Dynamic Cluster Structure for Object Detection and Tracking in Wireless Ad-Hoc Sensor Networks
Dynamic Cluster Structure for Object Detection and Tracking in Wireless Ad-Hoc Sensor Networks
Xiang Ji∗ , Hongyuan Zha∗ , John J. Metzner∗ , and George Kesidis†
∗ Department
of Computer Science and Engineering Pennsylvania State University, University Park Email: {xji, zha, metzner}@cse.psu.edu † Department of Electrical Engineering Pennsylvania State University, University Park Email: kesidis@engr.psu.edu leads to large amount of information delivery. In general, the cost for computation locally is much lower than that for communication for a sensor. In order to prolong the life of a wireless sensor network, it is desirable to minimize the communication costs in operating the sensor network. The continuous objects can be some poison gas or biochemical materials, which are released maliciously from some specific source and then slowly diffuse. Although they are usually in three dimension space in reality, it is generally more interesting to know their locations and spread in a two dimension plane of the earth’s surface. Figure 1 (a) illustrates the detection and tracking of three continuous objects. Sensors around the objects detect and track their boundaries and send the boundary information to the sink in hop-by-hop fashion along the dashed lines. Then, the sink relies the boundary information to outside computers or the Internet. The most efficient manner to identify the existence of the objects is to probe their boundaries. The boundary of a continuous object is consecutive and enclosing the continuous object, inside which the content of the target material per unit region is nearly homogeneous and higher than a threshold. We hope to find portion of objects’ boundaries that are inside the area with sensors deployed. Current signal processing techniques enable sensors to precisely detect the content of target material at their nearby region. However, it is desirable to propose some infrastructure to facilitate sensor collaboration. The continuous objects
Xiang Ji∗ , Hongyuan Zha∗ , John J. Metzner∗ , and George Kesidis†
∗ Department
of Computer Science and Engineering Pennsylvania State University, University Park Email: {xji, zha, metzner}@cse.psu.edu † Department of Electrical Engineering Pennsylvania State University, University Park Email: kesidis@engr.psu.edu leads to large amount of information delivery. In general, the cost for computation locally is much lower than that for communication for a sensor. In order to prolong the life of a wireless sensor network, it is desirable to minimize the communication costs in operating the sensor network. The continuous objects can be some poison gas or biochemical materials, which are released maliciously from some specific source and then slowly diffuse. Although they are usually in three dimension space in reality, it is generally more interesting to know their locations and spread in a two dimension plane of the earth’s surface. Figure 1 (a) illustrates the detection and tracking of three continuous objects. Sensors around the objects detect and track their boundaries and send the boundary information to the sink in hop-by-hop fashion along the dashed lines. Then, the sink relies the boundary information to outside computers or the Internet. The most efficient manner to identify the existence of the objects is to probe their boundaries. The boundary of a continuous object is consecutive and enclosing the continuous object, inside which the content of the target material per unit region is nearly homogeneous and higher than a threshold. We hope to find portion of objects’ boundaries that are inside the area with sensors deployed. Current signal processing techniques enable sensors to precisely detect the content of target material at their nearby region. However, it is desirable to propose some infrastructure to facilitate sensor collaboration. The continuous objects