Visualization is a powerful tool in understanding moving object data. There is, however, a lack of common open-source tools that can support users in this task. The main challenges are to provide rich transformations and visualizations through an interactive interface, to help users in exploring, understanding and presenting their moving object data. In this demo, we present MOVE (Moving Objects Visual Exploration), an open-source tool that integrates MobilityDB, a moving object database in PostgreSQL, and QGIS to visualize moving objects. MOVE is capable of querying and displaying moving object data through a simple interface and visualizing both static and animated spatial data in QGIS. We use Danish AIS data to demonstrate the capabilities of MOVE by presenting a set of example queries and visualizations.
Published in the Workshop Proceedings of the EDBT/ICDT 2022 Joint Conference (March 29-April 1, 2022), Edinburgh, UK $ maxime.schoemans@ulb.be (M. Schoemans); mahmoud.sakr@ulb.be (M. Sakr); esteban.zimanyi@ulb.be (E. Zimányi)
© 2022 Copyright for this paper by its authors. Use permitted under Creative CPWrEooUrckReshdoinpgs IhStpN:/c1e6u1r3-w-0s.o7r3g CCoEmmUoRns LWiceonsrekAstthribouptionP4r.0oIncteerenadtiionnagl s(CC(CBYE4U.0)R.-WS.org) 1www.qgis.org 2www.github.com/MobilityDB/MobilityDB 3www.github.com/mschoema/move are then added as layers in QGIS, marked as temporal, and can be animated using the QGIS built-in temporal controller. Additionally, indexes are built on the temporal and spatial columns of these tables. This allows users to scale up their database, without compromising the responsiveness of the display in QGIS.
It is worth mentioning that MOVE chooses by design that all the layers it creates for visualization use native QGIS representations. As such, the user has access to the whole range of styling, annotation, etc. that is available in QGIS to enrich the visualization as needed. The symbology of the created layers is always set to default and can be modified by the user. Each layer also contains all the non-geometry columns returned by the initial query, and can thus also be used by the user as usual. Specifically, columns with scalar temporal properties are also handled by the plugin and can be displayed with the use of the DataPlotly4 plugin. These columns are displayed as 2D line plots, with the time displayed on the x-axis and the scalar values on the y-axis.
of three parts: MobilityDB, QGIS, and the MOVE plugin linking these two systems.
MobilityDB [
QGIS is a geospatial visualization tool. It can connect visualizations, along with the queries that we used to to multiple spatial data stores, including PostGIS, and generate them. These queries and visualizations use a display static geometry objects, such as points, linestrings real dataset of AIS ship trajectories, that is published by and polygons. Using the Temporal Controller, it has the the Danish Maritime Authority5. The data covers one day, capability of displaying animated maps. As such, QGIS September 29, 2020. The file size is 1.9GB and contains has the capability of displaying both static and animated 8M AIS points of 1,788 diferent ships. representations of moving object data. Data is loaded into MobilityDB in the ta
MOVE allows users to write SQL spatio-temporal ble AIS(mmsi integer, trip tgeometry,
queries and to visualize the results in QGIS. These queries centroid tgeompoint). The mmsi attribute is a
are executed by MobilityDB, and the user has thus access unique ship identifier. The trip is a temporal geometry,
to the complete set of operations ofered by PostgreSQL, i.e., moving region, representing the ship movement.
PostGIS and MobilityDB. These include for instance pro- It represents the shape, orientation and movement of
jection, distance, speed, azimuth, temporal aggregations, the ship. The centroid attribute represents the ship
and temporal topological predicates [
The MOVE plugin, displayed in Figure 1, presents a we precompute the result of this operation in column simple but powerful interface to the user. When execut- centroid to simplify the queries. Next, we illustrate ing a SELECT query, the plugin inspects the types of the examples of possible visualizations on the AIS dataset. resulting columns and creates the appropriate layers in QGIS to visualize them. Columns storing PostGIS geometry types are displayed as they would have been by 3.1. Animated Points and Geometries default. If the query returns MobilityDB spatiotempo- Queries 1 and 2 can be used to display columns of temral types, that is, tgeompoint or tgeometry, MOVE poral geometries and points, respectively. Running these creates temporary database tables to approximate these spatiotemporal objects into a representation that uses native QGIS types, such as LinestringM. These tables
5www.dma.dk/SikkerhedTilSoes/Sejladsinformation/AIS/ Sider/default.aspx Query 1 (4s for 100 ships, 40s for 1000 ships): SELECT mmsi, trip FROM AIS; When working with mobility data, errors can occur at multiple stages of the pipeline, and the data received by the database is thus not free of errors. Visualizing this data is essential to not only determine the nature of the errors but also verify that the cleaning process was successful. With the use of the plugin, this cleaning and verification process can be done interactively.
After loading the data in table AIS, we can display the trajectories of the ships using Query 3. Figure 3a displays the result of running this query in the plugin interface. In this figure, we can see long straight lines passing through solid terrain, which are errors in the data. These lines are present because some data points are not placed on the actual trajectory of the ship, but rather on an incorrect location far from the actual position of the ship.
A possible way to remove these errors is to compute the speed of the ship, and remove the segments of the trajectory having a speed higher than a certain threshold, such as 30 m/s. This can be done through the use of multiple MobilityDB functions on temporal points as displayed in Query 4. Figure 3b shows the result of Query 4, and the erroneous lines visible in Figure 3a have indeed been removed. If this were not the case, a new query could have been run to continue this interactive cleaning process.
storing the geometries of a regular square grid enclosing the vessel trajectories. Using the QGIS layer styling, we can then color the grid cells with a high weight. This creates the visualization shown in Figure 4.
The lines shown in Figure 4 display grid cells that were traversed by multiple ships on the same day, which could indicate navigation routes or zones with high trafic.
queries in the plugin interface will automatically add temporal points and geometries, such as distance, intera layer in QGIS with the temporal property activated. section or restriction functions. An example of a complex This displays the ships as animated polygons and points operation on temporal geometries is traversedArea. respectively, and it is possible to interact with the anima- This function computes the area traversed by a temporal tion using QGIS Temporal Controller. geometry as a PostGIS polygon and can be used, for ex
Figures 1 and 2 display a snapshot of the layers result- ample, to compute the closest point of approach of a ship ing from Queries 1 and 2, respectively. Notice that the to a fixed point on land. Query 6 computes the traversed added layer is already marked as temporal, as indicated area of a ship, and the result of this query is displayed in with the clock symbol on the right of the layer names. Figure 6.
Using the Temporal Controller visible at the top of the image, this data can be explored and animated. 3.5. Temporal Attributes
Query 5 (> 1min): SELECT cell, count(*) as weight FROM Grid, AIS WHERE intersects(centroid, cell) GROUP BY cell; Query 6 (2s): SELECT mmsi, trip, traversedArea(trip) FROM AIS WHERE mmsi = 538090508; Query 7 (5s): SELECT port, geom, tcount(atValue(
tintersects(centroid, geom), True)), FROM Ports, AIS WHERE port = ’Skagen’
AND intersects(centroid, geom) GROUP BY port, geom; map and the count of the ships as a stepwise function in a scatter plot. As the last example, the MobilityDB tdwithin operation can be used to find out when two temporal points are within a given distance of each other. Query 8 uses this operation to restrict the trajectories of ships to the times when they are within 100 meters of each other. This can for example be used to display close encounters of two ships at sea. Figure 8 shows a zoom of one close encounter returned by Query 8. The described examples only form a subset of the visualizations possible using the plugin. More complex queries can be also executed to create more advanced visualizations, such as the flow map displayed in Figure 9. This can be achieved by combining both the wide range of operations in MobilityDB as well as the wide range of styling features in QGIS. Indeed, since the plugin adds QGIS layers to visualize the results of the queries, the users are then free to adapt the symbology of the data as they see fit. For example, Figure 1 displays every ship in a random color, and Figure 5 corresponds to Figure 3b with diferent styling, giving a better impression of the density of the trajectories.
integrates with MobilityDB and QGIS. MOVE presents a simple interface to interactively query and visualize spatial and spatiotemporal data. Using the extensive MobilityDB API and the large number of visualization options available in QGIS, the users can easily build rich visualizations. Especially, MOVE creates transformations of the MobilityDB temporal types to be able to build static and animated visualizations of moving objects in QGIS. MOVE delegates the data processing to MobilityDB, which can handle large data sets and ofers a wide range of transformation and aggregation operations in SQL. This solution builds on the open-source stack of PostgreSQL, PostGIS, MobilityDB and QGIS, and is thus available and ready-to-use for its large community of users and developers.