Database Reference
In-Depth Information
data is produced in the context of a wide variety of applications such as
the following:
A wide variety of mobile devices are now GPS-enabled. This has
lead to unprecedented opportunities in the context of several ap-
plications such as social sensing [4]. GPS data is also available in
the context of many location-aware devices and applications.
The decreasing cost of RFID tags has lead to tremendous volumes
of RFID data. The cost of an RFID tag is now in the range
of under 5 cents. This has allowed cost-effective deployment of
RFID tags on products of even modest price. RFID data poses
numerous challenges because of the tremendous amounts of noise
in the collected data [5].
Numerous military applications use a wide variety of sensors in
order to track for unusual events or activity. This could include
visual or audio cameras, or seismometers for tracking movements
of large objects [9].
Sensors are also deployed in the context of a wide variety of en-
vironmental applications, such as detecting weather and climate
trends [7], and tracking pollution levels in water networks [11].
Sensor data brings numerous challenges with it in the context of data
collection, storage and processing. This is because sensor data processing
often requires ecient and real-time processing from massive volumes of
possibly uncertain data. Some of these challenges may be enumerated
as follows:
Data collection is a huge challenge in the context of sensor pro-
cessing because of the natural errors and incompleteness in the
collection process. Sensors often have limited battery life, because
of which many of the sensors in a network may not be able to col-
lect or transmit their data over large periods of time. The errors
in the underlying data may lead to uncertainty of the data repre-
sentation [8]. Therefore, methods need to be designed to process
the data in the presence of uncertainty.
Sensors are often designed for applications which require
real-time
processing
. This requires the design of ecient methods for stream
processing [1]. Such algorithms need to be executed in one pass of
the data, since it is typically not often possible to store the entire
data set because of storage and other constraints.
