Refractive index matching (RIM) technique is widely used in the investigations of particle–liquid mixture flows. However, the determination of the volume fraction Φ in RIM is generally empirical and lacks theoretical or experimental basis. To reasonably reconstruct the internal Φ, this paper performs systematic calibration experiments to study the relationship between and the imaging indicators. The results show that the model based on area fraction Φ=ac_2D will induce large error for low due to the instability of the particle’s apparent size. The traditionally adopted proportional function of c_2D is only valid for the volume fraction near the calibration point. A prediction model is proposed based on the size of the Voronoi cell d_poly which presents to be more accurate and stable in RIM experiments. The d_poly-based model is well applied to the dense granular flows down an inclined plane to study the internal profiles of Φ and the bottom normal stress.