A Sensed-Point-Oriented Geographic Routing for Camera Networks Hiroki Ishizuka†, Yoshito Tobe†† † Department of Information and Media Engineering, Tokyo Denki University †† Department of Information Systems and Multimedia Design, Tokyo Denki University Email: † isi@u-netlab.jp , †† yoshito-tobe@osoite.jp [Authors are co-affiliated with CREST, Japan Science and Technology Agency] Abstract Camera networks have been paid much attention in sensor networks since those network are useful to visualize a phenomenon sensed by other sensors for temperature, humidity, and light. When a user selects a point of an area deployed cameras in an application, a geographic routing is best used as a routing of the camera network. In conventional geographic routing, the destination of a message should be the location of a node. However, this is not suitable for sensors that cover a wide or directed sensing area such as cameras. To cope with this problem, we propose a modified geographic routing scheme called SenriGan. SenriGan accommodates specifying a sensed point as a destination instead of using a location of a node. In this demonstration, we show a prototype system of SenriGan. 1. Introduction Multimedia sensors such as a Web camera have rapidly been developed and disseminated in recent years. They can be utilized in various scenarios such as life-saving at a disaster area, observation of abnormal weather, and security in our urban life. Queries issued by users of these applications are often associated with two attributes. One is time when a data is sensed. For example, researchers of weather need a time-series observation. The other is a location where data is sensed; a security guard of a building needs to identify the exact location about which a suspicious person is wandering. Time and location are important attributes to search user's required data in sensor networks. Therefore, a developer of sensor networks should design a specialized query routing protocol to enable the two attributes to be processed. Geographic routing is designed as a specialized query routing for a location attribute and has been paid much attention in query-based sensor networks. It assumes that sensor nodes are installed onto a field with Euclidean coordinates and can route messages from a sensor node to another only using their spatial coordinate. Each sensor node manages its own and its neighbors location information in its neighbor table. When a node receives a packet, the node finds and forwards the packet to the geographically nearest node to the destination in the table. Eventually, the packet is delivered to the location specified by a user, thus enabling the user to obtain the data at the location. In conventional geographic routing, the destination of a message should be the location of a node. However, there is also a case in which a user wants data at a location where no node exits. Assume that a region is covered by several camera nodes. A user may specify a sensed point which is observed by one or more sensor nodes. Conventional geographic routing does not accommodate such a case. To cope with this problem, we propose a modified geographic routing scheme called SenriGan. SenriGan is an enhanced geographic routing scheme which accommodates specifying not a location of a sensor node but a directly sensed point as the destination. A node of SenriGan calculates its own sensing area beforehand. Each node of SenriGan broadcasts its sensing area and own location to its neighbor nodes in its beacon packets and manages the area of neighbors in its neighbor tables. Basically, these nodes create a planar graph such as Gabriel graph and Relative nearest graph among their neighbor nodes using their locations. A source node sends a query packet including a sensed point as a destination. Like GPSR [1], SenriGan uses greedy and perimeter (or face) forwarding. However it differs from GPSR in that it does allow turning around a face routing surrounding the sensed point to investigate the existence of sensors that capture the sensed point. In this demonstration, we show our prototype system using personal computers and Web cameras. 2. Background and Goals 2.1. Problems in conventional geographic routing schemes First, unlike a sensor of temperature or pressure at point, a camera can obtain data covering some wide area. Conventional routing schemes cannot be applied to destinations inside a sensing area. Second, a camera is also associated with direction. Suppose that one desires to observe on image at some location. However, the image changes depending on the direction of sensing. 1-4244-1241-2/07/$25.00 ©2007 IEEE 209