BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//talks.staging.osgeo.org//foss4g-2024//talk//ACVK7R
BEGIN:VTIMEZONE
TZID:-03
BEGIN:STANDARD
DTSTART:20000101T000000
RRULE:FREQ=YEARLY;BYMONTH=1
TZNAME:-03
TZOFFSETFROM:-0300
TZOFFSETTO:-0300
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-foss4g-2024-ACVK7R@talks.staging.osgeo.org
DTSTART;TZID=-03:20241204T174500
DTEND;TZID=-03:20241204T181500
DESCRIPTION:Forest ecology research often requires detailed forest inventor
 y data at the individual tree level\, but such data are time-consuming and
  costly to collect using traditional ground-based manual survey methods. R
 ecent advances in uncrewed aerial vehicles (UAVs\, or drones)\, image proc
 essing\, and deep learning are enabling a new era of forest research in wh
 ich individual trees can be mapped\, measured\, and taxonomically identifi
 ed across broad areas without extensive ground surveys. The Open Forest Ob
 servatory (OFO\; openforestobservatory.org) is a new multi-institution org
 anization that makes cutting-edge forest mapping tools and data accessible
  to ecologists and practitioners without extensive specialized computing b
 ackground. Open-source OFO tools simplify and automate tasks including: (a
 ) processing drone imagery into 3D canopy models and stitched imagery mosa
 ics\, (b) performing individual tree detection\, geospatial crown delineat
 ion\, and height measurement from drone-derived canopy height models\, and
  (c) obtaining taxonomic classification of detected trees from raw drone i
 mages (including multiple views of each tree) using deep learning and 3D g
 eometric reasoning. The OFO also hosts an extensive public database of raw
  and processed drone imagery from western U.S. forests (> 35 km2) across b
 road gradients in forest structure\, species composition\, and disturbance
  history\, and > 100 field-based individual tree maps used for developing 
 and validating the drone-based mapping tools. The growing database is avai
 lable to host community-contributed datasets from forests globally. In rel
 atively challenging (dense and structurally complex mixed-conifer forest c
 onditions\, current OFO overstory tree detection algorithms achieve precis
 ion and recall of 70-90%\, and current tree height estimation achieves R2 
 of 0.95. In a challenging cross-site task\, preliminary tree species class
 ification using OFO multi-view computer vision tools achieved 76% accuracy
  across five species\, compared with 54% accuracy of a baseline using a si
 ngle top-down view from a stitched imagery mosaic. All tools and data are 
 free for use by anyone to address ecology questions or build on the tools\
 , and the OFO welcomes collaborations and contributions to data and code. 
 Some current development priorities include (a) expanding multi-view mappi
 ng tools to support tree detection using computer vision\, (b) optimizing 
 tree detection and species classification algorithms across broad gradient
 s of forest structure and species composition\, and (c) developing cloud-n
 ative workflows for automated cataloging and processing of contributed dro
 ne-based and field-based forest data.
DTSTAMP:20260505T124500Z
LOCATION:Room IV
SUMMARY:Open Forest Observatory: Open-source drone-based forest mapping too
 ls and data for ecologists - Derek Young
URL:https://talks.staging.osgeo.org/foss4g-2024/talk/ACVK7R/
END:VEVENT
END:VCALENDAR