Retrieval of Spectral Reflectance of High
Resolution Multispectral Imagery Acquired with an
Autonomous Unmanned Aerial Vehicle: AggieAir
™
Bushra Zaman, Austin Jensen, Shannon R. Clemens, and Mac McKee
Abstract
This research presents a new semi-automatic model for
converting raw AggieAir™ footprints in visible and near-in-
frared (
NIR
) bands into reflectance images. AggieAir, a new
unmanned aerial vehicle (
UAV
) platform, is flown autono-
mously using pre-programmed flight plans at low altitudes to
limit atmospheric effects. The
UAV
acquires high-resolution,
multispectral images and has a flight duration of about 30
minutes. The sensors on board are twin cameras with dupli-
cate settings and automatic mode disabled. A white Barium
Sulfate (BaSO4) panel is used for reflectance calibration
and in situ irradiance measurements. The spatial resolu-
tion of the imagery is 25 cm; the radiometric resolution is
8-bit. The raw images are mosaicked and orthorectified and
the model converts their digital numbers (
DN
) to reflectance
values. Imagery, acquired around local solar noon over
wetlands on the Great Salt Lake, Utah, is used to illustrate
the results. The model generates high quality images and the
results are good. The reflectance values of vegetation in the
NIR
, Green and Red bands extracted at the test locations are
consistent. The image processing, reflectance calculations,
accuracy issues, with the proposed method are discussed.
Introduction
In the recent past, various unmanned aerial vehicle (
UAV
)
platforms equipped with a myriad of devices have been used
to gather data in different bands for an array of applications. It
is an area of remote sensing that has become very active, and
UAVs
are rapidly becoming the preferred platform for develop-
ment of remote sensing applications (Watts
et al.,
2012). Earth
orbiting satellites and manned aircraft remote sensors have the
advantage of covering large areas, but the high operating cost
of such instruments limits the availability of timely informa-
tion for specific areas of interest (Hakala
et al.
, 2010). Remote
sensing applications require more sustainable, affordable,
user-friendly systems which are compliant with various levels
of changes in technology.
and
(2012) state that
low altitude remote sensing platforms, or
UAVs
address most of
these issues, and can be a potential alternative to satellite imag-
ery given their low cost of operation, high spatial and temporal
resolution, and flexibility in image acquisition programming.
UAV
imagery provides the ability to quantify spatial patterns,
and is used in rangeland monitoring and mapping to quantify
patches of vegetation and soil not detectable with piloted air-
craft or satellite imagery (Laliberte and Rango, 2009).
UAV
data
have been extensively used in forest fire applications (Merino
et al.
, 2006, Ambrosia
et al.
, 2003), wetland management and
riparian applications (Zaman
et al.,
2011, Jensen
et al.
, 2011),
precision agriculture (Primicerio
et al.
, 2012), agricultural
decision support (Herwitz
et al
., 2004). Field reflectance data
from
UAV
platforms are also increasingly being used for image
classification and predictive models (Berni
et al.
, 2009). But
the accuracy issues related with conversion of the information
acquired by the sensors on board these
UAVs
into useful data
remains a widely discussed topic. The small
UAV
systems have
low payload capabilities and are commonly equipped with
lightweight, low-cost digital cameras, which may complicate
the image processing procedures. Additionally, the chemical
basis for making a filter used on these cameras is proprietary
and there is variation in filter spectral transmittances among
various digital cameras (Hunt
et al.
, 2010), which calls for a
specific radiometric and geometric calibration (Hruska
et al.,
2012) to produce reliable data. Several
UAV
imaging systems
require custom designed applications for photogrammetric
processing and creation of orthomosaics to handle the large
number of small-footprint images acquired by the
UAVs
with a
rather unstable platform (Du
et al.,
2008; Laliberte and Rango,
2008; Wilkinson
et al.
, 2009). This paper discusses processing
of data obtained from a new
UAV
system, AggieAir
™
, which
uses off-the-shelf Canon PowerShot SX100 cameras as sensors.
UAV
imagery has spectral information in the form of digital
numbers which have noise arising from changing view, illu-
mination geometry, and instrument errors. Huang
et al.
, 2002
demonstrate the necessity of converting
DN
to at-satellite reflec-
tance when atmospheric correction is not feasible.
DN
is a func-
tion not only of land-cover, but also of the sensor calibration,
solar zenith angle, sensor viewing angle, seasonally variable
Earth/Sun distance, and diurnally variable atmospheric con-
ditions (Slater, 1980). Exposure settings on the digital camera
are chosen based on overall light intensity, which varies over
time with changes in solar elevation, atmospheric transmit-
tance, and clouds (Gates, 2003). Consequently, it is desirable to
convert
DN
to reflectance values that correct for these changes
(Hunt
et al
., 2005). Surface reflectance value has become the
vital measurement required for most remote sensing models
(Moran
et al.
, 2001). Laliberte
et al.
(2011) state that a UAS-
based image acquisition system produces hundreds of very
high resolution small footprint images that require geometric
and radiometric corrections and subsequent mosaicking for
use in a Geographic Information System (
GIS
) and extraction
of meaningful data. Similarly Jensen
et al.
(2010) discuss the
necessity of calibration of
UAV
imagery and navigation sensors.
UWRL 247, Utah Water Research Laboratory, College of Engi-
neering. 1600 East Canyon Road, Logan, UT 84321
).
Photogrammetric Engineering & Remote Sensing
Vol. 80, No. 12, December 2014, pp. 1139–1150.
0099-1112/14/8012–1139
© 2014 American Society for Photogrammetry
and Remote Sensing
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
December 2014
1139
