Loading presentation...

Present Remotely

Send the link below via email or IM

Copy

Present to your audience

Start remote presentation

  • Invited audience members will follow you as you navigate and present
  • People invited to a presentation do not need a Prezi account
  • This link expires 10 minutes after you close the presentation
  • A maximum of 30 users can follow your presentation
  • Learn more about this feature in our knowledge base article

Do you really want to delete this prezi?

Neither you, nor the coeditors you shared it with will be able to recover it again.

DeleteCancel

Pest Detection Using Unmanned Aerial Vehicles

International Society of Arboriculture 2017 International Conference and Trade Show, Washington DC, USA August 2 2017
by

Dan Staley

on 15 August 2017

Comments (0)

Please log in to add your comment.

Report abuse

Transcript of Pest Detection Using Unmanned Aerial Vehicles

INTRODUCTION
Big Picture
Up to Speed
RESULTS
Preliminary Findings
DATA ANALYSIS
More Complexity
CONCLUSION
Wrap-up, Next Steps
Pest Detection Using Unmanned Aerial Vehicles
Pest Detection Using Unmanned Aerial Vehicles
ISA International 2017
Washington D.C., USA
August 2, 2017 11:20 - 12:00
Dan Staley
Arbor Drone, LLC
http://arbordrone.net
dan@arbordrone.net
All presentations on Prezi.com (search 'Dan Staley')
survey, construction, volumetric. Use what you
have from TX, others, Wendy Booth.
Figure out how to fit nursery trade
maybe one here for aerial counts and
some other place for indoor flying? (several years
out for sure)
BACKGROUND
Dan Staley, Arbor Drone, LLC
Objectives of Project
Project Details
1. Early detection of Emerald Ash Borer (EAB) using
remote sensing and UAVs (drones)
3. Report to City and County of Denver
Aircraft & Equipment
2. Species identification using multispectral sensor
Key Limitations
Had to fly in appropriate airspace
Had to obtain permits for training flights in Denver parks
Certified pilots for operations
(Accurate) inventories for species ID testing
Weather - late leaf-out after two late freezes
No one has done this before
Spectral Analysis - Precision Agriculture
RedEdge Sensor
MicaSense RedEdge 5-band multispectral sensor
High end of small sensors for environmental applications
Current UAV standard (at 100 m/335 ft above
ground level (AGL)):
Resolution ~6-9 cm (4 in) ground sample
distance (GSD)
RedEdge Sensor
Each sensor is "tuned" to receive only specific
wavelengths of light in a narrow "band"
Multispectral Sensors
Hyperspectral Sensors
Early Detection
Remember keys to ground ID?
Laws and Regulations
Flying for income requires UAV pilot's license
FAA - 14 CFR Part 107 ("Part 107")
Practice under Part 101 (hobbyist)
Test administered at participating FAA locations
$150 fee, ~60 questions on maps, airfield ops,
weather, airspace restrictions, procedures
No flying test (!)
Current pass rate ~87%
Early EAB Detection
What we know
Early EAB Detection
Species ID
Species ID
Early goal is 80% accuracy
We are ~66% +/- today (time of year,
more samples, more backgrounds)
Finding inventory inaccuracies
High-resolution aerial identification is
around the corner
Thank You!
Early detection of EAB has been done
One standard vegetation index won't do it
Will soon make rapid deployment of high-
resolution remote sensing for pest
monitoring and action standard tool for
urban forest managers
We think - but are not 100% sure today - that
species ID can be done with 5-band sensors
Several multispectral sensors are available
to do this work
Likely more on the way, and aircraft to
carry them abundant
Thousands of pilots flooding marketplace
in USA
USA commercial drone laws inadequate for
large-scale tasks
Pest monitoring, inventories, assessment
Software platforms exist that allow you to
use a drone for bids, assessments, analysis
Maryland C
DATA COLLECTION
Complexities
Complexities
Unforeseen challenges
4. Process improvement for later business offering
Project for Denver Forestry to confirm techniques used in precision agriculture can be used to detect early onset of EAB
Using high-resolution imagery from drones flown at 100m
Using 5-band multispectral sensor mounted on custom-built 8-rotor UAV
Obtaining imagery of EAB hits on trees in Boulder, Colo
Project Details
Obtaining imagery of park, ROW trees in Denver along ~10-15 miles of roadway
Analyzing data with Spectrabotics software, including ENVI integration
Image analysis software developed by Harris Geospatial
Products: report to Denver, presentations of findings, scholarly paper, trade journal articles
Acknowledgements
Custom-built octocopter - Colorado College
Payne Jungblut: Ground crew,
spotter, mapping, gear shlepper
L-R:
- Loren Anderson, COO Spectrabotics
- Tim Haynie, CEO Spectrabotics
- Darren Ceckanowicz, Technical Dir.
Program on Environment, Colorado
College
Drones in ag. already $multibillion business in U.S.
Large software industry to service - mostly ready-made "use this tool in this box" uploading service
Services offer flight planning and control, rudimentary Vegetation Index (VI) analyses - usually NDVI, GNDVI (G = Green), OSAVI (reduces soil reflectance)
"Simple" visualization and reporting tool for farmers, agronomists
Urban Operations: Higher Level of Complexity
Fly over people and property
Cities often have controlled airspace
Trees have hundreds or thousands of pixels - sunlit vs shaded, over grass/pavement, deadwood etc. Corn may be 6 pixels
Current rules state UAVs must remain in visual line of sight of pilot - street with 70-year-old trees
Most large cities require permits to fly in parks
~10% of what hyperspectral equipment costs, much less
computer processing time
Lower weight allows larger range of delivery
Complexities
Foreseen challenges
Cloud-based software companies not good enough
for our analysis needs
Complexities
Foreseen challenges
Our processes much higher definition
Important in infrared for canopy segmentation and
delineation
Also important to automatically detect deadwood in

Fraxinus
- common in Intermountain West
Need as many pixels at Red Edge to perform math to
be able to determine individual species as well
Complexities
Our processes much higher definition + 5 band visual
Complexities
Detect deadwood in
Fraxinus

Visible 5-band 80 meters AGL
Training filter to detect vegetation, register images
False positives with NDVI
Google Earth overlay for comparison
Complexities
Appropriate pixels on 'red edge' to determine individual species?

Problem Statement
Can we use drones and remote sensing for early detection and monitoring of Emerald Ash Borer (EAB) to give urban forest managers high-resolution, timely data with which to make decisions?
Custom-built octocopter "The Tractor" Spectrabotics
DJI Phantom 3 Professional - Arbor Drone
Programming the flight path
Hyperspectral: hundreds of bands - processing time
immense, must determine which bands to keep,
weight in kilograms
Thinning from top
Leaves in thinning areas ~less green
Epicormic shoots
What will this look like in reflectance?
Early Detection
Important to understand tree physiology and reaction
to manipulate reflectance values in useful analysis
Reflectance values turned into numbers to
manipulate to look for what you want (what a
Vegetation Index does)

Looking for clues in chlorophyll, anthocyanin, carotene
for early detection
Early Detection
What will this look like in reflectance?
Chlorophyll index
Hi-res NDVI
Non-"standard" indices
Early Detection: NDVI Issues
"Standard" ag. vegetation index - NDVI - saturates at high Leaf Area Index (LAI)
Early Detection
"Standard" agriculture indices inadequate to task
Must understand what is visible in bands - we have to
manipulate reflectance values to understand
conditions
Fancy math needed to integrate multiple vegetation
indices (also needed for species ID)
Pest Detection Using Unmanned Aerial Vehicles
ISA International 2017
Washington D.C., USA
August 2, 2017 11:20 - 12:00
Dan Staley
Arbor Drone, LLC
http://arbordrone.net
dan@arbordrone.net
All presentations on Prezi.com (search 'Dan Staley')
Maryland C
Still collecting 'representative populations'
and understanding time of season, health
Available 'here is your box' ag software is OK, but
not suited to good urban forestry - due to limited
range of vegetation indices available
Will need new laws for widespread deployment
Amazing speed of technological advancement will
mean aerial urban forestry will be standard tool
soon
'Technology forester' a new staff position??
NDVI inadequate
Anthocyanin alone
is inadequate
Chlorophyll - anthocyanin
What we don't know
Whether our sample size today is sufficient
More flights scheduled
Why anthocyanins don't always show up in
lightly stressed trees (less chlorophyll...)
DBH will be a challenge for some time
without LiDAR flights
Work still needed to analyze and streamline custom
platforms
(simple up front for customer - complex
behind the scenes)
Can we make new processes for new applications??
Future wide open
Aerial urban forestry, landscaping will be
~common in ~5 years
Replication in more areas next
OK now for parks, landscapers, turf industry
Look for our articles, paper in the near future
for more detail and information
Full transcript