A Multiple Extended Object Tracker with the Gaussian Process Model Utilizing Negative Information
paper Menu
In multiple extended object tracking, the Poisson multi-Bernoulli mixture (PMBM) tracker is considered state-of-the-art. Originally, it was presented with the gamma Gaussian inverse Wishart (GGIW) target model, which is a random matrix model. When tracking larger objects using a light detection and ranging (LiDAR) sensor, measurements are generated by the contour rather than the whole target surface, and it is beneficial to model this with the target model. A target model that has this capability is the Gaussian process (GP) extent model. This paper presents a PMBM tracker using this target model. We also discuss considerations related to the use of the GP model in the PMBM framework. Secondly, we present improvements in the target model that increase the robustness of the model by dealing with the inherent non-linearities using the Gauss–Newton method. Furthermore, we incorporate an improvement to the tracker that utilizes the concept of negative information to generate virtual measurements that are then used in the Gauss–Newton optimization. In relation to this, we also present an occlusion model that utilizes the same negative information model to ensure that the state estimate is consistent in the presence of occluding targets.
