Geography Reference
In-Depth Information
4
Hyperspectral Imagery
in Fluvial Environments
Mark J. Fonstad
Department of Geography, University of Oregon, Eugene, OR, USA
hyperspectral instruments are designed so that each
'colour' or 'channel' collected are recording very narrow
band of wavelengths. This narrow, precisely-tailored
range of wavelengths for each band is as important as
the number of bands in distinguishing hyperspectral
imagery from broad-band passive optical imagery. This
band narrowness means that very specific information
about materials, such as biophysical and biochemical
properties, can be correlated and connected with very
specific optical wavelengths. This specificity is what
makes hyperspectral remote sensing so powerful.
As different hyperspectral instruments have different
spectral resolutions, wavelength numbers and ranges,
platforms, and advantages/disadvantages, this chapter
strives to be deliberately vague about the exact boundary
betweenwhat is multispectral imagery (described in detail
elsewhere in this volume) and hyperspectral imagery. The
typical rule-of-thumb is that hyperspectral imagery has
at least dozens of bands of light being collected. Another
way of defining hyperspectral imagery is the idea that the
word 'hyper' can mean 'too many'; hence, hyperspectral
imagery is imagery that contains 'too many' colours. This
may sound like a bad thing (and having a large amount
of data does have its problems), but what having 'too
many' colours really means is that you have too many
colours to worry about the technical quantity limitations
usually constraining colour remote sensing of environ-
ments. A more practical way to think of this concept
is thinking about classifying a river scene. If you wish
to classify instream river habitats into six different cat-
egories, then the number of colour bands to do the job
4.1 Introduction
One of the most conspicuous river tools on the rise dur-
ing the past decade has been the use of hyperspectral
imagery to characterise riverine environment compo-
nents such as aquatic habitats, bathymetry, and water
quality. Hyperspectral remote sensing, also known as
imaging spectroscopy, involves the use of instruments
that record electromagnetic radiation as narrow contigu-
ous spectra, often dispersing light into several dozen or
even hundreds of wavebands (often called either 'bands'
or 'channels' by the hyperspectral remote sensing commu-
nity), as opposed to traditional cameras or multispectral
imagers that capture radiation as broad and often discrete
spectral bands (3-20 bands at most). Scientific articles
extolling the virtues of hyperspectral imaging have shown
the improved capability of this type of remote sensing for
mapping instream habitats, classifying riparian vegeta-
tion, and extracting geometric information such as water
depth. As river managers planning applied river projects,
an important question should be the costs versus the ben-
efits of including hyperspectral imaging in a river survey.
This chapter focuses on that question.
The three colours sensed by the human eye, blue,
green, and red (of which all brain-processed colours are
derived), are actually each a mixture of several ranges of
wavelengths that centre on each of the three colours. In
optics, we refer to this phenomenon as being 'wide bands'
of light. The bands actually overlap considerably; in some
cases the eye can register something as more than one
colour even though it is a single wavelength. In contrast,
Search WWH ::
Custom Search