We have developed a method for using hyperspectral (HS) data to identify and locate chemical materials on arbitrary surfaces using the materials’ reflection or emission spectra that makes no prior modeling assumptions about the presence of pure pixels or the statistics of the background clutter and sensor noise. To our knowledge, this is the first time that surface detection without dependence on background information has been achieved. There are three main components to the method: (1) an HS unmixing algorithm based on the alternating direction method of multipliers that is applied over local subsets of the imaging to resolve the HS data into a set of linearly independent spectral and spatial components; (2) the fitting of those unmixing spectra to a set of candidate template spectra; and (3) a support vector machine classifier for chemical detection, identification, and location. The algorithm is illustrated on HS data collected by a Telops Hyper-Cam infrared camera on data resulting from the deposition of chemical agent simulants on various surfaces.
Over the past two years we have developed a new approach for detecting and identifying the presence of liquid chemical contamination on surfaces using hyperspectral imaging data. This work requires an algorithm for unmixing the data to separate the liquid contamination component of the data from all other possible spectral effects, such as the illumination and reflectance spectra of the pure background. The contamination components from S and P polarized reflectance data are then used to estimate the complex refractive index. We retain the index estimates within spectral windows chosen for each of a set of candidate contaminant materials based on their optical extinction. Spectral estimates within those windows are characteristic of the liquid material, and can be passed on to an algorithm for chemical detection and identification. The resulting algorithm is insensitive to the composition of the surface material, and requires no prior measurements of the uncontaminated surface. In a series of field tests, data from the Telops Hyper-Cam sensor were used to develop and validate our approach. We discuss our hyperspectral unmixing and index estimation approaches, and show results from tests conducted at the Telops facility in Québec under a contract with the U.S. Army Edgewood Chemical Biological Center.
Raytheon and the U.S. Army have been developing laser remote chemical sensors for the last decade. This has included advanced transmitter and sensor development, field testing, and concepts for spacecraft, aircraft, unmanned aerial vehicles, ground-mobile transports, and fixed sites. The WILDCAT sensor utilizes a wavelength agile CO2 laser with output energy of 1 J/pulse at a repetition rate of 100 Hz for ranges of 40 km. The receiver is composed of a 60 cm dia. telescope and HgCdTe detector, integrated into a gimbal system with full hemispherical scan. Algorithms allow for real-time data processing and concentration map display. An all solid state sensor breadboard has also been developed that is capable of 300 (mu) J output at 8-12 micrometers and 300 Hz repetition rate. The system is based on a Nd:YAG pump slab laser and two-stage, angle-tuned, optical parametric oscillator wavelength shifter. The system provides for power efficiency, compactness, and light weight that are consistent with manportability. Anticipated horizontal range is 3 km on the ground and 5 km vertically. Analysis of a space-based, low earth orbit system shows that chemical and biological species detection can be performed effectively by sensors derived from the laser components developed under these programs.
The CO2 laser is well suited for detecting a number of chemicals in the 9.3 - 10.7 micrometers band. However, there are several important species that require emission at 8 - 8.5 micrometers , which is not available from this laser. This has led to the development of a CO2 wavelength shifting technique that allows for interrogation of the 8 micrometers region. A two-stage process using the non-linear crystal AgGaSe2 accomplishes the shifting. In the first stage, the 10.6 micrometers CO2 laser pump is doubled to 5.3 micrometers by second harmonic generation which then pumps the second stage, shifting the wavelength to 8.3 micrometers by optical parametric oscillation. Energy conversion efficiencies of 40 percent have been obtained for the first stage shift and 10 percent in the second stage.
The US Army Chemical Biological Center and Raytheon Electronic Systems are developing a lightweight, compact sensor, known as the Standoff Handheld Real-time Early Warning Detector (SHREWD), for detection of airborne chemicals at ranges of 3-5 km by differential absorption lidar for manportable applications and for vehicles where sensor size and weight are restricted. Engineering analysis shows that the final deployable sensor size and weight would be 0.9 cu gt and 35 lb, respectively. The fieldable breadboard sensor now under development in phase 1 of the program is composed of independent transmitter and receiver sections mounted on either side of a single, 20 in. By 24 in. Optical table held vertically on a tripod. The transmitter is composed of an air-cooled Nd:YAG pump laser and a robust, two-stage OPO that shifts the pump laser output to the 8-12 micrometers band. The pump laser emits 20 mJ pulses at a repetition rate of 300 Hz in a 1.2 time diffraction limited beam; and the OPO overall conversion efficiency is 1.2% resulting in an output pulse energy of 240 (mu) J. The sensor receiver is based on a 12 cm diameter, off-axis paraboloid mirror and cryo-engine-cooled HgCdTe detector. Data acquisition is performed by 8 bit, analog- digital converters with 0.5 ns resolution and data processing/display are performed in real time.
The WILDCAT sensor was developed to provide chemical detection and identification at large standoff ranges on the order of 20 km for concentration-path length product measurements and 5 km for range-resolved measurements. The transmitter is a wavelength agile CO2 laser with output energy of 1 J/pulse at a repetition rate of 100 Hz. The laser wavelength can be shifted by a two-stage second harmonic generator and optical parametric oscillator to prove chemical absorption features outside the normal laser emission bands. The receiver is composed of a 60 cm dia. Cassegrain telescope and two-element HgCdTe detector that are integrated into a gimbal system for full hemispherical scanning. The laser/optical table and gimbal/telescope subsystems are connected by a rigid truss and all components are integrated into a transportable field test station. The data acquisition system is composed of 12 bit, 125 MHz analog-digital converters and a digital signal processor. Algorithms allow for real-time data processing and display of chemical concentration maps. All key transmitter and receiver components are capable of further development for compact, standalone sensors that can operate from fixed sites or mobile platforms, including aircraft, ships, and ground vehicles.
A lightweight, compact sensor breadboard demonstrator is being developed that will be capable of detecting chemical agents in the 8 - 12 micrometers band by DIAL for ranges on the order of 4 km. Engineering analysis shows that the overall sensor size and weight would be approximately 1 cu. ft. and 35 lbs., respectively. The sensor is composed of a 12.5 cm diameter, off-axis paraboloid receive telescope and a 300 Hz repetition rate solid-state laser transmitter. The transmitter is based on a diode-pumped, 1.06 micrometers laser with an output energy of 20 mJ and two cascaded OPO stages that shift the laser wavelength to the far IR bands. The first stage OPO is configured in an arrangement that shifts the laser output to beyond 2 micrometers . The second stage OPO completes the shift to the 8 - 12 micrometers band, giving an overall 1.06 (mu) $m yields 8 - 12 micrometers conversion efficiency of 2% and a sensor output power of 0.1 W.
The WILDCAT sensor is being developed to provide long range, laser standoff detection of chemical agent vapor and aerosol clouds by DIAL and range-resolved cloud mapping by DISC. The sensor is composed of two major subsections, including a telescope/gimbal assembly and a laser/beam diagnostics section that are held in alignment with a surrounding truss system and electrically-actuated, closed-loop mirrors. The entire assembly is integrated into a transportable, 30 ft long container that is field operable. The laser is a pulsed CO2 type with output energies of 1 J at a repetition rate of 100 Hz and it uses an agile grating to access approximately 60 lines over the 9.2 - 10.7 micrometers band, also at a 100 Hz rate. The telescope/gimbal subsection is composed of a 60 cm dia telescope mounted to a yoke gimbal which provides for full hemispherical scans. The sensor includes a data acquisition system composed of 12 bit, 30 MHz analog-digital converters and a digital signal processor that maintains a running average data stream for each range bin. Algorithms allow for real-time data processing and radar displays of chemical concentration.
A compact, wavelength agile laser and sensor have been developed for remote detection of chemicals. The laser is a computer- controlled, sealed TEA CO2 device with internal catalyst that operates at a maximum firing and wavelength shift rate of 200 Hz with 40% duty cycle. The wavelength shifter, composed of a fixed grating and galvanometer-mounted mirror, operates in repeating patterns with access to 55 lines in the CO2 spectrum in any order. Multi-mode output energy exceeds 125 mJ for all lines and the output pulse is composed of a 120 nsec wide gain-switched spike, followed by a 1.5 microsecond(s) ec tail. Environmental testing was successfully carried out under a 3 g shock, 2 g sine wave vibration, and a 0-40 degree(s)C temperature range. An operational lifetime exceeding 50 million shots was demonstrated with tests terminated by the operator, not a malfunction. The fully integrated laser weighs 100 pounds and requires only a source of 28 Vdc to operate. The laser was integrated with a sensor and in successful field trials the sensor noise was found to be 1-2%, measurement of atmospheric water vapor was within the accuracy of meteorological instruments; and measurement of concentration-path length product for chemicals in a vapor chamber approached the value expected for the noise figure.