Page 166 - Digital Analysis of Remotely Sensed Imagery

132 Cha pte r F o u r

support for its users. For instance, a tutorial could be added to
support novice users and to explain how the system should be run
properly. More resources such as freely available introductory
training materials and analysis examples should be provided to
ensure that the user can learn quickly how to operate the system
with confidence. With a minimum effort PCI could upgrade and
reorganize the e-mail discussion thread (Page, 2006). In this way
the geospatial community can exchange ideas with software devel-
opers and learn from each other. If the discussion becomes a well-
organized and moderated forum, the users can support one
another in using PCI products, such as how to find solutions to
their problems quickly. PCI can also benefit from this forum by
receiving feedback from its users about how to improve its prod-
ucts in future releases.
The strength of PCI lies in its high degree of automation and
customized workflows. However, this advantage turns to a limita-
tion at the same time as it can be applied to a narrow range of image
analysis functions such as image georeferencing, orthorectification,
mosaicking, and DEM generation. It does not apply to thematic
information extraction from multispectral imagery. Another obvi-
ous area for improvement is the componentization of functions,
even though they may be linked together into automated workflows
and run in batch processes. The flagship package contains only
seven modules, mostly for the extraction of elevational information.
There are too many add-on modules that are core components in
other similar systems. These modules are tailored toward perform-
ing specific tasks. For instance, the Optical module is designed to
analyze optical remote sensing data. It offers tools for spectral
unmixing, neural network classification, geometric and atmospheric
correction of AVHRR data, and surface temperature and vegetation
index extraction from AVHRR imagery. However, a separate add-on
module is required to analyze radar data. Contained in this add-on
module are a few programs for radar geometric correction and
despeckling filtering. Similarly, a separate module is essential for
analyzing hyperspectral images, such as visualization and spectral
libraries. One more module is needed to compress hyperspectral
data. In fact, a separate module is needed for atmospheric correc-
tion and another for pan sharpening. Thus, the organization of PCI
functions is not logical and self-obvious. The same function (e.g.,
radiometric correction) appears in more than one module and for
different satellite images. In this regard PCI is better suited to indus-
trial applications than teaching the concept of image analysis.
Geared toward extraction of digital information, PCI Geomatica
lacks capabilities of analyzing satellite data to produce accurate
land cover maps.