© 2019 Author(s). The study of flow and heat transfer through porous media or randomly packed beds is important as these configurations are widely used in many engineering applications, for example, heat energy storage, chemical catalytic reactors, and nuclear reactors. The flow mixing characteristics in a cross-flow plane of a facility with randomly packed spheres at an aspect ratio of 6.3 were experimentally investigated. The velocity fields at several regions of the cross-flow plane located in the vicinity of the wall and in the pores between spheres were obtained by applying the matched-index-of-refraction and time-resolved particle image velocimetry (TR-PIV) techniques for Reynolds numbers ranging from 700 to 1700. The TR-PIV results revealed various flow patterns in the transverse plane of the packed spheres, including swirling flow structures aligned with the axial flow direction, a strong bypass flow near the enclosure wall, and a circulation region created when the bypass flow ejected into a large spatial gap. When the Reynolds number was increased, the peaks of root-mean-square fluctuating velocities, urms′ and vrms′, were found to increase approximately at the same ratio as the increase in Reynolds number, and the magnitude of the Reynolds stress increased considerably. In addition, the characteristics of flow mixing in different flow regions were investigated via the two-point cross-correlation of fluctuating velocities. Using Taylor's hypothesis, the vorticity iso-surfaces were constructed. Thus, constructed iso-surfaces showed that shear layers generated from the bypass flow gaps were stretched, broken into smaller flow structures, and then evolved as vortex pairs when entering the neighboring gaps. The results obtained by applying proper orthogonal decomposition (POD) analysis to the velocity fields showed that the statistically dominant flow structures had approximately the same size and shape as those depicted by Taylor's hypothesis. Vortex characteristics, such as populations, spatial distributions, and strengths, for various spatial regions and Reynolds numbers were obtained by a combination of POD analysis and vortex identification.