摘要
传感器聚类状态的切换是多传感器数据融合的难点之一,也就是某个传感器在某一时刻应该向哪个方向融合数据的问题。文中采用粗糙集进行知识的获取,把1天内的54个传感器的可融合典型聚类分布作为数据样本空间形成决策表———"数据-融合分布"决策表;然后对一个月内的数据运用粗糙集的知识约简算法,去除冗余的属性和样本;根据神经网络聚类分析方法,形成多传感器数据的融合分布规则。仿真结果表明该模型的分类效率较好、实现传感器聚类分布的判断较快速。
The difficulties of fusing multi-sensor data lie in the switching of the state of sensor clusters. That is, at a given moment which direction the sensor should fuse data into. First the rough set is used for acquisition of knowledge. The typical clustering distributions of 54 sensors within one day are regarded as sample room for the decision-making table of the ＂ data-fusion distribution＂. Next,based on rough set of method of simplified knowledge,for one month date,remove redundant properties and samples. Then,the neural network is used to analyze clustering. And finally the patterns of multi-sensor data fusion distribution are formed. The model is proved experimentally to be efficient in classification and rapid in sensor clustering distribution decision.
出处
《计算机技术与发展》
2013年第4期221-225,共5页
Computer Technology and Development
基金
国家自然科学基金资助项目(61104216)
江苏省科技支撑计划项目(BE2011843)
南京邮电大学人才引进项目
