Center for Embedded Networked Sensing

1
Center for Embedded Networked Sensing
Eric Graham, Staff Ecologist egraham _at_
cens.ucla.edu http//cens.ucla.edu
2
Remote Sensing From Satellite to Digital Camera

• Acquisition of information about the earths
  surface by remote instruments
• Satellite imaging, daily time series of
  multispectral data.
• Aerial photography, less frequent but high
  resolution data.
• - Radar, Lidar sensing, surface elevation and
  canopy characteristics.
• - On-the-ground sensors, including
  multispectral, single spectra (infra red
  temperature), Lidar and digital cameras.

3
Remote Sensing Satellite
We have mostly used MODIS (Moderate Resolution
Imaging Spectroradiometer) satellite imagery to
create a satellite-based index of vegetation
greenness to compare to a daily greenness
measure from webcams located across North
America. Satellite data format is a 10 by 10
degree latitude/longitude tiles, in a sinusoidal
projection.
4
Remote Sensing Satellite Data

• Free (or low-cost) satellite remote sensing
  products
• Global Land Cover Facility (GLCF) University of
  Maryland USA (http//glcf.umiacs.umd.edu/index.sht
  ml)
• Warehouse Inventory Search Tool (WIST)
  (https//lpdaac.usgs.gov/lpdaac/get_data/wist)
• Data Pool online data holdings (USGS/NASA)
  (https//lpdaac.usgs.gov/lpdaac/get_data/data_pool
  )
• GloVis USGS Global Visualization Viewer
  (http//glovis.usgs.gov/)
• Free SPOT imagery (http//free.vgt.vito.be/)
• EOS-Webster (http//eos-webster.sr.unh.edu/home.j
  sp)
• ESA Earth Net (European Space Agency)
  (http//earth.esa.int/)
• Tropical Rainforest Information Center focus on
  Central Africa, SE Asia, and Amazon.
  (http//www.trfic.msu.edu/ (not free, but low
  cost))
• Satellite Imaging Corporation (Commercial)
  (http//www.satimagingcorp.com/)

5
Remote Sensing Satellite Data
The American Museum of Natural Historys (AMNH)
Biodiversity Informatics Facility has GREAT list
of resources for viewing and processing remote
sensing data, as well as guides and
tutorials. For a list of more public domain
tools https//lpdaac.usgs.gov/lpdaac/tools
http//biodiversityinformatics.amnh.org/index.php
6
Remote Sensing Satellite Data

• The native formats of some remote sensing
  products may not be readily read by GIS software.
  Converting files to more usable ones may be
  necessary
• Data Viewers
• HDFview http//www.hdfgroup.org/hdf-java-html/hdfv
  iew/
• HDF explorer http//www.space-research.org/
• HDF look http//www-loa.univ-lille1.fr/Hdflook/hdf
  look_gb.html
• PCI freeview http//www.pcigeomatics.com/
• ENVI Free Look http//www.ittvis.com/
• Geospatial image processing
• OpenEV http//openev.sourceforge.net/ (open
  source)
• GRASS http//grass.osgeo.org/ (open source GIS)
• Image J http//rsb.info.nih.gov/ij/

7
Image Capture Cheap to Expensive
Remote Sensing Digital Cameras
Casual monitoring of installations.
Nest boxes, pitfall traps, moss-cam 2.
Towers, mobile nodes, lab use.

• Classic webcam
• USB, power data
• 640 x 480
• About 30
• Easy

• Cellphone-type
• Wireless, battery
• Low-to high resolution
• Cheap to expensive
• Technical

• Pan-Tilt-Zoom
• Ethernet, line power
• 640 x 480
• Expensive (1000)
• Easy

8
Remote Sensing Digital Cameras
Images captured from digital cameras can be used
to detect changes in vegetation, similar to
satellite remote sensing products. One advantage
to using digital cameras is that the image can
also be interpreted by a human observer, and so
quality can be readily assessed.
Also, multiple images can be captured per day and
the best image can be extracted for use in a tme
series. Red, Green, and Blue components of each
pixel can be used for separating vegetation from
background.
9
Remote Sensing Digital Cameras
Not only green vegetation, but flowers and
senescing leaves can be readily detected.
Excess green is one method for removing
luminance to calculate changes in color of
pixles Excess Green 2G R B
10
Camera Acquisition satellite pixel
location Google searches identified 1100
cameras with visible vegetation across North
America.
A 30 camera test subset was selected and images
were captured twice daily (am, pm) from February
2008 January 2009. Satellite MODIS data was
collected for each of the 30 test sites for
comparison.
11
Camera 824 has a nice view, and is also a nice
example of green-up during spring using simple,
whole-image, color (excess green) averaging.
12
Satellite
Camera
NDVI (t) or ExG (t) (w1 w2) 0.5 (w1 -
w2)tanh(w3(t u)) tanh(w4(t v))
13
Camera
Points of inflection can be used as modeled
indicators of spring and fall. A visual estimate
of ground truth can be used to also detect
spring or fall from image sequences. Solid lines
to the left indicate 10 green and 10 fall
colors in the images. Dashed lines are the model
inflection points.
14
We can also manually segment the deciduous
vegetation from the evergreen and understory (or
grassland) and record changes in the green signal
through time for sections of webcam images.
15
(No Transcript)
16
Cameras vs. Satellite data in a latitudinal
gradient. For this subset of pixels, the
cameras performed better for predicting changes
in the onset of spring.
Camera Spring
Camera Fall
Satellite Spring
Satellite Fall
17
Remote Sensing Soil Energy Balance Project
18
Remote Sensing Soil Energy Balance Project
Panoramas constructed every 10 minutes allowed
the capture of sun and shade on the soil surface
in a forested environment
19
Remote Sensing Soil Energy Balance Project
Soil energy balance is used in ecosystem studies
and eddy flux measurements. Soil surface
temperature can be related to the air temperature
when the soil is in shade (or at night) and to
incident solar radiation during the daytime. The
rise in temperature during the day is dependent
on soil properties (moisture content,
composition), and models of thermal conductivity
can be used.
20
Remote Sensing Soil Energy Balance Project
21
Remote Sensing Soil Energy Balance Project
We can examine the amount of time that any area
is exposed to direct sunlight between season and
calculate a whole meadow energy balance for the
soil surface and subsurface.
22
Remote Sensing Soil Energy Balance Project
Each pixel in the panorama was followed through
time to create a temperature time series.
For a meadow of 645 m2 Heat stored in top 8 cm
per day 3.36 GJ. Max surface heat flux 0.289 kJ
m-2 s-1
23
Imagers as Environmental Sensors
Remote Sensing Infinite Possibilities

• We are working on several aspects of image
  processing applied to ecological imaging
• More hardware control for to extract more
  information during image capture of spatial
  patterns of summer annuals in a James Reserve
  meadow.
• More sophisticated image segmentation routines
  for separating similarly-colored plants in the
  same frame.
• Using computer vision algorithms and machine
  learning techniques to automatically separate
  images into components and scale the analysis to
  thousands of cameras.

24
Imagers as Environmental Sensors
CENS People

• Just some of the CENS people who have contributed
  to this work presented
• Deborah Estrin, CENS director
• John Hicks, CENS staff
• Yeung Lam, CENS staff
• Erin Riordan, Biology graduate student
• Phil Rundel, Biology department professor
• Tom Schoellhammer, EE graduate student
• Eric Yuen, CENS staff

25
(No Transcript)
26
High Dynamic Range Imaging in Plant Ecology
Remote Sensing Digital Camera Control HDR
PAR sensor
Multiple images of the same scene are captured
with different iris and shutter speeds. Pixel
values range from 0 to 255 for each color channel
in each image.
Combined to form a High Dynamic Range (HDR) image
with floating point values for pixels. Pixel
Luminance is directly related to the amount of
reflected light reaching the camera.
27
Remote Sensing Digital Camera Control HDR