IEEE Geoscience and Remote Sensing Magazine - March 2013 - 71

Signal and Image
Processing in Analysis
of Remote Sensing Data
Algorithm development for analysis
of image data has persisted through
the decades of remote sensing at Purdue. Two decades after its acquisition,
the Indian Pine hyperspectral data
acquired by AVIRIS in 1992 over an
agricultural watershed west of Purdue
for studies of conservation tillage continue to be used by the international
community as a testbed for development of classification and unmixing
algorithms (Fig. 5). Analysis of hyperspectral data is a focus at Purdue, par31
308
ticularly for scenarios where training
data are limited and spectral signaFIGURE 6. Visits to Purdue MultisSpec software in 2012.
tures vary across large scenes or over
time. Leveraging our historical work
Although MultiSpec focuses on research in remote sensing,
in semi-supervised learning, we (and others in the internait is also used in analysis of multiband medical imagery and
tional community) have recently adapted active learning
in K-12 and university-level educational activities. In 2012,
heuristics for classification of hyperspectral data. Greater
there were more than 28,000 visitors from more than 3000
availability of hyperspectral data has also motivated our
locations, and 2,900 registered downloads of the MultiSpec
research in nonlinear manifolds to better accommodate
software application (Fig. 6).
nonlinear scattering in hyperspectral for dimensionality
Finally, the Purdue Terrestrial Observatory (PTO) operreduction, resulting in improved representation of some
ates a geostationary ground station for NOAA's GOES data
data sets and ultimately higher classification results and
and a tracking ground station for polar orbiting satellites,
improved unmixing.
including NOAA's AVHRR and NASA's Terra & Aqua MODIS
sensors. The PTO works closely with other research centers
Education and Outreach at Purdue:
at Purdue, including LARS, the Center for the Environment
Having Local to Global Impacts
(C4E), and the Purdue Climate Change Research Center
Purdue's long history of research in remote sensing has
(PCCRC) to provide regional satellite data products.
also translated into education and outreach through LARS'
reports, spectral libraries, and computer software. ContinLooking Forward: Our Alliance to Tackle
ued demand for historical articles led Purdue Libraries to
Society's Grand Challenges
recently digitize many of the LARS reports. Originally creThe early days of LARS played a pivotal role in shaping
ated during the 1960s to share results of Purdue's research
the research that Purdue conducts today. In particular, the
in algorithm development for analysis of remotely sensed
tradition of interdisciplinarity is pervasive throughout our
data, the processing system referred to as the Laboratory for
remote sensing research and education initiatives. Current
Applications of Remote Sensing Image Data Processing Sysefforts to develop a multi-track graduate program are untem (LARSYS) served as the forerunner of a number of sysderway. As a Land Grant institution, Purdue also prides
tems in government laboratories, university research labs,
itself in Discovery with Delivery-transfer of knowledge
and several commercially offered products.
to practice. Long-standing alliances between Purdue's
The current processing system, called MultiSpec (Purdue
Colleges of Agriculture, Engineering, and Science are the
Research Foundation https://engineering.purdue.edu/~biehl/
foundation for new initiatives in climate change research,
MultiSpec/) was developed by Prof. David Landgrebe and
global food security, sustainable communities, and disaster
Mr. Larry Biehl. It provides capability to interactively analyze
response and mitigation, as well as advances in sensor techEarth observational multispectral image data such as that
nology, data visualization and analytics, and data archiving
produced by the Landsat series of Earth satellites and hyperand distribution.
spectral image data from airborne and spaceborne systems.
The MultiSpec developers knew their system must be financially accessible, easy to learn, easy to use (even for the infreReference
quent user) and operate in modern software environments.
[1] D. Landgrebe. (1986). A brief history of the Laboratory for AppliThey designed the software to import data in a variety of forcations for Remote Sensing (LARS). [Online]. Available: http://
mats and export results in both thematic and tabular form.
www.lars.purdue.edu/home/LARSHistory.html
GRS
march 2013

ieee Geoscience and remote sensing magazine

https://engineering.purdue.edu/~biehl/ http://http:// http://www.lars.purdue.edu/home/LARSHistory.html

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - March 2013