2. What is the characteristic spatial trend in brightness
response within image frames, how do these trends vary
with different exposure parameters, and how well can
within image trends be balanced or normalized?
3. How can between image differences in radiometric bright-
ness response due to noise be minimized and due to signal
be maximized for multi-temporal
RSI
pairs, by proper
selection of exposure parameters?
We find no previous studies in the literature that evaluate the
radiometric characteristics of consumer grade
DSLR
cameras
for capturing aerial imagery in support of change detection
applications, nor for determining best practices for optimal
radiometric fidelity through optimization of camera con-
trol parameters. In addition, this study is novel in both its
ground-based image capture procedures that emulate airborne
imaging, including simulation of vertical path radiance in the
horizontal plane, and near replication of solar geometry, and
its damage change detection metrics, which were developed
for measuring change detection using a type of signal to noise
metric, and two quantitative change detection evaluation
methods based on pixel transects/profiles.
Methods
Experimental Variables and Image Capture Strategy
The methods chosen for this study were designed for col-
lection of imagery from a terrestrial, rather than an airborne
platform. The rationale is that for each image collection of 86
unique camera exposure setting combinations, the capture
location and view geometry must be as close to identical
as possible. Such a dataset would have been infeasible to
generate
vis a vis
an actual aerial image acquisition with the
appropriate experimental control.
In order to simulate aerial image collection, various col-
lection parameters were set to approximate illumination,
atmospheric path radiance, and transmittance conditions.
The sites chosen for these image collections (Figures 1 and
2) have urban scene composition, and an image capture
azimuth nearly parallel to the solar azimuth, so as to create
illumination and shadow conditions similar to those for aerial
imaging. Figure 3 is the camera and tripod set-up for station-
ary oblique imaging. Additionally, images were collected on
a clear and a hazy day, so the dataset contains two different
atmospheric conditions.
The experimental variables that pertain to the research
questions in this study are image exposure value and light me-
ter measurement zones for automatically estimating the appro-
priate exposure level for particular ambient light magnitudes,
aperture size, and
shutter speed for controlling the amount
of light reaching the detector,
ISO
(a carry over acronym from
film speed metrics of the International Standards Operation)
for controlling the sensitivity of the detector and white bal-
ance for inter-band radiometric consistency. Specific variables
pertain to each of the research questions, as not all of the vari-
ables are relevant for addressing each research question.
The first research question pertains to maximizing dynam-
ic range of a given image frame, while achieving high image
acuity, for which the image exposure value
(
EV
), light meter
measurement zones, relative aperture,
shutter speed, and
ISO
are the five relevant variables. The frame-specific
EV
is a major
determinant of dynamic range, as it sets the camera exposure
controls to yield an image that is not over- or under-exposed,
thus ensuring a maximized dynamic range in the resulting
image. The light meter measurement zone configuration may
have an impact on a maximized dynamic range as well, as
the “overall” and “center weighted” settings for determin-
ing which pixels of the detector array (and therefore, which
parts of the scene) are used for light metering, may differ in
their consistency for maximizing dynamic range. Slow
shutter
speeds can create image blur in actual aerial image capture,
depending on the influence of apparent image motion (
AIM
).
AIM
was calculated by multiplying the speed of the platform
Figure 1. Oblique intensity image of La Jolla Business Park scene (San Diego, California), used for the light meter testing.
Specific image subset used for the
RMSD
testing is delineated by the black rectangle.
150
March 2018
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
