|
Estimation of Canopy Reflectance around Mountainous Areausing
Satellite Remotely Sensed Imagery.AbstractHow to estimate the
reflectance around mountainous area and interpretthe canopy
information with fidelity is one important task in the
remotelysensed application around mountainous area. The aim of
this paper is todiscuss that how to use Landsat TM, Digital
Terrain Model( DTM) and measuredatmospheric transmittance to
estimate the multi-directional reflectancearound mountainous
area. The final goal is to use these results to improvethe
classification accuracy of mountainous area, and provide the
constraintinformation for correcting atmospheric radiance/
transmittance computationmodels in the future.By computing the
radiant difference of same ground classes in the
directlyincident solar area and the shadow area, The reflected
solar irradianceobserved by satellite-borne sensors can be
gotten. With the aid of DTM andtransmittance data, we can
compute the reflectance around the directlyincident solar area.
In this computation, the radiance values around shadowarea could
be replaced with the average radiance of all shadow pixels in
thesame high range. The reflectance of mountainous canopy is
dependent on threedirection factors, incident solar direction,
satellite observed direction andsurface normal direction. In
this study, we assume the contributions of bi-directional
reflectance function and topographic effect( slope and
azimuthdirection) to reflectance are indepent, and the
topographic effect for thesame pixels in different bands are
same. The bi-directional reflectanceproperty can be gotten from
the analysis of the plain canopy area. So, wecan eliminate the
bi-directional effect and topographic by deriving thepercent
reflectance in each band. The percent reflectance is calculated
bydividing the reflectance by the sum of all bands' reflectance.
Owing to thepercent reflectance only dependent on the canopy
class' spectrum property,it is more proper to the classification
work than the original(absolute)reflectance. This study adopted
the Northern Taiwan mountainous area, including Taoyuan, Hsintzu
and Miauli mountainous area, as our test area. We used Lowtran7
model to compute the atmospheric transmittance, which the input
meteorological range for Lowtran7 was estimated from the
waterbody pixels' counts in TM images. Finally, this study used
unsupervise classification to classify the computed reflectance.
The result showed that 11 classes were classified. Justify by
the aid of the reflectance values in TM4 band, 10 classes in
them were vegetables and the other was soil canopies.
|