ICH
ie images.
and in the
imera was
-A image
id used to
corded as
geometry
ages. The
and forest
effects on
base. The
asible for
ined with
and large
=
c
condition,
, and the
quality of
fect is not
mificantly
'erent sun-
ost severe
ident light
5).
correcting
er data.
g HyMap
pendicular
gating the
improved
been used
rious types
ulen et al.,
ind sea ice
clockwise
ition over
resource
oped for
| with a
mes were
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part Bl. Istanbul 2004
corrected for bidirectional effects using sampled image digital
numbers for known natural objects (e.g. alpine meadows and
deciduous forests). Multiple video frames produced different
measuring geometries. The correction was performed as a
function of the scattering angle between instantaneous
illumination and measuring directions. The same method was
also applied to reduce bidirectional effects in aerial CIR
imagery taken on deciduous forest (Pellikka et al., 2000).
The present study aims to find an applicable method interposing
field-based goniometry and satellite-based BRDF observations.
Goniometry is most capable for fine and tiny objects, such as
sand grains, snow, grass, and tree branches, while it is
practically impossible to measure a whole tree. The resolution
of satellite systems is often too coarse to distinguish fragmented
natural objects, but is adequate for large homogenous areas. For
instance in Finland, forests and farmlands are generally
fragmented into relatively small units and large homogenous
areas are a rarity. This may lead to a large number of mixed
pixels, reducing the applicability of low-resolution instruments
and emphasising the need for a high-resolution airborne
measuring system. Also, mosaicking aerial images for the
production of three-dimensional models requires techniques that
are scientifically well-proven.
The objective of the study was to acquire directionally defined
reflectance data to be used in modelling and correction of
bidirectional effects on images and for establishment of a BRDF
database. This paper presents the detailed calculation of
geometric quantities of surface bidirectional reflectance. The
accuracy assessment and first tests for the data are presented in
the last part of the paper.
2. METHODS
2.1 Test sites
Multiangular images were acquired from two different test sites
in southern Finland. The Sjókulla test site, approximately 1 km“
in area, consists mainly of agricultural targets, e.g. barley, wheat
and oat fields, meadows and fallow and the forests in the area
are mainly broad-leaved trees such as birch and aspen, although
there are also coniferous trees such as pine and spruce, with
pine being the dominant species. The Finnish Geodetic Institute
also has a photogrammetric test field used for calibration of
aerial cameras at the test site. The Kuckuberg test site is mainly
covered with forests, with some hay and oat fields. The forests
are largely old coniferous forests, although there are also some
harvested and young areas with deciduous trees and coniferous
saplings. The intensive area of the test site was somewhat larger
than that of the Sjókulla area. Both are roughly limited to the
intersection area of the intersecting flight strips. The terrain in
both areas is slightly contoured. The mean terrain height is
about 40 metres above sea level.
2.2 HRSC-A data
A digital HRSC-A stereo camera was used for multiangular
image data acquisition. An HRSC-A camera measures with nine
different angles (+18.9°, £15.9°, £12.7°, +3.4° and 0.08°) in the
flight direction, producing the same number of overlapping
image strips. Five spectral bands in visible and near infrared
regions are arranged by mounting specific filters for the CCD
line sensors. The remaining channels function in a panchromatic
band of 585 to 765 nanometres. The radiometric resolution of
the HRSC-A was originally 10 bits, which was reduced to eight
211
bits for the images. More details in table 1 and in (Wewel et al.,
2000.)
Focal length 175 mm
Field of view 37.8°x11.8°
Number of CCD sensor lines 9
CCD-sensors per line 5184 (active)
Size of sensor cell 7 um
Radiometric resolution 10 bit, reduced to 8 bit
Spectral bands:
Panchromatic 585-765 nm
Blue 395-485 nm
Green 485-575 nm
Red 730-770 nm
Near infrared 925-1015 nm
Table 1. Technical parameters of the HRSC-A.
The spatial resolution of the images used for sampling was
interpolated to 0.5 metres for CIR images and 0.3 metres for
panchromatic stereo bands. Images were projected to UTM
zone 35 with a central meridian of 27°.
2.3 FGI’s HRSC-A flight mission and data acquisition
The HRSC-A flight mission was carried out during a three-day
period from 27 to 29 July 2001. Two different acquisition times
and four different flight directions were used for both test sites
to achieve representative angular range for the image-based
sampling of the bidirectional reflectance. First, three parallel
flight strips were flown with approximately 40% overlap. Then
the flight direction was adjusted approximately 30° clockwise
and three parallel lines were again acquired. Four different
flight directions were used in every session (see Figure 1). At
both test sites image acquisition was carried out twice to obtain
sufficient change in illumination due to the sun’s movement. As
a result, 24 image strips were attained for both test sites. In the
Kuckuberg area, the second flight headings of the strips were
redirected counter-clockwise.
Figure 1. Image acquisition principle.
The position and orientation data were provided by the GPS-
INS system. The position of the camera was recorded and
delivered at a second rate, whereas the attitude of the camera
was provided with a 10 Hz rate. The position data were at first
interpolated linearly to the same rate as that of the attitude data,
and the data were combined. Finally, the position and attitude
data were interpolated to a frequency of 100 Hz.