Geography Reference
In-Depth Information
Table 1.1
Simplified Electromagnetic Spectrum table (Modified fromWard et al., 2002).
Micro-waves
Visible (Optical)
Infrared
blue green red near middle thermal
far
K-band:
1.1-1.4 cm
X: 2.4-3.75 cm C: 3.75-
7.5 cm
L: 15-30 cm Radio
radiation, with frequencies ranging from approximately
0.4 to 0.7 microns, received by the sensor (e.g. the camera
film). A further example would be standard colour pho-
tography. In this case, it would clearly be impossible to
have a near infinite number of detectors each sensitive to
a specific wavelength in the continuous visible spectrum.
The solutionwhichwas therefore adopted in the early days
of colour photography was to emulate human vision and
to re-create colour by first sampling radiation in three dis-
tinct areas of the spectrum: red, green and blue (Lillesand
et al., 2008). Within each of these primary colour bands,
the total amount of radiation incident upon the sensor is
recorded. Therefore for the red band, the sensor detects
all the radiation with frequencies between approximately
0.6 and 0.7 microns. For the green band the sensor detects
all the radiation from approximately 0.5 and 0.6 microns
and for the blue band, detectable wavelengths range from
0.4 to 0.5 microns. It should be noted that the term
'band' mentioned earlier is one of the most fundamental
in the remote sensing vocabulary. Formally, a 'spectral
band' is a finite section of the electromagnetic spectrum,
recorded and stored in a raster data layer. In the examples
above, a greyscale image is a one band image and a colour
image is a three band image. The term 'multispectral'
therefore refers to a remote sensing approach or dataset
which has several bands. Strictly speaking, colour pho-
tography, with its three bands in red, green and blue, can
be considered as multispectral imagery. However, many
authors and practitioners reserve the term 'multispectral'
for datasets which have at least four spectral bands with
one of the bands usually covering the infrared portion
of the spectrum. It should be noted that the number
of available bands is not the only important character-
istic of a remotely sensed image. Potential applications
of remotely sensed data are often limited and one might
even say, defined, by four additional parameters: spectral
resolution, spatial resolution, temporal resolution and, to
a lesser extent, radiometric resolution.
The concept of spectral resolution is closely related to
the concept of a spectral band. It relates to the width,
expressed in linear units of radiation wavelength (nm or
μ
m), of the spectral bands of the imaging device. A clear
distinction must therefore be made between the number
of bands measured by a sensor which determines the
range of radiation wavelengths that is sampled and the
width (or narrowness) of an individual band which deter-
mines the sensors sensitivity to specific spectral features.
Arguably the most classic example of the use of spectral
features in remote sensing is the detection of vegetation.
In healthy green vegetation, chlorophyll absorbs over 90%
of incident radiation within the visible spectrum, albeit
with a slightly lesser absorption and higher reflection in
green wavelengths, which explains the colour of vegeta-
tion. However, in the infrared wavelengths, vegetation is
a strong reflector. Sensors designed to detect vegetation,
such as the classic Thematic Mapper sensor mounted
on Landsat satellites, therefore try to exploit these dif-
ferences by sampling red light (0.63-0
m) which is
strongly absorbed by vegetation and near infrared light
(0.76-0
.
69
μ
m) which is strongly reflected. Note the rela-
tively narrow width, in spectral terms of these bands. Our
ability to accurately detect vegetation from remote sens-
ing therefore depends not only on increasing the number
of bands beyond the visible spectrum, but also on an
improvement of the spectral resolution. If we follow this
line of thought to its logical conclusion, we realise that it
would be desirable to produce a sensor with a very high
number of bands each with a very narrow bandwidth.
.
90
μ
Search WWH ::
Custom Search