Image Processing Reference
The huge number of bands and information redundancy might also require
data optimization, such as MNF transformations. Relevant analysis approaches
include material detection techniques, spectral angle mapping, or spectral
mixture analysis. Hyperspectral data will further gain in importance in the
future with the advent of new sensors and dedicated analysis options.
Download the trial version of
The Environment for Visualizing Images (ENVI) :
http://www.ittvis.com/Academic/Students/ENVIStudentEdition.aspx. Use the online
help system to learn more about the different options to analyze hyperspectral data.
Read Goetz et al. from the list of references to get an insight on imaging
Read Section 4 (Urban and Land Use Applications) and Section 13 (Collecting
Data at the Surface - Ground Truth; The “Multi” Concept; Hyperspectral
Imaging Spectroscopy) in the NASA Remote Sensing Tutorial: http://rst.gsfc.
Discuss in how far multispectral and hyperspectral data differ. Why are these
differences particularly useful in analyzing urban remote sensing data?
Why is it mandatory to perform a radiometric pre-processing of hyperspectral
data to compare spectra with those from a spectral library?
Access the ASTER spectral library (see list of references). Compare the avail-
able list of urban materials with the prevailing materials of an urban area you are
familiar with. Evaluate the completeness of the database.
Use the trial version of ENVI to load the John Hopkins University spectral
library (choose: Spectral - Spectral library viewer - Open Spec Lib; change to
the folder jhu_lib and chose manmade2.sli; confirm your choice with ok). Click
the “red smooth-faced brick” (a plot window will open to visualize the spec-
trum) and then “Reddish asphalt roofing shingle”. Why would Spectral Angle
Mapping not necessarily be the best way to separate these materials?
A modeling approach for simulating the urban heat island requires an estimate
of the amount of different roofing materials. Which analysis methods could be
chosen to extract such information from hyperspectral data?