一种多传感器温度数据动态融合方法  

The Approach of Dynamic Data Fusion Based on Multi-sensor Temperature Data

作　　者：李时辉[1] 

机构地区：[1]义乌工商职业技术学院,义乌322000

出　　处：《科技通报》2015年第1期146-149,205,共5页Bulletin of Science and Technology

基　　金：浙江省2010高技能人才培养和技术创新项目(2010R30044);浙江省自然科学基金(Y1100219)

摘　　要：目前,多传感器温度数据融合方法中存在阈值选取过于绝对化和经验化的问题,并且难以动态反映被测对象的真实情况。针对此问题,提出了一种多传感器温度数据动态融合方法,首先运用模糊理论中相关性函数对各传感器支持度进行计算排序,再将支持度较低的数据认定为无效数据予以剔除,最后运用正交基函数神经网络与递推最小二乘法相结合的数据融合方法。在疫苗冷链温度检测中使用该动态融合方法,其均方误差0.0383和误差0.0381优于基于平均值法的均方误差0.1332和方差0.1371,检测精度上升了65%。因此,该方法克服了阈值选取过于绝对化和经验化的缺点,提高了多传感器温度数据的融合精度,从而满足疫苗冷链温度高精度检测的需要。At present, there are some limitations for multi-sensor temperature information fusion methodssuch as threshold values are too absolute and experience-based, which are hard to dynamically reflectthe real situation of the measured objects. To solve this problem, a new multiple sensor temperatureinformation fusion technique was proposed in current research, which was able to overcome thedisadvantages of lots of existing temperature data fusion methods. The correlation function in Fuzzytheory was applied to calculate and sort the support degree of each sensor. Sensor information with lowsupport degree would be identified as invalid data and be rejected. Then neural network algorithm withorthogonal basis functions and recursive least- squares method were employed to fuse the data. Theexperiment results of using the dynamic fusion method in the vaccine cold chain temperature detectingshows that the error of mean square of 0.0383 and the error of 0.0381 are better than that of the averagemethod which are 0.1332 and 0.1371. The detection accuracy is increased by 65%. So this method notonly overcomes the problems of related techniques, but also improves measurement accuracy oftemperature so as to meet the high standard of accuracy of the vaccine cold chain temperature detecting.

关 键 词：相关性函数 神经网络 动态融合 多传感器 疫苗 

分 类 号：TP212.9[自动化与计算机技术—检测技术与自动化装置]