Real Time Image Analysis for Infomobility Massimo Magrini, Davide Moroni, Gabriele Pieri, and Ovidio Salvetti Institute of Information Science and Technologies (ISTI), National Research Council of Italy (CNR), Pisa, Italy name.surname@isti.cnr.it Abstract. In our society, the increasing number of information sources is still to be fully exploited for a global improvement in urban living. Among these, a big role is played by images and multimedia data (i.e. coming from CCTV and surveillance videos, traffic cameras, etc.). This along with the wide availabil- ity of embedded sensor platforms and low-cost cameras makes it now possible the conception of pervasive intelligent systems based on vision. Such systems may be understood as distributed and collaborative sensor networks, able to pro- duce, aggregate and process images in order to understand the observed scene and communicate the relevant information found about it. In this paper, we inves- tigate the characteristics of image processing algorithms coupled to visual sensor networks. In particular the aim is to define strategies to accomplish the tasks of image processing and analysis over these systems which have rather strong constraints in computational power and data transmission. Thus, such embedded platform cannot use advanced computer vision and pattern recognition methods, which are power consuming, on the other hand, the platform may be able to ex- ploit a multi-node strategy that allows to perform a hierarchical processing, in order to decompose a complex task into simpler problems. In order to apply and test the described methods, a solution to a visual sensor network for infomobil- ity is proposed. The experimental setting considered is two-fold: acquisition and integration of different views of parking lots, and acquisition and processing of traffic-flow images, in order to provide a complete description of a parking sce- nario and its surrounding area. 1 Introduction The application of computer vision methods to the large and heterogeneous amount of information available nowadays is a tricky and complex goal. The first thing to take into account is the platform for which these methods need to be implemented. When a large and powerful platform is available (e.g. desktop PC-like, clouds,...) many different and more powerful methods and algorithms can be implemented. On the other hand, aiming to achieve a low-cost, low-consumption and pervasive implementation, platforms like embedded systems need to be considered. Recently, embedded systems have become widely available along with low-cost camera sensors, thus allowing to design sensor-based intelligent systems centered on image data [1]. These vision systems can be connected wireless in networks, forming the so called visual Wireless Sensor Networks (WSNs). Such systems, formed by a large E. Salerno, A.E. C ¸ etin, and O. Salvetti (Eds.): MUSCLE 2011, LNCS 7252, pp. 207–218, 2012. c  Springer-Verlag Berlin Heidelberg 2012