Short URL

Hyperspectral Remote Sensing

 

Other Remote Sensing areas:

Loading, please wait...

Hyperspectral?

Rather than dividing the electromagnetic spectrum into just a few bands (e.g. near-infrared, red, green, blue) as in typical multispectral imaging, hyperspectral sensors are able to capture a larger number (even hundreds) of spectral bands.  Geometric correction techniques seek to repopulate image pixels with data values that correspond to reflectance from the precise x, y locations of those pixels.  It is easy to understand how important geometric correction is when viewing high spatial resolution imagery (e.g. that found in GoogleEarth or Microsoft Virtual Earth).  Conversely, the nominal spatial resolution of hyperspectral data is necessarily lower (e.g. 5 x 5 m when using a typical aircraft platform) in order to maintain an acceptable signal-to-noise ratio.  There are simply fewer available photons (in narrow hyperspectral bands) to interact with a sensor's detector elements.  While geometric correction is still just as important for hyperspectral imagery, atmospheric correction is a major research area for enabling two hyperspectral images to be comparable.  The goal is to convert all images to a reliable measure of percent reflectance so that signatures from around the world may be compared.  This allows a spectral library to be used in automatically determining what materials are being remotely sensed.  Detection of component percentages of materials within a hyperspectral image pixel are possible using spectral mixture analysis techniques.

CAST uses a handheld spectroradiometer and a 17 m articulating boom for experiments that require assessing hyperspectral signatures in situ.  The boom is particularly useful when studying the effects of spatial aggregation on spectroradiometer measurements (aggregation is a natural effect of raising a remote sensor platform to higher and higher elevations).  CAST is particularly interested in techniques for efficiently developing hyperspectral-based biophysical maps.  For example, Environmental Dynamics (ENDY), CAST, and Entomology researchers are currently seeking to model forest biophysical variables (e.g. biomass) in the Boston Mountains using orbital Hyperion hyperspectral information content in conjunction with samples of small footprint LIDAR-derived biophysical surface maps, in situ, and ancillary data (see above figure showing study area).  The goal is to develop forest biophysical monitoring techniques that can then be used efficiently with a minimal temporal set of orbital hyperspectral imagery, thus avoiding significant data processing expenses.  While LIDAR is at the forefront of forest biophysical remote sensing, future space-based hyperspectral imagery has some economic and data continuity advantages if it can be efficiently processed.