Applied Analysis
Inc.
630 Boston Road
Suite 201
Billerica, MA 01821
USA
PH: 978-663-6828
FAX: 978-663-6389
|
|
|
|
 |
|
|
| FAQs:
Signatures and Atmospheric Correction |
|
16. How do I develop a signature for a Material of
Interest?
|
|
IMAGINE Subpixel Classifier is a
supervised classifier. You must first identify an Area of Interest
(AOI) containing training set pixels, which are believed to contain
the material of interest. The training set may contain either whole-pixel
(>90%) or subpixel occurrences of the MOI. In either case, the process
will produce a pure signature equivalent to a whole-pixel occurrence.
This signature can be used to classify both whole-pixel and subpixel
occurrences of the MOI. Most traditional classifiers only work with
whole-pixel training sets.
You may use your training set with
Manual Signature Derivation when you have a whole-pixel training set
or you are certain of the material pixel fraction. Otherwise, you use
Automatic Signature Derivation which automatically generates and evaluates
a number of signature candidates produced with different parameter
combinations. In that case you can supply AOIs covering areas where
the MOI is expected and where is should not be detected. The software
evaluates the candidate signatures and notifies you of the best performing
candidates.
With IMAGINE Subpixel Classifier
it is better to pick training sets in which the MOI is mixed with several
different backgrounds. The software automatically selects the spectral
component that is common to the majority of the training set pixels
at the subpixel level. Normally the common element to all of the training
set pixels will be the MOI, but at low fractions other materials may
be equally common, soil for example. By specifying AOIs for good and
bad detections, the software automatically selects the best performing
signatures. You can then select the best signature for your application. |
|
|
|
17. Can I use signatures developed with IMAGINE
Subpixel Classifier interchangeably with signatures developed with traditional
classifiers?
|
| No. An
IMAGINE Subpixel Classifier signature is different from other classifier
signatures in both information content and file format. It contains additional
information required for subpixel classification and scene-to-scene usage. |
|
|
|
18. Can I use laboratory spectra as a signature with IMAGINE Subpixel Classifier?
|
| Not currently.
Laboratory spectra differ from imagery-derived signatures and are generally
not usable by any classifier without modification. Laboratory
spectra do not adequately represent the MOI when viewed through the
atmosphere by a given sensor. Future enhancements to IMAGINE Subpixel
Classifier may overcome this limitation however. |
|
|
|
19. Can I develop a signature from a training set
that contains only subpixel occurrences of my MOI?
|
| Yes.
You can develop an IMAGINE Subpixel Classifier signature using either
a whole pixel or subpixel training set. It is best to use Automatic
Signature Derivation with subpixel training sets. The Automatic Signature
Derivation process extracts the subpixel component that is common to
all pixels in the training set for a number of signature parameter
combinations and then automatically evaluates the signatures performance
to avoid selecting false signatures. The resulting signature is the
equivalent to a whole pixel reference signature of that common material. |
|
|
|
20. Will IMAGINE Subpixel Classifier generate whole
pixel detections if I use a signature developed from a subpixel training
set?
|
| Yes.
Regardless of the training set you use to derive a signature using IMAGINE
Subpixel Classifier, whole pixel or subpixel, IMAGINE Subpixel Classifier
will use your reference signature to classify your MOI at both whole
pixel or subpixel levels. |
|
|
|
21. What is the IMAGINE Subpixel Classifier atmospheric
correction process and how does it work?
|
|
The IMAGINE Subpixel Classifier
atmospheric correction process allows an IMAGINE Subpixel Classifier
signature for a MOI developed in one image to be automatically used
in another image from the same sensor, but from a different image date
and/or location. The IMAGINE Subpixel Classifier atmospheric correction
algorithm differs from other processes in that only information from
the image itself is necessary to derive the correction factors. The
correction factors are derived from subpixel detections of particular
materials in the scene. No collateral calibration data is required.
The process automatically derives
two spectra from the image being processed. One of the spectra,
ARAD (n), is derived from subpixel occurrences of dark materials
in the scene. The other spectrum, SUN(n), is derived from subpixel
occurrences of bright materials. ARAD (n) is a scene-average
estimate of the contribution of atmospherically scattered radiance
to each pixel. SUN(n) is a scene-average estimate of atmospherically
attenuated source radiance along its entire path through the atmosphere.
To avoid the influence of bright clouds which are located well above
a significant portion of the atmosphere, the process provides an interactive
means of identifying cloud areas quickly. These areas are then excluded
from processing.
When using a signature in a scene-to-scene
mode, IMAGINE Subpixel Classifier creates a third spectrum, ENV(n),
from the SUN factors for the scene being processed and the scene from
which the signature was derived. ARAD (n) and ENV(n) are applied
to each image pixel during IMAGINE Subpixel Classifier classification
in a way that normalizes the pixel characteristics to those of the
signature source image, allowing the signature to be directly used
in the new scene. |
|
|
|
22. Can I use a different atmospheric correction with
IMAGINE Subpixel Classifier?
|
| No. Alternate
atmospheric correction schemes are not necessary and may conflict with
processing integral to IMAGINE Subpixel Classifier processes. |
|
|
| 23. Does IMAGINE Subpixel Classifier atmospheric correction
calibrate the image to units of reflectance? |
| Not currently.
The current atmospheric correction process is an image normalization
process. Future enhancements to IMAGINE Subpixel Classifier may include
a process to calibrate imagery to apparent reflectance using only
scene information. |
|
|
|
|
|
|
|
AAI's Alaskan Oil Spill Assessment Using Satellite Imagery. (more...)
|
|
Demo Web-based mapping using SVG (more)
|
|
