Image Processing Reference
In-Depth Information
queries. We advise the reader to refer to this table while reading Section .. which provides details
of the various projects/papers.
4.4.2 Essential Conceptual Building Blocks
While there are a host of features that are necessary to form a full-fledged distributed data extraction
system for WSNs, we highlight four essential conceptual building blocks in Figure . that have been
widely mentioned in the existing literature. Below, we provide a very brief overview of each building
block and describe their specific roles. his is followed by a discussion detailing how the various com-
ponents are related to each other thus enabling the reader to visualize how the various components
fit into the “bigger picture.”
•
In-network processing
: his involves moving various types of computation that are tradi-
tionally done on the server-side to within the sensor network itself. It is generally used
for filtering and processing of messages flowing within the network thus preventing the
transmission of unnecessary information.
•
Acquisitional query processing
: Energy consumption in sensor nodes depends on two
main factors: Operation of the transceiver and operation of the sensors. Acquisitional
query processing helps minimize energy consumption by targeting the sensors, i.e., sam-
pling of the various attached sensors is carried out in an energy-efficient manner. For
example, a user may be presented with sensor readings generated using certain statistical
methods rather than actually sampling sensors.
•
Cross-layer optimization
: Unlike conventional computer networks which can generally
be used to perform a wide variety of tasks, WSNs are usually designed for a particular
application. his makes it possible to design the various components of the WSN archi-
tecture, e.g., the routing and MAC protocols specifically for the application in mind. his
could mean that the MAC and routing protocols may be able to adapt to the changing
requirements of the application. This is fundamentally different from the conventional
OSImodelusedfortypicalnetworkswherethelower-layerprotocolsoperatecompletely
independently from the higher-layer protocols.
•
Data-centric data/query dissemination
: Unlike conventional routing protocols which do
not actually bother about the content of the data message being transmitted, the path
taken by a message being routed by a data-centric data/query dissemination protocol for
a WSN is completely dependent on the contents of the message. his allows messages to
be routed more efficiently.
While Figure . illustrates the pertinent features of every building block, it does not illustrate the
relationship between them. We present this relationship in Figure .. Probably the most notice-
able feature is that we have placed acquisitional query processing, cross-layer optimization, and
data-centric data/query dissemination all within the class of in-network processing. The reason
for this can be traced back to the way data is usually collected in conventional databases using
the warehousing approach. Under this model, data is initially collected from various sources (e.g.,
sensors with wireless transmitters) and stored at a central location. he data is then processed cen-
trally to extract the required information. This model is highly unsuitable for WSNs as it involves
excessive transmission overhead and also prevents users from accessing real-time, streaming data.
The only viable alternative is to migrate from off-line processing to processing the data within
thenetworkasclosetothedatasourceaspossible.hepracticeofprocessingdataatthesensor
nodes themselves is known as in-network processing and can result in a significant reduction in
the number of messages transmitted. he ability to perform in-network processing is a cornerstone
of WSNs.
