Edmonton Tree Project
Recently, the availability and quality of regional aerial images strongly improved. For geographers, this allows for new analysis methods to monitor bio-diversity and vegetation. In particular, researchers take interest into the presence of trees including their species, sizes and health status. Novel computer vision techniques like convolutional neural networks allow for accurate object detection and classification. a
The goal of the thesis is to evaluate existing object detection methods to identify trees and predict their characteristics. In particular, the vision of this thesis is to analyse the trees on an aerial image of a forest and automatically generate statistics on occurring tree species and tree characteristics. .
Contact: Prof. Dr. Matthias Schubert
The thesis will be part of the project “Forest restoration assessment using Deep Learning for image interpretation” which is pursued within the international research network Abby-Net.