PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
May 2019
331
conventional photogrammetric processing
software, which would be prohibitively
expensive.
Some traditional photogrammetric software
cannot easily handle the excessive sensor-
orientation angles that are usually caused by
the wind effect on the lightweight body of a
small UAS. Most traditional photogrammetric
software does not fully utilize the principle of
Structure from Motion (SfM), which is efficient
in estimating the geometry of a block of
imagery obtained by drones. The SfM method
solves the camera positions and scene geometry
simultaneously, using a bundle adjustment
of highly redundant measurements based on
matching features in multiple overlapping
images acquired from different locations. In
contrast, the conventional photogrammetric
approach requires ideally consistent camera
properties and prior knowledge of camera
positions or a ground control network.
Photogrammetry
versus SfM
Although it is unintentional in some cases,
incidents over the last few years reveal
clear abuses to map-making standards
and practices when it comes to UAS. Many
of the UAS-operator-turned-mapper are
lacking basic knowledge and fundamentals of
photogrammetric mapping processes. For many
of them, their knowledge does not go beyond
what they learned from operating these new
breeds of black-box processing software. They
are often excited and fascinated by the term
“SfM” and some agencies have started calling
the team or the unit involved in processing
of drone imagery “SfM Department.” Some
drone-operator-turned-mappers claim that
photogrammetric practices and methods are no
longer needed when using the SfM approach.
Many of this new generation of mappers
neglect the fact that drones are just a new
platform or aircraft that just happens to be
unmanned. The imaging process is executed by
cameras that are based on the same principle
and design of the cameras that have been
operated for decades. These cameras are no
different than the traditional ones flown on
“These innovations
have positively
impacted our
industry, but they
also have resulted
in some negative
implications when it
comes to standards
of practice.”
manned aircraft. Standard practices should still
be applied. As a matter of fact, consumer-grade
cameras, such as those used on drones, inherit
more problems than high-end metric mapping
cameras. Consumer-grade cameras are low-cost,
nonmetric cameras with low-quality lenses and
shutters. The global positioning system (GPS)
used with these cameras is frequently based
on the less-accurate, single-frequency receiver.
Performing camera calibration and modeling
GPS timing and positioning error techniques
were perfected by the photogrammetric
community over the last few decades.
Moreover, bundle block adjustment, least
squares, and adjusting products to fit datum
and coordinates systems were not the
inventions of the SfM scientists. These are the
results of decades of hard work by the extensive
community of geodesists, photogrammetrists,
and mapping scientists. Software based on
SfM principles is great for constructing a
3D scene and producing a 3D product, but
without borrowing some photogrammetric
principles in camera self-calibration, geo-
located bundle block adjustment, least squares
errors distribution, and modeling of GPS/
inertial measurement unit (IMU) shift and
drift, the derived products will be anything but
positionally accurate.
I plead to the new community of mappers
to embrace the new reality of merging
photogrammetric techniques and practices with
innovations by the computer vision community
to produce scientifically sound mapping
products. Relying on push-button, black-box-
based processing alone is a risky undertaking,
since the mapping process and imaging sensors
are complicated. Users eventually will run
into situations where only proper knowledge of
photogrammetric and mapping principles will
be able to produce a quality product.
The mapping community offers tremendous
opportunities for these new mapping
practitioners through many American Society of
Photogrammetry and Remote Sensing (ASPRS)
workshops and certifications. Several schools
around the country also offer good online
classes on geospatial and photogrammetric
sciences. To be able to build accurate mapping
