The Detection of Transient Optical Events at …...JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 20, NUMBER 2 (1999) 157 OPTICAL EVENTS AT NARROWBAND VISIBLE WAVELENGTHS visible spectrum

JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 20, NUMBER 2 (1999) 155

OPTICAL EVENTS AT NARROWBAND VISIBLE WAVELENGTHS

R

The Detection of Transient Optical Eventsat Narrowband Visible Wavelengths

Peter F. Bythrow and Douglas A. Oursler

emote sensing of optical transients represents a paradigmatic shift in approachto the detection and identification of anthropogenic terrestrial events. For the mostpart, short-lived optical events lasting from tens of milliseconds to a few seconds areeither undetectable or ignored by most current satellite remote sensing systems. Thework described in this article shows that by disregarding transient data, importantinformation about the event source is discarded. This oversight is significant, since thedesired information regarding the source may be gleaned within seconds of eventonset. These data give an observer the opportunity to rapidly evaluate and respond.Work to date has focused on high-speed, high-resolution imaging at narrowbandvisible wavelengths that simultaneously captures transient histories and suppressesbackground clutter from reflected sunlight. Experiments conducted at Cape Canaveral,Florida, have used a high-speed digital camera system and a narrow band-pass filtercentered at 589 nm. These experiments have resulted in characterization of theignition flash and initial plume signature from several large rocket boosters whilesuppressing daylight background clutter. (Keywords: Fraunhoffer filter, Opticaltransients, Remote sensing.)

INTRODUCTIONAs viewed from space, the Earth’s surface is dotted

by short-lived optical emission events. These eventsrange in intensity and duration from modest anthro-pogenic events such as rocket launches and the det-onation of high explosives to intense natural eventssuch as lightning strokes or the occasional meteorichigh-altitude explosion. With the notable exception oftwo NASA space sensors, Optical Transient Detectorand Lightning Imaging Sensor (both designed forlightning detection), and one Department of Defensesensor, the Earth remote sensing community has gen-erally overlooked optical events lasting from tens of

milliseconds to a few seconds. Because of sensor ormission design, those events are either undetectableor ignored by most current satellite remote sensingsystems.

As a result of discounting or overlooking transientdata, a wealth of information about an event is alsodiscarded. Most man-made systems operating in asteady state exhibit start-up transients characteristic ofthat system. Therefore, analysis of a transient recordedduring initialization can reveal specific characteristicsof the observed event. The desired information regard-ing the source can be gleaned from a transient within

156 JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 20, NUMBER 2 (1999)

P. F. BYTHROW AND D. A. OURSLER

seconds of event onset and may provide an observerwith the potential for rapid response.

We have focused our efforts in transient detectionon the visible portion of the spectrum between ≈400and 900 nm for four reasons. First, many signals ofinterest are robust at visible wavelengths. Second, thelow background afforded by night and by solar absorp-tion lines in daylight reduces clutter. Third, atmo-spheric transmission is high, and forward scattering oflight in water droplets through a multiple scatteringprocess is preferentially skewed toward the visible.Finally, advanced imaging technology of visible wave-length devices such as large-format, high-speed, frame-transfer focal plane arrays is readily available and israpidly improving owing to the development of thecommercial imaging market.

Our work to date has focused on conducting proof-of-concept experiments with a commercially availablehigh-speed, high-resolution imaging system operatingat narrowband sodium (Na) wavelength to simulta-neously capture transient histories and to suppressbackground clutter from reflected sunlight. We haveconducted three experiments of this type on threedifferent rocket boosters launched from Cape Canav-eral, Florida. A fourth experiment is planned with thecooperation of the Naval Surface Weapons Center atIndian Head, Virginia. These experiments used a high-speed digital camera system and an interference filtercentered at 589 nm. They have successfully character-ized the ignition flash and initial plume signature fromthe boosters while suppressing background clutter.These data provide the necessary inputs to models thatpredict the signature expected from an overhead sen-sor platform.

DAY/NIGHT DETECTIONAs shown in Fig. 1 (top), the Earth’s surface when

viewed from space at night in the visible and near-infrared portion of the spectrum is more than 6 ordersof magnitude dimmer than when seen at wavelengthslonger than about 3 µm. Simply put, at night it’s dark!Thus, from a signal-to-background considerationalone, the advantage of looking at visible wavelengthsat night for high-temperature optical transients isobvious. In this case, the clutter does not come fromthe entire Earth’s surface but from lightning, meteors,volcanoes, clouds, and man-made sources.

As opposed to darkness, one of the major stumblingblocks to using visible wavelength detectors whenviewing a sunlit scene is the tremendous backgroundsignal that must be overcome. This fundamentalsignal-to-background issue may be addressed in severalways, the simplest of which is to reduce the back-ground in a given pixel. For example, in daylight for

a conservative background albedo of ≈0.9, a broadbandpoint source radiating 120 kW in a 10-m2 pixel has abackground of ≈12 kW. Therefore, with a 10-m2 pixelthe signal-to-background ratio is about 10. Conversely,if the source remains constant but the pixel includes,for example, 100 m2, then the background signal inone pixel is increased by the area ratio to ≈120 kW andthe signal-to-background ratio is reduced to 1. If animager had a ground sample distance of a few meters,the signal-to-background ratio for 120-kW targetswould not be an issue.

For wide-area coverage at high framing rates, thisdegree of resolution is not practical. Fortunately, an-other means of addressing the issue is available throughnarrowband optical filtering. In general, this techniqueis most effective when the source signal emits in thedesired narrow wavelength band and the backgrounddoes not. For the visible portion of the spectrum,nature provides a set of narrow wavelength absorptionbands at which the solar emission is drastically re-duced. These absorption lines are called Fraunhofferlines, at which blackbody solar emission is reduced byover 90%. They are the result of the presence of spe-cific elements in the solar atmosphere.

The alkali metals such as sodium or potassium, withtheir univalent free electron in the outer shell, notablyproduce very deep absorption features. Figure 2 showsthe depth of absorption at various Fraunhoffer wave-lengths near 600 nm. The Na absorption line exhibitsa signal reduction of >98% with a line width of >0.05nm. Therefore, the use of a Na transmission filter witha narrow bandwidth should dramatically reduce theoverall solar background while retaining the Na signalfrom the source.

Since clouds of varying thickness often cover theEarth’s surface, cloud penetration by light of a specificwavelength is another consideration in the use of the

10–2

10–6

10–10

10–12

10–4

10–8

Bac

kgro

und

radi

ance Day

Night

EWS Seeker

1.00.80.60.40.2

0

Atm

osph

eric

tran

smitt

ance

1 2 3 4 5 6 7 8 9 10Wavelength (mm)

10 km

0 km

Figure 1. Earth’s day and night background radiance as seenfrom space (top) and atmospheric transmittance as seen fromaltitudes of 10 and 0 km (bottom). Absorption bands due to variousatmospheric constituents are seen as lowered transmittance.Bands marked as EWS (early warning system) and Seeker referto the wavelengths currently considered for missile early warningsensors and seekers.



visible spectrum for detection of ground-level opticaltransients. The scattering of light by water aerosolparticles is a function of wavelength and particle size.For particle sizes much smaller than the wavelength(l) of the source emissions such as air molecules,Rayleigh scattering dominates the process and thescattering angle is proportional to l4. Thus, longerwavelengths are least scattered. However, for particlesizes on the order of, or greater than, the wavelengthof the source of illumination, Mie scattering dominatesthe process. For water droplets of typical size in clouds,wavelengths in the visible spectrum are therefore pref-erentially forward scattered over longer wavelengths.Multiple scattering interactions with several particlesincrease the probability of visible-wavelength penetra-tion of clouds and atmospheric water vapor. Hence, asignificant portion of the source intensity may pene-trate through a cloud layer while retaining informationon time history and event location, thereby allowingan optical transient occurring below a cloud layer tobe detected from overhead. The clouds, in effect, actas a diffuse back-illuminated projection screen.

Recent experiments conducted by the Air ForceResearch Laboratory at Hanscom Air Force Base,Bedford, Massachusetts, have detected Na line emis-sions through overcast clouds. Low-pressure Na lamps,

with a total power emitted on the order of 100 W, wereused for these experiments. Measurements were madefrom an aircraft using an atomic line filter and a pho-tomultiplier tube detector.

GROUND-BASEDINSTRUMENTATION

The development of the concept of optical tran-sient detection by means of a narrowband visibleimaging system requires that both the sources of emis-sion and the background be characterized at the spe-cific wavelength selected for observation. For reasonsdiscussed earlier, our efforts have focused on the588.9973-nm line of Na. We have therefore acquireda Daystar temperature-controlled interference filtercentered at that wavelength. The APL Research andTechnology Development Center conducted charac-terization of the filter with a dye laser. The completeinstrumentation consists of the filter mated to a 7-in.Maksutov telescope manufactured by Meade Corpora-tion and a high-speed digital camera.

As shown in Fig. 3a, the camera and filter are at-tached to the telescope via a beamsplitter. A broad-band visible camera is attached to the other axis ofthe beamsplitter to aim and focus the telescope. In

Figure 2. Absorption features of the solar spectrum near 600 nm show Fraunhoffer absorption lines, in particular, the Na line at 589.6nm with >98% absorption.

1.000.980.96

0.900.920.94

1.000.980.96

0.900.920.94

589.0 589.2589.1 589.4589.3 589.9589.8589.7589.6589.5 590.1590.0

588.0 588.2588.1 588.4588.3 588.9588.8588.7588.6588.5 589.1589.0

Na 589.0 nm

1.000.980.96

0.900.920.94

591.0 591.2591.1 591.4591.3 591.9591.8591.7591.6591.5 592.1592.0

Wavelength (nm)

1.000.980.96

0.900.920.94

590.0 590.2590.1 590.4590.3 590.9590.8590.7590.6590.5 591.1591.0

1.000.800.60

00.200.40

1.000.800.60

00.200.40

1.000.800.60

00.200.40

1.000.800.60

00.200.40

Frac

tion

of tr

ansm

issi

on

Frac

tion

of tr

ansm

issi

on



addition, this configuration allows us to record theevent in the broadband visible spectrum. A low-pres-sure Na lamp is used to adjust the optical path lengthof the visible camera to match that of the filteredcamera. Therefore, objects in focus at the visible cam-era are also in focus at the filtered camera. The com-puter shown in Fig. 3b controls the digital camera.Data from the filtered camera are collected, stored inthe computer’s memory, and transferred to the harddrive. Broadband data are captured on tape. The sys-tem has recently been upgraded, and another digitalcamera with a cooled focal plane to reduce the detectornoise level has been added. The current digital camerawill replace the broadband visible camera and allow adirect comparison between broadband and narrow-band results. In addition, we have converted a labo-ratory spectrometer for field use and will include spec-tral measurements in future observations.

OBSERVATIONS/MEASUREMENTSThe optical transients produced by rocket boosters

and their detection from space are of significant inter-est to several government agencies. Hence, we selectedthree different large rocket boosters as candidates fordata-collection experiments. For an exceptionallylarge solid booster, we chose the launch of a Titan IV(T IV) rocket on 12 August 1998. The T IV hasroughly 3.4 million pounds of thrust from its solidstrap-on boosters. Although the vehicle exploded lessthan a minute into flight, our data collection was asuccess. (The central core hypergolic engine of theT IV is not ignited until well into flight. Therefore,launch signatures are entirely from the solid engines.)For a large medium-class solid booster, we chose theinaugural launch of the Delta III rocket on 26 August1998. Although this booster has a central core fueledwith rocket propellant-1 (RP-1), most of the initial≈800,000 lb of thrust is from the solid strap-on boost-ers. Our third experiment was the observation of theall-liquid Atlas 2A rocket with ≈460,000 lb of thrustfueled with RP-1 and liquid oxygen. Observation ofthe T IV was made from a range of 11 nmi, whereasfor the Delta III and Atlas 2A rockets the ranges were4 and 6 nmi, respectively. Figure 4 shows the Delta IIIlaunch pad from the experimental observation site atCape Canaveral. The flame ducts that direct muchof the energy away from this site extend away fromthe pad, and beach foliage obscures the very base ofthe pad.

Daylight observations of the launch pad and boosterwere made through the Na filter. These data werecompared with those taken when the telescope aper-ture cover was in place, and no discernible differencewas detected between the Na background and no

(a)

(b)

Figure 3. Ground-based instrumentation. (a) Setup showingtelescope, beamsplitter, Daystar Na filter, and SMD digital cam-era. The visible color camera is mounted at right angles to thetelescope. (b) Computer and monitor setup as configured for thelaunch of a Titan IV at Cape Canaveral.

Figure 4. A view of the Delta launch pads from the experimentalobservation site at Cape Canaveral. The Delta III is on the rightlaunch pad.



aperture illumination. Therefore, this sets an upperbound for the background signal level at equal or lessthan the instrument thermal and read offset levels.

The Delta III launch occurred in darkness at ≈2130EST on 26 August 1998. High-resolution imaging datawere collected at 7.5 frames per second for the first8 s of boost. Figure 5 is an excerpt from these datashowing six panels in both the narrowband (589 nm)and broadband. In each panel, the development of theplume is separated by approximately 0.5 s. The firstthree panels show the booster ignition sequence andplume development before the booster left the pad.From our horizontal vantage point, smoke obscurationof the plume is obvious. The data showed a bright,optically thick, resolved plume at the Na wavelength.The intensity increased with time from ignition andthen decreased as the plume left the imager’s field ofview. Despite the smoke obscuration, the measuredpeak scene intensity projected from a conical Lam-bertian surface was roughly 3 kW at 589 nm. Theplume length, radius, and cone angle were calculatedfrom the observations, as was the booster acceleration.From these data the total plume surface area wasdetermined as ≈850 m2, resulting in an average plumeintensity of ≈30 W/m2.

Another launch observation of some significancewas that of the Atlas 2A from pad 36 at Cape Canav-eral on 21 October 1998. The plume from Atlas 2Ais shown in Fig. 6 in both broadband and narrowbandemissions. Unlike the T IV and the Delta III, theAtlas 2A has no solid motors. Its three engines pro-duce ≈460,000 lb of thrust, and it is fueled with RP-1 (kerosene), with liquid oxygen as the oxidizer.

The key data collected in each of these launchobservations are the intensity versus time profiles.These data contribute to what would be detected as anoptical transient from a space or airborne sensor plat-form. Figure 7 shows that the observed intensities scalesomewhat as thrust for the solid boosters. The T IV hasroughly 4 times the thrust as the Delta III and isroughly 4 times as bright. The Atlas 2A has about halfas much thrust as the Delta III, yet is nearly as brightat 589 nm. We are examining reasons for these behav-iors and suspect that they may be attributed to adifference in fuel, contaminants, and oxidizer. Finally,each booster has a significant difference in intensityversus time profile that may be attributed to the dif-ference in thrust-to-weight ratios.

MODEL PREDICTIONSWhen observed from overhead, such as from an

Earth-orbiting spacecraft, the intrinsic brightness ofoptical transients produced by rocket boosters appearssignificantly amplified over the integrated plume

Figure 5. Launch sequence of the Delta III booster from CapeCanaveral on 26 August 1998, showing both broadband and Naemissions at 589 nm.

Broadband visible NaTime history



Figure 6. Launch of the Atlas 2A on 21 October 1998 showingboth broadband emissions on the left and Na emissions on theright at 589 nm. It is significant that the Atlas 2A is fueled with RP-1 and liquid oxygen.

Time history oflaunch intensity

(side observation)Atlas 2A (20 Oct)Delta III (26 Aug)Titan IV (12 Aug)

12000

8000

4000

0

Pow

er (

W)

at 5

89 n

m

–4 0 4 8 12

Time (s)

Figure 7. Time history profiles (1998) of intensity in watts at 589nm for each of the three boosters observed. Note that the peakintensity of the Atlas 2A is near that of the Delta III; however, itstemporal variation is quite different.

Figure 8. The three-stage process by which we have modeledthe launch flash of a rocket booster. Stage 1 is expansion of theplume at Vs the gas expansion velocity ≈330 m/s, stage 2 iscontraction at the velocity of the rocket, and stage 3 is the plumeexpansion with altitude (terms not defined in text are B = coneangle of booster plume, c = function of density and altitude,f = function of pressure and altitude, IL = intensity at any plumelength, and I Lmax = intensity at maximum plume length).

Broadband visible Na

Time history

emissions as viewed from right angles. This effect ispredominantly seen during the period when the rock-et’s plume is interacting with the ground.

We have applied a simple geometric model to themeasured data from our three booster observations toyield the signal expected when viewed from overhead.In this model, the plume is assumed to remain opticallythick and its surface irradiance to remain constantwhile it is within a few plume lengths of the ground.The model we use is the three-stage process depictedin Fig. 8. Stage 1 assumes a plume irradiance Ip andan expansion velocity Vs. The plume expands radiallyat a velocity Vs from rmin, the minimum plume radius,until it reaches a radial distance rmax roughly equal tothe length of the plume Lp, where rVs

2= Pa, thecompressed atmospheric pressure. From overhead, thelaunch flash is a result of a rapidly expanding circle

with a surface irradiance Ip. Once the plume hasreached its maximum expansion, it begins to collapseto rmin@sl (minimum radius of the plume at surfaceatmospheric pressure) at a rate determined by the ac-celeration of the rocket ar away from the launch pad,thereby decreasing the radius of the illuminating sur-face. In this model, once the plume is no longer in-teracting with the ground, the only light emissionscome from the projected surface area of the plume. Asthe rocket climbs to higher altitudes, this projectedarea increases as a function of rmin@h (radius of theplume at altitude h) owing to reduced ambient pres-sure, but is bounded by the optical thickness of theemitting surface area. Results shown in Fig. 9 weregenerated when we applied this crude model to thedata collected from each observed launch.

The results produced by this model show amplifi-cation of signal due to ground interaction and the rapidinitial rise of the signal with a fall in intensity to anearly constant level within a few seconds. Thesemodel results are comparable to observations madefrom space with a broadband visible sensor. Each plotin Fig. 9 is a function of booster fuel, thrust, and thrust-to-weight ratio. A more comprehensive model is likelyto reproduce engine characteristics and separate fuelcontributions from purely thrust factors.

SUMMARYThe work described in this article demonstrates the

viability of daytime background suppression using anarrowband optical filter centered at 589 nm andevent identification from a unique intensity temporal

rmin = rp rmax = Lp rmin = r(p@sl) r = f(p@h)

IL = Ip 3 p(Vs 3 t )2

IL = ILmax – Ip 3 p(a 3 t2)2

IL = Ip 3 p {tan B/e–c[(ar 3 t )2]}2

Projected plume surface area



and night detection and identification of these events.This phenomenon is significant, since by using a tran-sient event the desired information regarding thesource may be gleaned within seconds of event onset,and thereby may provide an observer with the poten-tial for rapid response.

Work to date has focused on high-speed, high-resolution imaging at narrowband visible wavelengthsthat simultaneously captures transient histories andsuppresses background clutter from reflected sunlight.Measurements conducted at Cape Canaveral, Florida,have successfully resulted in the characterization ofthe ignition flash and initial plume signature fromseveral large rocket boosters while suppressing day-light background clutter. Our initial modeling effortsare encouraging because we can reproduce the Na-linelarge-scale features observed from orbit in broadband.Our future work will extend to using an atomic linefilter and to the development of a prototype airbornesensor platform.

ACKNOWLEDGMENTS: The authors wish to acknowledge the APL Indepen-dent Research and Development Sensor Thrust Area for the support of theseefforts. We also wish to acknowledge the efforts of Dick Margoli and Donald Paulat APL’s NOTU facility in Florida for their work in supporting our observationsat Cape Canaveral.

Figure 9. Results of expected launch temporal signatures ob-served from overhead obtained from a simple geometric model.

THE AUTHORS

PETER F. BYTHROW joined APL’s Space Department in 1981. He is aPrincipal Staff physicist working on the development of new space-relatedcapabilities for the Department of Defense. Dr. Bythrow received his B.S. degreein physics from the University of Massachusetts at Lowell in 1970 (then LowellTechnological Institute). After serving as a pilot in the Air Force from 1970 to1975, he earned his M.S. and Ph.D. degrees in space physics from the Universityof Texas at Dallas in 1978 and 1980, respectively. For his first 10 years at APL,Dr. Bythrow conducted research on magnetospheric plasmas and electrodynam-ics. He was a co-investigator on NASA and DoD space plasma and magneticfield missions, notably HILAT, Polar BEAR, and UARS. In 1988, he beganwork with SDIO on missile launch detection and phenomenology, and was theproject scientist on the SDIO Delta 183 mission. Dr. Bythrow is currentlyengaged in developing new sensor technologies in surveillance, missile detec-tion, and tracking. His e-mail address is [email protected].

DOUGLAS A. OURSLER received a B.S. in electrical engineering andcomputer science from The Johns Hopkins University in 1987 and an M.S. inelectrical engineering from the University of Virginia. He returned to Hopkinsto receive an M.S. and later a Ph.D. in materials science in 1996. As apost-doctoral fellow, he continued his research in microwave and electromag-netic techniques of nondestructive testing. Dr. Oursler co-founded a smallbusiness that developed a novel laser contouring system to map automotivebody panels for the automotive and steel industries. In 1997, he joined APL’sSenior Staff in the Space Department. Since then, he has been involved withthe testing of the all-polymer and hybrid batteries, the development of theXylophone Bar Magnetometer for space applications, and the investigation ofoptical methods for missile launch point detection. His e-mail address [email protected].

modulation function. We have shown that the remotesensing of optical transients generated by rocket boost-ers at this wavelength may be practical for both day

1 3 105

Inte

nsity

(W

/sr)

0

1 3 104

1 3 102

1 3 103

2 4 6 8 10 12 14

Time from engine start (s)

Delta III

Titan IV

Atlas 2A

The Detection of Transient Optical Events at …...JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 20, NUMBER 2 (1999) 157 OPTICAL EVENTS AT NARROWBAND VISIBLE WAVELENGTHS visible spectrum

Documents