Using Computer Vision to Create 3D Forest Models from Thousands of Aerial Photos

Dr. Erle Ellis – Associate Professor, UMBC GES
Dr. Marc Olano – Associate Professor, UMBC CSEE
Dr. Matthias Gobbert – Associate Professor, UMBC Math & Stat
Jonathan Dandois – Graduate Student, UMBC GES
Yu Wang – Graduate Student, UMBC CSEE
Christopher Leeney – BS/MS Student, UMBC GES
Dr. Noah Snavely – Assistant Professor, Cornell CS


Computer vision structure from motion (SfM) algorithms make it possible to generate accurate 3D reconstructions of forest canopy structure and composition using thousands of overlapping aerial photos. This process has large computation demands and current reconstructions based on thousands of images require weeks of processing in serial, making it impractical for experimentation or widespread application. Working with HPCF will improve our ability to use SfM for reconstructing forest canopies to study forest carbon and diversity and will also make it possible to develop an understanding of the functioning of these algorithms in forested landscapes.

For more information please visit our website: