Probably the most important file on each CD to get started is the “geo”
file. This is a quicklook of the data in georeferenced form.
It can be found by looking for the file with "_geo" at
the end of the name. Open this file
and display the bands labeled "Band 14", "Band 8", and
"Band 2" (RGB) to see the spatial coverage of your data as a true
color image. An ENVI header is provided for convenience when opening and
displaying in ENVI, however, these data can easily be displayed using other
software. The key image information is contained in a text file with "_geo.hdr"
at the end of the name.
sponsor received their data and pointed out "I opened several quick look images that came with the data. With the bands
set to 14, 8 and 2, these "true color" images are very blue and I am wondering if this is something I did wrong or the way it supposed to be?".
This occurs because of the mask values on the image inserted by the
geocorrection. To get a good contrast stretch, make your main window small enough so that it fits inside the image itself, then select
Functions->Display Enhancements->Default (quick) Stretches->[Image] Quick 2%. This should produce a properly stretched image.
ordered 5m data, why does the CD say 4m? How do I determine true pixel
pixel spacing of the geometrically corrected image is marked on the CD, labeled
for example as "GSD: 4m" at the end of the 3rd line from the
bottom of the CD. This pixel size is set by AIG during the geometric
correction. You can find the true pixel resolution of the uncorrected data
by examining the "_report.txt" file on the CD by printing
or using a text editor. The "mean downtrack line spacing (m):"
parameter indicates the mean GSD in the flight direction. The "nadir across track spacing (m):"
indicates the actual GSD at nadir.
the Apparent Reflectance Data
Open the "_effort" file using ENVI or another display/analysis
program. Display a grayscale image (Band 14 is a good one), a true color image
(Bands 14, 8, 2), or a False Color Infrared image (Bands 28, 14, 8).
Spatially/Spectrally browse the image by panning the display with a spectral
profile displayed. Familiarize yourself with the spatial and spectral
characteristics of your data.
notice some spikes in the reflectance data plot.
Why are these in the data and what do I do about them? (Mark bad bands)
You may notice "spikes" in the reflectance data
around bands 62-63 and bands 94-95. These bands are on the edges of the
1.4 and 1.9 mm atmospheric bands, where most of the
signal is obscured by the atmosphere. The best thing to do is to mask
these out as "bad bands" prior to processing. Depending on your
actual atmospheric conditions, you may have to mask several bands on either side
of these bands as well to totally eliminate the spikes. This can be
accomplished by selecting "Edit Header" from the ENVI File menu,
choosing Edit Attributes->Bad Bands List, then using "control
click" (depressing the Ctrl key and clicking with the left mouse button) on
the bands to be excluded from the plots and future analysis. Be sure to save
your results (you may have to change file protection on the header if you copied
it from CD). The next time you plot a spectral profile, these bands will
not be displayed. Additionally, they will be removed from the dataset the
first time you complete an ENVI processing procedure that creates a new output
there seem to be two or more apparent reflectance CDs for each flightline?
Because we've elected to distribute the HyMap data on CD, we've been
forced to break flightlines into segments that will fit within the 650mb CD
limitation. Both the radiance data and the apparent reflectance data have
been broken up this way. You can tell which segments go together and how
by looking at the "Flightline Id:" on each CD. The last 3
characters of the Flightline Id indicate the scene number. Match the
flight number and run number, then look for the sequential scene numbers.
For example, a HyMap data set broken into 3 scenes would have the sequential
Only the CD with the first
HyMap data segment has all of the ancillary HyMap data files. Subsequent
CDs with additional segments contain only the effort data.
do I reconstruct multiple segments into a single file for analysis?
It may be easiest just to use the virtual mosaic to do the
processing, however, this treats the data as BSQ and therefore programs that
work best on BIL data (e.g. MNF Transform, PPI, etc) may be somewhat slower than
using a BIL file. You can convert the data to BIL by either converting the
virtual mosaic on-the-fly, or by first making an
output mosaic, and then doing a BSQ-BIL conversion. Our recommendation is
to convert the virtual mosaic directly to BIL before doing any additional
processing. Here are the options, the decision is yours.
the conversion, first copy each segment from the multiple CDs to your disk. Be sure to
copy the "_effort" and the "_effort.hdr" files
for each segment. If you want to just look at the data in one piece, the best
thing to do is to create a virtual mosaic using ENVI. If you want to
process the data as a whole, you really need to mosaic the data to new output
file. Unfortunately, ENVI 3.2's default mosaic output is a BSQ file, so
you should also convert the BSQ file to BIL (optimized for spatial/spectral
processing), otherwise all of the spectral processing will be painfully
slow. The next version of ENVI will allow you to directly create a
BIL mosaic from BIL data, but for now you've got to do the manual conversion.
the Virtual Mosaic: Open each segment of the dataset using ENVI.
Select Register->Mosaic Images->Pixel Based Images from the ENVI
main menu. Click on the Import menu item at the top of the Pixel
Based Mosaic dialog and choose "Import file without feathering".
Select scene 01 from the Mosaic Input File dialog and click OK. Click again on
the Import menu item at the top of the Pixel Based Mosaic dialog,
choose "Import file without feathering", and select scene 02
from the Mosaic Input File dialog and click OK. Repeat for as many scenes as
you have for the complete flightline. Now enter the "Y Size" for the
output mosaic in the appropriate box at the top of the Pixel Based Mosaic
dialog. This can be determined by adding together the number of lines
for each image included in the mosaic, or by examining the "_report.txt"
file and looking at the input file size. For example, a line in the "_report.txt"
file will say something like "Input File: 512 x 5512 x 128 [Integer]".
The 5512 dimension will be your output mosaic Y Size. Now click on each
image in the mosaic and enter the starting line for that image in the "YO"
box in the Pixel Based Mosaic dialog. The starting line of each segment can be
determined as "number of lines in the preceding image(s)+1". Once
all of the images have been imported and positioned, select "File->Save
Template" to create a virtual mosaic. Give the template a file
like rootname.mos, click OK to build the mosaic template, then go to the ENVI
File menu and open the template file using File->Open Image File.
The virtual mosaic can be viewed and processed as if all of the data were in
one input file.
the Virtual Mosaic directly to BIL
Once the virtual mosaic is created and opened as above, then the simplest,
and quickest way to convert to BIL is to convert directly from the virtual
mosaic file. Select "Utilities->File
Utilities->Convert Data (BIL, BSQ, BIP)". Choose the virtual
image and click OK. In the File Convert Parameters dialog, click
on the "BIL" radio button to select BIL output, enter an
output filename, and click OK to start the processing. The converted data will appear in
the Available Bands List when processing is completed. This step may a
moderate amount of processing time (around an hour for an average flightline), but will be well worth it because of the
time saved later when using the BIL data for spectral processing.
a mosaicked output file and converting to BIL (in-place): Alternatively,
follow the procedure above, then once all the images have been imported and
positioned, select File->Apply, enter an output file name, and click
OK to create the BSQ mosaic file. The individual bands will appear in
the Available Bands List when the processing is completed. Now, convert
the mosaic into BIL by selecting "Utilities->File
Utilities->Convert Data (BIL, BSQ, BIP)". Choose the mosaic
image and click OK. In the File Convert Parameters dialog, click
on the "BIL" radio button to select BIL output, click
on the toggle arrow to change "Convert in Place" to "Yes",
and click OK to start the processing. The converted data will appear in
the Available Bands List when processing is completed. These two steps may
take substantial processing time (several hours), but can be used if the
direct convert option above is not an option.
do I go from here? What kind of processing do I need to do?
AIG recommends using ENVI and a standard processing methodology to achieve best
results. This typically consists of the following:
MNF Transform to reduce spectral dimensionality
Pixel Purity Index (PPI) to reduce spatial dimensionality
Thresholding of PPI to further reduce spatial dimensionality to size that can be
n-D Visualization to select endmembers
Mixture-Tuned-Matched Filtering (MTMF) to map spatial occurrences and abundances
Geometric Correction of results and map output
We typically analyze the 0.4 ~ 1.3 mm
and the 2.0 - 2.5 mm regions separately, then combine
the results at the end. Please see the ENVI on-line or printed tutorials for
do I georeference my data or results?
We've provided an ENVI routine that you can install to do the georeferencing.
This uses the information in the "_geo_glt" file provided with
the data to allow precision georeferencing based on the HyMap ephemeris
data. If you're already running ENVI, you need to exit the program (and
IDL) to install the new functionality.
Place the file "hymap_geocorrect_from_glt.sav" in the "save_add" directory
of your ENVI installation and then restart ENVI. When you restart ENVI a new option will automatically appear on the
ENVI menu under "Utilities->Data Specific Utilities->Hymap->HyMap Geocorrect from GLT Lookup
the new option and open/select the "_geo_glt" file in the HyMap
GLT Input File dialog. Open/Select the "_effort"
data (the mosaicked BIL data of the full flightline) or an uncorrected output
product (e.g.: a MTMF or other mapping result) in the Uncorrected Data Input
File dialog. Spectrally subset the data in this dialog if desired.
Enter -9999 as the Background Value enter an output filename, and
click OK to perform the geometric correction.
While these images are visually
pleasing and map-correct, they do have several practical drawbacks. First they have null values (-9999) around their edges that must be
masked in processing. Secondly they are often inflated in size by replicated pixels as indicated in the
"_glt" files. These two disadvantages lead to our suggestion to acquire and process the
HyMap imagery in its raw spatial format, then apply the geocorrection to the derived final products.
We do not recommend geocorrecting the entire reflectance data cube (_effort).
How about a quick mosaic of
You can make a quick mosaic of overlapping flightlines by opening all
of the georeferenced (_geo) images, choosing Register->Mosaic
Images->Georeferenced Images from the ENVI main menu, and importing each
of the overlapping images sequentially, with the images you want on top
following the underlying images. Be sure to use "Import With
Feathering" and enter "-9999" as the "Background
Data Value to Ignore" in the Enter Parameters dialog. When all of
the images to be mosaicked have been imported, select File->Apply,
enter the output filename, enter -9999 into the Background Value
text box., and click OK to create the mosaic.
How accurate is the georeferencing?
When I look at the mosaic,
I notice some offsets across image boundaries and apparent repeat pixels.
The geometric transformation and georeferenced images provided are
the best that can be accomplished without use of a detailed digital elevation
model and orthorectification. The correction is "precise" based
upon what we know about the sensor geometry and aircraft attitude from the
ephemeris data. There may be some absolute accuracy offsets, however,
introduced by the topography. This may be result in slight offsets in
absolute position from a map and also may be visible as offsets across image
boundaries and apparent repeat pixels when putting together a mosaic.
These problems can typically be corrected by selecting a few ground control
points (GCPs) to tie the images to a map, and/or using GCPs to match mosaicked
images in overlap regions.
Frequently Asked Questions - Click Here
offers several options for clients needing further assistance:
Data Analysis and Image Processing Workshops
To maximize your hyperspectral processing capabilities plan on
attending one of AIG's hyperspectral workshops. The
next workshop is offered November 16-19, 1999. Participants gain
hands-on instruction in hyperspectral data processing
analysis, and interpretation. Additionally, the fundamentals and
principles of imaging spectrometry are presented and advanced methodologies are discussed and demonstrated. Classes are offered
in Boulder, Colorado.
spectral processing and analysis
With over 60 combined years
of hyperspectral analysis experience
we can help you extract
the most information from your HyMap data. Some of the spectral
processing options offered by AIG include target identification
and mapping; endmember extraction, identification and
quantification; surface material maps; production processing;
data merging and resampling; as well as many other services.
Contact AIG for more information.
Additional Help or Information about the HyMap Groupshoot data contact:
Imaging and Geophysics LLC
4450 Arapahoe Ave., Suite 100
Boulder, CO 80303
Copyright 1999, © AIG Limited Liability Company, 4450
Arapahoe Ave, Suite 100, Boulder, Colorado 80303, Phone: 303-604-2844, FAX:
303-665-6090, Email: firstname.lastname@example.org,
(Updated 21 June, 2000