An algorithm for the detection and characterisation of volcanic plumes using thermal camera imagery
Résumé
Volcanic plumes are turbulent mixtures of particles and gas which are injected into the atmosphere during a
volcanic eruption. Depending on the intensity of the eruption, plumes can rise from a few tens of metres up
to many tens of kilometres above the vent and thus, present a major hazard for the surrounding population.
Currently, however, few if any algorithms are available for automated plume tracking and assessment. Here,
we present a new image processing algorithm for segmentation, tracking and parameters extraction of convective
plume recorded with thermal cameras. We used thermal video of two volcanic eruptions and two
plumes simulated in laboratory to develop and test an efficient technique for analysis of volcanic plumes.
We validated our method by two different approaches. First, we compare our segmentation method to previously
published algorithms. Next, we computed plume parameters, such as height, width and spreading
angle at regular intervals of time. These parameters allowed us to calculate an entrainment coefficient and
obtain information about the entrainment efficiency in Strombolian eruptions. Our proposed algorithm is
rapid, automated while producing better visual outlines compared to the other segmentation algorithms,
and provides output that is at least as accurate as manual measurements of plumes.