A comprehensive performance metric for evaluation of multi-target tracking algorithms
Xiaobin Shi, Feng Yang, Fei Ming Tong, Hao Lian · 2017
The optimal sub-pattern assignment (OSPA) distance/metric is widely used to evaluate the performance of Multi-target tracking (MTT) algorithms based on random finite sets. Potential shortages of the original OSPA metric are discussed and a comprehensive metric for evaluating performance is established through incorporating the third error component, which represents the error of angles, into the original OSPA distance. Thus, the new metric can be viewed as the combination of three components including localization errors, cardinality errors and angle errors. Two computational approaches for computing new error component are described based on cosine similarity/distance. The simulation results verify the new metric.