IEEE
Geoscience and Remote Sensing
Society
Abbreviation: GRSS, S Code 29
|
IEEE
|
|
|
 |
HYPERSPECTRAL SENSORSJohn KerekesLincoln Laboratory Massachusetts Institute of Technology Lexington, Massachusetts I. IntroductionHyperspectral Sensors (also known as Imaging Spectrometers) are a systems technology whereby images of a scene are collected in tens to hundreds of narrow spectral bands nearly simultaneously. They represent the next step in the spectral dimension of the evolution of multispectral imaging radiometers currently represented by satellite sensors such as the Landsat Thematic Mapper which collects data in seven simultaneous bands. The term hyperspectral usually refers to an instrument whose spectral bands are constrained to the region of solar illumination, i.e., visible through shortwave infrared, and in the remote sensing context have an observing platform that is either airborne or spaceborne. The data collected are often termed an "image cube" where the two spatial dimensions are joined by the third spectral dimension.II. Technology StatusThe development of hyperspectral sensors has been enabled by two key component technologies. One is the spectral filtering technique by which the observed scene radiance is divided into narrow distinct bands. Examples of these techniques include dispersive gratings, prisms, multi-order etalons, interference filters, Michelson interferometers, and acoustic-optical tunable filters. The other key is detector array technology which allows multiple spatial and/or spectral samples through one- or two-dimensional arrays. These arrays are often made from silicon in the visible and near infrared and with indium antimonide, mercury cadmium telluride, or Schottky barrier platinum silicide in the shortwave infrared.Numerous airborne systems have been developed and flown over the past ten years. Recent review papers [1,2] list over twenty airborne imaging spectrometers that have flown since 1982. Improvements have come in signal-to-noise ratio, radiometric and spectral calibration accuracy, sensor size, swath width and number of spectral channels. A recent conference [3] included detailed descriptions of imaging spectrometers using several types of spectral selection techniques including gratings (AVIRIS, AISA, CASI, DAIS, MOS), prisms (HYDICE, HRIS), Fourier transform interferometers (SMIFTS, DASI) and liquid crystal tunable filters (LCTF). Another system recently developed by Hughes Aircraft Co. uses spectral separation filters deposited directly on the detector array to ease optical alignment (WIS). Each system has advantages and disadvantages depending on the platform and application. The first spaceborne hyperspectral sensor is expected to be the Hyperspectral Imager (HSI) on the Lewis satellite being developed by TRW for NASA with an expected 1996 launch. It will have 384 spectral bands covering 0.4 to 2.5 microns with 5-6 nm spectral resolution and 30 meter ground resolution across a 13 km swath. III. Application AreasApplications are being pursued in all areas of Earth science including land, water and atmospheric topics. Land applications include vegetation studies (species identification, plant stress, productivity, leaf water content, canopy chemistry), soil science (type mapping and erosion status), geology (mineral identification and mapping) and hydrology (snow grain size, liquid/solid water differentiation). Lake, river and ocean applications include biochemical studies (photoplankton mapping, activity), water quality (particulate and sediment mapping) and bathymetry. Atmospheric applications include parameter measurement (water vapor, ozone, and aerosols) and cloud characteristics (optical thickness, cirrus detection, particle size).Processing techniques generally identify the presence of materials through measurement of spectral absorption features. Often the hyperspectral data are post-processed to derive surface reflectance through the use of atmospheric radiative transfer models such as the Air Force Phillips Laboratory computer code MODTRAN. Various spectral matching or band ratioing techniques are then used to extract the presence and amount of the materials. Recent symposia [4,5] and a journal special issue [6] provide a sampling of the various potential applications and techniques. IV. User RequirementsWhen looking at natural objects spectral resolution around 10 nm in wavelength over the solar spectrum (0.4 to 2.5 microns) provides adequate resolution for detection of spectral features. Spatial resolution requirements for land use applications were found to be approximately 30 meters according to studies done in the 1970's in preparation for the Landsat Thematic Mapper. While greater spatial resolution would reduce the "spectral mixing" problem (each pixel has a contribution from many surface materials thereby mixing their spectra together), this problem would never completely disappear. Temporal update requirements are generally application specific, although weekly observations would satisfy most applications. Signal-to-noise requirements also vary with application and spectral region, but SNR's of approximately 100 to 1 for 5% reflectance targets yield very useful data.V. Recommendations for Further ActivityContinued data collection along with excellent supporting ground truth and the development of application algorithms are the highest priorities. The advent of HSI aboard the Lewis satellite is an exciting opportunity. This sensor, along with the numerous airborne sensors, will provide data that can be used for application development activities. Another focus should be ensuring wide access to the data and associated ground truth which have and are being collected. This would encourage algorithmic development. Technology activities should focus on developments that would lead to lighter and cheaper instruments such as lightweight optics and focal plane arrays with integrated spectral selection components. The development of future satellite hardware could possibly be deferred until a better appreciation of the utility of spaceborne data is gained. However, given the utility that airborne data have shown so far and the long lead time required for the procurement of operational satellite instruments, it is appropriate now to began development of a follow-on system.REFERENCES[1] Staenz, K., "A Decade of Imaging Spectrometry in Canada," Canadian Journal of Remote Sensing, Vol. 18, No. 4, October 1992, pp. 187-197.[2] Birk, R.J. and T.B. McCord, "Airborne Hyperspectral Sensor Systems," IEEE AES Systems Magazine, October 1994, pp. 26-33. [3] "Imaging Spectrometry of the Terrestrial Environment," edited by G. Vane, SPIE Vol. 1937, April 1993. [4] "Recent Advances in Remote Sensing and Hyperspectral Remote Sensing," edited by P.S. Chavez, Jr., C.M. Marino, and R.A. Schowengerdt, SPIE Vol. 2318, September 1994. [5] "Summaries of the Fifth Annual JPL Airborne Earth Science Workshop, Vol. 1. AVIRIS Workshop", edited by R.O. Green, NASA JPL, January, 1995. [6] Special Issue on Airborne Imaging Spectrometry, Remote Sensing of the Environment, Vol. 44, May/June 1993. |