The task of tracking dynamic objects is to estimate the 3D-pose and the extension of objects through recursive algorithms like Bayes filtering.
There are several multi-object filter, e.g, the PHD-Filter and single-object filter (Unscented-Kalman-Fiter, Extended-Kalman-Filter), which potentially need a track management.
Furthermore, there are extensions for extended object tracking. One modern method for extended object tracking is recursive Gaussian-Processes.
Moreover, the choice of the measurement as well as process model plays a significant role.
Applications at our Institute
Multi-Object Tracking
Challenges of the multi-object tracking are the unknown number of objects as well as the data association.
Single-Object Tracking for Convoy/Mule Scenario
A crucial part of the convoy-/mule scenario is the estimation of the pose, velocity and orientation of the leading object.
