Optimizing Radiometric Fidelity to Enhance Aerial
Image Change Detection Utilizing Digital Single
Lens Reflex (DSLR) Cameras
Andrew D. Kerr and Douglas A. Stow
Abstract
Our objectives are to analyze the radiometric characteris-
tics and best practices for maximizing radiometric fidel-
ity of digital single lens reflex (
DSLR
) cameras for aerial
image-based change detection. Control settings, exposure
values, white balance, light metering,
ISO
, and lens aper-
ture are evaluated for several bi-temporal imagery datasets.
These variables are compared for their effects on dynamic
range, intra-frame brightness variation, acuity, temporal
consistency, and detectability of simulated cracks. Test-
ing was conducted from a terrestrial, rather than airborne
platform, due to the large number of images collected, and
to minimize inter-image misregistration. Results point to
exposure biases in the range of −0.7 or −0.3EV (i.e., slightly
less than the auto-exposure selected levels) being preferable
for change detection and noise minimization, by achiev-
ing a balance between full dynamic range and high acu-
ity.
DSLR
cameras exhibit high radiometric fidelity and can
effectively support low-cost aerial image-based change
detection, such as for post-hazard damage assessment.
Introduction
Consumer-oriented digital single lens reflex (
DSLR
) cameras of
ever increasing spatial resolution and quality are an economi-
cal and readily available sensor option for airborne remote
sensing. Increasing coverage per image frame and higher im-
age fidelity, along with low barriers to entry (i.e., affordability
and ease of use) inherent to
DSLR
cameras, make them viable
for many remote sensing applications. Consumer
DSLR
cam-
eras have features and incorporate processes that differ from
traditional aerial cameras, as they are designed with photom-
etry in mind, rather than high radiometric fidelity associated
with many aerial imaging sensors.
Photometry seeks to measure light, as closely as possible
to how it is perceived by the human eye, whereas radiometry
seeks to normalize or even measure the absolute spectral radi-
ance. Having photometric accuracy as the primary concern of
camera manufactures, has led to the development of several
photometric materials and processes, including specific lens
or sensor spectral coatings, and onboard image processing
steps, to achieve greater photometric accuracy, sometimes at
the expense of radiometric linearity or greater radiometric ac-
curacy (Lebourgeois
et al
., 2008).
A dearth of research articles exist that explore remote
sensing based on
DSLR
digital cameras (Clemens, 2015). These
studies have tended to focus on vegetation remote sensing
(Dean
et al
., 2000, Ahrends
et al
., 2008, Lebourgeois
et al
.,
2008, Richardson
et al
., 2009), change detection for wide area
aerial surveillance (Coulter and Stow, 2008), and generation
of color indices for soil identification (Levin
et al
., 2005).
In investigating these topics, past studies have outlined the
benefits of replicating solar ephemeris (Coulter
et al
., 2012,
Ahrends
et al
., 2008), specific shutter and aperture settings
(Ahrends
et al.
, 2008, Lebourgeois
et al
., 2008), using
RAW
files (Dean
et al
., 2000, Coulter
et al
., 2012, Ahrends
et al
.,
2008, Lebourgeois
et al
., 2008), vignetting abatement (Dean
et al
., 2000), and maintaining intra-frame
white balance
(
WB
)
consistency (Richardson
et al
., 2009, Levin
et al
., 2005).
Through this study we seek to identify and determine how
to compensate and account for, the photometric aspects of im-
age capture and postprocessing with
DSLR
cameras, to achieve
high radiometric fidelity within and between digital multi-
temporal images. The overall goal is to minimize the effects of
these factors on the radiometric consistency of multi-temporal
images, inter-frame brightness, and capture of images with
high acuity and dynamic range.
The applications contexts for conducting this study are
detecting post-hazard damage and monitoring changes in
urban infrastructure. The technical context is Repeat Station
Imaging (
RSI
), where image capture over time occurs at nearly
the identical station points in the sky, and image registration
and change detection are performed on a frame-by-frame basis
(Coulter
et al
., 2003; Stow
et al
., 2003). Due to the dynamic
nature of urban scenes, a challenge is to minimize noise
sources due to variations in illumination characteristics and
scene conditions and features, sensor variability, and appar-
ent image motion (
AIM
) motion blur, through the selection of
an appropriate shutter speed. The goal is to automate image
processing and analysis as much as possible, but with the ex-
pectation that a human analyst will make the final analysis of
whether damage or other land surface changes have occurred.
This study was conducted in such a way that the charac-
teristics of the lens used, such as the specific or various focal
length(s) and relative aperture(s), had minimal bearing on the
replicability and adaptability to other users, can be performed
without specialized equipment, and can be tailored to specific
collection parameters. The study design also should allow for
future camera models to be tested, provided that they have a
variable aperture lens, bayer array, and image sensor similar
to a
CMOS
sensor. We captured high quality images with vary-
ing exposure parameters using consumer grade
DSLR
cameras,
and the resultant image sets were utilized to empirically
address the following research questions, in the context of
RSI
-based change detection.
1. With what combination of exposure settings can the
dynamic range of image brightness values be maximized,
while achieving high image acuity?
Storm Hall 307B, Department of Geography, San Diego State
University, San Diego, CA 92182-4493 (
).
Photogrammetric Engineering & Remote Sensing
Vol. 84, No. 3, March 2018, pp. 149–158.
0099-1112/17/149–158
© 2018 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.84.3.149
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
March 2018
149
