Download Characteristics of remote sensing systems and images (PPT-III)

REMOTE SENSING
Characteristics of Sensor Systems
Professor Ke-Sheng Cheng
Department of Bioenvironmental Systems Engineering
National Taiwan University
Outline

The Electro-Optical Sensor Systems
Detector characteristics
 Types of image sensing systems


Resolutions of Remote Sensors
Spatial resolution
 Spectral resolution
 Radiometric resolution
 Temporal

The Electro-Optical Sensor
Systems

Major elements of an electro-optical
sensor system include
an optical sensing system which is
composed of lenses, mirrors, apertures,
modulators and dispersion devices,
 detectors which convert the radiances
reaching its surface to electrical signals,
 a signal processor which transforms the
electrical signal to desired output data.

Schematic diagram of a multispectral electro-optic sensor
system



The optical sensing system receives radiances
from the atmosphere and the Earth surface. The
combination of lenses, mirrors, apertures,
modulators and dispersion devices separates the
total radiance into desired multispectral radiances.
These multispectral radiances reach the detectors
and their EM radiation energy is converted to
electrical signals (current or voltage) of the
corresponding channels. The advent of high
resolution Charged-Coupled-Device (CCD) has
made CCD detectors most widely used for remote
sensing.
Finally, the signal processor transforms the
analogue electrical signals to discrete outputs
based on predetermined transformation functions.
Detector characteristics


Modern remote sensing systems use
detectors which yield output signals in the
form of a measurable electrical signal
(current or voltage).
There are two broad types of electrooptical detectors
thermal detectors
 photon detectors

Thermal detectors


The thermal detectors absorb optical or
infrared energy of incident photons,
convert it to heat, and thus undergo a
temperature rise.
The thermal detector can be used over the
wavelength range of visible to thermal
infrared since the temperature rise is
dependent on the energy absorbed and not
the photon wavelength directly.


Any energy which is absorbed causes a response
in the detector. Therefore, it is possible to use a
thermal infrared detector at room temperature to
detect radiation from room temperature
blackbodies.
Thermal detectors need time to reach thermal
equilibrium before they can provide an accurate
measurement, their response to incident
radiation tend to be slow, and thus, are not
practical for most remote sensing applications.
Photon detectors


Photon detectors measure the rate at which
individual photons interact with the detector.
Photons incident on a photon detector interact
directly with the electronic energy levels
within the detector to produce free charge
carriers.
Individual photons produce significant change
in the detector material which in turn produces
a change in the current or voltage in an
external circuit.

Since photon detectors do not rely on the
conversion of incoming radiation to heat,
but convert incoming photons directly
into an electrical signal, photon detectors
are sensitive to the amount of energy
associated with each incident photon and
have a high speed of response and are
therefore are commonly used in remote
sensing applications.

Performance of detectors is determined
by characteristics of detectors including
spectral responsivity, speed of response,
signal-to-noise ratio (SNR), spectral
detectivity, dynamic range of response,
etc. Some of these parameters are defined
and discussed as follows.
Responsivity

The responsivity is defined as the detector
output per unit of input power. The units
of responsivity are either volts/watt or
amperes/watt, depending on whether the
detector output is a voltage or electric
current.

Let  be the radiant flux incident on the
detector and V be the output of the
detector. The responsivity can be
expressed as:
dV
R ( ) 
d 

For photon detectors, the responsivity
depends on wavelength, and the term
spectral responsivity is used.
Schematic illustration of the spectral responsivity curves
of thermal and photon detectors

Since photon detectors rely on the action of
photon to interact with electrons in the detector
material and to generate free electrons, the
incident photon must have sufficient energy to
free an electron. Thus, a cutoff wavelength (c)
appears in the spectral responsivity curves of
photon detectors. Photons with wavelength
longer than c do not have enough energy to
free the electron and the detectors yield no
response.

In remote sensing applications, incident
radiant fluxes are often restricted to be
within specific spectral bands (1, 2) and
thus the detector output (V) and the
bandpass responsivity (R) can be
calculated by
2
V   R( )  d
1
R


2 
V
 
 2
   R ( )  2
d

1
  d 

1  d

1



The spectral responsivity is a fixed
characteristic of the detector. However,
for a given detector, the bandpass
responsivity may vary with the spectral
distribution of incident flux.
dV
R ( ) 
d 
R


2 
V
 
 2
   R ( )  2
d

1
  d 

1  d

1


Signal-to-noise ratio (SNR)


There is uncertainty in the signal (voltage
or current) level within a fixed period of
time even when the detector is exposed to
a constant level of photon flux.
The standard deviation of the signal level
is defined as the photon noise (Nphoton) or
shot noise.




Noise can also exist even when there is no
incident flux at the detector. This may be
due to fluctuations in detector dark current
and readout-related noise.
The dark current is caused by thermally
generated electrons in the detector system.
The read noise is mainly caused by
thermally induced motions of electrons in
the output amplifier.
To achieve a good performance of the
detector, we would want the signal level to
be sufficiently higher than the noise floor.
Fluctuation of the signal and noise levels



Variation of the signal level arises from the
random arrival of photons on the detector
surface.
The number of photons incident on the detector
for a given period of time can be characterized
as a Poisson distribution. The standard deviation
of a Poisson distribution is equivalent to the
square root of its mean value. Thus, the photon
noise is numerically equivalent to the square
root of the average signal level.
Similarly, the dark noise equals the square root
of the average dark current.

The Signal-to-noise ratio (SNR) is defined as
the mean signal level divided by the total noise,
i.e.
SNR 

2
QE  N photon
2
2
2
QE  N photon
 N dark
 N read
Higher SNR values are desired to ensure good
performance of detectors, and there are several
ways to improve the SNR including



broadening the sampled wavelength bands
increasing the spatial resolution
increasing the dwell time for each pixel.
Noise equivalent power (NEP)

The noise equivalent power (NEP) is
defined as the required radiant power
incident on the detector that can produce
an output voltage (or current) equivalent
to the dark noise of the detector. By
definition, the NEP (in watts) is expressed
as
N total
NEP( ) 
[watts]
R ( )


NEP is another way of quantifying the noise in
terms of the minimum level or incident power
that can be detected by the detector.
For a detector with measurement band (1, 2),
the amount of noise is dependent on the
bandwidth of the measurement and that
bandwidth must be specified in expression of
NEP, i.e.
IA  Ntotal
NEP( )  2
12
N photon(f )
I is the irradiance incident on the detector of area A, f is the
frequency interval corresponding to measurement bandwidth
=21.


From the definition, it is apparent that the
lower the value of the NEP, the better are
the characteristics of the detector for
detecting a small signal in the presence of
noise.
The NEP is dependent on the area of the
detector. Since many detectors have NEP
proportional to the square root of their
area, another term called detectivity (D*)
is defined by
A
D ( ) 
NEP( )
*


D* is independent of the area of the
detector and provides a measure of the
intrinsic quality of the detector material
itself, independent of the area with which
the detector happens to be made.
A high value of D* means that the
detector is suitable for detecting weak
signals in the presence of noise.
Types of image sensing systems

Line scanner




A rotating scanner sweeps the ground along the cross-track
direction.
The cross-track angular extent of the area swept by the
system is called the field of view (FOV). Accordingly, the
angular extent of the detector is referred to as the
instantaneous field of view (IFOV).
The area projected on the ground by the scanner at any time
instant is called the ground sample or ground instantaneous
field of view (GIFOV).
The ground width of each scanning cycle is called the
ground swath.
A line scanning system


If the scanning system is flying at an
altitude of H, then GIFOV at nadir (the
point perpendicularly under the sensor)
can be calculated as GIFOV  H .
During each scan, the radiant energy from
ground surface is sampled at a constant
rate. The cross-track sampling rate and
the scanning speed of the sensor together
determine the ground sample interval
(GSI) in the cross-track direction.


Similarly, the along-track GSI is
determined by the combined effect of the
along-track sampling rate and flying
speed of the sensor.
Although a common practice is to design
the image sensing system such that both
GSIs (along-track and cross-track) equal
the GIFOV, there are systems (e.g.
Landsat MSS) that yield oversampling
effect along the cross-track direction.
Effect of oversampling

Oversampling reduces the cross-track
GSI, and thus yields an image of finer
details. However, it should be noted that
oversampling does not affect the GIFOV
which is a characteristic of the detector by
design.



Many line scanners also use a dispersing
element to spectrally disperse incident light. A
few discrete detectors or a linear array of
detectors placed at the focal plane are then
used to sample signals of individual spectral
channels.
Such systems are known as the multispectral
line scanners. Line scanners are mostly
mounted on aircrafts.
Aircraft platforms and flying speed are usually
not stabilized such that geometric distortions
may occur from one line to the other, or even
from pixel to pixel within a line.
Whisk-broom scanner


The line scanner design uses a single
detector to sweep the ground line by line.
As a result, the dwell time, the time span
for the detector to scan over a ground
sample, is usually very short.
In contrast, whisk-broom scanners
employ an array of several detectors
aligned in the along-track direction,
resulting in improvement in dwell time.


Similar to the multispectral line scanners,
multispectral whisk-broom scanners also
use dispersing element and discrete or a
linear array of detectors to sample signals
of individual spectral channels.
The Landsat multispectral scanner (MSS)
and Landsat Thematic Mapper (TM) are
two examples of multispectral whiskbroom scanners.
A whisk-broom scanning sensor
Push-broom scanner


Unlike whisk-broom scanners which have
discrete detectors aligned in the alongtrack direction, push-broom scanners
have a linear array of detectors in the
cross-track direction.
Instead of swing-scanning in the crosstrack direction, the detectors move
forward during flight without any swing
movement.
A push-broom scanning sensor


The flying speed of the scanner system
and the radiance sampling rate usually are
designed in a way such that both the
cross-track and along-track GSIs are
numerically the same as the GIFOV.
The swath width (in number of pixels) of
an image equals the number of detectors.
Area-array scanner

Framing-array scanners utilize a twodimensional array of detectors to acquire
the image of a frame-projected area on
the ground (see Figure 2.5(c)),
eliminating the distortions due to withinframe sensor motion.
An area-array scanning sensor
Outline

The Electro-Optical Sensor Systems
Detector characteristics
 Types of image sensing systems


Resolutions of Remote Sensors
Spatial resolution
 Spectral resolution
 Radiometric resolution
 Temporal

Resolutions of Remote Sensors




In order for a remote sensor to collect and record
energy reflected or emitted from a target area, it must
reside on a stable platform.
Platforms for remote sensors may be situated on the
ground, on an aircraft, or on a spacecraft or satellite.
The quality of remote sensing images acquired by
theses sensors is affected by many factors including
the orbit and speed of platforms and sensor
characteristics.
Four types of resolutions which are pertinent to
application of remote sensing images can be defined
as follows.
Spatial resolution


Spatial resolution is a measure of the spatial
detail in an image, which is a function of the
design of the sensor and its operating altitude
above the surface.
Each of the detectors in a remote sensor
measures radiant energy received from an area
within its ground sample. Thus, spatial
resolution is most commonly expressed as the
size of the ground sample.



The smaller the ground samples are, the
more detailed will be the spatial
information that we can interpret from the
image.
Shape is one visual factor that we can use
to recognize and identify objects in an
image. Shape is usually discernible only
if the objects are several times larger than
the spatial resolution.
On the other hand, objects smaller than
the spatial resolution of the image may be
detectable in an image (Smith, 2006).


If such an object is sufficiently brighter or
darker than its surroundings, it will
dominate the averaged brightness of the
image cell it falls within, and that cell will
contrast in brightness with the adjacent
cells.
We may not be able to identify what the
object is, but we can see that something is
present that is different from its
surroundings, especially if the
“background” area is relatively uniform.
Spectral resolution


Spectral resolution is referred to as the
wavelength range within which a sensor
is able to detect and measure radiant
energy.
It describes the ability of a sensor to
distinguish between wavelength intervals
(or bands) in the electromagnetic
spectrum.
Spectral resolutions of SPOT-5 and Formosat-2 satellites.
(F: Formoat-2, S: SPOT-5)


Sensors of higher spectral resolution result
in less radiant energy received by detectors,
which in turn yield higher signal-to-noise
ratios since the noise remains constant.
One way to overcome such disadvantage
without reducing the spatial resolution is to
choose proper scanning system. For
example, a push-broom sensor offers longer
dwell time than a whisk-broom sensor, and
thus more radiant energy from the target
area can be collected and the SNR is lower.
Radiometric resolution


Radiometric resolution represents the sensor’s
ability to discriminate very slight differences in
received radiance. It is defined as the minimum
amount of changes in received radiance that
can result in a change in detector’s output.
To digitally record the energy received by an
individual detector, the continuous range of
incoming energy must be quantized, or
subdivided into a number of discrete levels that
are recorded as integer values.


The more levels that can be recorded, the
finer is the radiometric resolution of the
sensor system (i.e. the more sensitive it is
to detecting small differences in energy).
The total number of output levels (grey
levels) that can be used by a sensor to
record radiant energy is referred to as the
dynamic range.

Let Lmax and Lmin be respectively the
maximum and minimum radiances that
can be detected by a sensor. Also, let  be
the radiometric resolution of the sensor
and NG the dynamic range. The dynamic
range and the radiometric resolution of
the sensor are related by
Lmax  Lmin

NG  1


The dynamic range is governed by the
number of bits used by the sensing system
to represent the output signal generated
by a single detector for a pixel.
For example, an 8-bit image can display a
total of 256 (=28) grey levels or digital
numbers (DN), i.e. the dynamic range NG
= 256.

The signal processor of the sensor system
converts the radiance received at sensor
(Lsensor,) to a corresponding digital
number by
Lsensor,   Lmin
DN 

Temporal resolution


Temporal resolution is the period of elapsed
time between images taken of the same object
at the same location. For satellite systems
temporal resolution is described as the revisit
time, which refers to the time it takes for a
satellite to return to the same area on
subsequent orbits.
The more frequent a sensor is able to return to
an exact specific location the greater the
temporal resolution. Several observations
over time can reveal changes and variations in
the object being observed.


Generally speaking, it is impossible to
simultaneously improve all resolution
properties. There are trade-offs between the
spatial resolution, spectral resolution, and
dynamic range.
For example, a sensor of high spatial
resolution has a small IFOV which in turn
reduces the amount of radiant energy
reaching the sensor. Thus, the dynamic
range (or the ability of the sensor to
distinguish small difference in radiant
energy from ground samples) is also
reduced.

On the contrary, to increase the amount of
energy detected (and thus, the radiometric
resolution) without reducing spatial
resolution, we would have to broaden the
wavelength range of a particular spectral
band which would reduce the spectral
resolution of the sensor.
Illustration of trade-off between
spatial and radiometric resolutions
High radiometric resolution, coarse spatial resolution
Fine spatial resolution, low radiometric resolution
Linear DN~Radiance Conversion

There exists a linear relationship between
the effective radiance received at sensor
from a ground sample and the digital
number of its corresponding image pixel.
K b  c1kRmax

In the above equation Rmax is the
maximum value of the R() function.
Thus, notwithstanding quantization, there
exists a linear relationship between the
sensor’s output signal, i.e. digital number,
and the band effective radiance Leff ,  .
b
An example of detailed correction
of atmospheric effects


(Forster, 1984. Derivation of
atmospheric correction procedures for
Landsat MSS with particular reference
to urban data. International Journal of
Remote Sensing, Vol. 5, 799-817.)
Landsat 2 MSS image,   0.8 – 1.1
m [Band 7].

Atmospheric condition at the time of
overpass
Temperature
29C
 Relative humidity
24%
 Atmospheric pressure1004mb
 Visibility
65km


The total normal optical thickness of
the atmosphere a = 0.15

The transmittances of the atmosphere
for a solar zenith angle of 38 and a
nadir viewing satellite
are

1  e
 0.827
 ( 0.15sec 0 )
2  e
 0.861
 ( 0.15 sec 38 )




The solar irradiance at the top of the
earth atmosphere is E = 256 w/m2.
The total diffusive sky irradiance ED =
19.77 w/m2.
The path radiance Lu = 0.62 w/(m2-sr).

Radiance reaching the sensor
Ls 
r

 2 E  1 cos    Ed   Lu


 256  0.827  cos 38  19.77 

r

 0.861  0.62
 51.13  r  0.62

For Landsat 2 MSS data,
Lmax  Lmin 39.1  1.1
2


 0.603 [W/(m  sr )]
NG  1
64  1
Ls  0.603DN  1.1  51.13  r  0.62
r  0.0118DN  0.0094  1.18DN  0.94%
 The reflectance-DN relationship varies
with band wavelength interval. For
example, for band 5 of Landsat 2 MSS
(  0.6 – 0.7 m [Band 5]) it yields
r  0.44DN  0.5%