分布式检测系统的混合融合算法  

Mixed fusion method for distributed detection systems

作　　者：严军[1] 关键[2] 彭应宁[1] 

机构地区：[1]清华大学电子工程系,北京100084 [2]海军航空工程学院电子工程系,烟台264001

出　　处：《清华大学学报（自然科学版）》2006年第1期46-49,共4页Journal of Tsinghua University(Science and Technology)

基　　金：国家"十五"预研项目(40106020201)

摘　　要：在采用局部检测统计量的分布式检测系统中,为克服传统的求和融合在局部信噪比差异较大时明显的性能损失,提出基于R统计量的混合融合准则,该准则同时采用OR融合准则和直接求和准则,对两个准则处理结果进行或逻辑融合。混合准则结合了传统OR融合准则稳健性较好,直接求和融合最佳性能较高的特点。给出了混合融合算法的系统虚警概率和检测概率表达式。对双传感器并行分布式检测系统的仿真表明,混合融合准则相对直接求和准则,能将局部信噪比差异明显时的检测概率提高0.05—0.12,更加适合分布式检测系统。When detecting with a distributed system based on local statistics, the conventional sum fusion method is ineffective if the local signal-to-noise ratios （SNRs） differ significantly. A mixed fusion rule developed for varying SNRs uses the sum rule and the OR rule to simuhaneously process observations. Therefore, the mixed fusion rule combines the robustness of the OR rule and the higher detection performance of the sum rule. Analytical expressions are given for the probability of false alarms and probability of detection （Pd）. Simulations based on a distributed system with 2 local sensors show that the mixed fusion rule improves the Pd by about 0. 05 - 0. 12 when local SNRs have large differences. For a distributed detection system, the mixed fusion rule is more suitable than the sum rule alone.

关 键 词：多传感器 分布式检测 检测统计量 融合 

分 类 号：TP212[自动化与计算机技术—检测技术与自动化装置]