Metrics for performance evaluation of elliptic extended object tracking methods
Shishan Yang, Marcus Baum, Karl Granström · 2016
The objective of extended object tracking is to simultaneously track a target object and estimate its shape. As a consequence, it becomes necessary to incorporate both location and shape errors in the performance assessment of extended object tracking methods. In this work, we highlight the difficulties of selecting a proper metric for this purpose and discuss currently used metrics from literature. Furthermore, we suggest the Gaussian Wasserstein metric for comparing elliptic shapes and illustrate its advantages.