摘要: 针对最优子模式分配(OSPA)度量标准只能适应测量轨迹距离的缺点,提出最优子模式分配概率(OSPAP)度量标准。通过计算真实轨迹与预测轨迹之间的基本概率和标签错误的惩罚概率评价跟踪算法的性能。在现实的跟踪系统中,无法获得跟踪目标的真实轨迹,但是基于概率的多目标跟踪算法的概率较易获得。实验结果表明,采用该标准得到的度量数据与实际情况一致,能够在实际的跟踪系统中实时地反映概率跟踪算法的性能。
关键词:
最优子模式分配概率度量,
性能评估,
基本概率,
本土化基本概率度量,
标签错误度量,
多目标跟踪
Abstract: In order to overcome the shortcoming that Optimal Sub Pattern Assignment(OSPA) only measures the distance, Optimal Sub Pattern Assignment Probability(OSPAP) metric standard is proposed. By introducing the probability, it measures tracking algorithm performance by computing elementary probability and label error probability between the ground truth tracks and the estimated tracks. It is impossible to obtain the ground truth tracks in the applications but the probability is easy to be got in the multi-target tracking algorithms based on the probability. Experimental result shows the measurement data obtained from this standard is accordance with the actual situation, and it can reflect the performance of probability tracking algorithm in actual tracking system in real-time.
Key words:
Optimal Sub Pattern Assignment Probability(OSPAP) metric,
performance evaluation,
elementary probability,
localization elementary probability metric,
label error metric,
multi-target tracking
中图分类号:
刘国营, 陈秀宏. 多目标跟踪算法的最优子模式分配概率度量[J]. 计算机工程, 2013, 39(5): 293-296.
LIU Guo-Ying, CHEN Xiu-Hong. Optimal Sub Pattern Assignment Probability Metric for Multi-target Tracking Algorithm[J]. Computer Engineering, 2013, 39(5): 293-296.