一种基于不确定分析的多传感器信息动态融合方法  

Uncertainty Analysis Based Dynamic Multi-Sensor Data Fusion

作　　者：罗本成[1] 原魁[1] 陈晋龙[1] 朱海兵[1] 

机构地区：[1]中国科学院自动化研究所高技术创新中心,北京100080

出　　处：《自动化学报》2004年第3期407-415,共9页Acta Automatica Sinica

基　　金：国家"863"高技术研究发展计划(2001AA422200)资助项目~~

摘　　要：提出了一种基于不确定分析的多传感器动态分布融合方法.首先引入贴近度的概念对传感器进行动态聚类,接着基于兼容测度实现了组内传感器信息的最优Bayesian估计融合;最后给出了一种基于一致测度的多传感器信息动态融合的方法.通过实验对比分析,证实了此方法具有较好的实效性和鲁棒性.The problem of modeling multi-sensor data fusion system under uncertainty is discussed. Taking advantage of the measure of relative proximity in terms of uncertainty, the authors firstly implement multi-sensor dynamic clustering. Based on Bayesian estimation technology and the measure of compatibility, an optimal fusion paradigm for multi-sensors data fusion in the same group is presented. By examining the mutual impact of sensor groups based on the measure of confidence, a novel model for dynamic multi-sensor fusion system is described. The efficient fusion of data from different sources enables the system to respond promptly to the uncertain environment. Finally, experiments demonstrate the model is of higher sensitivity and practicability, especially in uncertain environment for intelligent systems.

关 键 词：不确定分析 格贴近度 动态多传感器信息融合 最优Bayesian估计融合 

分 类 号：TP212[自动化与计算机技术—检测技术与自动化装置]