Remote Imaging of Controlled Gas Releases using Active and
`Passive Infrared Imaging Systems
`
`Thomas J. Kulp*, Peter E. Powers, and Randall Kennedy
`Sandia National Laboratories
`
`Livermore, CA 945 51-0969
`
`ABSTRACT
`
`The results of field tests of an active backseatter absorption gas imaging (BAGI) system and a passive imager based on a
`Ga:Si infrared focal-plane array are presented. Both imagers allow real-time video imaging of gas emissions. The former system
`images gases through their attenuation of backscattered laser illumination; the latter images gases through temperature or emissivity
`differences. The results represent the first side-by-side comparison of an active and passive imager and the first BAGI field trial
`involving the imaging of plumes of controlled concentration and dimension.
`
`KEYWORDS: Infrared imaging, lidar, gas detection, remote sensing, laser radar
`
`1. INTRODUCTION
`
`Gas plume visualization using infrared (IR) imaging is a powerful remote-sensing technique for the detection and location of
`fugitive emission sources. By its very nature, imaging allows the instaneous probing of a wide area and presents the results in a
`format that is easily interpreted by the system operator. This can be contrasted with widely used “sniffer”-type point sensors that
`measure gas concentration in a single-point mode and whose output must be numerically-interpreted. Leak-detection using IR
`imaging offers the possibility to greatly accelerate and simplify gas leak detection in a variety of industrial applications that include
`leak detection in natural gas transmission and distributions systems, detection of emissions in complex piping networks at petroleum
`refineries, and leak detection for emergency response applications.
`
`In recent years, we have focused on the development of laser-active IR imaging methods for gas plume visualization. The
`general technique is referred to as backscatter absorption gas imaging (BAGI), and is carried out by illuminating a scene with IR
`radiation as it is being imaged by a real-time IR video imager. When an IR—absorbing gas is present in the imaged region a portion of
`the scene illumination is attenuated, creating a plume image. BAGI systems were originally developed using continuous-wave (cw)
`C02 laser sources coupled to flying—spot-scanned IR cameras to image gases absorbing in the 9- to ll—um range of the 1R1. That type
`of device was later extended for imaging natural gas (and other hydrocarbons) using a 3.39 tun IR helium—neon laserZ. A modified
`long-range version of the flying—spot imager was assembled3’4 to allow C02 laser—illuminated imaging at ranges (>300 m) sufficiently
`large for use in a low-flying aircraft. Most recently, a new type of imager was developed for extended-range imaging of natural gas
`(and other hydrocarbon) leaks in the 3—3.5 um wavelength range5. That imager employed a pulsed laser source whose beam was
`expanded to flood—illuminate the field-of—view of a gated InSb focal—plane array camera. The continously-tunable nature of the laser
`allowed the first demonstration of two—wavelength differential imaging to enhance gas plume visibility?
`
`IR imaging of gases via their absorption or emission of passive thermal radiation is also possible. A passive gas imager is
`created by limiting the spectral response of an IR camera to only those wavelengths that are absorbed by the gas to be detected. Gases
`become visible with this instrument if there is a sufficient temperature/emissivity difference between the plume and the background
`scene to cause a scene radiance change that is above the noise floor of the camera. Spectral selection has been accomplished in a
`variety of ways that include the use of discrete bandpass cold filters7, cold tunable etalon filters", and imaging-mode Michelson
`interferometers9. The signal-processing time required for the latter precludes real—time video output of the imagery.
`
`In this paper, field evaluations of an active and a passive gas imaging system are presented. The active system was the long-
`range scanned C02 laser-based imager. The passive imager was a bandpass-filtered Ga:Si focal-plane array (FPA). The test
`conditions differed from earlier field trials in that the observed plumes were generated by a well-characterized plume source and that
`
`*corresponding author
`
`SPlE Vol. 3061 O 0277-786X/97/$10.00
`
`269
`
`FLIR Systems, Inc.
`Exhibit 1012-00001
`
`

`

`In earlier tests, plumes were generated by a point leak
`this was the first side-by-side comparison of an active and passive imager.
`source that created plumes that were highly variable in concentration and shape. Tests described here were performed under conditions
`allowing both aspects of the imaged plume to be carefully controlled. This allowed determination of the sensitivity of both systems
`at a variety of standoff ranges.
`
`Long-range C02 laser-based imager
`
`2. IMAGER DESCRIPTIONS
`
`A description of the long-range BAGI imager has been presented in the past3a4 and will be summarized here. The imager
`operates in a flying-spot raster-scanned mode to achieve real-time laser illuminated imaging at ~50 wavelengths in the 9-11 um tuning
`range of the line-tunable C02 laser.
`Its optical layout is shown in Figure 1. Scanning of both the 18-W continuous-wave (cw) C02
`laser beam and the instantaneous field-of—view (IFOV) of the single—element detector are accomplished using a pair of
`galvanometrically-driven mirrors. The scan rates of each mirror (3933 Hz horizontal; 60 Hz vertical) are compatible with interlaced
`real-time video rate operation. The scanned transmit and return optical paths are carefully isolated from each other to prevent crosstalk
`and to allow the insertion of a refractive ZnSe telescope in each path. The telescopic expansion of the paths reduces the scanned field
`of view to 3.6 x 2.6 degrees and increases the collection aperture by a factor of 25 over that of earlier scanners. The imager was
`developed to allow imaging at standoff ranges (>300 m) compatible with operation from a low-flying airborne platform (e.g., a
`helicopter).
`
`Ga:Sif0cal-plane array
`
`The passive imager was a 128x128 Ga:Si focal-plane array (FPA) that was obtained from Amber, Inc (Goleta, CA) and has a
`spectral response range of 3-18 um. Other specifications of the array are listed in Table 1. During operation, the camera was fitted
`with a cold-filter that passes a narrow band of wavelengths that overlap the absorption band of sulfur hexafluoride, as shown in
`Figure 2. The gas emission was imaged onto the FPA using an f/2 germanium lens.
`
`3. BAGI SENSITIVITY AND RANGE
`
`The sensitivity and range of an active gas imager has been described in the past“. This performance model has been
`modified recently to include speckle noise, laser amplitude noise, shot noise in the active photon return, and analog-to-digital
`conversion noise. The speckle noise power (Nsp) is calculated as
`
`
` —-— 42,
`N. =P
`L)
`A
`”dell
`A];
`
`1
`( )
`
`where PA is the active return power, XL is the laser wavelength, d is the receiver aperture diameter, and BL is the laser divergence. The
`laser intensity noise is assumed to be 5%, causing a noise power of 5% in the return signal. The shot noise power is given by
`
`N - (w)?
`
`(2,
`
`where vL is the laser frequency, Pp is the passive thermal power collected, Af is the receiver electronic bandwidth, and Q is the
`detector quantum efficiency. Only the active shot noise is calculated because the passive shot noise is included in the noise of the
`detector. The analog-to-digital conversion noise power is
`
`P
`_E.A_
`N =4 0.5 _
`64
`32(
`AD
`)
`
`3
`()
`
`assuming a return signal spanned to the mean of the 64-bit analog-to-digital converter and a noise level of 0.5 bits.
`
`270
`
`FLIR Systems, Inc.
`Exhibit 1012-00002
`
`

`

`Laser beam ‘npu
`Position sensor 'npu
`
`Vertical scan mi'ror
`
`Telescope objective elemerls
`
`
`
`
`L FItered HgOdTe deieclor
`
`Telescope objective elements
`
`Figure l - Optical layout of the long-range BAGI imager.
`
`750
`
`800
`
`850
`
`900
`
`950
`
`1000
`
`1050
`
`1100
`
`ISF6 in nitrogen
`5 cm paihiength
`
`IIII I
`
`\
`
`Absorbance
`
`80
`
`O) O
`
`A O
`
`I\.) O
`
`
`
`
`
`(%)uogssgwsueu19mg
`
`Filter bandpass @ 20K
`
`/
`
`IIII
`
`l|IlIlI
`
`l
`
`750
`
`800
`
`850
`
`1000
`950
`900
`Energy (wavenumbers)
`
`1050
`
`1100
`
`Figure 2 - Plot of the passive imager bandpass at 20K and the sulfur hexafluoride absorption band.
`
`Table 1 — Ga:Si FPA Snecifications
`
`
`
`
`
`
`D*
`
`
`
`
`
`
`
`271
`
`FLIR Systems, Inc.
`Exhibit 1012-00003
`
`

`

`Figure 3 shows plots of the predicted total noise power and the contributions to the total from each source as a function of
`range. The first plot excludes speckle and the second plot includes it. This distinction is made because two regimes of speckle noise
`exist that affect BAGI imaging differently. They are termed correlated and uncorrelated speckle. Correlated speckle appears as a
`temporally-invarient spatial intensity modulation in an image Uncorrelated speckle appears as a randomly-varying temporal
`modulation on each pixel.
`In the former, the return power at each pixel is fixed in time but the variation in return power among a
`group of pixels viewing a uniformly-reflecting object is given by (1).
`In the latter, the return power at each pixel varies randomly in
`time with a standard deviation dictated by (1). The correlated speckle is viewed as a source of fixed-pattern intensity noise, but not a
`temporal noise source. Thus, under correlated conditions, speckle is excluded from the noise term in BAGI calculations.
`
`Et.
`“3’
`g
`E
`2N
`.23
`U'
`I.”
`0
`.2o
`z
`
`—— Total
`-—-- Detector
`----- Laser
`- -- Active Shot
`—- - -ADC
`
`
`
`1 0 O
`
`2 O 0
`Range (m)
`
`3 O 0
`
`4 0 0
`
`z
`
`-- Total
`-—-- Detector
`----- Laser
`— - - Active Shot
`----- Speckle
`.
`
`EL
`
`.
`g
`o
`0.
`E
`2
`N
`.23
`o-u.I
`o
`.2O
`
`100
`
`200
`Range (m)
`
`300
`
`400
`
`Figure 3 - Calculated noise powers for the long—range BAGI imager. Total powers and the different contributions to the total are
`plotted. The top plot excludes speckle noise; the bottom includes it.
`
`272
`
`FLIR Systems, Inc.
`Exhibit 1012-00004
`
`

`

`Figure 4 shows the signal-to-noise ratio in the total backscatter signal predicted by the model. Three cases are shown that are
`derived assuming different noise contributions. The electro-optically limited curve shows the predicted signal-to-noise ratio expected
`in the video display electronic signal due to all electro-optic noise sources except speckle. .The visually limited curve is derived
`assuming1 that the eye is sensitive to contrast changes greater than or equal to an empirically—determined level of 1 part in 12 of the
`mid—greyscale. This introduces an effective noise of 1/12 of the mid-greyscale (or it limits the signal-to-noise to a maximum of 12)
`that is added to the electro—optic noise (speckle is still not included). The speckle—limited curve is determined by adding uncorrelated
`speckle noise to the other noise sources.
`In the graph, the two latter curves are shown to make an abrupt transition to the uncorrelated
`speckle limit at a range of ~250 m. This transition was determined arbitrarily and is shown because a transition similar to this was
`observed in many of the field measurements3~4. On several occasions an abrupt change from correlated to uncorrelated speckle was
`found to occur at a target range of 200-250 in. If the transition did not occur, the visually-limited curve would intersect the electro-
`optically-limited curve at some range.
`
`Figure 5 shows the noise-equivalent absorbance (NEA) that is derived from the signal-to-noise ratios for the conditions in
`Figure 4. Visually-limited resolution limits the NEA to near 0.1 at short ranges. Uncorrelated speckle would limit the NEA to a
`very low sensitivity (0.38). The arbitrary transition to uncorrelated speckle shown in the S/N plot is represented in the NEA plot.
`
`The maximum range of the imager has been defined1 as the point where the backscatter signal—to-noise ratio falls to a level at
`which the NEA just equals the visually-detectable threshold (i.e., where S/N=12). The model is used to calculate the laser power
`required to reach this point as a function of range. This is shown in Figure 6 for the long-range imager configuration. A curve is
`plotted for minimum and maximum zoom because the time-of—flight misalignment of the laser spot and the IFOV has not been
`compensated for in this imager. At mid-zoom, the model predicts a power requirement of about 14 W to achieve a range of 400 m.
`
`Electra-optically limited
`
`Ratio
`
`Visually limited
`
`Signal-to-Noise
`
`\\\—Uncorrelated
`
`\
`
`speckle onset
`
`0
`
`100
`
`200
`Range (m)
`
`300
`
`400
`
`Figure 4 — Calculated signal-to-noise ratio ofthe long-range BAGI imager.
`
`273
`
`FLIR Systems, Inc.
`Exhibit 1012-00005
`
`

`

`absorbance 0_3
`
`Visually limited
`
`if
`0 0.1
`.2
`
`Oz
`
`0.0
`
`o
`
`100
`
`200
`
`300
`
`400
`
`Range (m)
`
`Figure 5 - Calculated noise-equivalent absorption for the long—range BAGI imager.
`
`20
`
`’3‘ 15
`
`La
`
`) 3
`
`a? 10
`I—
`u)
`
`mm_
`
`1
`
`5
`
`0
`
`Maximum zoom
`
`0
`
`100
`
`200
`
`300
`
`Range (m)
`
`Figure 6 - Predicted laser power requirement versus range for the long-range BAGI imager.
`
`4. TEST CONDITIONS
`
`The field trials were carried out at the Remote Sensor Test Range (RSTR), located at the Department of Energy’s Nevada
`Test Site. A central feature of the RSTR is a large wind tunnel (see Figure 7) that has been adapted to create controlled plumes of
`volatile chemicals at its exit nozzle. The species to be emitted are introduced into the wind tunnel through an array of nozzles located
`near the air intake. The rate of chemical introduction and the flow rate through the tunnel are remotely-controlled to generate a 2-m-
`diameter region of laminar flow having the desired concentration of the diluted species at the exit point.
`
`274
`
`FLIR Systems, Inc.
`Exhibit 1012-00006
`
`0.4
`
`Speckle limited
`
`Uncorrelated
`
`speckle onset\
`
`,
`
`0.2
`
`E2m
`
`.2'3
`
`

`

`Anemometers
`
`
`I
`
`HH
`
`2-m plume diameter
`
`. Flow meter
`
`
`.L
`
`
`" Flow regulator
`IR absorption
`sensors
` Supply
`
`
`
`“‘.‘““‘adk‘m‘“
`
`
`
`Figure 7 — Diagram of the RSTR wind tunnel facility.
`
`During the tests, the imagers were mounted to view out of the rear of a mobile lidar vehicle, and were positioned to View the
`generated plume. The goal was to simulate a situation in which a plume would be viewed in a downward-looking fashion from an
`imager mounted in a helicopter. A 12-ft square panel was positioned behind the laminar region to serve as a backseattering surface.
`The panel was covered with two different grades of silicon carbide abrasive paper that have been determined to have reflectivities at
`10.6 um of 0.02-0.03 sr'l, respectively”). These reflectivities are comparable to that of an average terrestrial surface.
`
`Imaging was done at
`The imagers viewed releases of sulfur hexafluoride gas at concentrations between 1 and 140 ppm.
`standoff ranges between 90 and 360 m. As data were collected, the target temperature and theair temperature near the target were
`logged with thermocouples. This allowed correlation between the air-target temperature differential and visibility of the gas plume.
`
`5. TEST RESULTS
`
`Three tests were performed during the three-day duration of the field trial. These were (1) BAGI sensitivity determination
`(90-m standoff range); (2) BAGI/passive performance comparison (90-m range); and (3) BAGI/passive range evaluation (90-360-m
`ranges).
`In all tests, imagery was recorded on SVHS tape. At this time, only visual analysis of the results has been completed.
`Figure 8 shows examples of BAGI images under conditions of no gas, 3-ppm of SF6, and 40-ppm of SF6. Figure 9 shows a plume
`generated by the passive imager at 40—ppm SF6.
`
`The BAGI imager was found to image SFG at concentrations as low as 1.3 ppm at the 90-m location. Detection of 0.7 ppm
`was attempted, but was not visible. Thus, the sensitivity limit is somewhere between 0.7 and 1.3 ppm. Visualization of the 1.3-
`ppm release required considerable manual adjustment of the video gain and offset to accentuate the plume contrast. At longer ranges,
`imaging was only attempted at concentration of 2 ppm and above. Those plumes were visible at all ranges attempted, up to 360 In.
`Visually—adequate backseatter returns from the target panel were detected at all ranges as well, although the signal was noticeably
`fading at the longest distance. The observed speckle pattern remained moderately correlated over all ranges at which tests were made.
`This may be attributed to the very low wind speeds that were encountered during all test periods and to the very rigid mounting of the
`target panels.
`
`During the BAGI performance comparison, the passive imager generated discemable plume imagery at concentrations down
`to 4.5 ppm. At other times, however, the gas was invisible at concentrations as high as 40 ppm. The variability is the result of
`changes in the relative temperature of the air and target. Figure 9 shows a plot of the air and target temperatures that were measured
`during one of the test days. There was a clear differential of ~6-8 degrees during the early part of the day; however, the temperatures
`equalized later. The figure also shows imagery of two 40-ppm releases made during those two distinct time periods. Range
`performance of the passive FPA was limited to less than 250-300 In by spatial resolution due to the inability of the camera to zoom.
`The camera is otherwise able to operate at essentially unlimited range because it does not have the R‘2 signal intensity dependence of
`an active sensor.
`
`275
`
`FLIR Systems, Inc.
`Exhibit 1012-00007
`
`

`

`
`
`Figure 8 — BAGI plume images taken at a range of 90 m at the RSTR wind tunnel facility. Top left image is taken with no gas release
`top right image is taken with a 3-ppm release; lower image is taken with a 40-ppm release.
`
`6. DISCUSSION AND CONCLUSIONS
`
`The results represent the first BAGI data collected on plumes of controlled geometry and concentration. Although a more
`quantitative image analysis of the data should be carried out, the qualitative visual interpretation provides some useful conclusions:
`
`(1) The implemented design of the long-range imager meets the design goals of providing adequate mean active return signals at
`ranges of 300 m or more.
`
`(2) The sensitivity of the imager was found to be less than that assumed in the past--a concentration range of 0.8-1.3 ppm (1.2-2.6
`ppm-m roundtrip) should cause a round-trip attenuation of between 14% and 28%. This was seen only with some difficulty and is
`significantly higher than the assumed sensitivity threshold of 8%.
`
`(3) Laser speckle did not present as great a temporal noise source as in the past”. A complete transition to turbulent speckle did not
`occur, even at ranges as great as 360 m. This may be attributed, in part, to low winds and to more stable mounting of the target
`panels. The fixed-pattern speckle noise does, however, limit the dynamic range of the imager and the ability to accentuate weak
`absorptions. Partial speckle decorrelation may have caused the reduction of sensitivity from the predicted NEA of ~0.1 (Figure 5)
`to the observed NEA of 0.15-0.28.
`
`(4) The passive imager exhibited significant variations in its performance. Sensitivities about a factor of 2 worse than the BAGI were
`observed at moderate (5-7 C) air-target temperature differences. At lower differences the signal deteriorated substantially. These
`observations are relavant to an imager of this type--an imager using higher resolution spectral discrimination or differential
`processing may exhibit better sensitivity.
`
`276
`
`FLIR Systems, Inc.
`Exhibit 1012-00008
`
`

`

`fiassive imagery, 43 ppm §F5
`
`m ()1
`
`M C}
`
`
`
`Temperature(C)
`
`
`
`
`“is
`
`{)1
`
`“a
`
`(2)
`
`i8 M 22
`
`i3 $4
`
`“$5
`
`is
`
`Time at Say {art
`
`Figure 9 - Passive images collected during 40-ppm SF6 releases at two different times during the day. The target and air temperature
`during the day are plotted below. Arrows indicate the times at which the measurements were made.
`
`At this time, quantitative analysis of the images will allow further refinement of the BAGI sensitivity and explanation of the
`discrepency noted in (2). More detailed comparisons of the active and passive approaches should be made in the future. The latter is
`attractive because of its unlimited range and spectral bandwidth, and its simplicity and freedom from speckle noise.
`Its use must,
`however, be accompanied by the assumption that the required temperature and/or emissivity differences between the gas and
`background will always exist. This assumption is not necessary in active imaging.
`
`7. REFERENCES
`
`1. T.G. McRae and T]. Kulp, "Backscatter absorption gas imaging: a new technique for gas visualization" App. Opt. 32, 4037-4050
`(1993).
`
`2. T.G. McRae and LL. Altpeter, Proc. 1992 Int]. Gas Research (‘onf 2, 312 (April 1993).
`
`277
`
`FLIR Systems, Inc.
`Exhibit 1012-00009
`
`

`

`3. T.J. Kulp, R. Kennedy, D. Garvis, L. Seppala, D. Adomatis, and J. Stahovec, "Further advances in gas imaging: field testing of
`
`an extended-range gas imager,“ in Proceedings of the International Conference on Lasers ’90, D.G. Harris and J. Herbelin, eds.
`
`(Society for Optical and Quantum Electronics, McLean, Va., 1991), pp. 407-413.
`
`4. T.J. Kulp, R. Kennedy, M. Delong, and Darrel Garvis, “The development and testing of a backscatter absorption gas imaging
`
`system capable of imaging at a range of 300 m," in Applied Laser Radar Technology, Gary M. Kamerman and William E. Keicher,
`
`eds. Proc. Soc. Photo—Opt. Instrum. Eng. 1936, 204-212 (1993).
`
`5. T.J. Kulp, P.E. Powers, R. Kennedy, and U.-B. Goers, "The development of a pulsed backscatter absorption gas imaging system
`
`and its application to the visualization of natural gas leaks", in preparation for submission to Applied Optics.
`
`6. PE. Powers, T.J. Kulp, and R. Kennedy, "Differential absorption gas imaging", preceding paper in this proceedings.
`
`7. M. L. G. Althouse, "Chemical vapor detection and mapping with a multispectral forward-looking infrared (FLIR)," in Optical
`
`Instrumentation for Gas Emissions Monitoring and Atmospheric Measurements, J. Leonelli, D.K. Killinger, W. Vaughen, and
`
`MG. Yost eds. Proc. Soc. Photo-Opt. Instrum. Eng. 2366, 108-114 (1994).
`
`8. W.J. Marinelli and ED. Green, "Infrared imaging volatile organic carbon field sensor," in Optical Remote Sensing for
`
`Environmental and Process Monitoring, VIP-55, 245-254 (1995).
`
`9. CL. Bennett, M. R. Carter, and DJ. Fields, "Hyperspectral imaging in the infrared using LIFTIRS," in Optical Remote Sensing
`
`for Environmental and Process Monitoring, VIP—55, 267-275 (1995).
`
`10. H. Henshall and J. Cruickshank, "Reflectance characteristics of selected materials for reference targets for 10.6 um laser radars,"
`
`Appl. Opt. 27, 2748-2755 (1988).
`
`278
`
`FLIR Systems, Inc.
`Exhibit 1012-00010
`
`