Page 154 - Digital Analysis of Remotely Sensed Imagery

122 Cha pte r F o u r

Another way of achieving a high level of automation is through
the SPEAR toolbox. Its tools are intended for automatic spatial and
temporal change detection, pan-sharpening images, and terrain cat-
egorization. Nevertheless, it must be pointed out that this set of tools
is designed for defense and intelligence image analysts to perform
both common and advanced image processing routines. They are not
developed with the general users in mind. Neither are they available
for all kinds of analyses.
The DEM generation tools in ENVI 4.4 are tailored for extracting
elevational information from a pair of stereo aerial photographs or
satellite imagery. Additional information on this pair, such as the
rational polynomial coefficients for frame photographs and push-
broom sensor imager, must be supplied to construct a 3D model. Such
information is available for ASTER, Cartosat-1, IKONOS, OrbView-3,
and QuickBird data. The accuracy of the extracted DEMs depends on
the quality of ground control.
4.3.4 Processing of Hyperspectral and Radar Imagery
An extensive suite of functions designed specifically for processing
hyperspectral data are found in the Spectral toolbox. Some of these
functions are Pixel Purity Index, n-Dimensional Visualizer, and Spec-
tral Analyst. The Pixel Purity Index enables the identification of the
spectrally purest pixels in an image that serve as the spectral endmem-
bers in subpixel image classification. The n-Dimensional Visualizer
allows the interactive animation of the n-dimensional scatterplot,
through which to select the best endmember materials and their cor-
responding spectra. The Linear Spectral Unmixing function serves to
determine the relative spectral abundances of endmembers depicted in
multispectral and hyperspectral data. Spectral libraries may be built or
viewed through ENVI routines, and compared to image spectra. This
comparison with reference spectra is carried out at selected wavelengths,
based on the least-squares principle. A root-mean-square error is pro-
duced for each reference spectrum.
A wide range of radar imagery, such as EnviSat, ERS, JERS, Radar-
sat and Topsar, can be processed in ENVI with a full range of generic
or radar-specific methods. Some exemplary routines are antenna pat-
tern correction, slant-to-ground range correction, and generation of
incidence angle images. SAR-specific analysis functions include
review and reading of header information from CEOS-format data.
Other radar image analysis functions include adaptive and texture
filters, creation of synthetic color images, and a broad range of polari-
metric data analysis methods. EVNI 4.4 is unable to extract 3D infor-
mation from stereoscopic radar imagery, though.

4.3.5 Documentation and Evaluation
ENVI provides a comprehensive online document. This hyperlinked
text can be viewed sequentially or searched through an index. Keywords