DRONE MAPPING BASICS
– How to get started
Introductions
Daniel Murphy, Technical Support Engineer
daniel.murphy@sensefly.com
In this talk…
Topic Duration (1 hour total)
Introductions and Presentation of Company < 5 min.
Mapping Drones 15 min.
Processing Drone Data 10 min.
eMotion Demo 5 min.
Pix4D Demo 5 min.
Example Data / Use Case 10 min.
Questions 10 min.
About senseFly:
What we do
About senseFly
Founded in 2009 - spin-off of EPFL (Ecole
Polytechnique Fédérale de Lausanne)
Headquarters in Lausanne, Switzerland
Business & service office in Washington, DC
Ag Solutions / field office in Fort Dodge, Iowa
Integrated within the Parrot Group (publicly
traded in Paris, PARRO) since June 2012
Parrot Group: senseFly, Pix4D, MicaSense,
Airinov
our
applications
Surveying Agriculture Earthworks/monitoring
Urban planning & land Research / geodesy Quarries, aggregates & mining
management
6
and of course..
Forestry / Land Management
7
Mapping Drones
-Concepts
-Workflow
-Products
Remote Sensing
1. Aerial remote sensing is nothing new.
Balloons, kites, satellites and planes have
been doing it for a long time
2. It is very common to use remotely
sensed observations to aid decision
making
3. Fusing up-to-date maps and expert
knowledge of sites is excellent practice!
4. Remember: drones are a tool, not a
complete solution to any one problem.
Satellite Imaging
Filling the Gap Wait time: Days to Weeks
Typical resolution: 50+ cm
Manned Aviation
Wait time: Weeks to Months
Typical resolution: 10-30 cm
Resolution Matters!
Satellite – good for low
resolution applications.
Detecting large objects and
observing phenomena across a
landscape. Very accessible, but
not temporally flexible.
Google Earth - 5/30/2015 Zoomed in
~50-65 cm pixels
Drone – provides high spatial
resolution with flexible
acquisition schedule
Drone image – 3/3/2017 Zoomed in
2.16 cm pixels @ 91 meters altitude
Photogrammetry
from images to 3D points
Depth from stereoscopic vision 3D points from images with common features
12
Drone Expectations
Simple, easy & automatic flight
Portable/rapid deployment
Integrated payloads
Consecutive flights
High & low resolution
Reliable & serviceable
Unless it’s R&D you shouldn’t
have to “make it work!”
Mapping Drone Platforms
+ +
• Take off / land in tight locations • Longer flight times / greater coverage
• Hover • Handle stronger winds
• Fly close to objects
• Less parts
• Document inclined/vertical surfaces
• Video possible (heavier payloads) • Less damage (weight/gliding)
-
-
• More mechanical parts • Larger take off / landing area
• Shorter flight times / less coverage • No hover
• More damage / danger (weight) • Cannot fly close to objects
• Lower wind tolerance
Multi-Rotor Fixed Wing
Drone Mapping Workflow
Visual Inspection or
Flight planning
Analysis in third-
party software
Generation of Setting of
2D orthomosaic and 3D on-site GCPs
(if necessary, and no RTK/PPK available)
point cloud/DSM
Import images
Flight
(Flight Data Manager)
15
Platform:
eBee Plus
• Real-world flight time:59 min
• Up to 220 ha (540 ac) in a single 122 m
(400 ft) flight
Large Coverage
High Precision on Demand
Includes
Project Perfect Payloads eMotion 3!
16
Sensor:
S.O.D.A.
• 20 MP RGB
• Ultra light
• Compact
• Built-in dust and shock protection
• No external moving parts
• Global Shutter
Platform:
eBee SQ
More Precise
Larger Coverage
Workflow Compatible
Affordable
Sensor:
Parrot Sequoia
• Four filtered 1.2 MP sensors (NIR, RE, R, G)
• One 16 MP RGB sensor
• Upward-facing Sunshine Sensor
• Customized data capture
• ½ Res.
• MSP Only
• MSP + RGB
• Seamless integration
Processing Drone Data
• Concepts
• Workflow
• Products
Biological Solution
Two Cameras
Photogrammetry Solution
Multiple Cameras
Optimal Overlap
>1000 Automatic Tie Points <100 Automatic Tie Points
≈75% image overlap ≈20% overlap
Agriculture or Dense
Vegetation
• 75% frontal overlap
• 75% side overlap
• Image geolocation
• Avoid windy conditions
• Average GSD
• ≈10 cm/pixel
Visualization in the Pointcloud
Visualization the orthomosaic
Processing Time
Time required for processing:
• Depends on:
No. of images
Settings (ground resolution / point density)
• 15 min to several days
• Several PCs reduce time: split data sets by flight
Specific example:
• 1 field of corn-v5
• 1 flight of 160 acres
• 5 cm ground resolution (pixel size)
• 394 images
• Est. 1.5 hrs from landing to reflectance map
Specific example:
• 1 area of deciduous forest
• 2 flights covering of 175 acres (80% lateral overlap)
• 2.91 cm ground resolution (pixel size)
• 712 images
• Est. 14–20 hrs landing to orthomosaic, surface model, point cloud
Product Demonstrations
1. eMotion
2. Pix4D Mapper
eMotion 3:
Demo
Pix4DAg:
Demo
Nectarines in
California
Example Data / Use Cases
• Site overview, leaf off
• 3D mapping and visualization, leaf on
Site overview
Example
3D mapping and visualization
Example
Questions?
Thank You
Spectral Remote Sensing
Plants & Light
By monitoring the amount of NIR and
visible energy reflected from the plant with
a camera, it is possible to determine the
health of the plants:
• High NIR reflectance / Low visible
reflectance = Healthy
• Low NIR reflectance / High visible
reflectance = Unhealthy – stressed
Utah State University
Sensor:
thermoMAP
• Senses thermal radiation
• Auto–calibrates in flight
• Customized data capture
• Single frames
• Video
• Seamless integration
Photogrammetry Solution
Two cameras
Flavors of Remote
Sensing
Broadband
-cell phone cameras
-thermal imagers
Multispectral
-Landsat
Hyperspectral
Ultraspectral
