NIR botany images

[TERA EPA]

Hello,
We're working on a global warming study with 12 different plots of prairie
plants on which we have taken pictures. The pictures and plants are analyzed
using ImageJ. We have recently purchased a camera that is modified to be
more sensitive to near infrared (NIR) as well as filters that narrow the
range of wavelengths visible. We plan on using ImageJ to analyze these
images taken with this new camera. We want to look at the different plants
on an individual species basis, based on each plant's unique spectral
reflectance. We want ImageJ to pickout these differences, is ImageJ capable?
If so, are there plug-ins already written for this type of thing? What are
the capabilities of ImageJ in regard to NIR?

We would really appreciate any help/advice that this listserv can provide.
Thank you,
TERA (Terrestrial Ecosystem Research Associates)

[Bill Christens-Barry]

The Principal Component Analysis plugin from the BIJ group, found at:

    http://webscreen.ophth.uiowa.edu/bij/pca.htm

may be useful for this. Principal Component Analysis (PCA) is apt for sussing out spectral
"fingerprints" of different objects in a set of images differing in their acquisition wavelengths.

Also, the DStretch plugin from Jon Harman, available at:

    http://www.dstretch.com/index.html

is really nice for making subtle differences more visible.

If you could describe the camera, spectral bands, and imaging conditions more fully, I might have
some feedback on IR wavelength bands.

Bill Christens-Barry


On Tue, 26 Jun 2007 09:01:50 -0700, TERA EPA <[hidden email]> wrote:

>Hello,
>We're working on a global warming study with 12 different plots of prairie
>plants on which we have taken pictures. The pictures and plants are analyzed
>using ImageJ. We have recently purchased a camera that is modified to be
>more sensitive to near infrared (NIR) as well as filters that narrow the
>range of wavelengths visible. We plan on using ImageJ to analyze these
>images taken with this new camera. We want to look at the different plants
>on an individual species basis, based on each plant's unique spectral
>reflectance. We want ImageJ to pickout these differences, is ImageJ capable?
>If so, are there plug-ins already written for this type of thing? What are
>the capabilities of ImageJ in regard to NIR?
>
>We would really appreciate any help/advice that this listserv can provide.
>Thank you,
>TERA (Terrestrial Ecosystem Research Associates)

[TERA EPA]

Bill Christens-Barry and ImageJ listserve,

For a more detailed introduction to our research: we are working at the
EPA's Global Climate Change Research Facility in Corvallis, Oregon.  We're
conducting a study that compares different types of warming treatments
within growth chambers of Oregon prairie plants.  The chambers are 1 meter
by 2
meters and 1.5 meters tall.  They are very similar to a greenhouse with the
exception that all of the environmental variables (temperature, relative
humidity, CO2, etc.) are closely monitored and regulated to maintain our
treatments.



Initially, we were going to just use our "point and shoot" 5 mega-pixel
digital camera.  We have installed semi-permanent camera mounts in each of
the 12 chambers for taking weekly photographs of the plants and their
phenological changes. The camera was inexpensive and has served our purposes
well until now.

After taking a year's worth of photographs and observing the variability in
plants
between chambers, we've decided to further explore the photographic options
for differentiating species and quantifying plant growth.  Currently with
images from our Panasonic we are using ImageJ to identify and quantify
flowering species using simple color isolation algorithms.  The software is
great but with high variability in lighting it is very difficult to
differentiate species by foliage color.  For this reason, we've recently
purchased a converted FujiFilm S3Pro capable of photographing UV, IR, and
visible light images.  With the camera, we've also
purchased a Nikon 20mm lens.  Our options for filters from one supplier are
the 715nm, 780nm, 830nm and 850nm filter.  We haven't made a final decision
on which filters would be optimal, and would appreciate any insight on the
subject.  We hope to use UV filters to block out visible and IR light to
distinguish unique UV patterns on the flowers (to further highlight floral
differences in seemingly identical yellow flowers).  Our supplier of filters
and cameras presently is LDP LLC and can be found at maxmax.com.

 We are also curious about the Kodak Wratten 87 filter and its grayscale
images.  For ImageJ, this
grayscale would be ideal for differentiating variability in foliage IR
reflectance.  Do you have any experience with this filter?  We are also very
interested in capturing thermographic images of foliar temperatures.  Am I
mistaken in thinking that IR cameras are capable of this?    If you are
familiar with other scientific research that has utilized UV and/or IR
images of plants at a close scale (~1.5 meters from subject) I would be very
interested in reading about it.  I
have found a number of journal articles that utilize IR images for aerial
and
satellite photography but nothing at a scale similar to our research.  In
addition, I have found only one article that utilized foliage *shape* to
identify plant "functional groups" (grasses, shrubs, forbs, etc.) at a close
range, and it relied on software from the Definiens group called eCognition.
Am I just missing the other journal articles or has close range image
analysis for plant ecology not yet been proven robust enough?

If you have any suggestions on techniques or equipment (especially which
filters would be most useful) that would help us in
our study, we would be very appreciative.


Take Care,
Emily Cavaliere & Peter Stocking

Terrestrial Ecosystems Research Associates