The increase of maritime location-based systems broadcasting information about ship movements is providing large sets of positioning data. Detecting outliers that behave in an unusual way in such large amounts of data is an active research field linked to data mining, statistical analysis and geovisual analyt-ics. Assuming that moving objects following the same itinerary behave in a similar way, these behaviours can be derived by data mining on spatio-temporal databases (ST DB). This allows for the understanding of common trajectories (considered as the normality) over a long period of time from which location dissimilarities should raise attention. Such trajectory analysis tied with a geovisualisation process is essential for safety applications. This allows traffic operators to focus on possible outliers and reduce cognitive load in overcrowded areas. The following details a process to qualify the position and trajectory of a moving object both on spatial and temporal criteria and discusses the benefits of geovisualisation for pattern understanding. Figure 1 presents the functional process used to extract spatio-temporal patterns Spatio-temporal data mining Storage Spatio-temporal analysis Visualisation Acquisition Knowledge database Visualisation of qualified positions and trajectories Display Spatio-temporal qualification Patterns Qualified data Spatio-temporal patterns computation Patterns Similar trajectories cluster, filter, resample Area of interest Homogeneous group of trajectories Real time data integration (AIS positions) Spatio-temporal database Positions Positions Positions Area of interest Traffic monitoring operator 1 3 4 2 5 6 7 8 Figure 1: Functional process from the ST DB and qualify ship positions and trajectories [3]. A data integration step (Fig. 1, step 1) relies on a monitoring system providing real-time AIS data integration [2]. Trajectories derived from AIS data are clustered in order to get a Homogeneous Group of Trajectories (HGT) which is filtered and re-sampled (step 3). Every positions of each trajectory of the HGT are paired to a reference trajec-tory using Fréchet matching [5] to produce clouds of matched positions. A statistical analysis of these clouds gives the median trajectory of the HGT and spatio-temporal normality bounds (step 4) which are combined together to define the spatio-temporal pattern of the HGT. Every trajectory of the HGT is also compared to this pattern in order to compute similarity measurements statistics such as maximum or mean spatial and temporal distance. Finally, this pattern and its statistics are stored in a knowledge database (step 5). These spatio-temporal patterns can be used either for geovisual analysis or to qualify in real-time ship positions and trajectories. In the latter approach, every newly acquired position is matched to a spatio-temporal pattern. The selected spatio-temporal pattern are used to qualify the position and the trajectory (step 6) using fuzzy logic to obtain a combined spatio-temporal similarity rating and visualise it (step 7).