Human tracking and pose estimation for subsurface operations

Abstract— Human lives are particularly at risk in critical security situations in underground train stations compared to surface events. Due to the closed situation of such subsurface events, considerable obstacles to the safe and efficient evacuation of people after an attack must be taken into account. Thus, this work presents a computer vision system based on artificial intelligence that uses available surveillance cameras in the optical and the thermal spectrum to detect and track human beings, and to allow an activity classification based on a pose estimation. Those results are then transferred into a 3D common operational picture to assist subsurface operations.