 22 small-scale forest plots in the Ecuadorian Andes and the Mata Atlântica were captured with an UAV between October 2017 and February 2018. The fully remote image acquisition used automated flight plans consisting of crossed flights with different incident camera angles and did not require ground control points (GCP) as further ground reference. Photogrammetric Point Clouds (PPC), image meshes and crown height models (CHM) were created with Structure from Motion (SfM), combined with a Digital Terrain Model from the Shuttle Radar Topography Mission (SRTM), in order to analyze structural parameters of the canopy. The standard deviation of the height values of the PPC and the CHM differs statistically significant between succession stages. Investigation plots in climax or late succession show a mean PPCSD of 6.7 m with a standard error of 0,3 m. Plots in early succession show a PPCSD of 4.8 m with a standard error of 0.3 m and degraded plots show 2.8m with a standard error of 0.2m. This result proves the applicability of the proposed workflow under typical NGO conditions and other applied science research with limitations in budget and access to sophisticated survey equipment.
