Convolutional Neural Networks enable efficient, accurate and fine-grained segmentation of plant species and communities from high-resolution UAV imagery

• Published in: Scientific reports Vol. 9; no. 1; pp. 17656 - 9
• Main Authors: Kattenborn, Teja, Eichel, Jana, Fassnacht, Fabian Ewald
• Format: Journal Article
• Language: English
• Published: London Nature Publishing Group UK 27.11.2019; Nature Publishing Group
• Subjects: 631/158/2178; 631/158/670; 631/158/853; 704/158/1145; Algorithms; Computer vision; Data Collection - methods; Deep Learning; Humanities and Social Sciences; Image processing; Image Processing, Computer-Assisted - methods; Learning algorithms; Mapping; multidisciplinary; Neural networks; Neural Networks, Computer; Plant communities; Plant Physiological Phenomena; Plants; Remote sensing; Remote Sensing Technology - methods; Remote Sensing Technology - standards; Satellite Imagery - methods; Satellite Imagery - standards; Science; Science (multidisciplinary); Segmentation; Species; Unmanned aerial vehicles; Vegetation; Vegetation mapping; Visual system
• ISSN: 2045-2322; 2045-2322
• DOI: 10.1038/s41598-019-53797-9

Recent technological advances in remote sensing sensors and platforms, such as high-resolution satellite imagers or unmanned aerial vehicles (UAV), facilitate the availability of fine-grained earth observation data. Such data reveal vegetation canopies in high spatial detail. Efficient methods are needed to fully harness this unpreceded source of information for vegetation mapping. Deep learning algorithms such as Convolutional Neural Networks (CNN) are currently paving new avenues in the field of image analysis and computer vision. Using multiple datasets, we test a CNN-based segmentation approach (U-net) in combination with training data directly derived from visual interpretation of UAV-based high-resolution RGB imagery for fine-grained mapping of vegetation species and communities. We demonstrate that this approach indeed accurately segments and maps vegetation species and communities (at least 84% accuracy). The fact that we only used RGB imagery suggests that plant identification at very high spatial resolutions is facilitated through spatial patterns rather than spectral information. Accordingly, the presented approach is compatible with low-cost UAV systems that are easy to operate and thus applicable to a wide range of users.