ENHANCED GM-PHD FILTER FOR REAL TIME SATELLITE MULTI-TARGET TRACKING

ENHANCED GM-PHD FILTER FOR REAL TIME SATELLITE MULTI-TARGET TRACKING Camilo Aguilar⋆ Mathias Ortner† Josiane Zerubia⋆ ⋆ Inria, Universit ́e Cˆote d’Azur, Sophia Antipolis, France † Airbus Defense and Space, Toulouse, France ABSTRACT We present a real-time multi-object tracker using an enhanced version of the Gaussian mixture probability hypothesis den- sity (GM-PHD) filter to track detections of a state-of-the-art convolutional neural network (CNN). This approach adapts the GM-PHD filter to a real-world scenario to recover tar- get trajectories in remote sensing videos. Our GM-PHD filter uses a measurement-driven birth, considers past tracked ob- jects, and uses CNN information to propose better hypotheses initialization. Additionally, we present a label tracking solu- tion for the GM-PHD filter to improve identity propagation given target path uncertainties. Our results show competitive scores against other trackers while obtaining real-time per- formance.