(2015/09) Drone Imagery Economics

Drone Imagery Economics

Martin Scholl @zeit_geist

[email protected]

Mapping Cost will be zero.

Maps will be made for humans

machines.

• Started by Stephen Mather on Apr 2014

• self-described as “an Open Source Toolkit for Processing Drone Imagery”

• 31 contributors

• installation takes 10m on EC2

• dead easy usage: ./run.pl in images directory

Open-Drone-Map

Point Cloud

Ortho- photo.

…

(Textured) Digital

Surface Model

• OpenDroneMap makes drone imagery data processing as easy as running a single command

• ~80pics take 1h10m on a 16-core EC2 machine

• but: several steps are quadratic in runtime in # images and resolution

• performance improvements underway

• Where to host drone Orthophotos?

• OpenAerialMap is infrastructure for a “simple open way to host and provide access to imagery”*

• Re-Initiated by HOT in 2015; grant from Humanitarian Innovation Fund

• “supporting humanitarian organizations involved in emergency preparedness and response activities”**

* https://github.com/hotosm/OpenAerialMap ** https://docs.google.com/document/d/

1Cyj3x4MWzif3avAviELNaRThKhBdDoAA0sy38kwyH_g/edit

http://beta.openaerialmap.org/

http://beta.openaerialmap.org/

0. Search & Filter

http://beta.openaerialmap.org/

1. Select Area

0. Search & Filter

http://beta.openaerialmap.org/

1. Select Area

2. DownloadOrthoimages0. Search & Filter

OpenAerialMap• OpenSource Project & Infrastructure

• Solves rapid sharing of imagery data forhumanitarian purposes

• Clever data federation architecture by separating indexing from data storage leveragingObject Storage (Amazon S3)

• mostly satellite images & some drone imagery

• 1.1k images indexed yet

A Bigger Picture• OpenDroneMap: free drone image processing

• OpenAerialMap: free 2D image hosting, etc.

• OpenTerrain: free terrain model hosting, etc.

• ArduPilot: free flight control & automation

• DIY Drones, OpenUAV, et. al.: free UAV blueprints

Drone Imagery Economics

Drone Imagery Economics 101 w/ OpenDroneMap

• Drone w/ 4K camera: $1000 range

• Storing 250 high-res images: $0.01 / month (Google Cloud Storage)

• Processing those images: 2h or ~$0.40 (AWS EC2 c4.x4large spot instances)

• 4k Camera @ 1cm resolution & 50% overlap: 250 pictures cover 0.2km2

5 times a day, 1 drone surveying 0.2km2

Per Day Per Month Per Year

Drone* $0.92 $27.78 $333

Data Storage $0.05 $1.50 $117Data

Processing $2.00 $60.00 $720

Covered Area 1km2 30km2 360km2

Total Cost ~$3 $90 $1170

* 3y amortization

5 times a day, 1 drone surveying 0.2km2

Per Day Per Month Per Year

Drone* $0.92 $27.78 $333

Data Storage $0.05 $1.50 $117Data

Processing $2.00 $60.00 $720

Covered Area 1km2 30km2 360km2

Total Cost ~$3 $90 $1170

* 3y amortization

Friedrichshain- Kreuzberg: 20.16km2

3D Mapping Berlin*Refresh

Rate Year Month Week Day

# Drones 3 30 128 892

Monthly Cost $270 $2.700 ~$12.000 ~$80.000

* Berlin: 892 km2; Drone month: $90 / 30km2

… even more numbersQuadrocopter Airplane Landsat 8

Acquisition $1.000 $300.000 $850.000.000

Operations $600 $150.000 (pilot + …) $30.000.000

Amortization 3 years 15 years 30 years

Per Year $900 ~$170.000 ~$60.000.000

http://robohub.org/price-wars-counting-the-cost-of-drones-planes-and-satellites/ http://spacenews.com/40841nasa-official-a-landsat-8-clone-would-cost-more-than-650-million/

$19

$19

space for iPhone baseboard

Some Consequences• Trade very accurate, highly-priced gear with

super-inexpensive smartphone-scale chips

• $100 autonomous drones ⇒ mapping cost approaches $0, becomes a “Big Data’esk game”

• Data Acquisition Cost < Total Derivable Value

• Guiding questions: how stale may imagery data be? What else can we do with it?

Historic Mapping Shift?

• Maps used to be power tools of the world, for humans, by humans.

• Machine Vision for Autonomous Systems has different maps needs

• Maps will be built for autonomous systems, human-scale map only a side-effect

Open Questions• At some point, drone manufacturers will truly

leverage smartphone market economics

• When will the first company deliver the autonomous$100 drone with 8K+ video?

• How will legislation react to this?

• How will 100MP+ camera chips on $1000 drones change the quality in the data collection game?

• OMG Privacy??

Summary

• Drones are quite cheap already. Economics works.

• Historic power shift in maps and maps’ users.

• Mapping cost will approach $0.

• 3D maps: machines & robots the primary user.

Thank you

Meg

apix

els

0

4

8

12

16

2012 2013 2014 2015

iPhone Samsung Galaxy

Smartphone Cameras

Moore’s Law• 2x transistor count every 1.5y

• translated into equiv speed up every generation

• Now Moore’s law works in “opposite direction”

• same performance, half the size every 1.5y

• 8 Core CPUs w/ less than 4Watt power draw

• Drones will become smaller & cheaper keeping the overall performance level