Passive Thermal Imaging of Bullets in Flight

Austin A. Richards*a
David M. Risdall*a
a
Indigo Systems Corporation

ABSTRACT

An ultra-low-noise readout IC originally designed for low-background imaging when hybridized with
indium gallium arsenide (InGaAs) detectors has been combined with indium antimonide (InSb) detectors instead.
This novel focal plane array operates in the 3-5 micron waveband and is capable of imaging the very low
backgrounds encountered at extremely short exposure or integration times. Combining the FPA with specialized
support electronics that enable precision triggering has resulted in a commercially-available camera system that can
take stop-motion thermal images of explosions, supersonic bullets and other fast projectiles without the need for
rotating mirrors or other optomechanical assemblies that are required in a scanning or streak camera system1. The
camera system can be easily calibrated to measure the in-band radiance of these objects, as well as enabling
estimates of their surface temperature based on laboratory measurements of emissivity.

KEYWORDS: High-speed thermal imaging, InSb, bullet, thermography, radiometry, MWIR, explosion, ultra-short
integration time, aerothermal heating

1. INTRODUCTION

Recent developments in commercial infrared camera technology have made it possible to capture ultra-
high-speed thermal images using a snapshot-mode focal plane array (FPA) with integration times as low as 1
microsecond in a standard camera product. The sensor capable of this performance consists of two parts: indium
antimonide (InSb) photodiodes for high sensitivity in the MWIR band, and a special readout IC (ISC9809) that was
originally designed for low-background imaging using InGaAs detectors. The 9809 readout has a CTIA unit cell
and user-selectable on-chip gain. In the high gain mode of the 9809, the well capacity is 170,000 photoelectrons,
which is about 100 times smaller than typical well capacities of readout ICs used in InSb FPAs. To avoid saturating
the high-gain integration capacitor in the readout, this sensor needs to be operated at integration times around 20
microseconds or less when imaging scenes at terrestrial temperatures with a standard 3-5 micron lens. This is
approximately 100 times shorter than the 1.8 millisecond integration time we use when imaging terrestrial scenes
with a typical InSb camera whose focal plane array has much larger integration capacitors. At 1 microsecond
integration times, the camera is still able to discern objects that are only 10 degrees above ambient, such as a hand.

2. ULTRAFAST THERMAL CAMERA APPLICATIONS

One application for an ultra-high-speed infrared camera that immediately comes to mind is the
measurement of infrared emission from high-velocity ballistic objects. If the camera’s integration time is
sufficiently short in duration, even supersonic objects can have their motion stopped and their image captured with
very little motion-induced blurring during the exposure. If the object is hot enough, it is possible to acquire imagery
without actively illuminating the target with a bright light source, as one must do for high-speed imaging in the
visible spectrum. This fast IR camera system can go beyond simply imaging the object, however. Quantitative
information about the object’s in-band radiance and apparent temperature can be determined by analysis of the
resulting digital images if the camera is properly calibrated. Radiometric measurements of high-speed objects is
useful in various aerospace and military applications, including tracking ballistic-phase rockets, missiles, and
spacecraft in low-light conditions, determining the trajectory of cannon projectiles for active armor protection, and
automated counter-sniper systems that trace bullets back to their points of origin. These applications depend on
knowing the in-band radiance of the target, but other applications demand that the actual surface temperature be
measured. For example, supersonic and hypersonic objects may be subject to aerothermal heating that pushes
materials beyond their thermal limits, causing softening, melting or ablation, all of which can lead to shape changes

258 Thermosense XXVI, edited by Douglas D. Burleigh, K. Elliott Cramer,

G. Raymond Peacock, Proc. of SPIE Vol. 5405 (SPIE, Bellingham,
WA, 2004) · 0277-786X/04/$15 · doi: 10.1117/12.547406

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 07/05/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

 that result in flight instability2. It is also possible to use aerothermal heating measurements to check the validity of
computational fluid dynamics models of supersonic and hypersonic flow3. It should be noted that measuring the
temperature accurately is difficult unless the in-band emissivity of the target object is well-understood.
The most readily available supersonic object (other than the cracking tip of a bullwhip) is a rifle bullet,
which is the subject of the experiments described below. Rifle bullets are inexpensive and have very well-defined
trajectories. We used an acoustic triggering method to command the camera to take a picture when the bullet passed
through the camera’s field of view. This is a tried-and-true technique that locates the bullet along the axis of fire to
within a distance of several centimeters, making it possible to locate bullets in three dimensions and take close-up
thermal images where the bullet nearly fills the field of view of the camera.
Other noteworthy applications for ultra-high-speed IR imaging include the imaging of turbine blades in
land-based gas turbine power generation systems, airbag deployment, ejection seat performance, and heating effects
caused by explosives detonation. These applications will not be discussed here, but may be the subject of future
papers.
3. CAMERA SYSTEM CALIBRATION

The camera system used for these experiments is an Indigo Systems Phoenix InSb camera head with a 320
x 256 pixel InSb FPA and a bandpass of 3-5 microns (determined by a cold filter inside the dewar). The camera
head is connected to an Indigo Systems DAS (digital acquisition system) back end, which consists of a host PC
running Media Cybernetics Image Pro software, along with custom hardware and software to acquire the digital data
from the head and bring it into the Image Pro environment. The camera had its non-uniformities corrected using a
pair of Santa Barbara Infrared extended-area blackbodies set to temperatures of 100 C and 200 C, which generated
signals at approximately 10% and 40% of well capacity. All the power settings on the readout IC were turned up all
the way to improve performance at the desired 1 microsecond integration time, which is at the extreme lower
boundary of usable integration times for the readout IC.
We determined qualitatively that the detection threshold target temperature was ~35 degrees C at a 1
microsecond integration time, and that a blackbody set to 50 C or greater temperature was clearly visible against a
room-temperature background. This provided evidence that we might be able to passively image a bullet in flight,
provided that the radiance of some part of the bullet exceeded the radiance of a 50 C blackbody. In order to better
understand system performance, we measured NEdT, and found it to be 760 mK in the temperature range of 100 to
130 C. The NEdT was determined in the following manner: The temporal noise at 100 C was measured by
exposing the camera to a 100 C blackbody and taking a sequence of 64 frames using Image Pro software. The
standard deviation of each pixel across the 64 frame sequence was calculated, and the average taken. This value was
17 digital counts. The digital count value difference between the two blackbodies at 100 and 130 C was 673 counts,
giving a responsivity of 22.4 counts per degree C. The quotient of the temporal noise and the responsivity is the
NEdT. This result suggests that the accuracy of temperature measurement is on the order of a couple of degrees C
in the 100-130 C temperature range.

Figure 1. Camera calibration curve relating target radiance and camera signal

Proc. of SPIE Vol. 5405 259

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 07/05/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

 The Phoenix system was then calibrated in order to be able to measure radiance and temperature values
from targets. The calibration process involves measuring the camera’s response in digital count values to a
blackbody source at various known target radiances. We pointed the camera and lens combination at a cavity
blackbody capable of achieving temperatures up to 1000 C with ~5 C accuracy. We then measured the mean count
value within a region of interest centered on the aperture of the blackbody. Figure 1 is a plot of the calibration
blackbody radiance versus the mean digital count value produced by the blackbody in the region of interest. The
camera’s spectral response is assumed to have a top-hat profile, with a 3 micron cut-on and 5 micron cut-off, a good
assumption that greatly simplifies the analysis. The data show that the system has a linear response within a
radiance range that corresponds to blackbody temperatures between 70 and 200 C. Above and below that radiance
range, the calibration becomes non-linear because the radiance signal of the calibration target either goes into a non-
linear region of the FPA’s performance curve or else gets buried in the noise. The calibration was performed in
terms of radiance rather than temperature because we are looking at targets with emissivities much less than unity.
We therefore can calculate the bullet’s temperature by measuring its radiance and then calculating what temperature
a greybody with an equivalent emissivity would have to have in order to have the same radiance as the
measurement. This makes it far easier to adjust the emissivity to different values, a requirement for measuring the
temperature of the bullet over its varying topography (and emissivity), from the rough marks made by the rifle’s
barrel near the bullet’s waist to the smooth surface of the bullet’s nose.

High-powered rifle bullets move at speeds on the order of 800-900m per second and are heated by friction
within the rifle barrel, aerodynamic forces and the propellant charge that provides initial acceleration. Initially, we
were not certain if the heating of the bullet surface was sufficient to enable imaging at 1 microsecond integration
times. A quick search of the Internet did not reveal any information about how hot bullets get in flight. We
determined experimentally that the grooves cut into the side of a .30 caliber bullet by the rifle’s barrel have an
emissivity on the order of 0.3, or 30% of a perfect blackbody, while the smooth surfaces are about 10% emissive.
Using a blackbody calculator that numerically integrates flux under the Planck curve, it is easy to show that the
bullet grooves with 30% emissivity need to achieve a temperature of at least 93 C to have the same radiance as a 50
C blackbody and thus to be clearly visible above a cold background. The determination of the bullet groove
emissivity value is described in section 5 of this paper. In short, we estimated that if the bullet grooves had 90-100
C temperatures at the moment of the exposure, they should be quite visible to the camera system at 1 microsecond
integration times.

4. EXPERIMENTAL SETUP

The rifle was held in a rubber-jawed clamping mechanism so that the bullet would follow the same
trajectory shot-to-shot. A soldering iron run at low power (so as not to appear obtrusively bright in the image)
provided a spatial marker for the camera system; it can be seen in the lower left corner of Fig. 2a. Bullet images
were first acquired with a Janos Asio 25mm focal length lens designed for 3-5 micron imaging, giving the camera a
wide field of view in order to locate the bullets relative to the soldering iron tip. We varied the microphone position
relative to the muzzle to change the delay between the instant that the gun fires and the instant the image is acquired.
Moving the microphone away from the muzzle increases the time delay in the system and results in the bullet being
farther downrange relative to the soldering iron tip. The converse is true when moving the microphone closer to the
muzzle. We started with a microphone to muzzle distance of approximately 40 cm. Once the bullets appeared at the
desired point in space near the soldering iron tip, we moved the camera closer to the soldering iron tip and changed
over to a Janos Asio 100mm lens on a 1/2” lens extender. This allowed a close-up view of the bullet from a safe
distance. The horizontal field of view of the camera was approximately 5cm, or nearly twice the bullet’s length, and
the front of the camera lens was approximately 60cm from the bullet’s trajectory, ensuring that the bullet would not
accidentally strike the lens. The bullets were nearly 1 meter out of the muzzle when the camera triggered. This is
sufficient time for the bullet to emerge from the gun gas cloud that forms in front of the muzzle, and allows the
camera an unimpeded view. We had previously found that gun gas leaks around the bullet during the firing
sequence, which is not surprising given that the bullet does not form a gas-tight seal in this type of rifle barrel.
The Phoenix DAS can record images at 346 frames per second at the full 320 by 256 pixel array size. The
DAS can be externally triggered with a TTL signal level, which is generated by a special timing box that has a
microphone input. The muzzle blast wave hits the microphone and the timing box sends a falling-edge pulse to the
DAS system, which triggers the camera to acquire an image or sequence of images, with the first image acquired 7

260 Proc. of SPIE Vol. 5405

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 07/05/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

 microseconds after the trigger is asserted. In the case of the bullet close-up image, only the first frame is interesting,
as the bullet is several meters downrange by the time the next frame is acquired approximately 3 milliseconds later.

5. EXPERIMENTAL RESULTS

Figure 2a shows a close-up image of a .30 caliber rifle bullet that has traveled 1 meter out of the muzzle of
a Fabrique Nationale FAL rifle. The bullet is traveling at 840 meters/second (~1900 mph), yet travels only 0.84mm
during the 1 microsecond exposure, reducing the image blur to about 5 pixels. The image shows the aerodynamic
heating at the bullet’s nose, as well as frictional heating of the grooves cut into the bullet by the rifling in the gun’s
barrel. The nose heats up because it is a stagnation point – a location of high pressures and temperatures in a
supersonic flow. The glowing areas on the tail of the bullet appear to result from two effects. The bright spot on the
tail along the bullet’s centerline is caused by reflections of the muzzle flash off the bullet’s tail, and also from the
flash heating of the tail caused by the burning powder charge. Further experiments will attempt to suppress the
reflections of the muzzle flash from the bullet image by firing the bullet through a very thin sheet of metal that will
act as a light shield between the muzzle flash and point at which the bullet is imaged. In that scenario, only the very
small portion of the muzzle flash that shines through the bullet hole in the light shield can reflect off the bullet’s tail.
Figure 2b shows a reference image of an identical bullet imaged with a Vision Research Phantom v7.0 camera set to
operate with a 2 microsecond integration time. The Phantom is a high-performance silicon CMOS range camera
designed to image in the visible portion of the spectrum down to 1 microsecond exposure times. The orientation of
the bullets in the two images is identical – they both travel from left to right. The bright glow seen on the waist of
the image is a reflection of bright studio lights that were required to properly illuminate the bullet during the
exposure. Unlike the thermal image, the visible image required active illumination, since the bullet is not hot
enough to glow brightly in the visible region of the spectrum.

Figure 2a. Midwave infrared image of a .30 caliber bullet in flight with apparent temperatures

Proc. of SPIE Vol. 5405 261

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 07/05/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

 Figure 2b. Visible-light image of an identical .30 caliber bullet in flight

The emissivities of various points of interest on the fired bullet were determined by heating a sample bullet
on a temperature-controlled hot plate to various temperatures and measuring the emissivity using the CirPASS
emissivity module within Indigo Systems’ ThermaGRAM software. The image data from a thermographically
calibrated InSb camera is converted into temperature values assuming a user-selectable emissivity value. We
manually adjusted the emissivity value until the apparent temperature matched the actual surface temperature. Note
that the resulting emissivity values in the table below have a fairly high degree of uncertainty, as do the temperature
values obtained from the image data. Since the variations in radiance (and temperature) are significant within a
region of interest (such as a particular heated groove on the bullet surface) the radiance and temperature values are
given as mean values.

The measured radiance and temperature ranges for various locations on the bullets’ surface are given in
Table 1 below. The radiance values were determined by fitting a third-order polynomial to a portion of the data
plotted in Figure 1, then relating mean digital counts values within a region of interest to a radiance value via the
third-order polynomial expression. The temperature is determined by calculating what temperature a greybody with
an emissivity of 0.1 or 0.3 would need to be heated to in order to observe the measured radiance. As mentioned
earlier, the tail’s seemingly high temperature is mostly caused by a reflection of the very hot muzzle gas cloud off
the minor radius of the bullet’s tail.

Location In-Band In-Band Mean Radiance Estimated Mean

On Image Emissivity (photons/cm2/sec) Temperature
(Celsius)
Bullet Nose ~0.1 4 x 1016 170
Heated Grooves ~0.3 7 x 1017 320
18
Bullet Tail ~0.1 1 x 10 >500 (apparent)
Bullet Body ~0.1 <1016 <50
16
Soldering Iron ~0.3 3 x 10 130

262 Proc. of SPIE Vol. 5405

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 07/05/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

 6. SUMMARY
This paper has presented a brief description of a novel focal plane array for very high speed thermal
imaging in the 3-5 micron waveband. An application of particular interest is described in some detail: thermal
imaging of a rifle bullet in flight and radiometric measurements of the bullet’s surface, along with some estimates of
the temperature on various points of interest on the bullet surface.

REFERENCES

1. Stumpfel,C.R., "In-Flight Projectile Imaging by Infrared Emission/ Rotating Mirror Technique With
Temperature Measurement" ARL-TR-1148, Army Research Laboratory, Aberdeen Proving Ground,
Maryland, July 1996. (AD A313428)

2. Stumpfel, Charles R., "Imaging In-Flight Projectiles by Infrared Emission", Test Technology Symposium
XXIII, April 1997.

3. Guidos, B.J., and Sturek, W.B., "Computation of Hypersonic Nosetip Heat Transfer Rates for an M829Like
Projectile" ARLMR52, U.S. Army Ballistic Research Laboratory, Aberdeen Proving Ground, Maryland,
April 1993. (AD A263226)

Proc. of SPIE Vol. 5405 263

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 07/05/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx