Ontology-based Modeling, Annotation, Storage, And Query Of Semantic Trajectories For Travel Analysis

The accelerated urbanization in China has led to a rapid expansion of activity space for urban dwellers. Intra-city travel has become one of their major daily activities and imposed a great deal of pressure on transportation infrastructure, energy consumption, and residential environment. It is therefore important to understand travel behavior of urban residents for better transportation management and spatial planning. The progress in global positioning system (GPS) and modern mobile communication technologies have made the large-scale acquisition of movement data (trajectory) less of a problem in traffic and travel analyses, whereas the key problem lies in proper management, analysis, mining, and application of these data. The current practice, however, tends to focus on the detection of spatiotemporal characteristics and spatial distribution patterns from trajectory data, much less efforts for incorporating semantics from specific application domains. As a result, the derived patterns are usually too difficult to comprehend by domain experts, leading to inefficiency and lack of usage in domain applications. Breakthroughs are necessary in both theoretical contemplation and methodological development to help this research direction out of the current stalemate.The intention of this dissertation is to change the aforementioned situation by introducing the theory of geographic ontology into the study of trajectories and exploring useful ways of integrating semantic and spatiotemporal information for comprehensive travel data analyses. A three-tier task was attempted in the dissertation: (1) at theoretical level, a trajectory ontology framework was constructed to embrace both spatiotemporal and semantic entities; (2) methodologically, a set of designing tools and implementation procedures were provided for the modeling, annotation, storage and query of a semantic trajectory ontology; (3) at practical level, application examples and case studies related to trajectory analyses were presented for reference in fields of traffic management and spatial planning.Major research contents and conclusions are summarized as follows.Chapter 1 provides a comprehensive evaluation of current progress in the field of geo-ontology research domestically and abroad. Depending on the research domain, studies of geo-ontology can be grouped into three categories:geographic domain ontology, geospatial ontology, and geospatial and temporal ontology. A detailed literature review was conducted on the geospatial and temporal ontology, which was comprised of two sub-ontologies:generic spatiotemporal and domain spatiotemporal. The former emphasizes research on typology and hierarchical framework, without relating to any application domain, whereas the latter is only concerned with requirements of a specific application domain, lacking proper theoretic guidance and flexibility in model representation and extension. This study therefore adopted the strategy of "using generic spatiotemporal ontology as the top ontology and domain spatiotemporal ontology as domain ontology" in the construction of semantic trajectory ontology. Different from most of previous studies in which only the design of theoretical models is of their concern, the implement of the designed model was also presented as part of important work in this study.In Chapter 2, the architectural design of semantic trajectory ontology was presented. First, the ontological framework of SNAP-SPAN was adopted as the top ontology, from which trajectory analysis oriented classes and relation types were abstracted. The representation of spatial, temporal, and domain information as well as their integration were then explored. Due to the fact that the ontological expression can only support binary relations, a number of available spatiotemporal representation strategies were compared, and the four-dimensional (or perdurantist) view was deemed to be best conceptual model for spatiotemporal representation. Further analyses were conducted on the modularized construction of domain ontology, which consists of generic trajectory ontology, geographic domain ontology, and application ontology. Analytic results in this chapter indicated that with the introduction of the top ontology, different types of class (Endurants versus Perdurant) and relations (spatial, temporal, and contextual) can be clearly expressed, and temporal objects, spatial objects, and their relations can be uniformly abstracted. The domain ontology substantiated on this basis has greater potential for extension and sharing, and the modulation for ontological representation may facilitate independency and reusability of individual sub-ontologies.Chapter 3 is devoted to exploration of the annotation procedure for producing semantic trajectories. The general semantic annotation procedures for both personal travel and taxi travel were deduced from their application ontologies established in the last chapter. Activities were treated as the focal point of personal travel in the entire annotation process, following the procedure of "data cleaning→trip segmentation→ travel mode identification→trip purpose extraction". The focus of taxi travel, on the other hand, is on moves, so as to follow the procedure of "trip segmentation→data cleaning→map match→road-based segmentation→speed-based segmentation" for trajectory annotation. On the implementation side, computer algorithms and data workflows were designed and tested, including T-DBSCAN, a trip segmentation algorithm modified from DBSCAN by incorporating the temporal dimension for better results in both correctness and speed. In addition, a geospatially prompted recall interview platform for GPS-based travel survey was developed for quality acquisition of trajectory-related semantics. For the floating car data generated from taxi travel, on the other hand, an off-line map-matching algorithm was developed to handle the problems of GPS data under-sampling and incomplete road networks, and a post-processing procedure was proposed for data conversion to semantic trajectories.The database design and development for the storage of semantic trajectory ontology was presented in Chapter 4. A comparative analysis of four ontology storage models available in Oracle was conducted, and the vertical model was selected for modularized storage semantic trajectory ontology in this project due to its simplicity in tabular structure, easy operation for ontology updating, and support from a variety of semantics software. Oracle 11g was chosen as the semantic database design and implementation platform, and its resource description module was extended to accommodate spatiotemporal components. The projection schema of case data from the relational database model to the ontological database model was explored, and a set of indexing operations for the ontological database were presented.In Chapter 5, empirical cases of semantic queries related to personal and taxi travel are presented within the framework of Oracle semantic query technology. While the results indicated some great advantages in ontology-based representation and operations of semantic information, the query techniques lack support for spatiotemporal and route sequence information, an important aspect of trajectory-oriented queries. To overcome this drawback, a query extension was designed on the basis of Jena Adapter, and case analyses of spatiotemporal query and route sequence query were presented. The analytic results demonstrated that the extended query techniques were able to combine the merits of both quantitative analysis and semantic query. The new query scheme was proven more executable for map matcing of travel routes.