WHITE PAPER

The Truth about range data

How to assess thermal camera

range capability for site design purposes
Author: John Love, Senior Applications Engineering Manager
DRS Technologies
Commercial Infrared Systems

1,000 m
500 m
250 m
100 m
50 m

DRS Commercial Infrared Systems White Paper

The Truth About Range Data:

How to assess thermal camera range capability
for site design purposes
Author: John Love, Senior Applications Engineering Manager DRS Technologies, Commercial Infrared Systems
When laying out a site design for surveillance using
currently marketed uncooled thermal cameras, it is
important to understand the coverage capability of the
cameras under evaluation in terms of field of view (FOV)
and range. FOV is easy: the numbers in the specifications
are not ambiguous and can be relied upon at face value.
Range, on the other hand, is far more complicated. The
industry is flooded with conflicting information when it
comes to range performance. Every valuation has its merits
some more pertinent than others. Accurately gauging
range performance from the onset of the site design may
allow one to avoid unintended, and pricey rework costs
later. The following white paper explores the various
determinants that are needed for thermal camera range
capability assessment.

What determines range capability?

Range capability depends on many factors, some of which
are easily controlled, and others of which are not. A list of
the most prominent appear in Table 1.
Five considerations that determine range capability in the
face of all these variables are:
1. Range capability predicted from analysis
2. Range criterion based on displayed size of target
3. Range criterion when using video analytics
4. Impact of weather on range capability
5. Impact of image processing on range capability
For each, a discussion of the issue is followed by
information for practical application.

Factors we can control

Factors we cannot control

Task

Weather and atmosphere

What do we want to accomplish? E.g., detection, tracking,

Clarity of the air (transmission, humidity)

Obscurants (rain & mist, snow, fog, dust, smoke, and haze)
Atmospheric turbulence caused by solar heating

Design of the camera

Target characteristics

Optics focal length, f-number, aperture size, aberrations, and

Size
Structure
Temperature
Motion
Emissivity

Placement of the camera

Background characteristics

Distance from camera to intended target

Location of the horizon in the scene
Choice of background for intended target

Temperature
Texture and clutter

assessment?
With what confidence do we want to be able to accomplish
the task?
Probability of correct detection
Probability of false detection

transmission
Detector pixel pitch, intrinsic sensitivity, array size, pixel
uniformity
Image processing to reduce non-uniformity, suppress noise,
compensate for blur, and enhance contrast
Display type, size, resolution, viewing distance, gamma,
luminance

How the camera output is to be used

Will it be displayed on a monitor for a human to interpret?

Will it feed a video analytics processor for automatic detection
and tracking?

Table 1: Factors that determine range capability of thermal imagers

DRS Commercial Infrared Systems White Paper

Range capability predicted from analysis
Discussion
No one has more thoroughly researched and developed
modeling methods for thermal cameras than the US
Army Night Vision and Electronic Sensors Directorate at
Fort Belvoir, Virginia. Beginning as early as the 1950s,
this organization has produced a string of analytical
foundations and computer models of ever-increasing fidelity
for predicting range performance in militarily significant
scenarios. The latest model is called NVThermIP2009,
usually shortened to NVThermIP.
The Army models have consistently addressed a specific
set of target acquisition tasks, namely
Target detection determining that a target is present in
the sensors field of view
Target recognition for a detected target, discerning
whether it is a tank, truck, or armored personnel carrier
(this task is sometimes referred to as a 3 alternative
forced choice experiment).
Target identification for a recognized tank, truck, or
APC, discerning which of 8 different types it is (this task
is sometimes referred to as an 8 alternative forced
choice experiment).
Example target types are shown in Figure 1. For purposes
of our security surveillance needs, where we are looking
for people more than military vehicles, it is not intuitively
obvious whether the theory built around them applies, but
we use it anyway by convention and because it works
reasonably well most of the time.
In each case, what is being predicted by the model is the
performance of a human viewing a display monitor on which
is presented the output of a thermal imaging camera looking
through a specified atmosphere at a specified target in front
of a specified background scene. By performance what is
meant is the percentage of a group of humans attempting
the task (detection, recognition, identification) as a function
of target range. It is customary to interpret the result as the
probability that a single observer correctly performs the
task (ergodicity). The result is that the model predicts the
probability of successful task accomplishment as a function
of target range.
A fascinating and useful concept was tested and promoted
by the Army lab in the 1950s, becoming an essential
paradigm in all its models up to the present. The paradigm
has two key elements: a) that a target of any size, shape,
or type has a so-called critical dimension that can be
used for analysis, and b) that the probability of successful
task accomplishment (detection, recognition, identification)
correlates strongly with the number of bars in a bar target
pattern that can be resolved across the critical dimension
of the target. The concept is illustrated in Figure 2. As the
target moves farther and farther in range, the angular size
of the target gets smaller, as would the equivalent bar

(a) Example APCs

(b) Example Trucks

(c) Example Tanks

Figure 1: Examples of targets used in Army model testing

pattern. Just short of the range where the bars become

indistinguishable (unresolvable), the bar pattern is said to
be at the range limit for that bar pattern frequency
(bars per unit of angle). It is the characteristics of the
camera-observer combination that determine the range
where the bars become not resolvable.
It was this concept (equivalency of actual target to a bar
pattern) that led to the ubiquitous Johnson Criteria. A
3

DRS Commercial Infrared Systems White Paper

published paper by John Johnson described the
relationship between the number of resolvable bars and the
tasks of detection, recognition, and identification for an
image intensifier system (See Reference 1). Subsequent
analysis has yielded a version of the original criteria as it
applies to thermal imaging cameras, which appears as
Table 2. For each task, the criterion is the number of
resolvable bar cycles subtending the critical dimension of
the target, where one bar cycle is one bar and one space,
such that the probability of task accomplishment is as
indicated.
Resolvable Bar Cycles Required for
Indicated Probability of Successful Task
Accomplishment
Task
Detection

Probability =
50%

Probability =
70%

Probability =
90%

0.75

0.94

1.34

Recognition

3.75

5.37

Identification

7.5

10.7

Johnson Criteria Concept

(a) Equivalency between standard target and bar pattern
0.5 m
0.95 m
1.8 m

Six Bar Cycles

0.95 m
Equivalent
to

Standard
Human
Target

Equivalent
to

0.95 = 1.8 X 0.5

Critical Dimension:
Equivalent Target
Size for Analysis

Equivalent Target Bar

Pattern for Analysis
Criteria for 50% Probability:
6 Cycles for Identification
3 Cycles for Recognition
0.75 Cycle for Detection

Figure 2(a): Target equivalency to a bar pattern

Johnson Criteria Concept
(b) Determination of range by resolvability of bar pattern

Table 2: Johnson criteria for thermal imaging cameras

The utility of the bar pattern/target equivalency and the

Johnson criteria is two-fold. First, it simplifies the analytical
approach since bars are simple and defined whereas real
targets are complex and defy definition. Second, it enables
performance to be characterized in a laboratory with
relatively simple equipment (blackbody source, bar target
masks, collimator), instead of having to round up trucks,
APCs, and tanks every time one wants to characterize a
different camera. Vestiges of this approach are still present
in updated version of the Army model (NVThermIP), but the
theory has become more sophisticated mathematically to
yield more accurate results.
The use of NVThermIP entails specifying a variety of input
values characterizing the camera lens, the detector, the
framing and sampling electronics, the signal processing
electronics, the viewing display, the atmosphere, the target,
and the task. The outputs are tables of probability vs.
range for each of the tasks of detection, recognition, and
identification.
A drawback to NVThermIP is that it is not always possible
to know values for all the inputs needed. Furthermore,
NVThermIP is not transparent in that, with all its variables
and the scores of equations relating them, it is often not
clear what is really going on.
An alternative to NVThermIP used by many is an approach
we will refer to as pixels on target. In this approach we
make a fundamental assumption that the bar pattern is
resolvable when the width of a bar is equal to the angular
subtense of a detector pixel, which we call instantaneous
field of view or IFOV. Two adjacent pixels have the
angular subtense of a resolvable cycle (i.e., a bar and
a space) at the limit of resolvability. Calculations in this
approach are done with small angle trigonometry and

Increasing range increases bar chart

spatial frequency (bars per perceived
angle)

6 resolvable cycles required for

identification (at 50% probability)
Identification range is the range at which
the 6-bar pattern is just at the threshold of
resolvability
Similarly, for recognition (3-bar pattern)
and detection (0.75-bar pattern)

Figure 2(B): Range relationship to bar spatial frequency

are often simple enough to do in ones head, hence the

transparency of the approach.
The pixels-on-target approach requires only the most basic
information: target size, detector pixel pitch, and the
effective focal length of the camera optics. With these inputs
the following four equations are evaluated in sequence to
predict the range as illustrated in the example.
Equation 1:

Target critical
dimension (m) =

Equation 2:

Pixel IFOV (mrad) =

Target
Target
Height (m) x Width (m)
detector pixel pitch (m)
optics effective
focal length (mm)

DRS Commercial Infrared Systems White Paper

Equation 3:

No. of reqd
from
2 x cycles
Johnson
pixels reqd. in
criterion
No. of
plane of
meter target =PPM = Target critical
dimension (m)
Note: We use the term PPM throughout this article to mean
the number of required pixels per meter in the plane of the
target.
Equation 4:

1000 radians
mrad
Range (m) =
pixels x Pixel IFOV (mrad)
PPM
m
Using the example below and the equations, the results
shown can be determined.

example:
Variable

Example

Target height

1.8 m

Target width

0.5 m

Detector pixel pitch

17 m

Optics effective focal length

16.75 mm

Task:

Recognition

Johnson criterion for recognition:

3 resolvable cycles

results:
Variable

Example

Target critical dimension

0.95 m

Pixel IFOV

1.01 mrad

PPM

6.3 pixels/m

Range (50% probability)

157 m

It is clear that, though simple, the pixels-on-target approach

glosses over many important considerations, e.g., weather.
However, it works well in many cases. Specifically, it works
well when the assumptions listed in Table 3 apply.

Category

Assumption

Signal

There is plenty of signal. Ther target is sufficiently hotter or colder than the background scene such that the
temperature difference is large enough to provide ample signal at the camera aperture.
Applies well when the targets are people or operating ground vehicles in front of a bland earth or sky background
out to ranges not greater than 5 or 10 km when the air is clear and dry. Does not apply when ranges are longer,
the targets are weaker (such as when a vehicle has been sitting without running for a long time), or when the
background is cluttered or variegated. It also does not apply at so-called thermal crossover times whereby the
daily temperature cycling of earth temperature swings above and below the target temperature, if it does.

Atmosphere

There isnt much atmospheric attenuation. The air is assumed to be clear and dry.

Optics

The optics are assumed to be well designed, meaning that the blurring effects of diffraction and aberrations are
assumed to be substantially insignificant at the spatial frequency range supported by the pixel IFOV. Saying this
another way, the pixel IFOV sets the limit of resolvability for the camera, not other blurring effects.
Applies well to optics designed for currently marketed uncooled thermal cameras by reputable manufacturers.
The diffraction part of the assumption is met for cameras with f-numbers 1.7, which all are. The aberration part
of the assumption is harder to describe, but is generally valid for these cameras.

Detector

Detector noise is sufficiently low that it does not enter the problem (it does not limit range performance).
Applies well to currently marketed uncooled cameras because NETD values are typically at or below 50 mK.

Image processing

No performance lift from image processing is accounted for.

Currently marketed uncooled cameras sometimes use optional image processing to sharpen images, enhance
contrast, suppress large area contrast variability, suppress noise, etc. These are all good things that enhance
range, but pixels-on-target does not account for them.

Display

The display is considered perfect, meaning it does not contribute to performance loss.
The display can have subtle effects on range performance, but in todays world of high resolution flat panels, it is
rarely much of a consideration. The most significant part of the display assumption is that it doesnt matter how
close to, or away from, the display the observer is positioned. Actually, it does matter a little bit, and pixels-ontarget does not account for it.

Human observer

The human is considered perfect, meaning his/her eyesight is adequate, his/her training is adequate, and his/
her motivation (attention to task) is sufficient.

Table 3: Assumptions Required for Use of Pixels-on-Target Method

DRS Commercial Infrared Systems White Paper

Practical Application
Tables A1 thru A4 (located in Appendix A beginning on
page 12) list the predicted ranges for the DRS uncooled
camera product line using both NVThermIP and pixels-ontarget. The tables list predicted ranges from NVThermIP
at probabilities of 90%, 70%, and 50%, for the tasks of
detection, recognition, and identification, as the Army
defines them. Also listed for each task are the pixels-ontarget predictions, which are understood to be ranges for
50% probability. The products are listed by optics effective
focal length. The input parameters and assumptions used
are:
Target:
Human 1.8 m high x 0.5 m wide, 2C temperature
contrast to background
Vehicle 1.5 m high x 4.0 m wide, 5C temperature
contrast to background
Background: bland and at or near room temperature
Atmosphere: clear and dry, transmission factor is 0.9 per
km (Beers law)
Optics characteristics: f-numbers range from 1 to 1.6
depending on model; average transmission factors
range from 86% to 90%; focal lengths range from 3.8 mm
to 65 mm.
Detector characteristics: pixel pitch is 17 microns in
horizontal and vertical directions;
array sizes: 640 x 480 and 320 x 240;
spectral band of sensitivity: <8 microns to >13 microns;
sensitivity: [see system NETD].
System NETD: 0.05C
System gain: displayed dynamic range = 6 X target-tobackground contrast
Image processing: none for special enhancements
Display: flat panel, pixel size = .03175 cm
viewing distance = 53.34 cm
average luminance = 10 ft-lamberts
Human observer: standard eye model; standard training
The DRS products, whether Tamarisk cores or
WatchMaster IP cameras, are represented in the tables
only by their optics effective focal lengths, regardless of
whether they use 640 x 480 or 320 x 240 arrays. Since
pixel pitch is the same for all, focal length is the primary
performance differentiator, which is to say that the additional
field of view afforded by the larger array does nothing
to increase range performance. That said, however, all
pertinent optical properties (e.g., transmission, diffraction,
aberrations, f-number) for each product are accounted for in
the modeling, not just focal length.

Range criterion based on displayed size of target

Discussion
In the camera marketplace, range performance is nearly

always listed on data sheets in terms of maximum values.

But for the practitioner designing a surveillance plan for an
actual site, it is important to consider what maximum range
performance looks like and whether that is acceptable. If
not, additional design margin should be applied to get the
desired result.
Heres an example. From the tables in Appendix A, the
maximum listed human detection range for a camera with a
19 mm focal length lens is 850 m (NVThermIP-50%). The
height dimension for that person is 1.8 m, meaning that he
is only 2.4 pixels high at that range, which is 1% of the
vertical field of view of a 320 x 240 image as depicted on
the display. The target image is very tiny. This situation is
approximated by Figure 3 (the actual range in the figure is
1000 m and the person is 2 pixels high).

Figure 3: Image of human at 1000 m using 19 mm focal length

320 x 240 imager having 16 horizontal field of view

If this is an expected and acceptable image of a person,

there is no problem. If not, additional margin is needed in
the design range maximum when planning the surveillance
site.

Practical Application
If an adjustment needs to be made, a simple way to make
it is to estimate the desired minimum target height in the
image in terms of pixels and then to calculate the new
design range maximum using the following two equations.
Equation 5:

minimum target height

pixels
displayed (pixels)
PPM
=
meter
target height (m) x 100%
where, as earlier, PPM refers to the required number of
pixels per meter in the plane of the target. Target height, in
this case, is the vertical height of the target, not the critical
dimension calculated by Equation 1.
6

DRS Commercial Infrared Systems White Paper

Equation 6:

Pixel IFOV (mrad) =

detector pixel pitch (m)

optics effective
focal length (mm)

Equation 7:

1000 radians
mrad
Range (m) =
pixels x Pixel IFOV (mrad)
PPM
m
Note that these equations for Pixel IFOV and for range are
the same as used in the pixels-on-target calculations, but
with a different value for PPM. They are repeated here for
convenience.
example:
Variable

Example

Target height

1.8 m

Optics effective focal length

19 mm

Detector pixel pitch

17 m

Minimum target height displayed

20 pixels

Results:
Variable

Result

PPM

11.1 pixels/m

Pixel IFOV

0.89 mrad

Range

101 m

At this range, a 1.8 m high target spans 20 pixels head to

toe. This reduction in range limit, from the tabulated 850 m
down to 101 m, not only makes the target look bigger on the
monitor but provides much more definition of the targets
shape by increasing the number of pixels in one dimension
by a factor of 8. The resulting image would look more the
one in Figure 4.

It is very important to note that this example is somewhat

of an exaggeration for illustration purposes. It is not meant
to prescribe how much reduction should be taken in the
range limit, but only to illustrate the trade that may need to
be made.

Range criterion when using video analytics

Discussion
Video analytics is a wonderful concept for surveillance
systems. The idea that hundreds of camera feeds can be
tirelessly observed, examined, and archived 24 hours a
day without missing intruder alerts and without generating
false alerts is a truly remarkable ideal. Of course, the ideal is
elusive, but reality is fast approaching it.
For analytics, range criteria are different from those modeled
by NVThermIP. Analytics providers have developed
sophisticated proprietary algorithms to accomplish their
functions, making performance prediction and analysis
not at all straightforward. But in the functional common
denominator of detecting and tracking targets, some
generalities can be offered. A useful notion is that a target
can be detected and tracked if it spans a certain number of
detector pixels.
A survey of product data sheets from 5 prominent video
analytics providers reveals that the criterion for detecting
and tracking a human target ranges from 4 to 11 pixels/m
in the plane of the target, depending on provider, with 6-7
pixels/m being the norm. Since the critical target dimension
of a standing human is about 1 meter (see above), an
analytics criterion of 6 pixels/m is essentially tantamount to
the Johnson criterion for recognition discussed earlier for
the pixels-on-target model. However, the more generalized
approach is described below.

Practical Application
Compute the maximum range limit for video analytics using
the following equation:
Equation 8:

1000 radians
mrad
Range (m) =
pixels x Pixel IFOV (mrad)
PPM
m
where, PPM (pixels/m) is the criterion provided by the
analytics supplier for the required number of pixels/m in the
plane of the target; and, Pixel IFOV (mrad) is calculated as
before by the following:
Equation 9:

Pixel IFOV (mrad) =

detector pixel pitch (m)

optics effective
focal length (mm)

Figure 4: Image of human at 100 m using 19 mm focal length

320 x 240 imager having 16 horizontal field of view

DRS Commercial Infrared Systems White Paper

example:
Variable

camera represented in table 8 was a camera with a 16.75

mm focal length lens. The weather impact on range for
a longer focal length lens is greater. For example, for the
camera with 65 mm focal length the decrease in detection
range at the 50% probability level is 70%.
Table 4 doesnt address all weather possibilities. Besides
rain and humidity, there could be fog, haze, smoke, and
snow. All can be characterized in a manner similar to
Table 4.

Example

Detector pixel pitch

17 m

Optics effective focal length

50 mm

PPM for analytics

6.5 pixels / m

Results:
Variable

Result

Pixel IFOV

0.34 mrad

Range

452 m

Practical Application

The advantage of this approach is that pixel pitch and

analytics criterion are explicit inputs, which make it possible
to compare products from different suppliers. For example,
though all the current DRS products use detectors with 17
micron pixel pitch, other suppliers use pitch values of 25
microns for some cameras. Larger pixel pitch results in
shorter range performance, all else being equal.

Impact of Weather on Range Capability

Discussion
Weather conditions impact the range capability of thermal
imagers by attenuating the infrared signal as it propagates
from target to camera. Table 8 shows the impact of 3
different conditions that have 3 different levels of severity.
The impact can be characterized by the signal transmission
over a 1000 meter path, with the good weather value of 0.9
serving as the baseline. The other 3 conditions are high
absolute humidity, moderate rain, and heavy rain. The range
predictions in Table 8 are found using NVThermIP because
the pixels-on-target approach does not account for signal
attenuation from weather.
By examining the numbers in the table it is clear that most
of the impact is minimal. The largest impact occurs for
detection at 50% probability, where the decrease in range
from good weather to heavy rain is 40%. Corresponding
decreases in recognition and identification are essentially
nonexistent. The impact on range is itself a function of
range. The longer the range, the more the impact. The

If the site designer knew what weather conditions he faced,

he could apply margin factors to account for the weather
losses in range capability using an analysis similar to
Table 4. However, there are so many different possibilities,
all of which affect performance of the cameras differently
depending on their performance in good weather.
The recommended course for site planning would be the
following:
a) Consider the likely weather conditions at the site and the
frequencies they occur (e.g., dont worry about snow in
south Texas and dont worry about heavy rain in Denver,
etc.).
b) Decide which must be tolerated without loss of
surveillance capability, and which, if any, may be
ignored.
c) Use NVThermIP to examine the impact of the conditions
that matter to determine a derating factor to be applied
to the nominal range capability listed in Tables A1 thru
A4.
d) Derate tabulated performance accordingly.

Impact of Image Processing on Range Capability

Discussion
Multiple thermal imaging suppliers (including DRS)
incorporate image processing options as part of their
products. Some of these features have generic names, such
as histogram equalization, automatic gain control (AGC),
boost, local area processing (LAP), and multi-frame (MF)

Detection
IR Signal Attenuation
Over 1,000 meter path

P = 90%

P = 70%

Recognition
P = 50%

P = 90%

P = 70%

Identification
P = 50%

P = 90%

P = 70%

P = 50%

40

71

112

39

70

109

39

69

107

39

69

105

Weather Case: Nominal (good) weather Temp: 73F | RH: 40% | No rain, fog, haze
0.9

280

486

751

52

92

144

Weather Case: High humidity Temp: 90F | RH: 68% | No rain, fog, haze
0.5

263

0.3

255

0.05

242

437

637

51

90

140

Weather Case: Moderate rain Rain rate: 4 mm / h

413

582

51

89

137

Weather Case: Heavy rain Rain rate: 25 mm / hr

379

512

51

88

134

Table 4: NVThermIP Analysis of Example Weather Impacts on DRS Camera with 16.75 mm Focal Length Lens

DRS Commercial Infrared Systems White Paper

processing; others have trademarked names such as Image
Contrast Enhancement (ICE) reflecting their proprietary
nature.
Our modeling tools (NVThermIP, pixels-on-target) do not
enable us to easily assess the benefits of these features,
though it is obvious to a casual observer that they do
improve the images and extend range capability. They
make these enhancements by doing two things: a)
sharpening image detail; and b) reducing the large area
gray shade range across the field of view. Figures 5 and 6,
original and enhanced versions of the same scene, illustrate
both characteristics. Figure 6 has sharper detail and a more
uniform large area gray shade rendering. Edge enhancing
boost brings out detail in both the truck and the human.
Contrast enhancement removes the large area shading
difference between the sky and earth, allowing the contrast
gain to be higher, exposing more detail in both the earth
and sky regions.
Industry literature offers some, but limited, help in assessing
image processing benefit. Reference 2 indicates that boost

Figure 5: Human and vehicle shown at 25 meters and 40 field

of view using Automatic Gain Control

Figure 6: Human and vehicle shown at 25 meters and 40 field

of view using Image Contrast Enhancement (ICE)

(edge enhancement) when combined with multi-frame

processing can increase range performance by as much
as a factor of 2 under some circumstances (wavelength,
target temperature, etc.), but that if not used with multiframe processing the benefit is much lower. It also explains
that NVThermIP is not capable of directly illustrating that
effect. Reference 2 also indicates that local area contrast
enhancement can, in fact, be modeled by NVThermIP and
explains the technique. When we apply that technique to
our DRS imagers we find a range improvement that varies
from 17% to 20%.

Practical Application
Recommendation for the practitioner: Do not factor in the
benefits of image processing when designing an installation
for a site. There is too much variability that cannot be
adequately predicted. Do, however, favor cameras with
image processing and plan to use it, because it provides
additional margin to the design and it yields images that are
significantly easier to interpret.

Summary and Recommendations Putting it All

Together
Throughout this paper we have attempted to be brutally
honest about understanding range capability. We have
addressed how range capability is specified, how it is
determined, and the limits of utility of the results. We have
also addressed how range capability is different when video
analytics and image processing are used. We have tried to
dispel the notion that the range numbers offered on product
data sheets (even ours) are definitive and are suitable for
use in designing site installations even though they are
clearly useful for product comparisons.
We have examined 5 relevant areas:
How analysis using NVThermIP and/or pixels-on-target
can be used to predict range performance for a human
observer looking at a display
How to compute a maximum range based on a minimum
required target image on the display
How to compute a maximum range consistent with the
use of video analytics
The impact of weather on range capability
The impact of image processing on range capability
For the first 3 items, we cast the range determination
problem in terms of a required number of resolvable pixels
at the range of the target in the plane of the target. We
chose to do it this way so we could standardize the way we
get to a range prediction number. The log-log graph shown
in Figure 7 is a single tool that allows us to take all 3 items
into consideration at the same time, as they relate to the
DRS camera portfolio. The graph (Figure 7) plots range on
the horizontal axis and the PPM metric on the vertical axis.
The diagonal lines each represent a different DRS camera
as characterized by its focal length and pixel pitch.
9

DRS Commercial Infrared Systems White Paper

Figure 7: Graph for determining range capability of DRS' 17 Um pixel pitch cameras as a function of PPM metric

The basic use of the graph is the following: To determine

the range prediction for a specific lens given a value for
PPM, select the PPM value on the vertical axis, follow it
horizontally to where it intersects the diagonal line for the
lens of interest, and then follow it vertically to the horizontal
axis to determine the range value.

Example
For a PPM value of 5 (indicated by the red line), and for
the camera having the 65 mm focal length lens (indicated
by the right-most dark blue diagonal line), follow down the
vertical green line and read 760 m where it crosses the
horizontal axis.
To illustrate an actual problem solution, consider the
following example.
Given:
A surveillance system needs to be laid out for a long
straight fence line around a moderately sensitive site
(such as an industrial facility, as opposed to a nuclear
materials depot, which would have much greater
sensitivity).
Human observers will watch displays to detect intruders,
which are taken to be humans on foot.
It is judged that intruder detection requires recognition
and that the confidence level (probability of correct task
accomplishment) needs to be at least 70%.
It is judged that weather in the form of moderate rain will
be a factor but only 5% of the time or less.

It is noted that at some future date (under a future

budget, perhaps) video analytics will be added to the
installation and that the analytics choice would require a
PPM metric of 8 pixels/m in the target plane.
It is judged that the minimum size of a person in the
display should be no smaller than the equivalent of 15
detector pixels to provide good human comfort with the
images.

Query
What camera or cameras are suitable from a range
performance perspective and what maximum ranges
should we adopt when laying out the site?

Solution
Since we are watching for human intruders, we
assume the target height is 1.8 m and the target critical
dimension is 0.95 m (see Equation 1).
Since the system involves human observers looking at
monitors, and that the task is taken to be recognition at
70% probability, we calculate the PPM metric thus:
The Johnson criterion for recognition at 70%
probability is 3.75 resolvable bar cycles (see table 2).
The PPM is calculated to be 7.9 pixels/m in the plane
of the target (see Equation 3).
Since the minimum displayed target height (for a person
standing erect) is 15 pixels, we compute the PPM metric
for this condition to be 8.3 pixels per meter in the plane
of the target (see Equation 5).
10

DRS Commercial Infrared Systems White Paper

Since we anticipate the use of video analytics requiring a
PPM metric of 8 pixels/m, we simply take note of the fact.
Now, we look at the three values of PPM metric and
select the largest, which is the most demanding. In this
case it is the 8.3 pixels/m required by the minimum target
display height. For this problem, they are all almost
the same, so we dont really need to re-rationalize the
requirements.
Using the graph in Figure 7, we determine the following
ranges (for the cameras with the 5 longest focal lengths):
Camera EFL

Range

65 mm camera

420 m

50 mm camera

325 m

35 mm camera

225 m

25 mm camera

160 m

21 mm camera

130 m

REFERENCES
1. Johnson, J., (1958), "Analysis of Image Forming
Systems", Proceedings of the Image Intensifier
Symposium, 6-7 October 1958: AD220-160, U.S. Army
Engineer Research and Development Lab, Fort Belvoir,
VA, 249-273.
2. Vollmerhausen, Richard, and Van Hodgkin, Range
performance benefit of contrast enhancement,
Proceedings of SPIE Vol. 6543 65430B, 2007.

Since the primary weather issue is moderate rain

occurring infrequently, and because the impact is
generally light at these ranges (see Table 8), and
because the sensitivity of the site is not at the highest
levels, we choose to ignore the rain impact.
We will, however, plan to take advantage of optional
image processing to give ourselves additional design
margin, but we will not change our range predictions.

About DRS Commercial Infrared Systems

DRS Technologies, Inc. is a leading supplier of integrated products, services and support to commercial partners and prime
contractors worldwide. The company is a wholly owned subsidiary of Finmeccanica SpA (FNC.MI), which employs approximately
69,000 people worldwide. DRS Commercial Infrared Systems line of advanced electro-optical sensor solutions is developed and
manufactured in Dallas, Tex. and Melbourne, Fla. to include the WatchMaster family of thermal surveillance systems.

WatchMaster IP Elite

Fixed-mount, fully sealed configuration

Resolutions of 320 x 240 and 640 x 480
IP and Analog connectivity
Low Power over Ethernet (PoE), 802.3af

Tamarisk Thermal

17 m pixel pitch technology

Resolutions of 320 x 240 and 640 x 480
Analog and digital video outputs
Integrated shutter for flat field correction

WatchMaster IP Ultra

Pan/Tilt, fully sealed configuration

Resolutions of 320 x 240 and 640 x 480
IP and Analog connectivity
60 per second Pan Speed

New Tamarisk Radiometric

Dynamic Range (-40C to +550C)

Temperature isotherms with up to 8 regions
and user defined color parameters
Image Contrast Enhancement (ICE)

To get more information on DRS Technologies and our family of thermal cameras and cores, call 855.230.2372 to speak to a DRS
representative today or visit www.drsinfrared.com.

This information is furnished in confidence with the understanding that it will not, without the permission of DRS Technologies, be reproduced, used, or disclosed for any purpose
other than the purpose for which it was furnished. All Rights Reserved.

Copyright 2014 DRS Technologies, Inc. | Approved for Release MR-2014-10-683

DRS Technologies | Commercial Infrared Systems

13532 N. Central Expwy, Dallas, TX 75243, Tel. 855-230-2372
[email protected] | www.drsinfrared.com
11

DRS Commercial Infrared Systems White Paper

Appendix A: Predicted Range Tables for DRS Products for Human and Vehicle Targets (Meters and Feet)

Table A1: Predicted Ranges for DRS Products for Human Target, listed in meters

Table A2: Predicted Ranges for DRS Products for Human Target, listed in feet

12

DRS Commercial Infrared Systems White Paper

Table A3: Predicted Ranges for DRS Products for Vehicle Target, listed in meters

Table A4: Predicted Ranges for DRS Products for Vehicle Target, listed in feet

13