3-D Application-Oriented Visual Correlation Model in Wireless Multimedia Sensor Networks

摘要

Wireless multimedia sensor networks comprise a large number of sensor devices and obtain multiple types data from field of interest. In a visual sensor network, there exist visual correlation (VC) characteristics among images observed by camera-equipped sensors with overlapped field of views. To describe the characteristics, the conventional method is based on image matching. However, it is too complex to be applied to resource-constrained cameras. In this paper, based on 3-D sensing model and spatial geometric transformation, a 3-D application-oriented VC (3-D-AOVC) model is designed to quantitatively describe the correlation characteristics according to the specific application requirement. Furthermore, a 3-D VC coefficient (3-D-VCC) algorithm, which is aimed at the measurement of data redundancy is provided to elaborate the process of 3-D-AOVC model. Finally, utilizing the data redundancy measured by provided 3-D-VCC algorithm, a VC-based distributed energy efficient (VCDEE) scheme is devised. Experimental results show that the 3-D-AOVC model can accurately model the VC characteristics and the VCC algorithm has lower computational complexity and energy consumption than existing algorithm. Further simulation results demonstrate that the VCDEE scheme outperforms the state of the art in terms of energy conservation and balance.