from gigapixel timelapse cameras to unmanned aerial vehicles to smartphones: a review of emerging...

From gigapixel timelapse cameras to unmanned aerial vehicles: A review of emerging near remote sensing technologies for cross-scale monitoring

from organism to ecosystem

Limits to the rate of knowledge discovery in Ecology…

Understanding high-order complexity is hard!

1. The environment is complex and continuous but we usually can only measure limited data over limited snapshots in time

2. Difficult and expensive to track change on the ground at high spatial and temporal resolution

3. Long term data sets are rare, particularly of images– Typically only satellite-based tools and data sets for long term

ecosystem monitoring

Ecosystem science research now is like genetics used to be before PCR and high throughput sequencing (but this is changing)

The Future is arriving now

By envisioning the future, we can bring it into being and shape it to meet our needs– What kind of hardware/software systems do we need to build?

Imagine for any research site…– Live streaming access via laptop, phone, tablet– Multi-billion pixel resolution image feed of any study area– All data is archived for as long as there have been sensors

With these tools, what kind of science and education will we do?

Capitol Reef Webcam: http://www.uvu.edu/crfs/livefeed/index.html

Conventional ecosystem monitoring

Satellite - Remote Sensing • Macro scale / Regional• Multi-spectral cameras

– NDVI (Normalized Difference Vegetation Index), “Greenness”

• Image Pixel data is georeferenced to a known location• Reliable time-series

• But:– Limited resolution– Low sample rate (daily to annually)– Weather dependent

Conventional ecosystem monitoring

Land based - Field-level sampling

• “Micro” scale / local

• Stand-level sampling

• Short time scales– Infrequent visits– 1 season or a few years

How to scale between ground-based sampling and regional-scale satellite sampling

“Near remote” Sensing – Intermediate scale

• Conventional aerial surveys (good data but expensive and infrequent)

• New imaging technologies can bridge the data gap

• Numeric data extraction is a challenging problem for image-based datasets

• Map pixels from ground-based image sources onto the landscape (just like satellite data)

Emerging technologies for ecosystem monitoring• Phenocam networks and towers

– Plant Cams / Trail Cams

– IP-based cameras

• Gigapixel timelapse cameras (gigavision)

• Unmanned Aerial Vehicles (UAV/UAS’s)

• Kite & balloon photography– http://www.geospectra.net/kite/equip/camera_rigs.htm

– http://www.grassrootsmapping.org

• Socially networked science / Citizen Science– Smartphone applications and geotagged images

Other enabling technologies

• Wireless Internet– Satellite Internet– 802.11 wireless– Cellular: A 4G cell phone has about the same bandwidth as

MODIS (and more processing power)

• Wireless microclimate sensor arrays

• Google Earth

Near remote sensing: “Phenocam” Towers

• Work well, particularly when coupled with ground-based sensor systems.

• Enable capture of visible & infrared images of research areas– Ground-truth satellite data

• But limited sampling area: FOV < 150m2

Rio Mesa sensor systems

• 30’ towers• Solar powered• 5 megapixel network

cameras• Infrared cameras (NDVI)• Satellite images• Weather station• Tamarisk & cottonwood

sap flux• Wireless network

– 802.11 Satellite Internet

Camera Systems

Tamarisk year timelapse in Vis and IR: http://www.youtube.com/watch?v=Wd4PBEmHU2o&hd=1

Rio Mesa “Gauge Site” - Results

Tamarisk NDVI measured from tower based Infrared cameras at Rio Mesa Field Station, 2008-2009. [Nagler, Brown, Hultine, et al, in prep]

Sap Flux

Beetle Arrives

NDVI

Online data visualization tools

http://riomesa-data.biology.utah.edu/station.aspxhttp://riomesa-data.biology.utah.edu/station.aspx?id=ws1&ubm=74

Multi-billion pixel imaging system– Pan/Tilt camera system– Resolution of 1 pixel / cm over 7 hectares

• (~600 million times the resolution of MODIS)– Monitor every individual plant within large area– Embed into Google Earth (should be pixel mapped)– Timelapse record of phenology & long term

ecosystem change

Camera Specifications– Cellular (3G) or 802.11g wireless access– Automated capture up to 1 image / hr– 180 - 360° degree view– Power consumption: ~12w– Solar powered

Gigavision: gigapixel timelapse camera

Image Visualization and Data Collection

http://www.gigavision.org

Area: 6-7 ha

http://www.youtube.com/watch?v=-_vzxzpJVjc&hd=1

Unmanned Aerial Vehicles (UAVs) for Ecosystem Monitoring• GPS Autopilot can repeat sampling with +/- 3m accuracy

• Flies @ 400’; 30-40 mph

• Survey 1 km2 area in < 10 minutes @ 5 pixels /cm

• 12 Megapixel consumer camera + 3MP Infrared Vegetation Camera (NDVI)

• Each 12mp image covers 274x366 m2 area [MODIS = 1 pixel / 250x250 m2]

• Resolution = ~100 pixels/m2

• Commercial version:– $9,000 - $23,000, hardware (Total cost with training, software, etc: $30,000-$50,000)

• DIY version: < $1,000 (http://www.DIYdrones.com)

• In the long term: Automated high resolution aerial imagery at long term sites like NEON

• Main current challenge is regulatory

RP Flight Systems http://www.mikrokopter.us/

76 imagepanorama

QuadroCopter Video: http://www.youtube.com/watch?v=L1rQJ7cjfuA&hd=1

High Altitude “Satellite Replacement” systems

NASA Helios project• Prototype long endurance solar powered aircraft• Flies at 15,000 – 22,000m• 6 months flight time (solar + hydrogen cells)• 200 lb payload

http://www.nasa.gov/centers/dryden/news/FactSheets/FS-068-DFRC.html

Kite / Balloon Aerial Photography

Grassrootsmapping.org

Kite / Balloon Aerial Photography

Grassrootsmapping.org

~40 image panorama of Chandeleur Island, LA, taken from a balloon

3D track of the balloon in the air

High resolution image layer from balloon’s images, in Google Earth

Other tools for data collectionGigapan• Low cost pan-tilt head for gigapixel repeat photography• http://www.gigapan.org• Repeat photography• Embed in Google Earth• Example: Early detection of bark beetle infestations at Alta Ski area, Alta, UT• See also: http://www.xrez.com/yose_proj/yose_deepzoom/index.html for an amazing example

(72 gigapans shot on the same day to build a pixel-accurate 3D model of Yosemite)

Game cameras / Plant cams– Wingscapes plant / birdcams – http://www.wingscapes.com– Buckeye game cameras -- http://www.buckeyecam.com/ – $300 with solar panel– Smithsonian game camera projects website: http://siwild.si.edu/

Canon CHDK– Convert a $50-$100 Canon into a 8MP timelapse camera– http://chdk.wikia.com/

Low cost GPS• Eye-fi SD card

– Add GPS/WiFi capabilities to any camera– http://www.eye.fi/

• GPS loggers– $100 / Many brands, search amazon for “gps logger”

• iPhone / Android apps– EveryTrail – http://www.everytrail.com

Citizen Science

• Picture Post– http://picturepost.unh.edu/

• Smartphone apps (See OOS22 session, Wed AM)

– Virtual picture post– Panorama software (try the “360panorama” app for iPhone)– Phenology apps:

• PhenoMap; BudBurst; others

• Geo-tagged photos– Flickr: ~4 million images geotagged per month

• 6 billion photos, 156 million are geotagged (Aug 2011)

– Facebook:

• 2.5 billion images uploaded / month

– National Parks – 280 million visitors / yr (http://www.nature.nps.gov/socialscience/stats.cfm)

Geotagged Chicago

© Eric Fischer (http://www.flickr.com/photos/walkingsf/ )

Photosynth

Microsoft and U. Washington PhotoTourism and Photosynth projects– http://grail.cs.washington.edu/projects/rome/ – http://phototour.cs.washington.edu/findingpaths/– http://photosynth.net/

Creating a 3D map of Dubrovnik with 3.5 million flickr photos:http://www.youtube.com/watch?v=sQegEro5Bfo&hd=1

Ecology’s PCR moment is coming…• Computers, hard drives and all technologies are getting faster, smaller and cheaper

• New monitoring tools– Exponentially increase rates of data collection

• Wireless technologies– Live data streams

• Long term data collection projects (NEON, LTER, etc.)– Large, long-term standardized data sets– DataONE

• Internet based data sharing– Collaborative, open-source science – Multi-disciplinary research– Increases in level of complexity that can be examined– Enhanced science education

• Crowd-sourcing / Citizen science– Improved public education– Improved citizen engagement– Improved landscape monitoring– Increase rate of data collection and knowledge discovery

• Improved programming and visualization tools– Larger data sets– Enhanced visualization and data management tools– Automated data processing– Easy to use analysis tools

The seasons of flickr

• http://hint.fm/projects/flickr/

Links and Credits

More Info• Gigavision Gigapixel camera project: http://Gigavision.org• TimeScience: http://www.time-science.com• Interactive online data visualization: http://riomesa-data.biology.utah.edu/• Gigapan pan/tilt camera head: http://Gigapan.org Search TimeScience• Kite photography: http://grassrootsmapping.org• Online mapping/image warping tool: http://Cartgen.org• Hyperblimp: http://www.hyperblimp.com/ • DIY Drones: http://diydrones.com/

Credit and thanks to• Programming and project support:

– Christopher Zimmermann, TimeScience

• USGS Tamarisk survey project: – Pamela Nagler, USGS– Kevin Hultine, University of Utah / N. Arizona University

• Gigavision Project: – Justin Borevitz, University of Chicago

CONTACTS

Camera System design and Data Visualization ToolsTim BrownTimeSciencetim@time-science.com801.554.9296http://www.time-science.com

Gigavision Camera Project / High Throughput PhenotypingJustin BorevitzUniversity of Chicago / Australian National Universityborevitz@uchicago.edu

Tamarisk Sap Flux and Research Projects Pamela NaglerUSGS, Sonoran Desert Research Station, Tucson, AZpnagler@usgs.gov

Kevin HultineDesert Botanical Garden, Phoenix, AZKevin.Hultine@nau.edu

• Gigapixel Yosemite– http://www.xrez.com/yose_proj/yose_deepzoom/index.html