一种基于似然比判决规则的检测融合算法及其鲁棒性分析  

Distributed Detection Based on Likelihood Ratio Decision Rules and Analysis of Robustness of System Performance

作　　者：相明[1] 王炜[1] 郭飞[1] 

机构地区：[1]西安交通大学电子与信息工程学院,陕西西安710049

出　　处：《探测与控制学报》2008年第4期4-7,共4页Journal of Detection & Control

基　　金：国家自然科学基金项目资助(60472005)

摘　　要：研究了一种传感器观测相关条件下分布式并行检测融合系统的检测性能优化方法,即限定各部传感器采用似然比判决规则,并在该约束条件下联合优化融合规则及各部传感器的判决门限,以使系统检测性能达到最优。推导了融合规则及各部传感器判决门限联合最优化的必要条件,得到了最优传感器判决门限满足的系统方程,并在此基础上给出了求解最优融合规则及传感器判决门限的数值迭代算法。仿真实验结果表明,该融合算法鲁棒性较好,且系统检测性能明显优于单部传感器的检测性能。An algorithm for optimization of detection performance for distributed parallel detection systems with correlated sensor observations is considered. The sensor decision rules are constrained to be likelihood ratio decision rules, and the fusion rule and sensor decision thresholds are jointly optimized for optimal system performance. The necessary condition for joint optimization of fusion rule and sensor decision thresholds is derived, and system equations for optimal sensor decision thresholds are obtained. A numerical iterative algorithm is also proposed to solve for the optimal fusion rule and sensor decision thresholds. Computer simulation shows that the performance of the fusion system is robust and that the system performance is much better than that of single sensor system.

关 键 词：并行检测系统 检测融合 似然比判决规则 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]