PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
December 2016
935
Bundle Adjustment of Spherical Images
Acquired with a Portable Panoramic Image
Mapping System (PPIMS)
Yi-Hsing Tseng, Yung-Chuan Chen, and Kuan-Ying Lin
Abstract
Thanks to the development of mobile mapping technologies,
close-range photogrammetry (
CRP
) has advanced to be an effi-
cient mapping method for a variety of applications. A compact
CRP
system equipped with multiple cameras and a
GPS
receiver
is one of those advanced portable mapping systems. A porta-
ble panoramic image mapping system (
PPIMS
) was specially
designed to capture panoramic images with eight cameras and
to obtain the position of image station with a
GPS
receiver. A
PPIMS
can be considered as a panoramic
CRP
system. The coor-
dinates of an object point can be determined by the intersec-
tion of panoramic image points. For the implementation, we
propose a new concept of photogrammetry by using panoramic
images. Eight images captured by
PPIMS
forms a spherical
panorama image (
SPI
). Instead of using the original images,
PPIMS
SPIs
are then used for photogrammetric triangulation
and mapping. Under this circumstance, one
SPI
is formed for
each station, and it is associated with only one set of exterior
orientation (
EO
) parameters. Traditional collinearity equations
are not applicable to
SPI
triangulation and mapping. Therefore,
a novel bundle adjustment algorithm is proposed to solve
EO
of multi-station
SPIs
. Because
PPIMS
SPIs
are not ideal
SPIs
, a
correction scheme was also developed to correct the imperfect
geometry of
PPIMS
SPI
. Two test studies were performed for the
data collected at a campus test field of National Cheng Kung
University (
NCKU
) and at a historical site of Tainan. Both cases
demonstrate the feasibility of
SPI
bundle adjustment and apply-
ing corrections for
PPIMS
SPIs
necessary for effective for bundle
adjustment. Furthermore, the experiment’s results also confirm
that
SPIs
can replace original images for
PPIMS
triangulation.
Introduction
With the development of mobile mapping systems (
MMS
),
multiple sensors such as cameras,
GPS
receivers, inertial
measurement units (IMU) and a barometer are integrated in a
variety of platforms.
IMU/GPS
integration is especially useful
to achieve rapid, direct georeferencing (Chiang, 2004). An
MMS
typically uses a vehicle as its platform (El-Sheimy, 1996),
which is convenient for the data collection along roads. These
data such as image sequences and panoramic images can be
applied to the moving objects detection (Sun
et al
., 2008), 3D
road line extraction (Li, 2009), establishment of a street view
(Guo, 2010), traffic sign detection (Iam, 2011). However, many
areas are prohibited for a vehicle to enter, such as rugged
terrain, forested areas, heavily damaged disaster areas, and
crowed areas. Consequently, alternative systems are needed
for mapping those areas.
Motivated by the backpack
MMS
proposed by Ellum (2001),
we developed a portable
MMS
with panoramic imagery. This por-
table
MMS
is a specially designed platform equipped with eight
cameras and a
GPS
receiver. To capture the panoramic image of
the surroundings, cameras are mounted on a round platform
facing in eight outward directions. In addition, a
GPS
antenna
can be mounted at the top of the platform for positioning. This
system is called portable panoramic image mapping system
(
PPIMS
).
PPIMS
can capture eight images simultaneously with
GPS
po-
sitioning. When images are taken from multiple stations, a large
number of images are captured. These images can be used for
bundle adjustment to solve for the exterior orientation (
EO
) pa-
rameters of each image and coordinates of ground points (Tsai,
2012). However, a user frequently needs to handle a large num-
ber of images for visualization and measurement, and sometimes
could be confused with finding corresponding target image. This
difficulty has motivates us to develop a better system to handle
and measure a large amount of images captured with
PPIMS
.
First, we need to learn how to combine eight images into
a panoramic image. A number of methods have proposed for
panoramic image stitching. Levin
et al
. (2004) proposed an
approach to seamless image stitching in the gradient domain
rather than in the intensity domain. Brown and Lowe (2006)
proposed an approach using invariant features. And Bhat
et
al
. (2013) proposed an approach using template matching. Al-
though the present approaches of image stitching can generate
a beautiful and seamless panoramic image, the geometric rela-
tionship between the image and object spaces are not record-
ed. In other words, image points can no longer be treated as
a bundle of light rays projected from the object space. Some
special cameras can capture a panoramic image directly with-
out image stitching. Such as the linear-array-based terrestrial
panoramic camera (Schneider and Maas, 2006) or the spher-
ical camera (Li, 2006). However, how well they retain the
geometric relationship needs a further investigation.
Instead of applying image stitching, this study develops a
method to form a spherical panorama image (
SPI
) with eight
images captured with
PPIMS
. The geometric relationship be-
tween cameras will be maintained in the
PPIMS
SPI
, although
the collinearity condition (Wolf and Dewitt, 2004) is not rigor-
ously recovered due to the unknown value of object range.
When the images captured with
PPIMS
are transformed into
SPIs
, how the
SPIs
can be used for mapping still remained in
question. Suppose we have ideal
SPIs
, i.e., each
SPI
pixel is a
direct projection of the corresponding object point through
the
SPI
center, it would be possible for us to perform photo-
Yi-Hsing Tseng and Kuan-Ying Lin are with the Department of
Geomatics, National Cheng Kung University, No.1, University
Road, Tainan 701, Taiwan
).
Yung-Chuan Chen is with the Department of Civil Engineer-
ing, De Lin Institute of Technology, No.1, Ln. 380, Qingyun
Road, New Taipei City 236, Taiwan
Photogrammetric Engineering & Remote Sensing
Vol. 82, No. 12, December 2016, pp. 935–943.
0099-1112/16/935–943
© 2016 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.82.12.935
