A survey of visual sensor network platforms
Bulent Tavli & Kemal Bicakci & Ruken Zilan &
Jose M. Barcelo-Ordinas
# Springer Science+Business Media, LLC 2011
Abstract Recent developments in low-cost CMOS cameras have created the opportunity
of bringing imaging capabilities to sensor networks. Various visual sensor platforms have
been developed with the aim of integrating visual data to wireless sensor applications. The
objective of this article is to survey current visual sensor platforms according to in-network
processing and compression/coding techniques together with their targeted applications.
Characteristics of these platforms such as level of integration, data processing hardware,
energy dissipation, radios and operating systems are also explored and discussed.
Keywords Visual sensor networks . Embedded systems . Vision platforms . Image
acquisition . System integration . IEEE 802.15.4 . IEEE 802.11 . Embedded system interfaces
1 Introduction
Wireless Sensor Networks (WSNs) are becoming a mature technology after a decade of
intensive worldwide research and development efforts. WSNs primary function is to collect
and disseminate critical data that characterize physical phenomena around the sensors [7,
31, 55]. Availability of low-cost CMOS cameras has created the opportunity to build low-
cost Visual Sensor Network (VSN) platforms able to capture, process, and disseminate
visual data collectively [44]. Emerging applications such as visual surveillance and vehicle
traffic monitoring can be enriched with visual data.
Multimed Tools Appl
DOI 10.1007/s11042-011-0840-z
B. Tavli (*) : K. Bicakci
TOBB University of Economics and Technology, Ankara, Turkey
e-mail: btavli@etu.edu.tr
K. Bicakci
e-mail: bicakci@etu.edu.tr
R. Zilan : J. M. Barcelo-Ordinas
Technical University of Catalonia, Barcelona, Spain
R. Zilan
e-mail: rzilan@ac.upc.edu
J. M. Barcelo-Ordinas
e-mail: joseb@ac.upc.edu

High-resolution images and videos require complex compression and coding algorithms and
high bandwidth usage. These requirements imply that visual sensor nodes dissipate significantly
higher power than scalar sensor nodes. Furthermore, sensors used in relaying data traffic have to
be capable of buffering large number of packets. To reduce the bandwidth utilization, vision-
processing techniques capable of reducing the amount of traffic by intelligently manipulating
the raw data can be used at the cost of allocating more computation resources.
Main motivations of this study are two-folds. First, we discuss and compare VSN platforms
based on processing and vision computing techniques. VSNs are distinguished fromWSNs by
their capability to present information-rich visual data, however, effective use of visual data
depends on the use of appropriate processing techniques. There are differences in the
techniques implemented or realizable in current VSN platforms because technical challenges
involve many design trade-offs and decisions are usually enforced by hardware limitations and
energy constraints. Second, we investigate the capabilities and limitations of VSN platforms in
detail. We believe that protocol, algorithm, and application development for VSNs cannot be
of practical use unless the underlying enabling technologies and platforms are well
understood. Other surveys (such as [5, 44]) give a more general perspective on VSNs.
We refer readers interested in specific VSN topics which are not in the scope of our survey to
other relevant surveys in the literature written on various aspects of VSNs (general overview
[44], challenging issues [5], multimedia streaming [29], security [16], cross-layer design [51]).
The rest of this paper is organized as follows: Section 2 presents the reasons for
designing special platforms for VSN applications. Section 3 provides an overview of image
compression, video coding, and vision computing techniques used in VSN platforms.
Section 4 presents an overview of the currently available VSN platforms. Section 5
categorizes these platforms according to integration level, data processing hardware, energy
dissipation, radios, operating system and software architecture, hierarchy, and applications.
We present a critical evaluation of VSN platforms and discuss open problems in Section 6.
Section 7 presents conclusions of this survey.
2 Platform design for VSNs
The main differences between WSNs and VSNs are in acquiring, processing, and
transferring data. Visual data requires higher bandwidth usage due to the amount of data
to be transmitted and higher power consumption due to the complexity of coding and vision
processing algorithms and high volume data acquisition. The spectrum of WSN platforms
ranges from lightweight platforms (e.g., Mica2) through intermediate platforms (e.g., Yale
XYZ) to PDA-class platforms (e.g., Intel Stargates) [20]. Table 1 presents the features of
several WSN platforms. Lightweight platforms (e.g., Mica2, Mica2Dot, MicaZ, and Telos)
are highly resource-constrained; thus, they are only suitable for simple sensing and
detection tasks. The Yale XYZ platform, which is a typical example of intermediate
platforms, has more memory and processing resources than the lightweight platforms. At
the higher performance set, there is PDA-class platforms (e.g., Stargates) which are more
powerful than the intermediate platforms but also consume more power.
Vision computing techniques can reduce the amount of visual data to be transmitted at
the cost of more computation. To illustrate this fact, let us consider an application that
controls registration plates of vehicles traveling in an urban area. Such application is useful
for the implementation of local administration policies to reduce the amount of pollution,
by alternating the circulation of vehicles based on the parity of their number plates (this
kind of a policy has been introduced for the Beijing Olympic Games). It is possible to use a
Multimed Tools Appl