In a pseudo-2D fluidized bed, a new color particle tracking velocimetry (PTV) method, integrating the Voronoi tracking algorithm and the relaxation probability tracking algorithm, is proposed to measure velocities of individual particles with different sizes marked by distinguishable colors. Thorough verification of the color-PTV methodology is conducted based on synthetic images generated from CFD-DEM simulation data, in terms of segmentation bias, segmentation ratio, recovery ratio and error ratio. The novel color-PTV methodology shows very good performance in simultaneously tracking a large number of different individual particles in poly-disperse systems. Experimental results are used to compare with CFD-DEM simulations using different drag models. Simulations with correction for the effect of particle size dispersity in the drag model show significantly better agreement with respect to the mixing index, time-averaged volumetric particle flux, distributions of individual particle velocity as well as distributions of particle granular temperature. (C) 2018 Elsevier Ltd. All rights reserved.