The AHI (Airborne Hyperspectral Imager) is a system for the detection of buried land mines from the air.  The system was designed to exploit a long wave IR observable associated with mine installation discovered by DARPA’s Hyperspectral Mine Detection (HMD) program.  The system is a helicopter-borne LWIR hyperspectral imager with real time on-board radiometric calibration and mine detection.

Hyperspectral Mine Detection Program

The DARPA Hyperspectral Mine Detection (HMD) program has been investigating, developing, and demonstrating a hyperspectral infrared capability for remote buried mine detection.  These mines are typically buried a few centimeters below the surface, in accordance with Army doctrine.  The primary hyperspectral infrared phenomenon that is being addressed is a spectral signature due to soil/sub-soil differences, allowing infrared detection of buried mines via the disturbed soil. Other phenomena that are currently being evaluated include temperature signature discrimination of targets.

Since late 1994, the program has collected extensive non-imaging and imaging data of buried mines and mine surrogates in the 0.4 to 14 micron wavelength region.  The most useful and robust discriminants occur in the 8 to 12 micron infrared region.  The data has been used to develop algorithms that have discriminated between undisturbed and disturbed soils, indicative of a buried mine.  Data has been taken over a wide range of soil types and locations to better define the utility of this technique.  Test results indicate that disturbances can be detected from days to months later, even after severe weathering has removed all visible clues (tests in process to fully quantify this).  A Phenomenology Report was issued which explains the spectral discriminants, and the extendibility of hyperspectral techniques to different geographical regions.
 
 

Table 1. Programmatic Requirements
Airborne
Real-time detection
Realistic demonstration
Path to an operational system
Parallel use as phenomenology data collection platform

Technical Requirements Derived from Phenomenology

Before the AHI sensor was designed, an extensive measurement program was undertaken to both determine the nature of the mine detection phenomena and to determine the method to detect mines.  The spectral signature of the soil disturbed in the process of mining proved to be the best mechanism for detecting buried mines using a hyperspectral approach.  This measurement program used point spectrometers, as well as existing imaging spectrometers from the visible to the thermal infrared.

From the point spectrometer studies, it was determined that the long wave infrared region from 8 to 12 micrometers provided the most robust phenomena, with the quartz reststrahlen feature at 9.2 micrometers the most important phenomena.  A study of the strength of this feature at many geographically and geologically diverse sites was the initial determining factor in setting the maximum allowable sensor noise level.  A sensor was needed that could spectrally resolve that feature and extend far enough into the longer wavelengths to get away from that feature.  The requirement was thus to cover the spectral region from 8.3 - 11.0 micrometers with a spectral resolution of 16-32 wavenumbers.
 
 

Table 2.  Phenomenology Driven Requirements
Spectral Range	8.3-11.0 microns
Spectral Resolution	16-32 wavenumbers
NESR	0.01 watt/m2-micron-sr
Pixels per target	30

The point spectrometer data and the imaging spectrometer data also provided data for detection algorithm studies.  These studies set the allowable sensor noise level to .01 watts/m2/sr/micrometer.  These studies also showed that the mine detection observable was not totally distinct from other false alarm sources in the scene.  Thus to detect the mines, multiple pixels were required on each disturbed soil target.  A minimum number of 30 pixels on target was set.

AHI System

AHI (Airborne Hyperspectral Imager) is the demonstration sensor built in response to the requirements listed in Tables 1 and 2.  AHI is an LWIR hyperspectral sensor with gyroscopic image stabilization and radiance calibration contained in an aerodynamic pod for external mount to a helicopter  (Figures 1 and 2).  The system includes a boresighted digital 3-color CCD linescan camera.  The on-board data collection/processing system includes a real-time data processor producing data calibrated to radiance and a near-real time algorithm processor producing mine detections.  The on-board data storage system is a RAID disk with 12 Gbyte capacity.  The system includes post-processed differential GPS.  The system also includes a ground data handling, archive and analysis system.  The differential GPS coordinates are

Figure 1.  University of Hawaii AHI sensor pod mounted on a commercial Bell JetRanger helicopter.
incorporated into the data by the ground data system.  A block diagram of the flight system is shown in Figure 3.
 

Figure 2.  Annotated cutaway of AHI sensor in flight pod.  Sensor system is vibration isolated from the housing and interior frame.
The sensor per se consists of four subsystems:  telescope, spectrograph, background suppressor and FPA and associated electronics.  The telescope is a 3 element diffraction limited transmission lens with 111 mm focal length and an 35 mm clear aperture.  The spectrograph is an uncooled commercial f/4 imaging spectrograph with gold coated optics.  For mine detection AHI uses an HMD nominal slit width of 225 microns corresponding to an along track (cross-slit) IFOV of 2.02 mrad and 125 nm spectral resolution.  Crosstrack angular resolution is .81 mrad  (Table of angular and spectral resolution for slits and binning modes).  We are able to use an uncooled spectrograph and still achieve high performance because of the AHI background suppressor which uses a cold Lyot stop and a cold linear variable filter to reject background.  The Lyot stop confines the FPA FOV to active low emissivity optics, and the linear variable filter bandlimits the radiance passed to each FPA detector to a 1% spectral bandpass (10 wavenumbers at 10 microns), which is higher resolution than the HMD nominal requirement of 16-32 wavenumbers.  These components are contained in a vacuum dewar 11 inches long and 6 inches in diameter which is cooled to 90 kelvins currently with a small supply of liquid nitrogen which will be replaced with a Magnavox split Stirling cryocooler.  The AHI uses a 256x256 element Rockwell TCM2250 HgCdTe focal plane array (FPA) mechanically cooled to 56K.  In addition to having high quantum efficiency and pixel yields, it has excellent residual nonuniformity after correction (0.08%), crucial to our application. 

Some of the system characteristics are shown in Table 3.
 

Table 3.  AHI System Characteristics
Spectral Range	7-11.5 ?m
Spectral Resolution	125nm (32 bands)
Angular Resolution	.8 by 2 mrad
Swath width	11.7 degrees (256 pixels)

The flight data processing system inside the helicopter is linked to the pod with a 200Mbit/s fiber optic data link.  The processor receives the raw data from the FPA and passes the raw data to a Xilinx field programmable gate array (FPGA) which spectrally bins the data  into a 256x32 data array.  This binning excludes pixels which have been previously defined to be bad” due to poor correctability.  The Xilinx passes the binned data to a Sharc digital signal processor (DSP) which applies a set of previously defined set of calibration coefficients to output data calibrated to radiance (calibration coefficients are derived from observations of the on-board calibration sources).  These data are passed to a second Sharc DSP which computes the principle components of the data.  The principal components are passed to a third Sharc DSP which applies the detection algorithm to the principle components.  The raw (binned) and calibrated data, and the principal components are passed via an output FIFO on a fourth Sharc to the Pentium PC and RAID array for storage.

Along with the hyperspectral data, output from a boresighted CCD linescan camera is also placed into the data stream.  The 3-band color CCD camera has twice the swath width of the LWIR hyperspectral data, twice the resolution, and is sampled at twice the rate (300Hz) to provide a very high quality color context video.  Both the optical and IR hyperspectral systems view the scene through a 3-axis (roll, pitch and yaw) gyroscopically stabilized mirror system to remove the effects of aircraft rotations.

The AHI on-board computer has custom software to control the sensor, control the digital signal processors, save data to disk and to provide a real time display.  The sensor control functions include control of the pod environmental shutter, the blackbodies, as well as all measurement modes.  A series of sensor functions is also continually monitored in real time.  The on board software controls the functions of the DSP boards, including downloading programs and switching modes.  The real time calibrated and uncalibrated  be displayed in a series of waterfall plots.  These plots show the infrared or color CCD data as it is being collected.  Several plot functions are also available to determine the performance of the sensor as the data is collected including noise histograms, and radiance plots and histograms.

System Performance and Current Status

In recent field tests data were collected for mine detection experiments and for remote sensing applications.  The airborne SNR curve measured using the methods described in Winter and Lucey (1997, this volume) is shown in Figure 4.  While the results are still being evaluated, the HMD demonstration system called AHI achieved sufficient performance to allow mine detection.

Figure 4.  AHI system flight performance signal-to-noise plot from a 45?C blackbody.

An example AHI data product not involving mine detection is shown in Figure 5 which was derived from a data collection run over the Mustang Mountains near Ft. Huachuca, AZ.  The AHI data are easily able to distinguish different types of rocks based on the strength of the silicate reststrahlen feature and can be used to subdivide these rocks into subclasses.

Conclusions

The AHI system, still in the process of optimization, has demonstrated that the phenomenology discovered by the HMD program associated with the installation of mines can be detected and characterized with an airborne system.  This system demonstrates a clear path to an operational mine detection system through the use of primarily off-shelf components.  The system will also be extremely useful for further phenomenology studies for mine detection or other applications.

References

Winter, E. M. and P.G. Lucey, Requirements for calibration of focal plane arrays for imaging spectrometers, this volume.

Acknowledgments

This research was sponsored by DARPA under the direction of Dr. David J. Fields  and Craig Sayre at the U. S. Navy SPAWARSYSCEN, San Diego.

 Figure 5.  A mosaic of images derived from an AHI overflight of the Mustang Mountains near Sierra Vista, AZ. The bottommost image is a boresighted simultaneous CCD guide video image of the flightline with twice the swath width and resolution of the AHI IR data. The second image strip from the bottom is the average thermal emission of the flightline. The image is a ratio of 11.0 to 11.5 µm showing the distribution of the spectral feature due to the presence of carbonates.  A distinctive limestone unit is evident. The second image from the top is the strength of the 9 µm silicate emission feature. The third image from the top is the relative strength of the silicate feature near 9 and 10 µm, showing differences in silicate mineralogy. The fourth strip down is a composite of the limestone ratio (red), the silicate strength (green), and the silicate mineralogy (blue).