-
Remote sensing satellites,sensors and data
3.1. Digital image storage and display
The primary data used in remote sensing are images of the earth
surfaces,which have a digital form so that they can be stored and
processed incomputers. This means that an image is in fact a set of
numbers, whichrepresents a quadratic record just like a traditional
analog image. Thedigitalization is achieved by dividing the
depicted area into small squarepicture elements or pixels, of equal
area, arranged in rows and columns.Thus the mathematical model for
a digital image is a matrix of R rowsand C columns, containing CRN
×= elements, as many as the imagepixels. The matrix element ijx of
the i
th row and jth column, is the valueof the corresponding pixel,
representing a level of light intensity or “toneof gray” by means
of a real number within a predefined range. Thesmaller possible
value corresponds to “no intensity” or black color andthe largest
one to “full intensity” or white color. To achieve economy
incomputer storage, where the binary number system is used, only
integervalues are used; with k bits disposable for each value the
range ofpossible k2 values is between 0 and 12 −k . For example, in
the mostcommon choice of 8-bit storage, the integer values are
25628 = and liebetween 0 (black) and 255 (white), which are stored
in binary form as00000000 and 11111111, respectively.
A digital image viewed as a computerfile consists of all the
image values to-
gether with some additional necessary information, which is
storedin a specific initial part of the file (or sometimes in a
separate file) and iscalled the header. The header contains all the
necessary informa-
Chapter 3
Figure 3-1
-
CHAPTER 3 A. DERMANIS: REMOTE SENSING
2
tion in order to reconstruct the image from the file, such as
the number ofbits used for each value, the order with which the
values are stored andthe number of columns in the image. It may
also contain other auxiliaryuseful information about the particular
image. Although analog imagescan be digitized using digitizing
scanners, the vast majority of imagesused are captured directly in
digital form, using sensors such as digitalcameras or scanners.
These consist of sensors, which are eithersimultaneously “looking”
at different parts of the recorded scene (pixels),or they are
sequentially directed to different pixels, using a
mechanicalscanning procedure, or they combine both possibilities.A
digital image is always a “black and white” image. Color images are
infact a combination of three different images of exactly the same
scene,each one depicting the intensity of one of the three basic
colors (redgreen and blue), which combined can reproduce all other
possible colors.Since red, green and blue refer to different areas
of the electromagneticspectrum, a color image is in a certain sense
a “multispectral image”.A more generalized type of multispectral
images is used in remotesensing. A multispectral image is an
assembly of images, each onereferring to exactly the same scene and
with identical partitions intopixels, where each one is a record of
the intensity of electromagneticradiation only within a different
area of the spectrum or spectral band.Thus a multispectral image
consists of related images in differentspectral bands, which are
not confined in the visible part of the spectrum.The bands used in
remote sensing are in the visible and infrared part ofthe spectrum.
The ultraviolet part is also rarely used, while microwavebands are
used for active remote sensing, where the electromagneticenergy
reflected from the earth surface and recorded, originates from
aradar system aboard an airplane or satellite.The file for a
multispectral image contains the values kijx for all rows
Ri ,,1= , all columns Cj ,,1= and all bands Bk ,,1= of the
image. Depending on the order with which the values kijx are
stored wemay distinguish between three fundamental generic formats:
BSQ, BIPand BIL.
In the Band Sequential format (BSQ) thebands are stored one
after the other in their natural order (increasingwavelength),
while each band is stored in the way a page is read, line byline
from left to right. Thus the order of the pixel values in BSQ is
thefollowing
11
11
111 Cj xxx …
1111 iCiji xxx …
1111 RCRjR xxx
kC
kj
k xxx 1111 …kiC
kij
ki xxx 1 …
kRC
kRj
kR xxx 1
BC
Bj
B xxx 1111 …BiC
Bij
Bi xxx 1 …
BRC
BRj
BR xxx 1
Figure 3-2
Figure 3-3
-
Remote sensing satellites, sensors and data
3
In the Band Interleaved by Pixel (BIP) format the values for all
bands foreach pixel are stored together, while pixel arrangement
follows that ofpage reading:
Bk xxx 1111111 …
Bj
kjj xxx 11
11 …
BC
kCC xxx 11
11
Bi
kii xxx 11
11 …
Bij
kijij xxx
1 … BiCkiCiC xxx
1
BR
kRR xxx 11
11 …
BRj
kRjRj xxx
1 … BRCkRCRC xxx
1
In the Band Interleaved by Line (BIL) format, priority is given
to thelines, which are stored band by band. Once a line in all
bands is storedthe next line follows:
11
11
111 Cj xxx …
kC
kj
k xxx 1111 …BC
Bj
B xxx 1111
1111 iCiji xxx …
kiC
kij
ki xxx 1 …
BiC
Bij
Bi xxx 1
1111 RCRjR xxx …
kRC
kRj
kR xxx 1 …
BRC
BRj
BR xxx 1
The order of storage for the three formats is depicted in figure
3.4.Unlike an analog image, a digital image has no dimensions of
its own.When depicted on a computer screen, or printed on paper, it
attains aparticular scale depending upon the resolution of its
presentation, i.e. onthe size allocated to each pixel. The number
of pixels along a unit oflength usually expresses the resolution of
image recording media. Theterm dpi (dots per inch) is typically
used to refer to the number of pixels(dots) per inch. For example
on a screen of typical resolution 80 dpi,
64008080 =× pixels are depicted within a square inch. For a
printer theminimum requirement for reasonably good printing quality
is a resolutionof 300 dpi, where 90000300300 =× pixels are printed
within a squareinch.A single-band image is displayed by assigning
different tones of grayfrom black to white in direct analogy to the
pixel values. On a color dis-plays using the RGB (R = red, G =
green, B = blue) system the tones ofgray are achieved by combining
the same amount of the three basic col-ors. When image bands are
available in the red, green and blue parts ofthe spectrum, they can
be naturally combined in order to display a “truecolor” image. For
a multispectral image any three of the available bandscan be used
with an arbitrary assignment of the RGB colors to produce a“false
color” display of the image. Color displays certainly convey
moreinformation than black and white ones and appropriate “false
color”
-
CHAPTER 3 A. DERMANIS: REMOTE SENSING
4
Figure 3-4: Formats of digital multispectral images: BSQ (=band
sequentilal), BIP (band interleaved by pixel andBIL (band
interleaved by line).
-
Remote sensing satellites, sensors and data
5
combinations of bands may be found which enhance certain
characteris-tics of a particular scene. Any color is uniquely
define by the three valuesR, G, B assigned to the red, green and
blue colors, respectively. For ex-ample in the 8-bit case, where
the values from 0 to 255 are used, blackcolor corresponds to R = 0,
G = 0, B = 0, white to R = 255, G = 255, B =255, while the various
tones of gray to choices of common values R = G= B between 0 and
255.If the three colors are viewed as axes in 3 dimensions, each
color corre-sponds to a particular point having the R, G, B values
as coordinates. Inthis way all colors are contained inside the
color cube (fig. 3.5) definedby the limits 2550 ≤≤ R , 2550 ≤≤ G ,
2550 ≤≤ B . The tones of grayare located on the diagonal of the
cube joining black (0,0,0) with white(255, 255, 255). The
combination of red and green gives yellow (R=255,G=255, B=0), the
combination of red and blue gives violet (R=255, G=0,B=255) while
the combination of green and blue gives cyan (R=0,G=255,
B=255).Another color system used in computer image processing
software is theCMYK system (fig. 3.6), which is related to color
printing. It producesall colors by combining cyan (C) with magenta
(M), yellow (Y) andtones of gray (K). Color in printing is achieved
by printing dots of vary-ing size in these four colors against the
white background of the paper.The values for each colors are
percentages of the maximum size, thus thezero values 0KYMC ====
corresponds to white. In order toachieve good printing quality it
is necessary to separately introduce theblack color K, rather than
to try to create it by combining other colors, asdone in the RGB
system. The %100K = value creates black independ-ently of the
values of C, M, and Y on which it is superimposed. Red iscreated by
combining magenta and yellow; green by combining cyan andyellow;
blue by combining cyan and magenta.A third color system the HSI (or
HSB) appeals more to perception (fig.3.7). Hue (H) refers to the
various colors arranged in cyclic order withcharacteristic values
at red (H=0°=360°), yellow (H=60°), green(H=120°), cyan (H=180°),
blue (H=240°) and magenta (H=300°). Satu-ration (S) refers to the
percentage of the color defined by hue, while in-tensity (I) to the
percentage of “white” where zero refers to the blackcolor. The
value %0I = gives black independently of the values of Hand S. %0S
= (no color) and %100I = (no black) gives white for anyvalue of H.
The pure color corresponding to a specific value of H isgiven by
setting in addition %100S = and %100I = . The values of grayare
reproduced by varying the values of intensity I, while holding
%0S = and any value of H.
-
CHAPTER 3 A. DERMANIS: REMOTE SENSING
6
Figure 3-5: The color cube associated with the RGB (red, green,
blue) color system
-
Remote sensing satellites, sensors and data
7
Figure 3-6: The CMYK color system. Color cubes for the CMY axes
(cyan, magenta, yellow) for two different tonesof grey (K=0 and
K=40)
-
CHAPTER 3 A. DERMANIS: REMOTE SENSING
8
Figure 3-7: The HSI color system. Combination of intensity I (%
of white) and saturation S (% of color) for variousangular values
of hue H (color)
3.2. Satellites sensors
The first images used for extracting information about the earth
surfacewere the usual analog aerial photographs, which record
electromagneticradiation over the whole visible part of the
spectrum. The related disci-pline, photointerpretation was based
mostly on texture information in or-der to identify the various
land cover classes, as well as other types of in-formation.
Although texture is not irrelevant, the main source of infor-mation
in remote sensing comes from variations in different spectralbands.
Thus the beginning of remote sensing could be traced back to theuse
of aerial photographs sensitive only to a particular band through
the
-
Remote sensing satellites, sensors and data
9
Band 1 Band 2 Band 3
Band 4 Band 5 Band 7
Band 6 R=3, G=2, B=1 R=4, G=3, B=2
R=7, G=4, B=2 R=7, G=5, B=4 R=5, G=7, B=4
Figure 3-8: The 7 bands of the TM sensors and various color
combinations in the RGB system
-
CHAPTER 3 A. DERMANIS: REMOTE SENSING
10
use of filters and in particular to the use of film sensitive to
infrared ra-diation.Eventually instruments for digitally recording
multispectral images havebeen developed for use on airplanes.
However airborne data collectionfor remote sensing remains a very
costly operation that can be under-taken only by large research
institutes and governmental agencies. Whatreally gave a boost to
the discipline was the placement of instruments onsatellites, which
produced data in a routine way, made available to awide range of
users at reasonable cost through distribution agencies.Thus the
beginning of modern remote sensing could be set at thelaunching of
the first Landsat satellite on July 23, 1972. The use of a se-ries
of Landsat satellites by NASA has guaranteed a continuous flow
ofdata, followed later from the French SPOT satellite series.
Individual sat-ellite missions are also of importance, but the
majority of practical workis in fact based on the sensors aboard
the Landsat and SPOT satellites.In addition the development of
various commercial software packagesmade remote sensing an
expanding discipline, with a wide variety of ap-plications, such as
thematic cartography, agriculture, forestry, geology,hazard
assessment, environmental monitoring, etc. Today remote sensingis a
major information provider for Geographical Information
Systems(GIS) which have are an indispensable tool for the
administration oflands and natural resources.Remote sensing
satellites are placed in orbits around the earth, with
char-acteristics that best fit their purpose. Circular orbits
guarantee a constantdistance from the earth surface, while
placement of the orbital planeclose to the poles guarantees a sun
synchronous operation. This meansthat the orbital plane slowly
changes position in inertial space, rotatingaround the earth axis
with the same speed as the sun, so that half of theorbit is always
over the sun-illuminated part of the earth. The height ofthe
satellite affects the period of its revolution around the earth, as
wellas the spatial resolution of the recorded images, which
diminishes withheight. Due to the eastward rotation of the earth,
the satellite after onerevolution does not return over the same
part of the earth, but over a newlocation towards the west. This
fact is used to “tune” the orbit so that itpasses over all parts of
the earth, with the exception of two polar cups,within a fixed
number of days say k days. Consequently the satellite re-visits the
same location every k days and this repeat cycle determinesthe
temporal resolution of the recorded data.We must distinguish
between the “sensor” as an image recording instru-ment and
individual sensor elements which at any moment record theamount of
electromagnetic radiation within a specific spectral band ar-riving
from a particular location (pixel) on the earth surface. A
sensorhas zero sensitivity outside a spectral interval but its
sensitivity withinthe band is not uniform. It typically has a
maximum and gradually di-minishes towards the limits of the
interval (fig. 2.2). The number of theavailable bands and the total
part of the spectrum covered determine thespectral resolution of
the sensor. Sensors with a number of bands in the
-
Remote sensing satellites, sensors and data
11
order of ten are characterized as multispectral, those with
bands of theorder of hundred as hyperspectral, while ultraspectral
sensors with thou-sands of bands are anticipated as a future
development.When a sensor element “looks” at a particular location,
it does not recordin a homogeneous way the radiation arriving from
a square pixel on theground. It is more sensitive at a central
direction corresponding at thepixel center and less at the edges of
the pixel. The area viewed is thus notprecisely determined, but it
is typically taken to be the part where sensorsensitivity is more
than 50% of its maximum. It is this circular essentialextend of the
viewed area, as well as the strictly related angular
distancebetween neighboring pixel centers, which determine the
spatialresolution of the sensor. The actual pixel size (and shape)
depends on theinclination of the direction of view with respect to
the (horizontal) earthsurface. For this reason, for satellite
sensors flying at a constant height,the spatial resolution is
associated with the pixel size at nadir direction.For airborne
sensors, where the height of flight is not constant, as in
thesatellite case, pixel size at nadir is inversely proportional to
the height offlight. For this reason spatial resolution is
characterized by the angularopening corresponding to a pixel at
nadir (fig. 3.9), which is calledInstantaneous Field of View
(IFOV).As the satellite flies over the earth surface it collects
data from a certainregion to the left and right of the trace of its
orbit on the earth; the lateralextend of the covered region is
called the swath of the sensor (fig. 3.9).The total angle of
coverage corresponding to the swath is called Field ofView (FOV).
The distance between neighboring satellite tracks on theearth,
which becomes maximum at the equator, must not exceed theswath, in
order to secure that no land strip between tracks is left
uncov-ered.The detail used for the description of the amount of
received electromag-netic radiation constitutes the radiometric
resolution of the sensor, whichis directly related to the total
number of values used. Since the number ofvalues is always of the
form k2 , radiometric resolution is typically ex-pressed by the
number k of bits used for storing a single value, which issometimes
called dynamic range.Among the various technical characteristics of
the various images wewill only discuss shortly those related with
how the scanning of the areapixels is achieved. In the most simple
case (fig. 3.10a) a different pixel isobserved at each instant,
while pixels in a row perpendicular to the di-rection of satellite
motion are sequentially scanned with a help of a ro-tating mirror
which alters the direction of observation. When scanning ofa row is
completed the scanning is repeated but a new image row is ob-served
because of the satellite motion. The electromagnetic energy is
di-verted by means of varying refraction through a prism, towards
separatesensors for each spectral band. A different possibility
(fig. 3.10b) is thesimultaneous observation of a whole row at a
single instant, by means ofa corresponding array of sensors, one
for each pixel. This is the “push-broom” scanning system used in
the HRV sensors aboard the SPOT sat-
Figure 3-9
-
CHAPTER 3 A. DERMANIS: REMOTE SENSING
12
ellites. A more advanced arrangement (fig.3.10c) allowing the
simultane-ous observation of a number of rows, by means of a matrix
of sensors, isthe “stepstare” scanning system. A sort of
combination of the two firstsystems is the “whiskbroom” system,
used on the MSS sensor aboard thefirst five Landsat satellites. An
array of six sensors is observing simulta-neously six pixels in the
direction of satellite motion (fig.3.10d). A singlescanning in the
perpendicular direction allows the coverage of six imagerows at the
same time. Repetition of the scanning allows the observationof the
next six rows, and so on, along the direction of satellite motion.
Inolder data when calibration of the six sensors was not very
successfuland their sensitivity varied, a “stripping effect” could
be seen in raw un-processed data.
Figure 3-10: Various types of scanning sensors.
Other technical characteristics is the capability of diverting
the imagingsensor in different directions providing nadir and
off-nadir images. Off-nadir viewing is selectively used in the SPOT
satellites to provide cover-age of the same area from two different
directions during different satel-lite passes. This produces stereo
images that can be used in digital pho-togrammetry for the
determination of terrain elevations.
-
Remote sensing satellites, sensors and data
13
Among the various sensors we present in more detail those on the
Land-sat and SPOT satellite missions, because of their importance
in remotesensing applications. Other satellite missions and sensors
as well as air-borne sensors are only summarized in tables.
Figure 3-11: The Landsat 7 satellite with its solar panel
(right) and the ETM+sensor (left).
The NASA Landsat missions were the first series of dedicated
missionsthat operate continuously since 1972, despite the
unfortunate loss ofLandsat 6 on launch. They have remained for
years the essential sourceof widespread data, at least until the
beginning of the SPOT series in1986, which constitute the main
alternative source at present. The firstthree missions carried two
sensors, the most important one being theMultispectral Scanner
(MSS) with four bands (4 at green, 5 at red, 6 and7 in the near
infrared) with pixel size of about 80 m and a swath of 185km. An
additional band at the thermal infrared was included only on
theLandsat 3 mission, with smaller spatial resolution (pixel size
about 240m).
Table 3.1: The Landsat satellite missions and their sensors.
THE LANDSAT SATELLITES
Satellite in orbit since until sensorsLandsat 1 23 July 1972 6
Jan. 1978 MSS, RBVLandsat 2 22 Jan. 1975 25 Feb. 1982 MSS,
RBVLandsat 3 5 March 1978 31 March 1983 MSS, RBVLandsat 4 16 July
1982 July 1995 MSS, TMLandsat 5 1 March 1984 … MSS, TMLandsat 6 5
Oct. 1993 lost on launch ETMLandsat 7 15 April 1999 … ETM+
The second sensor on Landsat 1, 2, and 3, was the Return Beam
Vidicon(RBV). On Landsat 1 and 2 the RBV had three bands (1 at
blue, 2 in red,3 in infrared) with the same spatial resolution as
the MSS. On Landsat 3
-
CHAPTER 3 A. DERMANIS: REMOTE SENSING
14
two identical RBV sensors were placed, each with a single
panchromaticband. Their spatial resolution was 40 m and the swath
98 km, which forthe two sensors overlapped into a total swath of
185 km. Following thebands 1, 2, 3 of the RBV, the MSS bands were
named 4, 5, 6, 7 (and 8) inthe first three Landsat missions. Since
the Landsat 4 mission the The-matic Mapper (TM) and its successors
Enhanced Thematic Mapper(ETM) on the lost Landsat 6 and ETM+ on
Landsat 7 replaced the RBV.In the newer missions the MSS bands were
renamed from 4, 5, 6, 7 to 1,2, 3, and 4. The MSS data gave rise to
extensive research in many appli-cation areas. It is characteristic
of this development that techniques suchas the vegetation index
(see chapter 9) originally tailored for the MSSwere later modified
to fit the TM and the SPOT sensors.
Figure 3-12: The spectral regions of the 7 bands of the Landsat
TM and the 4bands of the SPOT HRVIR sensor.
The Thematic Mapper aboard Landsat 4 and 5, has 7 bands (1 at
blue, 2at green, 3 at red, 4, 5, at near infrared, 7 at mid
infrared and 6 at thethermal infrared), a spatial resolution of 30
m (except for the thermalband 6 which has a lower resolution of 120
m) and a swath of 185 km.Note that band 7 is out of its natural
(increasing wavelength) order be-cause it was added later in the
original sensor design. The ETM+ oper-ating on Landsat 7 has more
or less the same characteristics as the TM,with the spatial
resolution of band 6 improving from 120 to 60 m and anadditional
panchromatic band with 15 m spatial resolution. The TM andETM+ have
a radiometric resolution of 8 bits (0-255), double the 7 bit(0-127)
resolution of MSS (6 bits for band 4, formerly 7). Among
variouswidely used sensors the TM and ETM+ are the ones that cover
the threebasic colors (RGB) in the visible and the relevant bands
can be used toproduce “true color” imagery. Another difference
between the first 3Landsat missions and the following ones concerns
the orbit design: Low-ering the satellite altitude from 920 to 705
km resulted in a little smallerperiod and a small improvement of
the temporal resolution from a repeatcycle of 18 to one of 16
days.
-
Remote sensing satellites, sensors and data
15
Table 3.2: Characteristics of the Multispectral Scanner.
MSS (Multispectral Scanner)Landsat 1, 2, 3, 4, 5
Spectral resolution:Band wavelength
Landsat1,2,3
Landsat4,5 (µm)
4 1 0.5 − 0.6 green5 2 0.6 − 0.7 red6 3 0.7 − 0.8 near infrared7
4 0.8 − 1.1 near infrared
8 (Landsat 3) 10.4 − 12.6 thermal infraredSpatial resolution:
79×79 m (1, 2, 3)
81.5×81.5 m (4), 82.5×82.5 m (5)Band 8: 237×237 m
Swath: 185 kmRadiometric resolution: 128 values (7 bits)
Band 8: 64 values (6 bits)Temporal resolution: 18 days (Landsat
1,2,3),
16 days (Landsat 4,5)
Table 3.3: Characteristics of the RBV sensor.
RBV (Return Beam Vidicon)Landsat 1, 2, 3 (pan.)
Spectral resolution:Band wavelength
(µm)1 0.475 − 0.575 blue2 0.580 − 0.680 red3 0.689 − 0.830 near
infrared
Pan 0.505 − 0.750 panchromatic(Landsat 3)
Spatial resolution: 79×79 mPan, Landsat 3: 40×40 m
Swath: 185 kmPan: (2×) 98 km
Temporal resolution: 18 days
-
CHAPTER 3 A. DERMANIS: REMOTE SENSING
16
Table 3.4: Characteristics of the Thematic Mapper.
TM (Thematic Mapper)Landsat 4, 5
Spectral resolution:Band wavelength
(µm)1 0.45 − 0.52 blue2 0.52 − 0.60 green3 0.63 − 0.69 red4 0.76
− 0.90 near infrared5 1.55 − 1.75 mid infrared7 2.08 − 2.35 mid
infrared6 10.4 − 12.5 thermal infrared
Spatial resolution: 30×30 mBand 6: 120×120 m
Swath: 185 kmRadiometric resolution: 256 values (8 bits)Temporal
resolution: 16 days
Table 3.5: Characteristics of the ETM+ sensor.
ETM+ (Enhanced Thematic Mapper)Landsat 7
Spectral resolution:Band wavelength
(µm)Pan 0.52 − 0.90 panchromatic
1 0.45 − 0.515 blue2 0.525 − 0.605 green3 0.63 − 0.690 red4 0.75
− 0.90 near infrared5 1.55 − 1.75 mid infrared7 2.09 − 2.35 mid
infrared6 10.40 − 12.50 thermal infrared
Spatial resolution: 30×30 mBand 6: 60×60 mPan: 15×15 m
Swath: 185 kmRadiometric resolution: 256 values (8 bits)Temporal
resolution: 16 days
The series of the French satellites SPOT (Système Probatoire d’
Obser-vation de la Terre) began in 1986 with 4 satellites flown up
to now and afifth planned for 2002. The sensor on SPOT 1, 2 and 3
is the High Reso-lution Visible (HRV) with three bands (1 at green,
2 at red and 3 at nearinfrared), with 20 m spatial resolution, 60
km swath (per sensor) and 8bit radiometric resolution. An
additional panchromatic band has a spatial
-
Remote sensing satellites, sensors and data
17
Figure 3-13: The SPOT 4 satellite with two HRVIR sensors
(left)
resolution of 10 m. On SPOT 4 (fig. 3.13) an improved version
wasplaced, the High Resolution Visible InfraRed (HRVIR). The most
notablechange was the addition of a fourth band at the mid infrared
region. Anew sensor on SPOT 4 is the Vegetation (VGT), with low
spatialresolution (1 or 4 km) but wider field of view (FOV = 101°),
whichallows the fast collection (1 to 2 day resolution depending on
latitude) ofaveraged global data (40° S – 60° N). The VGT has four
bands (1 atblue, 2 at red, 3 at near infrared and 4 at mid
infrared), a swath of 2250km and operates at two modes, the
“direct” or “regional” (pixel size 1km) and the “recording” or
“world-wide” mode (pixel size 4 km). On theplanned SPOT 5 mission
the High Resolution Geometry (HRG) will havethe same
characteristics as HRVIR, but an expected higher spatialresolution
of 10 m and 3 or 5 m in panchromatic band. This will becomparable
to the already available 4 m resolution multispectral imagesof the
Ikonos satellite launched on September 24, 1999, at an altitude
of680 km. Ikonos also provides panchromatic images of 1 m
resolution inaddition to the 4 bands of multispectral data, which
are identical to thebands 1, 2 and 4 of the Landsat Thematic
Mapper.Each SPOT satellite carries two identical HRV or HRVIR
sensors (fig.3.14), so that the individual swaths of 60 km are
combined with a 3 kmoverlap into a total swath of 117 km, when both
sensor observe in thenadir direction mode. The characteristic
difference of the sensors is thepossibility to change the direction
of view from ground control “oncommand”, with the help of a
steerable mirror (fig. 3.15). In addition tothe usual nadir mode
used for “thematic” mapping, oblique views (at 0.6°intervals up to
±27° away from nadir) are used to cover the same areafrom two (or
more) different directions taken at different satellite passes(fig.
3.16). These converging views are necessary for the
geometricmapping of the area, using digital photogrammetric tech
niques.
Figure 3-14
Figure 3-15
-
CHAPTER 3 A. DERMANIS: REMOTE SENSING
18
A typical product is a Digital Terrain Model (DTM), i.e. a file
of eleva-tions in a grid over the earth surface.
Figure 3-16: Stereometric view from different passes of the SPOT
satellite forthe photogrametric determination of the terrain
anaglyph.
Table 3.6: The SPOT satellite missions and their sensors.
THE SPOT SATELLITES(Système Probatoire d’Observation de la
Terre)
Satellite in orbit since until sensorsSPOT 1 22 Feb. 1986 1994
HRV (×2)SPOT 2 22 Jan. 1990 … HRV (×2)SPOT 3 26 Sept. 1993 14 Nov.
1997 HRV (×2)SPOT 4 24 March 1998 … HRVIR (×2), VGTSPOT 5 planned
2002 HRG (×2)
-
Remote sensing satellites, sensors and data
19
Table 3.7: Characteristics of the HRV sensor.
HRV (High Resolution Visible)SPOT 1, 2, 3
Spectral resolution:Band wavelength
(µm)Pan 0.51 − 0.73 panchromatic
1 0.50 − 0.59 green2 0.61 − 0.68 red3 0.79 − 0.89 near
infrared
Spatial resolution: 20×20 mPan 10×10 m
Swath: 60 kmRadiometric resolution: 256 values (8 bits)Temporal
resolution: 26 days
Table 3.8: Characteristics of the HRVIR sensor.
HRVIR(High Resolution Visible InfraRed)
SPOT 4
Spectral resolution:Band wavelength
(µm)Pan 0.61 − 0.68 panchromatic
1 0.50 − 0.59 green2 0.61 − 0.68 red3 0.79 − 0.89 near infrared4
1.58 − 1.75 mid infrared
Spatial resolution: 20×20 m(Band 2 also 10×10 m)
Pan 10×10 mSwath: 60 kmRadiometric resolution: 256 values (8
bits)Temporal resolution: 26 days
Table 3.9: Characteristics of the Vegetation sensor.
VGT (Vegetation)SPOT 4
Spectral resolution:Band wavelength
(µm)1 0.43 − 0.47 blue2 0.61 − 0.68 red3 0.78 − 0.89 near
infrared4 1.58 − 1.75 mid infrared
Spatial resolution: 1×1 km(also 4×4 km mode)
Swath: 2250 kmRadiometric resolution: 256 values (8
bits)Temporal resolution: 26 days
-
CHAPTER 3 A. DERMANIS: REMOTE SENSING
20
Table 3.10: Characteristics of remote sensing satellites.(np =
near polar, gs = geostationary)
REMOTE SENSING SATELLITES AND SENSORS
Satellite from until sensorsaltitude
h(km)
inclin.i
(deg)
periodT
(min)
repeatcycle(days)
Landsat 1, 2, 3 1972 (1) 1983 (3) MSS, RBV 920 99 103 18Landsat
4, 5 1982 (4) ... MSS, TM 705 np 98.9 16Landsat 7 1999 ... ETM+ 705
np 98.9 16SPOT 1, 2, 3 1986 (1) HRV 832 np 101 26SPOT 4 1998 HRVIR,
VGT 832 np 101 26NOAA 10 - 14 1986 ... AVHRR 861 / 845 np 1ADEOS
Aug 1996 July 1997 AVNIR, OCTS 797 np 101 41Nimbus 7 Dec 1986 CZCS
1000IRS-1A Mar 1988 LISS-1 904 101 22IRS-1B Aug 1991 LISS-2 904 101
22IRS-P2 Oct 1994 LISS-2 101 24IRS-1C 1995 LISS-3, WIFS 817 101
24IRS-1D Sep 1997 LISS-3, WIFS 736 / 825 101MOS-1 Feb 1987 MESSR,
VTIR 908 99.1 17MOS-1b Feb 1990 MESSR, VTIR 908 99.1 17RESURS-01
1985 MSU-SK 678 98 21Orb View-2 SeaWiFS 705 np 98.9 1GMS VISSR
35900 0 gsGOES GOES Imager 35900 0 gsSpace Shuttle 1983
MOMS-01Space Shuttle 1993 MOMS-02Mir Space Station 1996
MOMS-02PERS-1 July 1991 ATSRERS-2 Apr. 1995 ATSR-2ENVISAT-1 1999
AATSREOS-AM-1 1999 MODISIkonos Sept 1999 681 98.1 98
-
Remote sensing satellites, sensors and data
21
Table 3.11: Characteristics of multispectral satellite
sensors(Resolutions and swath)
SENSOR No. ofbandsIFOV(m)
swath(km)
repeatcycle(days)
dynamicrange(bits)
MSS 4 79 185 18 / 16 7 / 6RBV 3 79 185 18TM 7 30 185 16 8ETM+ 7
30 185 16 8HRV 3 20 60 26 8HRVIR 4 20 60 26 8VGT 4 1000 2250 26
8ATSR 4ATSR-2 7 1000 500AATSR 7 500
1000500
AVHRR 5 1100 2394 10AVNIR 4
pan168
80 41 87
OCTS 12 700 1400 41 10CZCS 6 825 1566 6 8VISSR 2 1250
500068
GOESImager
5 100040008000
10
MESSR 4 50 100 17 8VTIR 4 (900)
27001500 17 8
LISS-1 4 73 146 22 7LISS-2 4 36 146 22 7LISS-3 4
pan2310
142-14670
24 7
WIFS 2 188 774 24 7MSU-SK 5 160 600 21SeaWiFS 8 1100 2800 1
10Ikonos 4 11 11
-
CHAPTER 3 A. DERMANIS: REMOTE SENSING
22
Table 3.12: Spectral characteristics of satellite multispectral
sensors.
MSS RBV TM ETM+ HRV HRVIR VGT ATSRBlue 0.475 − 0.575 0.45 − 0.52
0.45 − 0.515 0.43 − 0.47Green 0.5 − 0.6 0.52 − 0.60 0.525 − 0.605
0.50 − 0.59 0.50 − 0.59Red 0.6 − 0.7 0.580 − 0.680 0.63 − 0.69 0.63
− 0.690 0.61 − 0.68 0.61 − 0.68 0.61 − 0.68
Near IR 0.7 − 0.80.8 − 1.1
0.689 − 0.830 0.76 − 0.90 0.75 − 0.90 0.79 − 0.89 0.79 − 0.89
0.78 − 0.89
Mid IR 1.55 − 1.752.08 − 2.35
1.55 − 1.752.09 − 2.35
1.58 − 1.75 1.58 − 1.75 1.6
ThermalIR
10.4 − 12.6(Landsat
3)10.4 − 12.5 10.40 − 12.5
3.710.812
Pan 0.505 − 0.750(Landsat 3)
0.52 − 0.90 0.51 − 0.73 0.61 − 0.68
ATSR-2 AATSR AVHRR AVNIR OCTS CZCS VISSR GOES Im.Blue 0.42 −
0.50 0.490 ± 0.01 0.433 − 0.453
Green 0.555 0.55 0.58 −0.52 − 0.60 0.520 ± 0.01
0.565 ± 0.010.510 − 0.5300.540 − 0.560 0.55 − 0.55 −
Red 0.659 0.67 − 0.68 0.61 − 0.69 0.670 ± 0.01 0.660 − 0.680
Near IR 0.865 0.86 0.725 − 1.10 0.76 − 0.89 0.765 ± 0.020.865 ±
0.02
0.700 − 0.800 − 0.75 − 0.75
Mid IR 1.6 1.6
ThermalIR
3.710.812
3.71112
3.55 − 3.9310.3 − 11.311.4 − 12.4
3.55 − 3.888.25 − 8.8010.3 − 11.410.4 − 12.7
10.5 − 12.5 10.5 − 12.53.80 − 4.006.50 − 7.0010.2 − 11.210.5 −
12.5
Pan 0.52 − 0.72
MESSR VTIR LISS-1,-2 LISS-3 WIFS MSU-SK SeaWiFS MODIS
Blue0.45 −
0.62 − 0.680.402 − 0.4220.433 − 0.4530.480 − 0.500
36bands:
0.4 –
Green 0.51 − 0.59 0.5 − − 0.520.52 − 0.59 0.52 − 0.590.5 − 0.6
0.500 − 0.520
0.545 − 0.565Red 0.61 − 0.69 − 0.7 0.62 − 0.68 0.62 − 0.68 0.6 −
0.7 0.660 − 0.680
Near IR 0.73 − 0.800.80 − 1.10
0.77 − 0.86 0.77 − 0.86 0.77 − 0.86 0.7 − 0.80.8 − 1.1
0.745 − 0.7850.845 − 0.885
Mid IR
ThermalIR
6.0 − 7.010.5 − 11.511.5 − 12.5
1.55 − 1.70 10.4 −12.6 – 14.5
Pan 0.50 − 0.75
-
Remote sensing satellites, sensors and data
23
3.3. Airborne sensors
Although remote sensing from sensors aboard airplanes has been
usedlong before the era of satellite missions, today their use is
relatively lim-ited. Satellite sensors are typically flown on
airplanes for testing beforebeen placed on satellites. The cost of
airborne remote sensing is its maindisadvantage, because the data
produced are of use to a limited numberof users, unlike the
satellite data which are addressed to a worldwide anddiverse user
community. Another disadvantage is the large image distor-tions
resulting from the instability of the sensor platform during
flight, aswell as to the very large field of view, which produces
larger pixels awayfrom the nadir direction. The large FOV is
necessary to achieve a reason-able swath from the relatively small
flight height. The spatial resolutiondirectly depends on the fixed
IFOV and the variable flight height.Smaller heights give smaller
pixel values (higher spectral resolution) butat the same time
smaller swath and therefore longer flight missions tocover the same
land area, i.e. higher cost. Pixel size at nadir is approxi-mately
the product of IFOV and height. Expressing IFOV in mrad andheight
in kilometers, their product gives pixel size in meters. For
exam-ple IFOV = 2.5 mrad corresponds to a 2.5×2.5 m pixel with a
height of 1km. For the same IFOV a satellite-like resolution of
25×25 m corre-sponds to a height of 10 km. The advantage of
airborne missions is thepossibility to produce data, which fit very
well the particular needs of aspecific user, especially with
sensors where the available band can bepreset to cover particular
spectral regions. In addition they provide theonly means to record
with high temporal resolution sudden events, suchas volcanic
eruptions or spread of oil spills. Therefore they play an
im-portant role in hazard assessment through remote sensing.The
characteristics of three airborne sensors are presented in table
3.13:the Daedalus 1240/1260 Multispectral Line Scanner, the
Airborne The-matic Mapper (ATM) with bands similar to those of the
Landsat TM andthe Thermal Infrared Multispectral Scanner (TIMS).A
number of airborne sensors with a large number of bands are
designedto provide hyperspectral data; they will be examined in the
next para-graph.Airborne sensors are also the radar sensors of
active remote sensing,which are not discussed in this book.
-
CHAPTER 3 A. DERMANIS: REMOTE SENSING
24
Table 3.13: Characteristics of three airborne sensors.
Airborne Sensors
Daedalus ATMS TIMS
bands 0 0.32 – 0.381 0.38 – 0.422 0.42 – 0.453 0.45 – 0.504 0.50
– 0.555 0.55 – 0.606 0.60 – 0.657 0.65 – 0.698 0.70 – 0.799 0.80 –
0.89
10 0.92 – 1.1011 3.0 – 5.012 8.0 –14.0
1 0.42 – 0.452 0.45 – 0.523 0.52 – 0.604 0.605 – 0.6255 0.63 –
0.696 0.695 – 0.757 0.76 – 0.908 0.91 – 1.059 1.55 – 1.75
10 2.08 – 2.3511 8.5 – 13.0
1 8.2 – 8.62 8.6 – 9.03 9.0 – 9.44 9.5 – 10.25 10.2 – 11.26 11.2
– 12.2
FOV (degrees) 86 86 76IFOV (mrad) 2.5 2.5 2.5dyn. range (bits) 8
8 8
3.4. Hyperspectral sensors
Imaging spectrometers with many more bands than the
multispectral sen-sors are characterized as hyperspectral. The high
spectral resolution al-lows a detailed recording of the form of the
spectral signature of eachimaged pixel, though with some distortion
due to the atmospheric effectsof radiation absorption and
scattering. The main difference in this case isnot in the sensor
technology but rather in the data processing. The prob-lem of
classification of the pixels in different classes of land cover
needsa fundamentally different approach, where classification is
rather re-placed by identification of the pixel class by means of
comparison withan available library of spectral firms. This problem
is discussed more inchapter 12.Most hyperspectral sensors are
airborne. The characteristics of several ofthem are shortly
presented in table 3.14.. These are the Geophysical En-vironmental
Research Imaging Spectrometer (GERIS) with 63 bands de-veloped by
the Geophysical Environmental Research Corporation, theCompact
Airborne Spectrographic Imager (CASI) with 288 bands fromITRES
Research - Canada, the Airborne Visible and Infrared
ImagingSpectrometer (AVIRIS) with 224 bands from NASA’s Jet
PropulsionLaboratories, MIVIS with 102 bands from Daedalus
Enterprises, theChinese MAIS with 71 bands, the Hyperspectral
Digital Image Collec-tion Experiment (HYDICE) with 206 bands from
US Naval ResearchLaboratories and HYMAP with 128 bands from
Integrated Spectronics.A satellite hyperspectral sensor is the
Moderate Resolution ImagingSpectrometer (MODIS) with 36 bands,
aboard the EOS (Earth Observing
-
Remote sensing satellites, sensors and data
25
System) satellites. It has variable spatial resolution with
pixel size of 250m, 500 m and 1 km.
Table 3.14: Characteristics of hyperspectral sensors(imaging
spectrometers).
spectralregion
No. ofbands
spectralresolution
(nm)
IFOV(mrad)
pixelsperline
dynamicrange(bits)
GERIS 0.40 – 1.081.0 – 2.02.0 – 2.5
247
32-------63
25.412016.5
2.5, 3.3, 4.6 512 or1024
16
CASI 0.4 – 0.9 288 1.8 1.02 – 1.53 512 12AVIRIS 0.4 – 0.72
0.69 – 1.301.25 – 1.871.84 – 2.45
31636363-------
224
9.79.68.8
11.6
1 550 12
MIVIS 0.433 – 0.8331.15 – 1.552.00 – 2.50
8.20 – 12.70
208
6410-------
102
2050
≤500
2 765 12
MAIS 0.45 – 1.11.40 – 2.508.2 – 12.2
32327
71
2030
400-800
varies 8
HYDICE 0.4 – 2.5 206 7.6-14.9 0.5 320 12HYMAP 0.44 – 0.88
0.881 – 1.3351.4 – 1.81
1.95 – 2.94
total128
16131216
2.5×2.0 512 12
MODIS 0.4 – 14.5 36
3.1. Digital image storage and display3.2. Satellites
sensors3.3. Airborne sensors3.4. Hyperspectral sensors