目标跟踪滤波算法精度与可信性指标研究

Research on Evaluating Accuracy and Credibility for Target Track Filtering Algorithm

滤波器性能评估是跟踪系统评估的重要内容。研究了精度和可信性指标的构建,考虑到采用单一误差指标进行精度评估时存在固有缺陷,引入DES指标并将其与层次分析法结合,提出了权重计算的思路,解决了精度综合评估问题。针对ANEES在可信性评估中的不足,给出了GANEES和HANESS的公式,进而提出了DNEESS的形式化定义以对可信性进行综合评估。仿真表明采用新指标评估时可以挖掘出滤波器的更多误差信息,满足评估者的不同需求。

Evaluating filter is important for performance evaluation of target tracking system. Track accuracy measure are supplemented by introducing various average error and a new method is proposed by combining the analytic hierarchy process in effectiveness measure field with the DES. In addition,to measure credibility of error covariance matrix provided by the tracking filter,similar to geometric average error and harmonic average error respectively,the approaches for calculating both geometric and harmonic average normalized estimation error squared( NEES) are obtained. Furthermore,NEESS and DNEESS are defined,taking advantage of which to evaluate credibility comprehensively. The simulation shows new indexes could help the evaluator to obtain more information in error samples and serve for more evaluation requirements.