We present a novel framework for measuring the body motion of multiple individuals in a
group or crowd via a vision-based tracking algorithm, thus to enable studies of humaninduced
vibrations of civil engineering structures, such as floors and grandstands. To overcome
the difficulties typically observed in this scenario, such as illumination change and
object deformation, an online ensemble learning algorithm, which is adaptive to the
non-stationary environment, is adopted. Incorporated with an easily carried and installed
hardware, the system can capture the characteristics of displacements or accelerations for
multiple individuals in a group of various sizes and in a real-world setting. To demonstrate
the efficacy of the proposed system, measured displacements and calculated accelerations
are compared to the simultaneous measurements obtained by two widely used motion
tracking systems. Extensive experiments illustrate that the proposed system achieves
equivalent performance as popular wireless inertial sensors and a marker-based optical
system, but without limitations commonly associated with such traditional systems. The
comparable experiments can also be used to guide the application of our proposed system.
2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY