物联网数据融合研究

发布时间：2018-04-22 14:09

本文选题：物联网 + 数据融合　；参考：《石家庄铁道大学》2014年硕士论文

【摘要】：物联网的产生提高了生产的效率、增加了生活的便利,随着数据源的成倍增加、传感识别层的设备也随着增加,传感端的数据量级也不断增大。利用数据融合技术对传感器及处理设备的架构进行合理的组织、对数据进行二次处理有很大的必要。同时,由于“云”服务、新一代网络等新技术的引入,物联网中数据融合的研究已经成为一个重要的课题。因此,本文做了以下工作：介绍了物联网中数据融合的起源、优势、应用及发展现状,以数据融合模型与数据融合算法两方面的角度对国内外研究物联网数据融合的成果进行总结,并指出数据融合在物联网中发展现状和面临的挑战。对融合算法的概念、背景、级别分类等内容进行了具体的介绍,并对数据融合模型与数据融合算法进行了详细描述。它们决定了系统的性能,其实现的基础功能对上层的高级服务提供支持。简要的介绍了物联网大数据出现的背景,并分析了物联网中大数据量融合处理的需求,接着对传统的压缩融合算法优缺点进行了分析和总结。经典的压缩融合算法实现过程与运行简单、处理开销小。但是其中的绝对增量技术不具有一般的纠错功能,对于异常的读数处理起来比较困难。而对于相对增量技术,原始数据的变化范围受到融合数据空间的限制,如果相邻数据差异过大将无法被记录,将未做任何的判断就被当成异常点进行丢弃。针对经典压缩算法的不足,提出了嵌套式的增量压缩融合方法。试验结果表明,在数据量相同的情况下,嵌套增量压缩算法的压缩比优于经典的增量压缩算法,四种算法在压缩时间上性能相差较小,适合实时存储。而在解压上,嵌套式的方法时间较长,尤其是时间嵌套的方式,这是因为在压缩后,为了突出分布式处理的特点,数据进行了打包存储,占用了时间,而数据间隔的方法具有简单的纠错功能,增加了时间的开销。分析了芜湖大桥的健康监测系统,对整体架构、传感器布设等进行了介绍,对预警系统的需求进行了分析：将监控的变量锁定在纵向位移上,将监控时间分为列车通过时断和无列车上桥时段。对整体融合模型和存储结构进行了设计,针对系统的分析,将融合算法分为两个模块：预警模块与监控模块进行实现,通过试验证明,设计的系统所模拟得到的纵向位移曲线与现实监测的震动曲线非常相似,并且相对误差与震动位移的大小呈负相关。最后对全文的工作做了总结,并对下一步的研究工作进行了展望。
[Abstract]:The production of the Internet of things improves the efficiency of production and increases the convenience of life. With the increase of data sources, the equipment of sensor recognition layer increases, and the data level of sensor end increases. The data fusion technology is used to organize the structure of sensor and processing equipment reasonably, and it is necessary to process the data twice. At the same time, due to the introduction of new technologies such as cloud services and new generation networks, the research of data fusion in the Internet of things has become an important subject. Therefore, this paper does the following work: This paper introduces the origin, advantage, application and development of data fusion in Internet of things, and summarizes the research results of data fusion of Internet of things at home and abroad from two aspects: data fusion model and data fusion algorithm. The status quo and challenges of data fusion in the Internet of things are also pointed out. The concept, background and classification of the fusion algorithm are introduced in detail, and the data fusion model and the data fusion algorithm are described in detail. They determine the performance of the system, its implementation of the basic functions to support the upper level of advanced services. This paper briefly introduces the background of the emergence of big data in the Internet of things, and analyzes the requirements of mass data fusion in the Internet of things, and then analyzes and summarizes the advantages and disadvantages of the traditional compression fusion algorithm. The classic compression fusion algorithm is simple to implement and run, and the processing overhead is small. But the absolute increment technique does not have the general error correction function, so it is difficult to deal with the abnormal reading. For the relative increment technique, the range of the original data is limited by the fusion data space. If the difference between the adjacent data is too large, it will be discarded as the outlier if the difference between the adjacent data is too large to be recorded. A nested incremental compression fusion method is proposed to overcome the shortcomings of classical compression algorithms. The experimental results show that the compression ratio of the nested incremental compression algorithm is better than that of the classical incremental compression algorithm when the amount of data is the same, and the performance of the four algorithms is small in compression time, which is suitable for real-time storage. In decompression, the nested method takes a long time, especially the time nesting method, because after compression, in order to highlight the characteristics of distributed processing, the data is packed and stored, which takes up time. The method of data interval has simple error correction function and increases the time cost. This paper analyzes the health monitoring system of Wuhu Bridge, introduces the whole structure and sensor layout, and analyzes the requirements of the early warning system: locking the monitored variables on the longitudinal displacement, The monitoring time is divided into two periods: the train is broken and the train is not on the bridge. The whole fusion model and storage structure are designed. According to the analysis of the system, the fusion algorithm is divided into two modules: early warning module and monitoring module. The longitudinal displacement curve simulated by the designed system is very similar to the vibration curve monitored in reality, and the relative error is negatively correlated with the magnitude of vibration displacement. Finally, the work of the full text is summarized, and the next research work is prospected.
【学位授予单位】：石家庄铁道大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP391.44;TN929.5

【参考文献】