环境监测中多传感器数据融合研究

Research on multi-sensor data fusion in environment monitoring

[目的]针对环境监测中单一传感器测量数据精度低、可靠程度低的问题,本文提出在无线传感网络监测系统中,通过改进自适应加权融合算法并利用模糊神经网络算法实现多传感器数据融合,来提高环境监测的准确性。[方法]基于多传感器同一时段采集的数据,先采用欧式距离及相关函数改进的自适应加权算法进行同质传感器数据融合,再设计模糊神经网络分类器把异质传感器的数据转化为环境质量等级信息。[结果]仿真实验显示出本文提出的同质传感器数据融合算法融合精度较高于其他几种算法、模糊神经网络算法通过对350组训练样本的学习后能够对96%的验证样本的环境等级进行正确分类且预测曲线基本可以拟合实际输出。[结论]本文的同质传感器数据融合算法提高了数据融合精度,异质传感器数据融合算法能够对整体环境质量得出较可靠的评价。

[Objective] For the problem of low accuracy and reliability of single sensor data in environment monitoring, the paper proposed that based on the WSN monitoring system use an improved self-adaptive weighted fusion algorithm and fuzzy neural network to increase the reliability of environment monitoring. [Methods] Based on the data collected by multi sensors in the same period, applied the self-adaptive weighted algorithm improved by Euclidean distance and correlation function on the data fusion of homogeneous sensors, we designed a fuzzy neural network to translate the data from heterogeneous sensor into environment quality grade. [Results] The simulation experiment showed that the accuracy of the proposed homogeneous sensor data fusion algorithm was higher than some other kinds of algorithms, the fuzzy neural network algorithm could correctly classify 96 % verification samples after learning 350 training samples and the prediction curve could roughly match the real outputs. [Conclusion] The proposed homogeneous sensor data fusion algorithm increased the accuracy of data fusion, and the heterogeneous sensor data fusion algorithm could give a relatively reliable evaluation to overall environment quality.