Modelling and Analysis of Semantically Enriched Simplified Trajectories Using Graph Databases
Geospatial databases are utilized in modelling the huge volume of spatial-temporal data generated by tracking moving objects equipped with positioning devices. This data can be used in performing trajectory analysis such as optimum path finding or identification of collision risk. At the same time, this massive data becomes difficult to handle using traditional databases as raw trajectories contain a lot of unnecessary data points. Thus, trajectory simplification techniques are applied to reduce the number of vertices representing a trajectory. However, elimination of intermediate points by simplification process leads to a loss of semantics associated with the trajectories. These semantics are dependent on the application domain. For example, a trajectory of a moving vessel can convey information about time, distances travelled, bearing, or velocity. This research proposes a graph data model that enriches the simplified geometry of trajectories with the semantics lost in the simplification process. Raw trajectories, initially modelled and stored in a PostgreSQL/PostGIS database, are simplified according to both their spatial and temporal characteristics using the Synchronized Euclidean Distance (SED), while the Semantically Enriched Line simpliFication (SELF) data structure is adopted to preserve the semantics of the vertices eliminated in the simplification process. Then, enriched simplified trajectories are transferred to a Neo4j database and modelled in terms of nodes and edges using graphs. Trajectories can then be further processed using Cypher query language and Neo4j spatial procedures. A visualization tool has been developed on top of Neo4j graph database to support the semantic retrieval and visualization of trajectories.