Calibration of Uncooled Thermal Infrared Cameras

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

J. Sens. Sens. Syst.

, 4, 187–197, 2015
www.j-sens-sens-syst.net/4/187/2015/
doi:10.5194/jsss-4-187-2015
© Author(s) 2015. CC Attribution 3.0 License.

Calibration of uncooled thermal infrared cameras


H. Budzier and G. Gerlach
Technische Universität Dresden, Electrical and Computer Engineering Department, Solid-State Electronics
Laboratory, Dresden, Germany
Correspondence to: H. Budzier ([email protected])
Received: 26 March 2015 – Revised: 30 April 2015 – Accepted: 6 May 2015 – Published: 2 June 2015

Abstract. The calibration of uncooled thermal infrared (IR) cameras to absolute temperature measurement is
a time-consuming, complicated process that significantly influences the cost of an IR camera. Temperature-
measuring IR cameras display a temperature value for each pixel in the thermal image. Calibration is used to cal-
culate a temperature-proportional output signal (IR or thermal image) from the measurement signal (raw image)
taking into account all technical and physical properties of the IR camera. The paper will discuss the mathemat-
ical and physical principles of calibration, which are based on radiometric camera models. The individual stages
of calibration will be presented. After start-up of the IR camera, the non-uniformity of the pixels is first corrected.
This is done with a simple two-point correction. If the microbolometer array is not temperature-stabilized, then,
in the next step the temperature dependence of the sensor parameters must be corrected. Ambient temperature
changes are compensated for by the shutter correction. The final stage involves radiometric calibration, which
establishes the relationship between pixel signal and target object temperature. Not all pixels of a microbolome-
ter array are functional. There are also a number of defective, so-called “dead” pixels. The discovery of defective
pixels is a multistep process that is carried out after each stage of the calibration process.

1 Introduction the temperature sensor was a requirement, nowadays cur-


rent arrays manage without this. The use of microbolome-
ters without temperature stabilization allows for compact,
In recent years, thermography has had a dramatic develop- energy-efficient camera designs.
ment with annual growth rates of over 20 % (Mounier, 2011). An uncooled IR camera consists of the following main
This development will have an even more dynamic impact in components (Fig. 1) (Budzier and Gerlach, 2011):
the future. Enabling this huge market success are infrared
(IR) image sensors based on microbolometer arrays, which – IR optics
have excellent thermal and spatial resolution (Kruse, 1997; – sensor arrays
Vollmer and Möllmann, 2010). Also, because no cooling is
required, they have low power consumption and have a low – processor-based camera electronics.
entry-level price. Both these components and the calibration process play an
Mainly due to the rapid development of micro- and especially crucial role in ensuring the quality of an IR cam-
nanotechnology, microbolometers have become significantly era. The calibration is implemented both in the hardware and
cheaper and more efficient. While early in 2000 the maxi- the software (firmware). For the two device concepts
mum image size was 320 × 240 pixels, nowadays commer-
cial sensor arrays are available with a maximum size of – vision display device and
1024 × 768 pixels. IR image sensors with full HDTV res-
– temperature-measuring image device
olution (Black et al., 2011) are already being advertised.
The development trend is towards even smaller pixel grids there are different calibration concepts (Budzier, 2014).
(< 17 µm), to lower power loss and compact ceramic pack- Vision devices display the measured radiation distribution
ages. While up until a few years ago, the stabilization of of the scene qualitatively. They serve primarily as a night

Published by Copernicus Publications on behalf of the AMA Association for Sensor Technology.
188 H. Budzier and G. Gerlach: Calibration of uncooled thermal infrared cameras

Figure 1. Structure of an uncooled IR camera with microbolome-


ters.

Figure 3. Camera model for the calculation of the IR image (expla-


nation in text).

housing. In addition, the pixel gives out emissions in the en-


tire half-space. The irradiance EP of the pixel (m, n) is

EP,mn = LO,mn ωFOV,mn + LC,mn ωC,mn − LS π, (1)

with the radiance LO,mn of the object, LC,mn of the camera


interior and LS of the pixel and the reduced solid angle of the
Figure 2. Optical channel in an uncooled IR camera. ωx : reduced object ωFOV,mn and the camera interior ωC,mn (= ωcase,mn +
solid angles; ϑx : temperatures. ωFront,mn ).
The reduced solid angle ωFOV,mn of the object depends
on the position of the pixel in the microbolometer and the
vision device. This class of devices is produced in very large f-number k of the optics used. For the central pixel and an
quantities. They are used in military applications, security f-number k = 1
technology and increasingly in automotive technology (night 1
driving aid). The aim of calibration is to produce as closely as ωFOV = π (2)
5
possible an optically flawless image. This process is referred
to generally as “smooth out”. applies (Budzier and Gerlach, 2011).
Temperature-measuring IR cameras also display a temper- Since the pixel receives radiation from the entire half-
ature value for each pixel. Here, after the “smooth out”, a space, the reduced solid angle ωC for the central pixel of the
radiometric calibration must also be carried out. Practically camera is
every pixel of an IR camera is a separate pyrometer. The main 4
problem is that the calibration parameters of microbolome- ωC = π − ωFOV = π. (3)
ters depend on both the ambient temperature and the camera 5
temperature (Budzier and Gerlach, 2011). Thus, the centre pixel receives 4 times as much radiation
In the following text the calibration of temperature- from the camera interior as from the object! In off-centre pix-
measuring, uncooled IR cameras will be described for mi- els, the ratio is even worse (Budzier, 2014).
crobolometer arrays1 . This is based on a radiometric camera In order to determine the radiance LC of the object from
model, which is described in Sect. 2. The individual steps Eq. (1), the radiance LC of the camera interior must also be
of the calibration process will be discussed in the following known. Since each pixel sees different elements in the cam-
sections. era interior, the radiance LC must be measured with an opti-
cal shutter. For this purpose, the shutter, which has the cam-
era temperature ϑC , is closed. The irradiance ESh with closed
2 Radiometric camera model
shutter now is
The theoretical basis for the calibration is based on a radio- ESh,mn = LSh,mn − LS π.

(4)
metric model of the thermal uncooled IR camera. In this case,
the sensor array including the sum of radiant fluxes from the Assuming that the camera shutter represents the interior cam-
object and from inside the camera will be considered (Fig. 2). era space (same mean temperature and simplified: LC =
A pixel “sees” inside the camera essentially the edge and LSh ), then from Eqs. (4) and (1) the irradiation intensity of
the bracket (front panel) of the optics and its own sensor the pixel through the object is obtained for the deviation be-
tween the shutter image and the measured object:
1 This article is a summarized presentation of the habilitation the-

sis of Budzier (2014). EO,mn = LO,mn − LSh,mn ωFOV,mn . (5)

J. Sens. Sens. Syst., 4, 187–197, 2015 www.j-sens-sens-syst.net/4/187/2015/


H. Budzier and G. Gerlach: Calibration of uncooled thermal infrared cameras 189

For calculating the object temperature the sensor temperature


is not required. Equation (5) is valid as long as the camera
internal temperature and the sensor temperature are constant.
If there is a change in any one of these temperatures then
the shutter must be activated again. The raw image VP,mn is
then obtained by multiplying the irradiance EO,mn with the
voltage sensitivity RV and the pixel area AP :
VP,mn = RV EO,mn AP . (6)
In the above considerations, a single DC (direct current) bias
(offsetmn ) and a uniform sensitivity (gainmn ) were assumed
for all pixels. Due to the complex manufacturing technology,
Figure 4. Extension of the camera model from Fig. 3 for the calcu-
that is not the case. While the sensitivity can vary by up to
lation of the IR image for TEC-less microbolometers.
±20 %, the deviation of the offset lies in the range of the
signal. Due to the large offset differences between the pixels
it is often the case that no signal is visible in the raw image. (10)
These differences are eliminated by a conventional two-point
correction and apply to the IR image VO,mn :
∗
VO,mn = VP,mn − VSh,mn − offsetmn gainmn , (7)
3 Calibration
with the pixel-specific variables offsetmn and the two-point
correction gainmn . For this purpose, when the shutter is A radiometer, where each individual pixel of an IR cam-
closed, an image is stored in an image memory and then sub- era can be determined, measures the radiant flux of the ob-
tracted from the current online IR image (Fig. 3). In order to ject and generates an output signal which, as a result of the
save an arithmetic operation per pixel in real time, the offset calibration, is proportional to the temperature of the object
and shutter correction are thereby summarized. (DeWitt and Nutter, 1989). A radiometric IR camera dis-
Modern microbolometer arrays do not contain Peltier el- plays as accurately as possible the true temperature of a black
ements in a vacuum housing. These are called TEC-less body. The calibration is used here to calculate a temperature-
microbolometers (TEC: thermo-electric cooler). The mi- proportional output signal (IR image) from the measurement
crobolometer is no longer stabilized to a constant tempera- signal (raw image) taking into account all technical and phys-
ture, i.e. the sensor temperature ϑS varies with the tempera- ical properties of the IR camera. The steps necessary for this
ture ϑK of the camera. However, since the sensitivity and the are summarized in Fig. 5.
operating point (offset) of a pixel are dependent on the sen- In the following section, the non-uniformity correction
sor temperature, this must therefore be measured and taken (Sect. 3.1), the temperature-dependent correction (Sect. 3.2),
into account in the camera model, as opposed to temperature- the defective pixel correction (Sect. 3.3), the shutter correc-
stabilized microbolometers. tion (Sect. 3.4) and the radiometric calibration (Sect. 3.5) will
The dependence of the offset and the sensitivity of a mi- be presented in detail. There will be no further discussion of
crobolometer array cannot be derived from the physical prop- the operating point setting which depends significantly on the
erties of a bolometer resistance without information concern- microbolometer used and would correspond to the example
ing the signal processing. Since the internal signal processing provided by the manufacturer’s procedure.
of a microbolometer array is not known in detail for reasons
of company in-house security, the array must be regarded as 3.1 Non-uniformity correction
a black box. In general, the following polynomials can be as-
sumed for the temperature dependence of the offset OV and Because of the technology, the individual pixels of a mi-
the sensitivity GV : crobolometer have uniquely different operating points (DC
bias values) and sensitivities and, thus, differing characteris-
OV (ϑS ) = o3 ϑS3 + o2 ϑS2 + o1 ϑS + o0 , (8) tics. During the correction of this non-uniformity all pixels
are converted onto a single characteristic curve, the so-called
standard characteristic curve. This process is referred to as
GV (ϑC ) = g2 ϑS2 + g1 ϑS + g0 . (9) “smooth out” because the IR image now with uniform illu-
mination has no structure and so is smooth. According to this
The sensitivity and the offset of the pixel must therefore be characteristic adjustment, all pixels behave the same and sub-
corrected using the temperature sensor ϑS (Fig. 4): sequent calibration steps can be exemplarily performed on
 ∗ any pixel or on any group of pixels.
UO,mn = UP,mn − USh,mn − offsetmn gainmn /GV (ϑS ) − OV (ϑS ) .

www.j-sens-sens-syst.net/4/187/2015/ J. Sens. Sens. Syst., 4, 187–197, 2015


190 H. Budzier and G. Gerlach: Calibration of uncooled thermal infrared cameras

Figure 6. Schematic of the two-point correction procedure, using


the example of two pixels. Dashed curves: the value “offset” is
shifted parallel to the standard curve.

tic curve. A significant simplification of the correction algo-


rithm is achieved by the subsequently described separation of
non-uniformity correction and radiometric calibration. This
means that the smooth out of the image is performed first
and then subsequently the temperature of connectivity of the
now common pixel characteristic curve. This approach also
leads to a reduction of the computational effort and correc-
tion allows for ease of data processing in real time.
Firstly, it is assumed that both the sensor temperature and
the ambient temperature are constant.
The linear relationship between the voltage of the pixel
Umn and the radiant flux 8 can be described with a linear
equation:

Umn = amn 8 + bmn , (11)

with the slope of the straight line represented by amn and the
intercept by bmn . Slope and intercept behave pixel-specific
and must be calculated so that a standard characteristic curve
is obtained for all pixels:
Figure 5. Flow chart of radiometric calibration.
Unorm = anorm 8 + bnorm , (12)

The description of the pixel characteristic curve is often a where anorm is the slope and bnorm is the intercept. For each
function of the object temperature ϑO . The function Umn (ϑO ) pixel (m, n), a constant pair offsetmn and gainmn must be de-
of the pixel at location (m, n) is not linear. Therefore, the termined so that
characteristic curve is generally described with a second-
order polynomial (Schulz and Caldwell, 1995). For radio- Ukorr,mn = (Umn − offsetmn )∗ gainmn + Unormoffset (13)
metric IR cameras this regression is not sufficient. Here, an
exponential regression (Horny, 2003) is provided, which will applies. The voltage Ukorr,mn is then corrected, i.e. on the
be described in Sect. 3.5. standard curve traced back to the voltage value of the pixel.
The characteristic curve corrections described in the liter- Figure 6 shows the principle of this two-point correction pro-
ature (Schulz and Caldwell, 1995, and Wallrabe, 2001) also cedure. First, the pixel graph is shifted in parallel (−offsetmn )
refer to photon sensors, whose function Umn (8O ) is often and then the slope is corrected (*gainmn ). Finally, for all pix-
not linear, as well as IR vision equipment without radiomet- els a valid constant voltage can be added. The standard curve
ric adjustment. In contrast, the relationship between the ra- is thus shifted back (+Unormoffset ). This last step is not nec-
diant flux 8 and the pixel voltage Umn of a microbolome- essary in every case but guarantees that the corrected pixel
ter is linear and can be used for describing a characteris- voltages are in the same range of values as the measured pixel

J. Sens. Sens. Syst., 4, 187–197, 2015 www.j-sens-sens-syst.net/4/187/2015/


H. Budzier and G. Gerlach: Calibration of uncooled thermal infrared cameras 191

values. This is important, for example, if the correction is im-


plemented in hardware (16 bit fixed-point arithmetic) and the
range of values is limited. offset∗mn = −dmn . (25)
In order to reduce the calculation during real-time correc- Due to various measurement errors, in particular the tempo-
tion, Eq. (13), ral noise of the pixel voltage, the determination of the coef-
ficients is faulty. This shows that a small deviation remains
Ukorr,mn = Umn · gainmn + offset∗mn , (14)
between the pixels in the corrected image. This can be inter-
can be simplified to preted as spatial noise. Now, in order to assess the quality of
the correction, the corrigibility parameter C was introduced
offset∗mn = −offsetmn · gainmn + Unormoffset . (15) by Gross et al. (1999) and Horny (2003):
s
To determine the standard curve and the pixel-related coeffi- USN2
cients, all pixel values Umn must be measured with T2 > T1 C= 2
, (26)
UTN
on two radiant fluxes: 8(T1 ) and 8(T2 ) (Eq. 12). Then the
coefficients of the standard curve (Eq. 12) may be initially with the noise voltages UTN and USN for the temporal noise
calculated: and spatial noise, respectively. The corrigibility C is ideal
hU2 (82 )i − hU1 (81 )i for correcting the characteristics of 0. It is equal to 1 if the
anorm = , (16) spatial noise has the same value as the temporal noise. If the
82 − 81
corrigibility is greater than 1, then the spatial noise dominates
hU1 (81 )i 82 − hU2 (82 )i 81
bnorm = , (17) and the non-uniformity is clearly visible in the IR image.
82 − 81
with the mean pixel voltage Uk (k = 1, 2) as the average 3.2 Temperature dependence correction
value for all pixels J :
If the temperature of the microbolometer is not constant, as is
1X J the case for TEC-less microbolometers, the sensor sensitivity
hUk (8k )i = Uj (8k ) . (18) and offset parameters change with temperature (Eqs. 8–10).
J j =1
To correct this temperature dependence, the polynomials of
Eqs. (8) and (9) must be determined. For this purpose, the
The standard curve is now the mean curve for all pixel char-
pixel voltages UP must be measured at various ambient tem-
acteristics. Each pixel now deviates at the measurement point
peratures ϑA . The pixel voltage is the sum of the pixel offset
k with 1Umn,k from the mean:
US , the signal from the camera interior UC and the constant
1Umn,k = Umn,k − hUk i . (19) object voltage UO :

Using the method of least squares, a function, UP (ϑA ) = UO + UC (ϑC ) + US (ϑS ) . (27)

1Umn = cmn Umn + dmn , (20) It is important that both the camera temperature ϑC and the
sensor temperature ϑS can be distinguished from the ambient
can now be determined so that all voltage differences 1Umn,k temperature ϑA . They are always higher by a few kelvin. If
are minimal. Using a simple regression the pixel-specific co- the ambient temperature changes, then the camera and sensor
efficients from Eq. (20) can now be obtained: temperatures change, in fact, with different time constants.
This makes them distinguishable in the output signal. The
1Umn,1 − 1Umn,2 temperature dependence of the pixel offsets US (ϑS ) is given
cmn = , (21)
hU1 i − hU2 i by Eq. (8). The signal voltage UC , resulting from the temper-
ature ϑC of the camera interior, is calculated with a quadratic
hU1 i 1Umn,2 − hU2 i 1Umn,1 polynomial:
dmn = . (22)
hU1 i − hU2 i UC (ϑC ) = aC ϑC2 + bC ϑC + cC , (28)
The corrected pixel voltage Ukorr,mn is then calculated from where the polynomial coefficients aC , bC and cC are initially
the current measured value Umn of the pixel and the differ- unknown. If a measurement is made at the point in time ti
ential voltage 1Umn : with the temperatures ϑC and ϑS i , then, by superposition of
Eqs. (27) and (28),
Ukorr,mn = Umn − 1Umn . (23)
3 2 2
Ui = UO +o3 ϑS,i +o2 ϑS,i +o1 ϑS,i +aC ϑC,i +bC ϑC,i +c (29)
The values for Eq. (15) result from a comparison of the fol-
lowing coefficients: applies using the combined constant
gainmn = (1 − cmn ) , (24) c = o0 + cC . (30)

www.j-sens-sens-syst.net/4/187/2015/ J. Sens. Sens. Syst., 4, 187–197, 2015


192 H. Budzier and G. Gerlach: Calibration of uncooled thermal infrared cameras

By making a series of measurements with i = 1, 2, . . ., So now the desired polynomial coefficients are known, in
i.e. measurements under varying ambient temperatures, then particular the sensor offsets o3 , o2 and o1 . The missing con-
the following vectors are obtained: stant o0 can be chosen arbitrarily. When the calculated value
is subtracted from the sensitivity corrected signal according
 
1 to Eq. (8),
 o3    
UPixel = Ukorr /GV (ϑC ) − o3 ϑC3 + o2 ϑC2 + o1 ϑC + o0 ,
  
 o2 
3 2 2
  
Ui = UO ϑC,i ϑC,i ϑC,i ϑK,i ϑK,i 1  o1 
.
 (39)

 ak 

 bk  the camera will behave like a camera with temperature-
c stabilized microbolometers.
(31)
3.3 Defective pixel correction
This series of measurements is taken with three different ob-
ject temperatures, but they are constant in each case. In the Due to the difficult manufacturing process for microbolome-
IR image three emitters are thus shown with different tem- ter arrays, all pixels have different parameters such as operat-
peratures ϑO1 , ϑO2 and ϑO3 : ing points, characteristic curves and noise. Pixels that either
do not work or whose parameters vary greatly from the mean
Un = An X, (32) are defined as non-functional or defective. Defective pixels
are generally referred to as “dead” pixels.
with n = 1, 2, 3, and the vectors Pixel defects manifest themselves as defective pixels in the
IR image. Their actual value can only be estimated with the
3 2 2
 
An = UO ϑC,i ϑC,i ϑC,i ϑK,i ϑK,i 1 , (33) help of neighbouring pixels. The measured value at this point
of the IR image is not reconstructable. Therefore, the number
of dead pixels is an important quality characteristic of mi-
  crobolometers. Normally not more than a maximum of 1 %
1 of all pixels should be defective.

 o3 
 A pixel is considered defective if any of the following con-

 o2 
 ditions is met.
X=
 o1 
. (34)

 ak 
 – The operating point is outside of the previously defined
 bk  voltage range 1UAP of the offset value dispersion.
c
– The sensitivity differs more than ±10 % from the mean
The series of measurements can now be represented in matrix value.
notation: – The noise voltage is 1.5 times greater than the average
noise voltage of the array.
U = AX, (35)
In addition, a group of defective pixels exist, which, although
with they do not meet the above criteria, behave differently and are
  classified as defective. These are, for example, short circuits
U1 between adjacent pixels or non-linear characteristics of indi-
U =  U2  , (36) vidual pixels. Figure 7 shows a raw image with a plurality of
U3 defective pixels (black dots).
Defective pixels occur not only individually but also in
  clusters. A cluster of dead pixels is a group of at least two
A1 defective pixels that are adjacent or gather together in a cor-
A =  A2  . (37) ner. Clusters are characterized by their size, that is, by the
A3 number of defective pixels. In the image section in Fig. 7b
clusters are clearly visible. Particularly critical are defective
The vector X includes the desired polynomial coefficients.
rows or columns, because, despite a correction in the IR im-
The solution to this over-determined system of equations
age, they are always conspicuous. A column or row is usually
with I  3 can be calculated with the Gaussian standard
considered defective if more than 50 % of the pixels do not
equation:
work.
 −1 Since the number of defective pixels of a microbolome-
X = AT A AT U. (38) ter is an important quality attribute, the manufactures will

J. Sens. Sens. Syst., 4, 187–197, 2015 www.j-sens-sens-syst.net/4/187/2015/


H. Budzier and G. Gerlach: Calibration of uncooled thermal infrared cameras 193

Table 1. Specification of allowable defective pixels.

Type Zone Defective pixel Example in Fig. 7


A None with more than 2 defective pixels 1 cluster with 2 pixels
Cluster
B Maximum cluster size of 9 pixels 4 clusters with 4 pixels
22 clusters with 3 pixels
90 clusters with 2 pixels
Columns and rows A 0 0
B 1 row or column 0
Functionality A 99.5 % 99.7 % (59 pixels)
B 99.0 % 99.3 % (1867 pixels)

trol range and, thus, affect the position of the standard


curve. Furthermore, all pixels which are too noisy are
eliminated.

2. The second defective pixel detection is performed us-


ing the calculated gain and offset values according to
the characteristic curve correction. Here defective pix-
els are identified that have too great a deviation from the
standard curve.

3. The third defective pixel detection is carried out at the


end of the calibration. In this case, all defective pixels
a) which have not yet been detected are recorded. This is
done by considering the IR image with different adjust-
ments and richly contrasting scenes.

While the first two defective pixel detections can be per-


formed computationally, the final detection is carried out
manually. This also means to verify the correction of the clus-
b)
ter and, if necessary, to change the correction method.
Figure 7. Raw image with 1926 defective pixels (0.6 %): (a) total There is no reading at the location of a defective pixel. This
microbolometer array and (b) the section marked in the top right- can only be estimated from the surrounding area. The aim of
hand corner (90 × 70 pixels). Defective pixels are shown in black. the recalculation of the pixel value is always to produce a
Microbolometer array with 640 × 480 pixels. high-quality visual image, i.e. so that an observer of the IR
image may not notice any defective pixels. The calculation
of the pixel value is carried out by methods of image pre-
always indicate in their specifications the maximum permis- processing, such as with median operators.
sible number of defective pixels. In the centre of an IR im-
age defective pixels are particularly noticeable. Therefore, 3.4 Shutter correction
the image area is divided into at least two zones (Fig. 8).
In the central region (zone A) higher demands are placed on In order to calculate the radiance and the object temperature
the functionality of pixels than of those on the edges. Table 1 from Eq. (5), the radiance LC of the camera interior must be
shows a specification of permissible defective pixels. known. Should the camera temperature ϑC vary as a result
The detection of defective pixels proceeds in three steps of a change in the ambient temperature, the radiance value
(flow chart in Fig. 5). of the camera interior needs to be adjusted accordingly. This
process is called shutter correction.
1. The first defective pixel detection must be done before The starting point for consideration is that the radiance
the uniformity correction. Here all defective pixels are of the camera interior space was determined using a known
detected that are located outside of the previously de- ambient temperature and thus a known camera temperature
fined range of variation of the pixel operating points. ϑC0 (ϑC0 = ϑSh0 ); therefore, according to Eq. (5) the object
These are primarily pixels which are outside the con- radiance and the object temperature can be calculated.

www.j-sens-sens-syst.net/4/187/2015/ J. Sens. Sens. Syst., 4, 187–197, 2015


194 H. Budzier and G. Gerlach: Calibration of uncooled thermal infrared cameras

Format AH AV

BH xBV

384 x 288 80 60

640 x 480 160 120

1024 x 768 256 192

Figure 8. Typical subdivision of infrared microbolometer arrays in the mid (zone A) and marginal zones (zone B).

When the shutter is open, the pixel voltage is

Uopen (ϑC ) = UO + ωC UC (ϑC ) , (40)

with the voltage equal to

UC (ϑC ) = RV AP LC . (41)

When ωC UC is known, then the ambient temperature correc-


tion can be carried out:

UO = UO (ϑC ) − ωC UC (ϑC ) . (42)

The voltage UO in Eq. (42) is now independent of the am-


bient temperature as well as the camera temperature and is
used to calculate the object radiance LO .
Figure 9. Histogram of the raw image at the operating point show-
The projected solid angle ωC is known theoretically. With
ing a black body with a temperature of 150 ◦ C (microbolometer
an f-number k = 1, then, for the centre pixel with 384 × 288 pixels).
4
ωC = π = 0, 8π (43)
5
The projected solid angle ωC of the camera interior can be
is obtained. read from Eq. (46):
For each pixel the voltage UC is determined with the aid of
shutter correction. For this purpose, the shutter is closed and ωC = aSh π, (48)
the pixel voltages (shutter image USh ) are measured. When
the shutter is closed, the pixel sees the entire half-space (ωc = with the increase aSh in the shutter curve. Now the signal
π ): voltage of the object can be calculated independently of the
ambient temperature using the shutter image:
USh = π UC (ϑC ) . (44)
UO = UO (ϑC ) − aSh USh (ϑC ) . (49)
Therefore, the shutter signal USh must be multiplied by a fac-
tor ash . This results from the so-called shutter characteris-
tic curve. The shutter characteristic curve is the ratio of the
shutter-open signal Uopen to shutter-off signal USh : 3.5 Radiometric calibration
Uopen ωC UC (ϑC ) + UO The previous corrections in Sect. 3.1 and 3.4 lead to all pixels
= . (45)
USh π UC (ϑC ) of the IR image having the same behaviour and the IR image
It follows that not being dependent on the ambient temperature. Finally, the
radiometric calibration calculates the temperature of the ob-
Uopen = aSh USh + UO , (46) ject to be measured from the grey values of the pixel. It works
with a voltage object temperature characteristic, so for each
with grey value UD a temperature value TO is assigned.
ωC
aSh = . (47)
π

J. Sens. Sens. Syst., 4, 187–197, 2015 www.j-sens-sens-syst.net/4/187/2015/


H. Budzier and G. Gerlach: Calibration of uncooled thermal infrared cameras 195

Figure 10. Effect of two-point correction. (a) Raw image (colour bar chart in grey values) and (b) thermal image (colour bar chart in ◦ C).

Figure 12. Temperature dependence of (a) the offset values and


(b) the sensitivity of the observed TEC-less microbolometer arrays.

Figure 11. Histograms of (a) the offset values and (b) the gain val-
ues of the microbolometer arrays under consideration. The coefficients allow for a physical interpretation. The value
of O is a general offset. Using this value, the characteristic
curve along the ordinate can be displaced in parallel. The co-
According to Horny (2003) it is possible to approximate
efficient R represents the system response of the IR camera
the sensor output signal with a Planck curve:
and is the counterpart to the system sensitivity. Planck’s ra-
R diation law can be applied for the coefficient B:
UD = B
+ O, (50)
e TO
−F c2
B= . (52)
λB
where in B, F , O and R are the regression coefficients to be
determined. Then the inverse function from Eq. (50) serves It thus describes the spectral behaviour of the system. The
to calculate the object temperature TO : wavelength λB is the effective wavelength of the IR camera.
B The coefficient F allows for an alignment of the non-linearity
TO =  . (51) of the system. The coefficients can be determined using a
R
ln UD −O +F non-linear regression analysis.

www.j-sens-sens-syst.net/4/187/2015/ J. Sens. Sens. Syst., 4, 187–197, 2015


196 H. Budzier and G. Gerlach: Calibration of uncooled thermal infrared cameras

By analysing the offset and gain values calculated in the


two-point correction (Fig. 11), in a second defective pixel
detection procedure more abnormal pixels that lie outside a
defined scatter band can be sorted out. Fixed limits of ±20 %
have been proven here by the gain values.
For TEC-less microbolometers, the temperature depen-
dence of the offset values and the sensitivity has now to be
determined (Fig. 12). The regression analysis includes the
known relationships between camera or sensor temperature
and the respective measured variables. These are for the sen-
sitivity of a polynomial of the second order (see Eq. 9) and
the offset values for a third-order polynomial (see Eq. 8).
Figure 13. Shutter characteristic curve. Parameter: camera temper- After correction of the non-uniformity, the correction of
ature of the respective measurement points. the ambient temperature dependence is carried out. The so-
called shutter characteristic curves are recorded, which rep-
resent the ratio of the shutter-open signal to shutter-off signal
as a function of the temperature of the camera (Fig. 13). For
this, the camera temperature (in Fig. 13 from 3.9 to 47.3 ◦ C)
varies and in each case the signal Vopen of the object and the
shutter Vsh is measured in a temperature chamber. Both sig-
nals are directly proportional to each other.
Subsequently, the radiometric calibration is carried out by
determining the signal voltage VD object temperature charac-
teristic (Fig. 14). The inclusion of this characteristic is done
with a black body as a measuring object. Its temperature TO
will vary within the measurement range, e.g. from −10 to
150 ◦ C. With the recorded measurement points, a regression
is carried out in accordance with Eq. (50).
Figure 14. Object temperature ϑS vs. signal voltage VD char- Finally, a third defective pixel detection takes place during
acteristic. Regression according to Eq. (50) where B = 1514.3 K; the final check. Here visually conspicuous pixels are rejected
F = 0.920; R = 1 967 454; and Off = −23 392. by visual inspection.

5 Summary
4 Example of application
The calibration of an uncooled IR camera is a complex and
In the following, the calibration process will be illustrated by lengthy process, which significantly affects the cost of an
an example. IR camera. Depending on the measurement technology used,
The calibration process begins with the setting of the op- such as black bodies, references and climatic chambers, the
erating point. Figure 9 shows a histogram of the raw image proposed calibration allows for the measurement of absolute
with optimized operating point (dynamic range of −32 000 temperatures with a maximum measurement uncertainty of
to +32 000 grey values). To demonstrate the modulation, a about ±1 K. This is only true for steady ambient tempera-
black body with a temperature of 150 ◦ C has additionally tures, i.e. when the camera is at a constant temperature. When
been shown. changing the ambient temperature a shutter cycle is always
After the operating point has been fixed, in the first de- required. With sudden changes of the camera temperature be-
fective pixel detection procedure, all pixels which lie out- tween two shutter cycles, e.g. as a result of a jump in the am-
side of the dynamic range are defined as defective. Figure 10 bient temperature, strong variations in measurements can ap-
shows the effect of a two-point correction of the pixel graph, pear. The behaviour of the IR camera is not ergodic. An addi-
in which the variation of the values of the individual pix- tional measurement of the ambient temperature is not usually
els occurring in the raw image (Fig. 10a) is eliminated. The possible. In commercial IR cameras, this problem is solved
recognizable characteristic stripe structure is formed by the by predictive models which estimate an expected camera in-
column-wise arrangement of “blind” bolometers. The optical ternal temperature from the previous temperature change of
image resulting from the natural vignetting of the optical sig- the camera. Solutions to these problems are not known or
nal and delivered to the image edge is also always present in published.
raw image where in Fig. 10a it is hardly recognizable by the
content of the thermal image.

J. Sens. Sens. Syst., 4, 187–197, 2015 www.j-sens-sens-syst.net/4/187/2015/


H. Budzier and G. Gerlach: Calibration of uncooled thermal infrared cameras 197

For rapid changes in temperature of the camera, the shut- Gross, W., Hierl, T., and Schulz, M.: Correctability and long-term
ter needs to be frequently operated, e.g. several times within stability of infrared focal plane arrays, Opt. Eng., 38, 862–869,
1 min. Since the operation of the shutter always causes 1999.
an interruption of the measuring process, the user should Horny, N.: FPA camera standardization, Infrared Phys. Technol., 44,
operate the shutter as little as possible or even completely 109–119, 2003.
Kruse, P. W.: Uncooled Infrared Imaging Arrays and systems, Aca-
avoid using it. For the shutterless operation of IR cameras,
demic Press, San Diego, 1997.
however, much more complex calibration algorithms are Mounier, E.: Technical and market trends for microbolometers for
required, which build on those described here. Such an thermography and night vision, Proc. SPIE, 8012, 80121U-1–6,
approach is pursued by Tempelhahn et al. (2014). 2011.
Schulz, M. and Caldwell, L.: Nonuniformity correction and cor-
Edited by: R. Morello rectability of infrared focal plane arrays, Infrared Phys. Technol.,
Reviewed by: three anonymous referees 36, 763–777, 1995.
Tempelhahn, A., Budzier, H.; Krause, V., and Gerlach, G.: Devel-
opment of a shutterless calibration process for microbolometer-
References
based infrared measurement systems, International Conference
on Quantitative Infrared Thermography, 7–11 July 2014, Bor-
Black, S. H., Sessler, T., Gordon, E., Kraft, R., Kocain, T., Lamb,
deaux, QIRT-2014-060, 2014.
M., Williams, R., and Yang, T.: Uncooled detector development
Vollmer, M. and Möllmann, K.-P.: Infrared Thermal Imaging,
at Raytheon, Proc. SPIE, 8012, 80121A-1–12, 2011.
Wiley-VCH, Weinheim, 2010.
Budzier, H.: Radiometrische Kalibrierung ungekühlter Infrarot-
Wallrabe, A.: Nachtsichttechnik, Vieweg, Braun-
Kameras, TUDpress, Dresden, 2014.
schweig/Wiesbaden, 2001.
Budzier, H. and Gerlach, G.: Thermal Infrared Sensors, John Wiley
& Sons, Chichester, 2011.
DeWitt, D. P. and Nutter, G. D.: Theory and Practice of Radiation
Thermometry, Wiley, New York, 1989.

www.j-sens-sens-syst.net/4/187/2015/ J. Sens. Sens. Syst., 4, 187–197, 2015

You might also like