Data Management
Figure 2 illustrates our approach to
managing both live and persistent smart
farm sensor data.
GSN is fed with the sensor data
collected from the farm. In addition,
Taggle cattle-location data is fed in
by polling a website. Virtual GSN
sensors are used to provide enhanced
streaming output for further processing.
Through these virtual sensors,
• summaries of sensor data are translated
to RDF and persisted in a
Virtuoso triple store;
• various algorithms, implemented in
Java or R, are deployed to consume
real-time sensor data and produce
value-added streams; and
• semantic event descriptions are
processed to generate alerts.
The smart farm control portal lets
farmers develop personalized event descriptions
in terms of an ontology designed
for their farm. The descriptions
are composed and handled via an extension
of Taylor and Leidinger’s method,6
whereby users are guided to create class
descriptions and class and property instances
that together describe an event of
interest. Event descriptions can include