Measuring aerosol plume velocity distribution using Sobel filter

Abstract

The standard approach to characterizing aerosol plume dynamics in a frame sequence is particle tracking. However, it fails in systems with a large number of poorly resolved particles. Previously, we proposed an image processing scheme that applied column-wise summation to each frame and stacked the resulting one-dimensional intensity profiles to construct a spatiotemporal map in one spatial dimension of the evolving aerosol distribution. In that work, we used only manual calculation of the edge angle to attain the aerosol plume velocity. Here, we use Sobel filtering to detect and calculate the edge angle, which gives a more information-rich velocity distribution of the aerosol plume. We test this method on an experimental setup consisting of a humidifier placed 60 cm away from an air purifier. We quantify the aerosol velocity by manual estimation and our method and compare their measurements. We also validate the results by using a hot-wire anemometer to obtain the flow velocity. When the air purifier was active, the Sobel method yielded a median velocity of 11.9 cm/s (IQR 4.5−25.0), compared to 13.5 cm/s from manual estimation and 16.4 cm/s (IQR: 9.4−23.4) from the anemometer.  We also found that velocities under 7 cm/s become indistinguishable from apparent velocities due to noise and smooth gradients in the spatiotemporal map. Overall, the Sobel method provides a more information-rich and objective complement to manual estimation.