Monitoring the relative density of static or moving powder inside a process line is essential for manufacturing high-quality products. The aim of this study was to predict density variations in a moving powder bed using terahertz reflection technology. We systematically investigated three grades (varying true density and particle size) of two materials: lactose and silicified microcrystalline cellulose (SMCC). These six powders specifically differ in their compressibility, which can be applied to assess the sensitivity and applicability of our method. The powders were filled into a round container, and terahertz reflection measurements were acquired continuously during the container's rotation. The setup allowed to adjust the relative density by compacting the powders into specific powder bed heights. Each powder was compacted to various relative densities (compression pressures up to 100 kPa). We calculated the surface refractive index based on the in-line terahertz measurements acquired during rotation, which has a linear dependence on the relative density of the powder. This was confirmed by correlating the refractive index values with the theoretical relative densities based on the bulk and true densities of the powder. The coefficient of determination (R-2) was larger than 0.962 (Lactohale 100) for all six powders, with the highest coefficients for Lactohale 220 (R-2 = 0.996) and SMCC 50 LD (R-2 = 0.995). The results suggest that the proposed method can resolve relative densities averaged across the powder bed that are as small as 0.3% (Lactohale 100). The high acquisition rate of the terahertz system (15 Hz) made it possible to determine the powder density in 230 positions uniformly distributed throughout the container, facilitating the investigation of the relative density uniformity in the container as a function of the powder bed height. It was observed that SMCC powders undergo a smaller change in the relative density variations upon compaction than the lactose powders. Moreover, the relative density maps clearly indicate local density differences in the powder bed for all powders. The relative density variations that were introduced by packing of the container prevailed throughout the compaction process for all samples with the exception of Lactohale 220. The presented approach allows a precise resolution of the spatial distribution of relative density, which facilitates an in-depth analysis of powder behavior upon compaction. (C) 2018 Elsevier B.V. All rights reserved.