The volcanormal density for radar-based extended target tracking
Peter Broßeit, Bharanidhar Duraisamy, Jürgen Dickmann · 2017
This paper presents a novel shape model for radar-based vehicle tracking. This model is a special probability density function (pdf), designed to approximate the radar measurement spread of a vehicle more adequately than widely used models. While this pdf is used to determine the likelihood of the spatial measurements, additionally a recent Doppler measurement model is integrated in a maximum likelihood estimator. This formulation allows the estimation of a vehicle's pose, motion and axis lengths based on the radar's spatial and Doppler dimension. On this basis, an efficient computation method is derived. An evaluation of actual radar and ground truth data demonstrates the precise estimation of a vehicle's state.