Spatio-temporal trajectory alignment for trajectory evaluation

Abstract Assessing the quality of a trajectory is a prerequisite for correctly interpreting and using the trajectory in applications such as kinematic laser scanning. Evaluation is for example done empirically by comparison with a ground-truth trajectory recorded simultaneously. Depending on the measurement and sensor configuration, both trajectories have to be aligned before comparison. Usually, either a similarity transformation or a rigid-body transformation is used for this purpose. We propose an extended spatio-temporal alignment, which additionally supports the estimation of a lever arm and a time offset between both trajectories. Our method can be used to align two trajectories of the same vehicle captured simultaneously by different sensors. We apply the approach on several recorded data sets and evaluate it empirically. We show that although real-world data sets can lead to high correlations between parameters, they can still be successfully aligned and evaluated using the methodology. To enable replication of our research, we publish the code which is available here: https://github.com/gereon-t/trajectopy .