CASIA OpenIR  > 毕业生  > 博士学位论文
多目标识别RFID系统优化关键技术研究
其他题名Research on Key Technologies in Multi-object Identification RFID System Optimization
程虹
学位类型工学博士
导师杨一平
2012-05-29
学位授予单位中国科学院研究生院
学位授予地点中国科学院自动化研究所
学位专业计算机应用技术
关键词多目标识别rfid系统优化 Rfid系统建模 混合现实rfid仿真系统 多天线部署优化 多标签部署优化 Multi-object Identification Rfid System Optimization Rfid System Modeling Mixed-reality Rfid Simulation System Multi-antenna Placement Multi-tag Placement
摘要RFID技术作为物联网环境下的典型前端信息载体,在生产制造、物流管理、身份标识等领域均有着广阔的应用前景。但是,受到复杂应用环境的影响,RFID大规模应用部署仍然具有挑战。RFID系统优化技术是RFID应用关键技术之一,旨在解决RFID实际应用部署中多目标识别的可靠性问题。 本文首先从RFID应用系统建模出发,以门型RFID应用环境为例,给出一种针对特定应用约束条件的多目标识别系统建模方法。区别于以往在可控环境中进行的RFID基准测试,本文尝试将复杂的环境因素整体作为系统扰动,基于支持向量机(SVM)利用尽量少的学习样本建立起一个泛化能力较强的多目标识别模型,在一定扰动范围内都可以获得较为准确的预测效果,为实现应用部署优化提供了理论支撑。 其次,本文基于该多目标识别模型建立了混合现实RFID仿真方法,可替代实际硬件完成部署方案设计及应用系统可靠性验证工作。仿真系统采用边采集边训练的迭代学习方法,可在缩短90%测试时间和降低90% 存储空间的前提下,实现超过90%的预测准确度,为提高应用部署效率提供了技术支撑。 再次,本文进一步将RFID系统模型应用于RFID系统多天线部署中,提出了基于遗传算法的RFID系统多天线部署优化方法,优化多目标识别系统的识别范围。本文根据多目标识别RFID系统模型的预测结果,基于启发式搜索算法优化搜索空间,进而对门型三天线RFID系统进行基于遗传算法的求解。相比于传统的试错法和枚举法,算法可以在短时间内收敛到全局极值,为实施应用部署方案提供了方法指导。 最后,本文还针对多目标识别应用中的单目标识别优化策略进行了研究,通过在恶劣应用环境中部署冗余标签提高整体目标识别率。通过选择单位时间内的有效识别次数作为评价指标,采用正交试验方法对冗余标签的部署位置进行测试,试验结果表明双标签冗余部署方案的识别效果显著高于单标签部署方案,为提高系统识别性能提供了新的解决思路。
其他摘要Radio Frequency Identification (RFID), as typical information carrier in Internet of Things, is widely applied in advanced manufacture, logistics management and identification. However, according to the complex environment of practical applications, there are many challenges in the deployment of large-scale RFID application. The RFID system optimization is one of the important topics in the deployment of RFID applications, which aims to solve the low read rate problem in RFID applications, especially in the applications which need to identify multiple objects simultaneously. Firstly, by using a scene of portal RFID application, a novel method of modeling the RFID system in the context of multi-object identification under the application constraints is put forward. Different from the RFID benchmarking test, in the proposed method, all the complex effects of the environment factors are treated as the disturbance of the RFID system, then Support Vector Machine (SVM) is employed to learn the model of multi-object identification RFID system. The proposed method can build a performance prediction model by using as less RFID data as possible. And the model built above can bring high accurate prediction data under a range of disturbance, which gives the theoretical support for the optimization of RFID system in real applications. Secondly, a Mixture-Reality (MR) RFID simulation system based on the learnt RFID model is contributed. This simulation system can substitute the real RFID system to be used for designing the deployment of the real RFID application system, as well as for testing the reliability of the RFID application software system. The simulating method learns the model online. It can achieve 90% prediction accuracy with 90% reduction in both time and storage space compared with the test performance of system with real hardware and environment. It provides technical support to enhance the efficiency of RFID system deployment in real applications. Thirdly, this paper applies the RFID model learnt above to the multi-antenna placement of the RFID system. A redundancy antenna placement method based on genetic algorithm (GA) is proposed, which is used to optimize the read range of the RFID antennas. In this algorithm we use the prediction results of learnt model as the fitness of the individual, and optimize the search-space using heuristic search method. Then the designed GA is applied to optimize the RFID portal system with three reader antennas. C...
馆藏号XWLW1757
其他标识符200918014629079
语种中文
文献类型学位论文
条目标识符http://ir.ia.ac.cn/handle/173211/6444
专题毕业生_博士学位论文
推荐引用方式
GB/T 7714
程虹. 多目标识别RFID系统优化关键技术研究[D]. 中国科学院自动化研究所. 中国科学院研究生院,2012.
条目包含的文件
文件名称/大小 文献类型 版本类型 开放类型 使用许可
CASIA_20091801462907(3823KB) 暂不开放CC BY-NC-SA请求全文
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[程虹]的文章
百度学术
百度学术中相似的文章
[程虹]的文章
必应学术
必应学术中相似的文章
[程虹]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

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。