【作者】 王标；

【导师】 王波涛；

【作者基本信息】 东北大学 ， 计算机应用技术， 2012， 硕士

【摘要】 近年来,随着无线通信技术、定位技术和智能移动终端的快速发展,基于位置信息的服务(Location-Based Services, LBS)在医疗、物流、交通、军事等诸多领域得到了广泛应用,它能够根据移动对象的位置信息为用户提供个性化的服务。LBS使用有效的移动对象时空索引技术来高效地处理查询请求服务,在这种环境下,移动对象将它们当前位置的信息传递给服务器,服务器以时空查询的形式为用户提供服务,比如范围查询、近邻查询、远邻查询等。这些形式的查询很大程度上需要维护移动对象当前位置的信息,因此针对这些查询,提出高效的时空索引结构是至关重要的。移动对象索引结构除了能够有效地支持查询操作外,还要能够有效地支持更新操作。由于大量移动对象位置的频繁改变,导致基于移动对象位置的索引执行效率非常低下。如何降低移动对象位置改变而带来的索引结构频繁更新成为一个新的研究热点。本文的一个主要工作就是提出了提高移动对象索引更新效率的两种策略,首先就是将整个区域划分为很多网格,将存在移动对象的区域的统计信息合并为特征向量,并提出了一种新型的移动对象索引结构,即基于ELM区域分类的移动对象索引结构。在这种体系结构中,使用R树索引那些存在移动对象的网格,而不是直接对移动对象索引,利用ELM对区域进行分类；然后本文给出了基于ELM区域分类移动对象索引的更新算法,并在此算法的基础上提出了更新策略；最后通过实验对比分析,验证了本文提出的索引结构和更新策略能较好地降低索引更新的频率。尽管与位置相关的查询及相应处理技术的研究是当前研究的热点之一,但当前查询技术仍然不能满足用户的需求。基于这方面的考虑,本文提出了一种基于距离的移动对象间的状态查询方式,例如渐近查询和渐远查询等,并提出了朴素的基于距离的移动对象间状态查询算法,从而计算得到移动对象间的状态信息。朴素算法简单易行,但是执行过程会很繁琐,而极限学习机(ELM)的分类速度非常快,本文将移动对象间的距离作为特征值,利用ELM对移动对象间的状态进行分类,并通过实验将朴素算法和基于ELM分类的查询算法进行对比,验证了基于ELM的移动对象间状态查询算法能够在很大程度上提高查询的运行效率。更多还原

【Abstract】 Recently,with the rapid development of wireless communication technologies, positioning technologies and smart mobile devices, Location Based Service (LBS) is widely used on fields like health care, material flow, traffic, military, etc. LBS can provide personalized service for users according to moving objects’location information. LBS uses effective indexing technology in spatial and temporal to process query request service. In such environment, moving objects send their current location information to the server, then the server provides users service in the form of spatial-temporal query, such as range query, nearest/farthest neighbor query, etc. Such kinds of queries rely heavily on the maintenance of the current locations of the mobile objects. Aiming to these queries, providing effective spatial-temporal index structure is of curial importance.Moving objects index structure should effectively support update operation as well as effective query operation. Due to the frequent change of abundant of moving objects’location information, it causes the bad indexing efficiency of moving objects. How to reduce the update frequency of index structure caused by the change of moving objects’location has become the new research topic. One of the main jobs is proposing several strategies to enhancing the update efficiency of moving objects. Firstly, the entire region is divided into lots of girds, then we combine the statistic information of regions with mobile objects into feature vector. Secondly, we propose a novel index structure for mobile object, called index structure on moving objects based on ELM. In such index structure, R tree is used to index occupied regions instead of the mobile objects themselves and Extreme Learning Machine(ELM) is used to classify the regions. Thirdly, the paper provides an update algorithm on mobile objects based on region classification using ELM, and then presents several update strategies. Lastly, by analysis of comparing experiments, we verify that the proposed index structure and update strategies better reduce update efficiencyThe research about the relevant location based query and the corresponding processing technology is one of the current hot topics, but the current query technique still can’t satisify user. Based on such consideration, our paper presents a state query style among moving objects, such as asymptotic query, etc.We then proposes naive algorithm about the state query on moving objects, which can calculate state information among mobile objects. Becoming of the lowness of efficiency of execution on naive algorithm, the fastness of classification speed on ELM, the paper makes the distances between moving objects as feature values, then uses ELM to classify state among mobile objects; Lastly, by contrastive experiment between naive algorithm and query algorithm using ELM classification, the state query algorithm among moving objects based on ELM can largely improve the performance efficiency of query.更多还原