Icon Legend

This session is not in your schedule.

This session is in your schedule. Click again to remove it.

119 Views

Organized Oral Session

Session: : The Future of Ecological Monitoring Is Collaboration with Artificial Intelligence

OOS 10-5 - Automating forest monitoring with high resolution remote sensing and AI

Monday, August 7, 2023

4:30 PM – 4:45 PM PDT

Location: 253

Presenting Author(s)

EL

Emily R. Lines

University Lecturer
Department of Geography, University of Cambridge
Cambridge, England, United Kingdom

Co-author(s)

PR

Paloma Ruiz-Benito, PhD

Profesora Titular Universidad
Departamento de Ciencias de la Vida, Universidad de Alcalá
MA

Matt Allen

PhD Student
Department of Geography, University of Cambridge
SG

Stuart W. D. Grieve, PhD

Senior Lecturer
School of Geography, Queen Mary University of London
HO

Harry J. F. Owen, PhD

Postdoctoral research Associate
Department Of Geography, University Of Cambridge

The recent explosion in availability of high resolution remote sensing technologies producing high resolution 3D data (such as LiDAR) and imagery has led to significant interest in using them for forest monitoring. The level of detail contained in the entire 3D shape of trees, fully captured within these data, can generate a wide range of metrics of interest to ecologists far beyond what can be captured with traditional plotting methods. However, the labour intensive nature of data processing presents a series of hurdles that lead to large amounts of 3D scanning data remaining unprocessed or underused. Artificial intelligence approaches may hold the key to unlocking the value of these data, but their effective application requires cross-domain collaboration to fully exploit both what the data can tell us, and what these techniques can achieve. Here we present results of work applying semantic and instance segmentation, object detection, and classification to automate the extraction of ecologically relevant properties from remote sensing data from a range of diverse forests across the major European biomes. We discuss data challenges and transferability of models across sensors and ecosystem types, and how these will be solved to move towards fully automated monitoring for ecologists.