Page 146 - Digital Analysis of Remotely Sensed Imagery

Image Pr ocessing Systems 115

accuracy assessment routine. This tool enables evaluation pixels to be
selected either randomly or via a combination of random with strati-
fied sampling. Once the true identity of all evaluation pixels is speci-
fied, indices such as overall accuracy and Kappa (see Sec. 12.4.4) are
then generated automatically.
The Subpixel Classifier module in Imagine is designed to classify
images at the subpixel level. With this module it is possible to identify
ground features that occupy a fraction of a pixel, or to discriminate
materials with similar spectral characteristics. However, this add-on
module is not iusually available in the standard package of Imagine.

4.2.5 Spatial Modeler
The flexibility and efficiency of undertaking image analysis in ERDAS
Imagine is considerably improved with the use of Spatial Modeler. In
this environment repetitive and complex image analysis operations
are lumped together sequentially to form a deliberately built model.
This complex image analysis model may be constructed from a num-
ber of fundamental built-in functions in a pick-and-mix manner using
an internal modeling language without the user having to possess
any programming knowledge. After a model is constructed, the ana-
lyst just needs to modify the names of import and output files in the
first and last building blocks while the system sets the import and
output file names in all other blocks in the model automatically. Apart
from high efficiency, another added advantage of using Spatial Mod-
eler is the disappearance of the necessity to save a huge amount of
intermediate results.

4.2.6 Radar
Contained in the Radar toolbox is a suite of functions designed to pro-
cess radar imagery specifically. Some of the key functions in this module
are Interpreter, IFSAR, StereoSAR, and OrthoRadar, the last three being
add-on modules. The Radar Interpreter module is able to perform many
analyses on Synthetic Aperture Radar (SAR) images, such as speckle
removal, texture analysis, image merging, slant range adjustment, radio-
metric and geometric calibration. A correlating pair of SAR images can
then be used to generate DEMs. DEMs can also be created from stereo-
scopic Radarsat imagery based on the principles of radar interferometry
in IFSAR (ERDAS, 2003). In generating a DEM, the system is able to
select the pair of images that have the best geometric configuration.
OrthoRadar is a module designed for accurate correction of terrain-
caused geometric distortions in SAR imagery. The output orthoimages
are produced by modeling the SAR sensor from flight path parameters.
4.2.7 Other Toolboxes
In addition to the aforementioned modules, ERDAS Imagine contains
several more toolboxes (e.g., 3D, Vector, Stereo, VirtualGIS, AutoSync,
and DeltaQue) for performing specialized analytical tasks. These