Herbert M. Runciman. "Thermal Imaging."
Copyright 2000 CRC Press LLC. .
© 1999 by CRC Pre
All objects a
makes use o
selected regi
generate a p
such a way t
Thermal
together wit
thermal ima
can be used
detect, recog
favored over
35.1 E
All thermal
waveband, a
detector arr
(a “staring a
can give goo
single row o
across the d
Herbert M
Pilkington Op
ss LLC
Thermal Imaging
35.1 Essential Components
35.2 Thermal Imaging Wavebands
35.3 Emission from Source
35.4 Atmospheric Transmission
35.5 Detectors
Photon Detectors • Thermal Detectors • Detector
Performance Measures • Detector Cooling
35.6 Electronics
35.7 Optics and Scanning
35.8 Temperature References
35.9 Imager Performance
SNR and NETD • Minimum Resolvable Temperature
Difference
35.10 Available Imagers
35.11 Performance Trade-offs
35.12 Future Trends in Thermal Imaging
t temperatures above absolute zero emit electromagnetic radiation. Radiation thermometry
f this fact to estimate the temperatures of objects by measuring the radiated energy from
ons. Thermal imaging takes the process one stage further and uses the emitted radiation to
icture of the object and its surroundings, usually on a TV display or computer monitor, in
hat the desired temperature information is easily interpreted by the user.
imagers require no form of illumination to operate, and the military significance of this,
h their ability to penetrate most forms of smoke, has been largely responsible for driving
ger development. Although thermal imagers intended for military or security applications
for temperature measurement, they are not optimized for this purpose since the aim is to
nize, or identify targets at long ranges by their shape; thus, resolution and sensitivity are
radiometric accuracy.
ssential Components
imagers must have a detector or array of detectors sensitive to radiation in the required
nd optics to form an image of the object on the detector. In modern thermal imagers, the
ay might have a sufficient number of sensitive elements to cover the focal plane completely
rray”), in the same way as a CCD television camera. Some of the most recent staring arrays
d performance without cooling. In other imagers, the detector might take the form of a
r column of elements, in which case a scanning mechanism is required to sweep the image
etector array. If a single-element detector, or a very small detector array, is used, a means of
. Runciman
tronics
© 1999 by CRC Pre
providing a
scanning im
performance
Although
the detector
reference bo
and accessed
or by deflec
35.2 T
The optimu
emitted rad
technology.
The powe
its surface. If
It then also
(given later)
temperature
at a wavelen
about 3
m
m,
objects. The
to the energ
It is impo
observed. Th
at 4.2
m
m du
FIGURE 35.1
scanning with
two-dimensional scan is required. (Figure 35.1 shows these options schematically.) For
agers, it is necessary to cool the detectors (usually to about 80 K to 120 K) to achieve adequate
.
in principle it would be possible to deduce target temperature from the absolute value of
signal, it is necessary in practice to estimate temperature by comparison with one or more
dies of known temperature. The temperature references are usually internal to the equipment,
by mechanical movement of the reference (which may take the form of a rotating chopper)
ting the optical path using a mirror.
hermal Imaging Wavebands
m waveband for thermal imaging is determined partly by the wavelength distribution of the
iation, partly by the transmission of the atmosphere, and partly by the chosen detector
r radiated from a given area of an object depends only on its temperature and the nature of
the surface absorbs radiation of all wavelengths completely, it is referred to as a “blackbody.”
emits the maximum amount possible, which can be calculated using the Planck equation
. Figure 35.2(a) shows the way in which blackbody emission varies with wavelength for several
s. It will be seen that for objects near normal ambient temperature, maximum output occurs
gth of about 10 m m, or about 20 times the wavelength of visible light. At wavelengths below
there is generally insufficient energy emitted to allow thermal imaging of room-temperature
emissivity at any wavelength is defined as the ratio of the energy emitted at that wavelength
y that would be emitted by a blackbody at the same wavelength.
rtant that the atmosphere should have sufficient transparency to permit the target to be
ere are two important “atmospheric windows” — one between 3 m m and 5 m m (with a notch
e to carbon dioxide absorption) and one between 7.5 m m and 14 m m. These are commonly
Thermal imaging options: (a) 2-D scanning for small detector array or single element; (b) 1-D
linear detector array; (c) staring array without scanning.
ss LLC
© 1999 by CRC Pre
referred to a
For thermal
but most ins
and 10 m of
Emissivity
lower and m
in the MWI
favor the LW
detector tec
either band,
give similar
staring array
For temp
scale derived
temperature
FIGURE 35.2
sufficiently hi
s the medium-wave infrared (MWIR) and long-wave infrared (LWIR) windows, respectively.
measurement over short ranges in the laboratory, it is possible to work outside these bands,
truments are optimized for either the MWIR or LWIR. Typical transmissions through 1 km
a clear U.S. Standard Atmosphere are shown in Figures 35.2(b) and 35.2(c).
for most naturally occurring objects and organic paints is high (>0.8) in the LWIR, but is
ore variable in the MWIR. Metallic surfaces have low emissivity in both bands. Solar radiation
R is significant, and can cause errors in measurements made outdoors. These considerations
IR for quantitative imaging, but the band chosen can also be influenced by the chosen
hnology, the latter frequently being determined by cost. Scanning imagers can be used in
but are more sensitive for a given detector architecture in the LWIR. Cooled staring arrays
sensitivity in either band, but are currently more readily available in the MWIR. Uncooled
s work well only in the LWIR band.
erature measurement, the electronics can be used to encode signal level as false color, a color
from the thermal references being injected into the display to allow the user to identify the
of the object under examination. For general surveillance, a conventional gray-scale image
Factors determining thermal imaging wavebands. Imager must operate in regions where radiance is
gh (a) and atmospheric transmission is good (b) and (c).
ss LLC
© 1999 by CRC Pre
is usually preferred. Imagers for thermography can also include emissivity compensation. If accurate
results are r
anything tha
the object is
temperature
35.3 E
The spectral
equation [1
frequently s
making the
of photon d
the radiance
photon flux
equations ar
where: Num
c
1
=
c
2
=
c
3
=
The unit of
The abov
per steradian
are obtained
some cautio
temperature
equired for an object of low emissivity, it is important to ensure that the temperature of
t might be reflected by the object is known and that the emissivity is accurately known. If
accessible, another object placed beside it with the same surface characteristics but known
can be used for calibration.
mission from Source
radiance W(l ,T) of a blackbody at temperature T and wavelength l is given by the Planck
]. For temperature differences between the target and the reference of a few degrees, it is
ufficiently accurate to assume a linear dependence of radiance on temperature difference,
temperature derivative of the blackbody equation, dW(l ,T)/dT, more relevant. In the case
etectors, the detector output is proportional to the photon flux, which can be derived from
using the fact that photon energy E(l ) = hc/l , where h is the Planck constant. The total
N(l ) and its derivative with respect to temperature are thus relevant in this case. The
e as follows:
(35.1)
(35.2)
(35.3)
(35.4)
erical values of the constants are:
3.742 · 108
1.439 · 104
1.884 · 1027
wavelength is chosen for convenience to be the micrometer (m m).
e values are for radiation into a hemisphere. The intensities (watts per steradian, photons
, etc.) are obtained by dividing the above values by p . The actual radiances for real targets
by multiplying by the spectral emissivity e (l ); but since target reflectivity r (l ) = 1 – e (l ),
n is required. For example, a target at temperature T surrounded by a background of
Tb will appear to emit W(l ,T)e (l ) + W(l ,Tb)r (l ) = [W(l ,T) – W(l ,Tb)]e (l ) + W(l ,Tb).
W T
c
e
c
T
λ
λ λ
, , ( ) =
−
µ− −1
5
2 1
2
1
W m
N T
c
e
c
T
λ
λ λ
, , ( ) =
−
µ− − −3
4
1 2 1
2
1
photons s m m
d
d
W m m K
W T
T
c c e
T e
c
T
c
T
λ
λ
λ
λ
,
,
( )
=
−
µ− − −1 2
6 2
2 1 1
2
2
1
d
d
photons s m m K
N T
T
c c e
T e
c
T
c
T
λ
λ
λ
λ
,
,
( )
=
−
µ− − − −3 2
5 2
1 2 1 1
2
2
1
ss LLC
© 1999 by CRC Pre
Provided tha
acts as an iso
The differen
a small temp
The spectral
A major d
the contrast
the LWIR.
35.4 A
Provided th
transmission
atmospheric
The standard
transmission
the backgro
LWIR is sev
operation in
absorptions
35.5 D
There are tw
of photon d
Chapter 6.1.
Photon D
In photon d
semiconduc
potential dif
of not requi
40% lower
recombinati
electronics. T
of the spectr
widely used
is the quantu
(in either wa
the LWIR, b
also an exce
uniformity o
Detectors
over the sam
t the background surrounds the target and that the target is reasonably small, the background
thermal enclosure, which can be shown [2] to behave as an ideal blackbody (i.e., e (l ) = 1).
tial spectral radiance against background D W(l ) is thus [W(l ,T) – W(l ,Tb)]e (l ), which for
erature difference D T is simply:
(35.5)
emissivity of a wide variety of natural and man-made objects is also given in [2].
ifference between thermal imaging and visual imaging is the very low contrast. In the MWIR,
calculated from Equation 35.1 due to 1 K at the target is about 4%, falling to about 2% in
tmospheric Transmission
at the absorption bands shown in Figures 35.2(b) and 35.2(c) are avoided, atmospheric
can frequently be ignored in the laboratory or industrial context. For longer ranges, an
transmission model must be used or calibrating sources must be placed at the target range.
atmospheric transmission model is LOWTRAN [3], currently at version 7. The atmospheric
Ta(l ) reduces the differential signal from the target proportionately, but has no effect on
und flux if the atmosphere is at background temperature. Atmospheric transmission in the
erely affected by high humidity, making the MWIR the band of choice for long-range
the Tropics. (Many gases and vapors such as methane or ammonia have very strong
in the infrared, making thermal imaging a possible means of leak detection and location.)
etectors
o main types of detector — photon (or quantum) and thermal. A more detailed discussion
etectors is given in this handbook in Chapter 8.1.1 and 8.1.2, and of thermal detectors in
8, so only aspects unique to thermal imaging are discussed here.
etectors
etectors, the response is caused by photons of radiation that generate free carriers in a
tor, which in turn increase the conductivity (for photoconductive detectors) or generate a
ference across a junction (for photovoltaic detectors). Photovoltaic devices have the advantage
ring a bias current (important to reduce the heat load on the cooling system), and they have
noise because the electric field at the junction separates the carriers, thereby eliminating
on noise. Whether or not the lower noise is achieved in practice depends on the read-out
he photon energy in the LWIR is only about 1/20th of that of a photon in the visible region
um, so a semiconductor with a much smaller bandgap than silicon must be used. The most
material is a compound of mercury, cadmium, and tellurium (MCT or CMT) since not only
m efficiency excellent (70% or more), but the bandgap can be tuned to the desired wavelength
veband) by altering the composition. Cooling of the detector to about 80 K is desirable for
ut about 120 K is acceptable for the MWIR. For the MWIR, indium antimonide (InSb) is
llent material; and since it is a true stoichiometric compound, it is easier to achieve good
f response, but cooling to 80 K is required.
for use in scanning systems are frequently arranged so that several elements are scanned
e part of the scene in rapid succession, the output of each element being delayed and added
∆ ∆W
W T
T
Tλ ε λ
λ( ) = ( ) ( )d
d
,
ss LLC
© 1999 by CRC Pre
to the previ
serial scanni
to the squar
material itse
elongated st
electrode ne
the signal bu
while diffusi
Large arr
bonded to a
is the Schot
process, and
limited to th
Vol. 3, p. 246
The detec
a “cold shiel
FIGURE 35.3
at which the i
a detector ma
FIGURE 35.4
match U.S. TV
ous one to enhance the signal-to-noise ratio (SNR). This approach (Figure 35.3) is termed
ng or “time-delay and integrate” (TDI) mode, and gives a theoretical gain in the SNR equal
e root of the number of elements in TDI. It is also possible to perform TDI in the detector
lf. In the SPRITE detector (Signal Processing In The Element), the sensitive element is an
rip of CMT. Photons incident on the device generate carriers that drift toward a read-out
ar one end. If the image is scanned along the detector at the same speed as the carrier drift,
ilds up along the length of the device. The useful length is limited by carrier recombination,
on of the carriers limits spatial resolution.
ays of photon detectors are generally of hybrid construction, the sensitive elements being
silicon CCD or CMOS addressing circuit using indium “bumps” (Figure 35.4). An exception
tky barrier detector (e.g., platinum silicide), which can be manufactured by a monolithic
thus tends to be lower in cost, but quantum efficiency is much lower and operation is usually
e MWIR band. Detector arrays and read-out architectures are discussed in depth in [2]
-341 and [4].
tor assembly is encapsulated in a Dewar as shown in Figure 35.5. In front of the detector is
d” that limits the acceptance angle of the radiation to match that of the optics.
Use of serial scanning to enhance signal-to-noise ratio. The delay times are chosen to match the speed
mage is swept along the detector array. Serial scanning is usually combined with parallel scanning using
trix.
Typical hybrid detector construction. A typical element size is 30 m m. A large array of this type to
standard would have 640 · 480 elements.
ss LLC
© 1999 by CRC Pre
Thermal
Thermal det
a change in
generally fai
(although it
is typically 1
The essentia
blackened el
dielectric ma
in electrical
is the therma
and for good
the dielectric
causes the ca
detailed desc
change in te
technique us
vanadium d
elements are
thermal insu
chopping is
Detector
The wavelen
current that
at such a h
A photon–1 s
FIGURE 35.5
of compresse
Detectors
ectors rely on the heating effect of the incoming radiation, the change in temperature causing
resistance, capacitance, or electrical polarization that might be detected electrically. They are
rly slow in response (several milliseconds) but have the advantage that cooling is not essential
can be of considerable benefit with some types). The detectivity of uncooled thermal detectors
/100 that of cooled photon detectors, so real-time imaging requires the use of staring arrays.
ls of a pyroelectric array are shown in Figure 35.6. Incoming radiation is absorbed by the
ectrode, and the heat generated is transferred to the pyroelectric layer, which comprises a
terial that has been polarized by means of a high electric field during manufacture. The change
polarization with temperature gives the electric signal. One of the most important parameters
l isolation of the sensitive elements, so some kind of insulating support structure is necessary;
performance, the device must be evacuated to prevent convection. In a variant of this approach,
bolometer, the rapid variation of the dielectric constant at temperatures near the Curie point,
pacitance of the sensitive element to change, and hence the voltage for a constant charge. A
ription of this approach is given in [5]. In both techniques, the detector responds only to
mperature, so it is necessary to modulate the incoming radiation using a chopper. In the
ed initially by Wood [6] (now licensed to several manufacturers), the sensitive elements are
ioxide coatings that undergo a large change in resistivity for a small temperature change. The
supported by silicon strips that are micromachined from the substrate and give excellent
lation of the element. Changes in resistivity are read out by circuitry on the substrate, and no
required, but the array must be maintained at a precise and uniform temperature.
Performance Measures
gth-dependent power responsivity of a detector R(l ) is defined as the output potential or
would result from 1 W of radiation at wavelength l , assuming that linearity was maintained
igh flux level. The units are V W–1 or A W–1. Photon responsivities in V photon–1 s–1 and
–1 are similarly defined.
Construction of typical cooled detector. Cooling can be by liquid nitrogen, Joule-Thomson expansion
d gas, or a cooling engine.
ss LLC
© 1999 by CRC Pre
A therma
of response
element or t
In an ide
per photon)
i.e., the pho
given numb
responsivity
cut-off is spr
antireflectio
The sensi
background
incident on
ment is mad
effect being
of detector, t
B, so that
defined as s
usually given
measure det
under the co
to 1/R(l ); so
wavelengths
the blackbod
FIGURE 35.6
bolometer, bu
l detector has a power responsivity that is essentially independent of wavelength, the limits
being determined by the transparency of the window and the absorption spectrum of the
he material used to blacken it.
al photon detector, the quantum efficiency h (defined as the number of carriers generated
would be constant at all wavelengths for which the photon energy is greater than the bandgap,
ton responsivity is independent of wavelength up to the chosen cut-off wavelength. Since a
er of Watts corresponds to a number of photons proportional to the wavelength, the power
(V W–1 or A W–1) would increase linearly with wavelength until the cut-off. In practice, the
ead over about 0.5 m m and shortwave performance is modified by window transmission and
n coatings.
tivity of a detector is limited by noise that may be due to the detector itself or due to the
radiation (as is discussed later). Noise-equivalent power NEP(l ) is defined as the power
the detector at wavelength l , which gives a signal equal to the rms noise when the measure-
e with a 1-Hz bandwidth. The NEP depends also on the modulation frequency, the latter
large for thermal detectors, but frequently negligible for quantum detectors. For many types
he noise is proportional to the square root of the sensitive area Ad and the electrical bandwidth
/NEP is constant. A performance figure that is proportional to sensitivity can then be
pecific detectivity D*(l ) = /NEP(l ). For historical reasons, specific detectivity is
in units of cm W–1, so it is important to remember to convert this to SI units or to
ector area in square centimeters. Since noise is an electrical quantity particular to the detector
nditions for which D* is defined and is independent of wavelength, NEP( l ) is proportional
if the value of detectivity D*p at the wavelength of peak responsivity Rp is known, for other
D*(l ) = D*
本文档为【CH35热成像测量】，请使用软件OFFICE或WPS软件打开。作品中的文字与图均可以修改和编辑，
图片更改请在作品中右键图片并更换，文字修改请直接点击文字进行修改，也可以新增和删除文档中的内容。
该文档来自用户分享，如有侵权行为请发邮件ishare@vip.sina.com联系网站客服，我们会及时删除。
[版权声明] 本站所有资料为用户分享产生，若发现您的权利被侵害，请联系客服邮件isharekefu@iask.cn，我们尽快处理。
本作品所展示的图片、画像、字体、音乐的版权可能需版权方额外授权，请谨慎使用。
网站提供的党政主题相关内容(国旗、国徽、党徽..)目的在于配合国家政策宣传，仅限个人学习分享使用，禁止用于任何广告和商用目的。
浙公网安备 33021202002483