A novel approach for extracting spatial correlation of visual information in heterogeneous wireless multimedia sensor networks

In applied wireless multimedia sensor networks, heterogeneous camera nodes with different sensing capabilities are usually deployed due to their role in enhancing the overall network performance and lifetime. Exploiting the correlation characteristics of the overlapping fields of view of different camera nodes would enable very efficient collaborative in-network processing algorithms. This paper introduces a novel geometrical model to extract the spatial correlation characteristics of heterogeneous camera nodes in wireless multimedia sensor networks, taking into consideration the different sensing radii and the angles of view of the camera nodes. The novelty in the proposed model is in using virtual cameras at the two far ends of the camera’s field-of-view. In order to provide better coverage of the field-of-view and hence better estimation of the correlation characteristics, key points in the observed scene are projected at the virtual cameras; in addition to the physical camera. This is shown to significantly improve the estimation of the spatial correlation characteristics to be almost identical to that extracted by well-known image processing techniques. An analytical closed-form solution of the proposed model is derived and validated and its performance is evaluated and compared against the state-of-the-art models; in terms of correlation characteristics estimating accuracy, visual information gain, and distortion ratio. The experimental and simulation results demonstrate that, compared to similar existing models, the proposed model achieves very accurate estimation of the correlation characteristics and significant improvement on the overall network resource utilization for a negligible increase in the camera node’s computational cost.