RFID路径数据清洗与聚类研究及其在农产品追溯系统中的应用

发布时间：2018-08-15 17:43

【摘要】：RFID以其扫描快、体积小、形式多、穿透力强及安全性好等特点被广泛运用于物流供应链系统、生产制造行业、交通运输管理等。带有RFID标签的物品在移动过程中会产生大量的RFID路径数据,如何从中提取有用的信息和知识,已经成为目前研究的重点。本文详细的介绍了RFID技术及其原理,并且结合RFID农产品追溯系统,分析了RFID路径数据特点,设计并且实现RFID技术在农产品追溯系统各个环节中的应用。由于射频识别技术容易受到外界环境的影响,导致RFID阅读器的准确率较低,造成了数据的不确定性,需要在数据存储之前对RFID数据进行预处理。针对自适应滑动窗口清洗算法SMURF(Statistical sMoothing for Unreliable RFid data),本文提出基于动态标签的RFID不确定性数据清洗算法DSMURF(Dynamic tags-based SMURF),通过构建真实的实验平台来分析阅读器准确率与标签距离、天线角度、标签速度之间的关系。自适应滑动窗口算法SMURF对于动态标签窗口设置过大,导致读取值与实际值之间存在较大误差。DSMURF算法通过设定标签的速度、阅读器的时隙、阅读器的阅读范围来构建滑动窗口的大小。此外,SMURF算法没有对冗余数据进行处理,本文提出了RFID数据冗余清洗框架,有效地降低了数据的冗余量,并通过实验平台和仿真数据集验证DSMURF算法的性能。本文提出基于在线微聚类和离线宏聚类的RFID数据流聚类算法RCluStream。路径对象的相似性度量是聚类分析的基础,首先提出RFID路径数据相似性定义。RFID路径数据具有流的特征,在聚类过程中,簇会因为数据的流入而不断地发生变化。在线微聚类设立簇的聚类特征,当簇发生改变时,记录簇的聚类特征；同时,离线宏聚类可以根据用户输入的时间参数,通过返回簇的聚类特征,获得这段时间簇的变化情况,从而有利于决策和分析。最后,开发了RFID农产品追溯系统和信息处理系统,将路径数据可视化,直观地表现路径数据的分布和挖掘结果,实验表明RCluStream算法的有效性。
[Abstract]:RFID is widely used in logistics supply chain system, manufacturing industry, transportation management and so on because of its fast scanning, small size, many forms, strong penetration and good security. RFID tagged goods will produce a large number of RFID path data in the process of moving. How to extract useful information and knowledge from it has become the current research. The focus of the study.
This paper introduces RFID technology and its principle in detail, and combines with RFID traceability system for agricultural products, analyzes the characteristics of RFID path data, designs and implements the application of RFID technology in every link of agricultural products traceability system. The uncertainty of data requires pre processing of RFID data before data storage.
Aiming at SMURF (Statistical Moothing for Unreliable RFid data), this paper proposes a dynamic tags-based algorithm DSMURF (Dynamic tags-based SMURF) for RFID uncertain data cleaning, which analyzes the accuracy of the reader and tag distance, antenna angle and tag speed by constructing a real experimental platform. The size of the sliding window is constructed by setting the speed of the tag, the time slot of the reader and the reading range of the reader. In addition, the SMURF algorithm does not process the redundant data. In this paper, a redundancy cleaning framework for RFID data is proposed, which effectively reduces the redundancy of data. The performance of DSMURF algorithm is verified by experimental platform and simulation data sets.
In this paper, we propose an RFID data stream clustering algorithm RCluStream based on on-line micro-clustering and off-line macro-clustering. The similarity measure of path objects is the basis of clustering analysis. Firstly, the similarity definition of RFID path data is proposed. Linear micro-clustering sets up clustering features and records clustering features when the clusters change. At the same time, off-line macro-clustering can get the changes of the clusters by returning the clustering features of the clusters according to the time parameters input by users, which is beneficial to decision-making and analysis. Finally, RFID agricultural products traceability system and information are developed. The processing system visualizes the path data and visualizes the distribution and mining results of the path data. Experiments show that the RCluStream algorithm is effective.
【学位授予单位】：浙江工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP391.44

【参考文献】