基于偏最小二乘法的多传感器信息融合研究  

Multi-Sensor Information Fusion Study Based on Partial Least Squares Technique

作　　者：康健[1] 唐力伟[1] 左宪章[1] 李浩 张西红[1] 

机构地区：[1]军械工程学院 [2]63880部队

出　　处：《火炮发射与控制学报》2009年第3期40-43,共4页Journal of Gun Launch & Control

摘　　要：针对先验知识未知的情况,且多传感器间存在时空相关性以及小样本等问题,提出了一种基于偏最小二乘法的多传感器信息融合模型。首先对数据进行标准化处理,消除数据间的量纲效应;然后进行成分的提取,并通过交叉验证确定提取的成分数;最后通过还原运算得到信息融合方程。通过模型验证和某型地炮复进机压强信息融合相关性能系数分析,基于偏最小二乘法的多传感器信息融合能够对多传感器进行相关性分析,剔除重叠和无意义的信息,并且融合精度优于基于支持度和相对距离的多传感器信息融合,也为多传感器信息融合及提高融合性能提供了一种新途径。Because prior knowledge is unknown, and spatiotemporal correlation between the multi-sensors and the data is difficult to obtain, a new kind of multi-sensor information fusion model was proposed based on partial least squares （PLS） technique. Firstly, origin data were standardized to eliminate the dimensional effect between different data. Then, components were extracted and the number of components were determined by means of cross validation. Finally, the information fusion model was obtained through reduction operation. By use of model verification and information fusion performance analysis of a certain type of ground gun counter-recoil mechanism, multi-sensor information fusion model based on partial least squares technique was proved to analyze spatiotemporal correlation, eliminate the overlapping and meaningless information. Compared with multi-sensor fusion based on support degree and relative distance, the information fusion performance based on partial least squares technique is better than multi-sensor fusion based on support degree and relative distance. This study can provide a new means for multi-sensor information fusion and fusion performance improvement.

关 键 词：信息处理技术 偏最小二乘法 信息融合 传感器 

分 类 号：TP212[自动化与计算机技术—检测技术与自动化装置]