一种基于最优聚类中心与权重欧式距离的多源异质传感器数据融合方法  

A Multi-Source Heterogeneous Sensor Data Fusion Method Based on Optimal Clustering Center and Weighted Euclidean Distance

作　　者：蔺万科 宋华 南新元[1] 李燕 黄家興 LIN wanke;SONG hua;NAN xinyuan;LI yan;HUANG jiaxing(School of Electrical Engineering,Xinjiang University,Urumqi Xinjiang 830047,China;Xinjiang Architectural Design and Research Institute,Urumqi Xinjiang 830002,China)

机构地区：[1]新疆大学电气工程学院,新疆乌鲁木齐830047 [2]新疆建筑设计研究院机电所,新疆乌鲁木齐830002

出　　处：《传感技术学报》2022年第1期49-56,共8页Chinese Journal of Sensors and Actuators

基　　金：新疆自治区自然科学基金(2019D01C079)。

摘　　要：针对多源传感器协同监测森林火灾时对于早期火灾识别准确度不高的问题,本文提出了一种基于最优聚类中心与权重欧式距离的多源异质传感器数据融合方法。将温度、烟雾和CO传感器的数据进行融合得到明火、阴燃、无火三种火情的概率估计,从而实现及时识别林火的目的。仿真实验结果表明:本文提出的方法可以实现各阶段火情早期特征的检测,有效识别早期森林火灾;与相关文献提出的方法相比,本文方法能够得出更为理想的林火概率,辨识准确性更高,可以有效降低误报风险。Aiming at the problem of low accuracy of initial fire recognition when multi-source sensors are cooperatively monitoring forest fires, the paper proposes a multi-source heterogeneous sensor data fusion method based on the optimal cluster center and the weighted Euclidean distance. The data of temperature, smoke and CO are fused to obtain the probability estimates of three fire conditions: open flame, smoldering and no fire, so as to realize the purpose of identifying forest fires in time. Experimental results show: The method proposed in this paper can achieve the detection of the early characteristics of the fire at each stage and effectively identify the initial forest fires;Compared with the methods proposed in the related references, the method in this paper can obtain more ideal forest fire probabilities with higher identification accuracy and effectively reduce the risks of false alarms.

关 键 词：早期森林火灾识别 多源异质传感器 数据融合 最优聚类中心 权重欧式距离 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]