An Improved Subpixel Phase Correlation Method
with Application in Videogrammetric Monitoring
of Shaking Table Tests
Zhen Ye, Xiaohua Tong, Yusheng Xu, Sa Gao, Shijie Liu, Huan Xie, Peng Chen, Wensheng Lu, and Xianglei Liu
Abstract
Videogrammetry is an extension to close-range photogram-
metry which provides a convenient way to capture the
three-dimensional (3D) dynamic response of moving objects
in a non-contact manner. The performance of tracking in
image sequence is crucial to the measurement accuracy.
This paper proposes an improved subpixel phase correlation
method for tracking in image sequence, which combines the
advantages of gradient representation, robust plane fitting
and robustness iteration. By the use of the proposed phase
correlation based tracking method and other advanced image
processing algorithms, a practical videogrammetric measure-
ment system is presented to provide an alternative means to
monitor and analyze 3D structural vibration deformations.
Two simulated tests as well as one static test are performed
to demonstrate the superiority and reliability of the proposed
phase correlation method. The empirical experimental results
of the monitoring of large-scale shaking table tests with a
landslide dam model, including coordinate comparisons
with the total station and acceleration comparisons with
the original waveform and acceleration sensors, confirm the
effectiveness of the videogrammetric measurement system
based on the improved subpixel phase correlation method.
Introduction
Videogrammetry is an indispensable branch of close-range
photogrammetry, which expands the methods and models of
photogrammetry from still imagery to multiple time steps,
enabling the measured object to be defined dynamically (Lin
et al.
, 2008). With the development of charge coupled device
(
CCD
) and complementary metal oxide semiconductor (
CMOS
)
technology, sensors are now available with very high reso-
lutions, very high frame rates, and different sensor formats
(Luhmann, 2010). The huge range of available cameras and
imaging sensors, which can adequately meet the requirement
of high-accuracy and high-speed measurements, provide new
application prospects to videogrammetry. As a result, the
videogrammetric technique has been widely applied in recent
years in several academic and practical domains, including
movement analysis, deformation measurement, and material
testing (Maas and Hampel, 2006; Remondino, 2006; Lin
et
al.
, 2008; Barazzetti and Scaioni, 2010; Sładek
et al.
, 2013).
Especially in the field of civil engineering, vision-based
systems and the videogrammetric technique are drawing
increasing attention for different applications in both static
and dynamic measurement, such as structural monitoring and
bridge measurement (Jiang
et al.
, 2008; Kohut
et al.
, 2013; Liu
et al.
, 2015).
The shaking table test in civil engineering is an indis-
pensable tool, through which the dynamic behaviors of the
simulated models are measured under different conditions
to provide reliable information for performance assessment
and method verification (Severn, 2011). In the vibration-based
applications, the dynamic data are normally collected by tra-
ditional acceleration and displacement sensors, which can be
generally classified into contact and non-contact types. The
traditional contact transducers include displacement gauges,
accelerometers, linear variable differential transducers (
LVDTs
),
and linear potentiometers (Choi
et al.
, 2011; Wittich and
Hutchinson, 2015), while the non-contact sensors mainly rely
on laser technology, radar technology, and Global Navigation
Satellite System technology (Nassif
et al.
, 2005; Gentile and
Bernardini, 2010). However, measurement using the contact
sensors can be both expensive and inefficient when detailed
observations at a large number of locations are desired, as
they can only provide local one-dimensional measurement
and are vulnerable to damage or failure, while the uses of
non-contact sensors are costly, dangerous to human health or
have a limited precision and acquisition frequency (Ribeiro
et
al.
, 2014; Feng
et al.
, 2015; Liu
et al.
, 2015).
In comparison, videogrammetry is a cost-effective tech-
nique for obtaining full-field measurements remotely and
flexibly, and it is able to realize 3D multipoint and repeti-
tive measurement with a high degree of accuracy. Several
previous studies of vision-based systems in vibration-based
applications have been reported. Lee and Shinozuka (2006)
developed a single-camera vision-based measurement system
for the health monitoring of bridges, in which a plane pattern
was designed for simple calibration and target tracking. The
measurement accuracy was verified by a laboratory test using
a small-scale shaking table. A similar vision-based system for
the real-time monitoring of the dynamic responses of large-
size civil engineering structures was proposed by Fukuda
et
al
. (2010). The 2D displacement of each point was captured
by a low-cost digital camcorder, and multipoint measurement
was realized using multiple subsystems and a time synchro-
nization system. The efficacy of the system was demonstrated
by a seismic shaking table test and a field bridge experiment.
Choi
et al.
(2011) applied a similar vision-based system using
Zhen Ye, Xiaohua Tong, Yusheng Xu, Sa Gao, Shijie Liu,
Huan Xie, Peng Chen, and Wensheng Lu are with The State
Key Laboratory for Disaster Reduction in Civil Engineering
and College of Surveying and Geo-Informatics, Tongji
University, 1239 Siping Road, Shanghai 200092, China
(
).
Xianglei Liu is with the Key Laboratory for Urban Geomatics
of National Administration of Surveying, Mapping and
Geoinformation, Beijing University of Civil Engineering and
Architecture, 1 Zhanlanguan Road, Beijing 100048, China.
Photogrammetric Engineering & Remote Sensing
Vol. 84, No. 9, September 2018, pp. 579–592.
0099-1112/18/579–592
© 2018 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.84.9.579
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
September 2018
579
