| 题名 | 面向物联网的数据分级处理技术研究 |
| 作者 | 高嵩 |
| 学位类别 | 硕士 |
| 答辩日期 | 2011-05-27 |
| 授予单位 | 中国科学院沈阳自动化研究所 |
| 授予地点 | 中国科学院沈阳自动化研究所 |
| 导师 | 胡琨元 |
| 关键词 | 物联网 RFID 分级处理 数据清洗 标签事件 |
| 其他题名 | Study of Multilevel Data Processing Technology for Internet of Things |
| 学位专业 | 模式识别与智能系统 |
| 中文摘要 | 随着物联网（IOT，Internet of Things）技术在国民经济与社会信息化领域的推广应用，从感知层捕获、采集的数据表现出异构性、海量性和不确定性等特点，使得数据处理面临一系列的挑战，如何有效地处理和利用成倍增长的物联网数据将成为制约其发展的瓶颈问题。射频识别（RFID，Radio Frequency Identification）技术是物联网技术体系的重要组成部分，可实现物品身份标识和时空状态描述。在传统的RFID应用开发中，采用哑读写器的工作方式，读写器对采集到的数据不作任何处理，对标签的数据清洗和事件检测等处理均由RFID中间件实现，这种工作方式会造成较大的网络负荷，对中间件的运算能力要求较高。 本文提出一种对RFID数据进行分级处理的技术，核心思想就是在RFID读写器和RFID应用之间部署智能数据采集预处理终端，对数据采集和预处理过程进行融合，将一部分数据过滤、聚合、缓冲、分发工作转移到智能终端，从而降低网络负荷、提高系统实时性、减轻上层系统的运算压力。同时在主机上开发了原子事件处理系统，数据采集预处理系统经过将清洗后的标签数据传输给原子事件处理系统进行标签原子事件检测，原子事件处理系统产生的RFID原子事件为上层的复杂事件检测提供了数据输入。 为了实现数据分级处理技术，本文分别在嵌入式系统和台式主机上设计开发了数据采集预处理系统和原子事件处理系统，并详细制订了两者之间的通信协议。对嵌入式操作系统的Bootloader、内核和根文件系统分别实现了裁剪和定制。为了实现对RFID标签数据的清洗和原子事件检测，采用协议标准检验算法解决标签的脏读问题，采用平滑窗口过滤算法解决标签的漏读问题，采用基于权重的噪点抑制算法解决标签的多读和冗余问题，并实现对原子事件的检测。 在上述研究的基础上，本文开发了数据分级处理系统，系统具有设备管理、数据采集、网络通信、数据清洗和事件检测等功能。本文对解决物联网中海量数据处理问题进行了有益的尝试，根据本技术所开发的系统具有良好的伸缩性、可重用性和扩展性，在未来的物联网应用中有着广阔的前景。 |
| 索取号 | F253.9/G27/2011 |
| 英文摘要 | Along with the spread and application of Internet of Things (IOT) in national economy and society informatization, the data acquired from the sensor layer has present the characteristics of heterogeneity, mass and uncertainty, which will challenge the data processing in IOT, the exponentially increasing data is also going to become the bottleneck of development of the IOT. Radio Frequency Identification (RFID) is the major object identification technology in IOT. In the traditional development of RFID applications, the dumb RFID reader is used for reading of RFID tags, the tag data will not be processed in the reader while the data cleaning and event detection are all completed by the RFID middleware, which will cause heavy load to the network and require more computing capability. In this thesis, a design of technology of multilevel data processing of RFID data is proposed, which place an intelligent embedded data acquisition terminal between the reader and RFID application, the data cleaning is transferred to the intelligent terminal to settle the miss-reading, over-reading and dirty-reading of the tags. The data after data cleaning will be transferred to the data processing system to detect the atomic tag event, which will become the source data of complex event detection. Embedded data acquisition preprocessing system and data processing system are developed to implement the multilevel data processing technology. The communication protocol is also designed for the communication of command and data between the two systems above. The bootloader, kernel and root filesystem have been custom cut on the basis of particular requirements of development. The Electronic Product Code (EPC) check algorithm is used to solve dirty-reading, the smoothing algorithm is used to solve miss-reading, the weight-based noise suppression algorithm is used to solve over-reading and redundancy and detect atomic events. On the basis of researches above in this thesis, a multilevel data processing system is developed, which has the functions of device management, data acquisition, network communication, data cleaning and event detection. This thesis has proposed an alternative solution of data processing in IOT, the system refined from the technology has the characteristic of flexibility, reusability and expansibility, and therefore it has bright prospects in the IOT application in future. |
| 语种 | 中文 |
| 产权排序 | 1 |
| 公开日期 | 2012-07-27 |
| 分类号 | F253.9 |
| 内容类型 | 学位论文 |
| 源URL | [http://ir.sia.ac.cn/handle/173321/9364]  |
| 专题 | 沈阳自动化研究所_工业信息学研究室 |
| 推荐引用方式 GB/T 7714 | 高嵩. 面向物联网的数据分级处理技术研究[D]. 中国科学院沈阳自动化研究所. 中国科学院沈阳自动化研究所. 2011. |
| 个性服务 |
| 查看访问统计 |
| 相关权益政策 |
| 暂无数据 |
| 收藏/分享 |
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。
修改评论