| People spend most of their time in indoor spaces, e.g., office buildings, shopping centers, airports, subway stations, etc. With the rapid development of wireless communication technologies and novel positioning technologies like Wi-Fi, bluetooth and RFID, obtaining the positions of indoor moving objects has become realistic and feasible. At the same time, the wide use of mobile phones, PDA, and other smart devices, which usually provides a bluetooth and Wi-Fi module, accelerates the development of indoor location-based services. Therefore, how to effectively manage the rapidly increasing indoor moving objects and their locations data has become an important issue in many areas including public security and business services, so that we can support efficient query processing for indoor moving objects and further satisfy the diverse needs of indoor location based services.As the trajectory data of indoor moving objects is a typical type of spatio-temporal data, indoor location based services rely on the effectiveness of spatio-temporal data management. Generally, spatio-temporal data management involves four key issues including spatio-temporal semantics, spatio-temporal data modeling, spatio-temporal data query processing, and spatio-temporal indexing. Existing researches on spatio-temporal data management, which are mainly focused on the moving objects in outdoor spaces, are not suitable for indoor moving objects, because they differ in many aspects such as space constraint, positioning techniques, and distance measurement.In order to effectively support indoor location-based services, it is urgent to study spatio-temporal data management issues for indoor space. A lot of issues need to be investigated, including spatial data model, spatial query processing, spatial indexes, etc.(1) Regarding spatial data modeling, as indoor moving objects are moving in constrained indoor spaces, indoor space ortiented spatial data models that represent the elements in indoor spaces as well as their spatial relationships are critical for indoor location-based services.(2) Regarding spatial query processing, efficient query processing techniques for indoor spaces are the foundation for supporting diverse indoor location-based services. As previous indoor space data models fail to effectively represent indoor spaces, existing studies on query processing for indoor spaces pay little attention to semantically related queries and indoor trajectory analysis.(3) Regarding spatial indexes, as indexes are important for boosting the efficiency of complicated queries on massive data, it has been a research focus that how to build efficient index structures to accelerate the efficiency of indoor query processing.Based on the above-mentioned research issues in indoor spatio-temporal data management, this dissertation conducts an in-depth study on spatio-temporal data management for indoor space. The study is based on the collection of indoor space layout information and moving object location data. We first give a comprehensive introduction to the research background and related technologies, and then analyze the related work including indoor space modelling, indoor query processing, and indoor object indexing. After that, we point out existing problems of previous researches and introduce the main research topics of this study.Particularly, this dissertation first analyzes the semantic features of indoor spaces, and establishes an indoor-space representation model to describe the complicated spatial semantics of indoor spaces. Based on this model, we exploit unified modeling approaches for locations and trajectories of indoor moving objects so as to support various indoor positioning techniques and indoor location-based services. Next, this dissertation studies the effective query processing methods for indoor spaces from two views, namely a location-based service view and a trajectory analysis view. Finally, we explore the indexes for indoor trajectories on flash memory. We propose to use an update buffer to improve the index-update performance.The main contributions of this dissertation are summarized as follows:(1) As existing indoor space representation models lack of the abilities to represent complicated semantic features, this dissertation proposes a novel indoor space multi-granularity grid graph model InMGG_Model. The model introduces a grid type as indoor primitive geometry to express indoor geometric features. It divides a physical indoor space into a collection of indoor grids and builds grid-based directional connected graphs, and then uses a semantic representation layer in addition to the grid graph structure. Further, we give the definitions of indoor direction and constraint rules. The model is able to offer a unified description for complicated spatial semantics of indoor spaces.(2) As previous researches on indoor space query processing cannot support indoor semantic features, this dissertation proposes a new semantic association based searching algorithm for Top-k spatial keyword queries in indoor spaces. Two kinds of query optimization methods are also presented. Based on the in-depth analysis of query associations between indoor spatial semantic objects, our algorithm defines the associated semantic similarity for indoor spatial objects, and combines the query distance with the associated semantic similarity to calculate the query relevance and rank for indoor spatial semantic objects. The query optimization methods propose to prepare an offline table that records the semantic weights for indoor spatial objects. This table is helpful to reduce the time costs to calculate the semantic similarities among objects. Then, we propose to prune the objects based on the indoor object index tree IKB-tree, so as to avoid the redundant calculation for irrelevant objects and therefore satisfy the effectiveness demands of processing indoor spatial semantic Top-k nearest neighborhood queries.(3) As few of previous researches on indoor space query processing are towards trajectory analysis, this dissertation studies the pattern analysis methods for the trajectories generated by indoor moving objects. Based on the special features of indoor trajectory data, our method first gives formulated definitions on indoor trajectory location pattern and trajectory spatio-temporal pattern, and then proposes a novel indoor trajectory location pattern mining algorithm based on pattern growth strategy, thereby avoiding the repeated access to trajectories. Further, we propose an indoor trajectory spatio-temporal pattern mining algorithm based on a two-phase temporal clustering process of the spatio-temporal projected trajectory database, so that we are able to get the indoor trajectory spatio-temporal pattern and support diverse indoor spatio-temporal applications.(4) As existing indoor space indexes for moving objects have poor update performance, this dissertation also proposes a novel indoor moving object trajectory index called OIR-tree. The basic structure of OIR-tree is based on the indoor moving object trajectory model of InMGG_Model, as the InMGG_Model model can support the uniform processing of various indoor positioning data streams. OIR-tree is designed for flash memory. It employs an overflow node and an update buffer schema to improve the update performance.This study is expected to present effective and efficient solutions to a few key techniques in spatio-temporal data management for indoor space. The results of the study can provide technical supports to solve the basic theoretical issues in indoor location-based services and indoor spatio-temporal data management. Our study is also valuable for meeting the diverse needs of indoor location-based services, as well as for improving the efficiency and effectiveness of indoor space related applications. It can also bring some new insights and theoretical clues for the future development of spatio-temporal data management. |
PDF Full Text Request