摘要： Due to energy and throughput constraints of visual sensing nodes, in-node energy conservation is one of the prime concerns in visual sensor networks (VSNs) with wireless transceiving capability. To cope with these constraints, the energy ef?ciency of a VSN for a given level of reliability can be enhanced by recon?guring its nodes dynamically to achieve optimal con?gurations. In this paper, a uni?ed framework for node classi?cation and dynamic self-recon?guration in VSNs is proposed. The proposed framework incorporates quality-of-information (QoI) awareness using peak signal-to-noise ratio-based representative metric to support a diverse range of applications. First, for a given application, the proposed framework provides a feasible solution for the classi?cation of visual sensing nodes based on their ?eld-of-view by exploiting the heterogeneity of the targeted QoI within the sensing region. Second, with the dynamic realization of QoI, a strategy is devised for selecting suitable con?gurations of visual sensing nodes to reduce redundant visual content prior to transmission without sacri?cing the expected information retrieval reliability. The robustness of the proposed framework is evaluated under various scenarios by considering: 1) target QoI thresholds; 2) degree of heterogeneity; and 3) compression schemes. From the simulation results, it is observed that for the second degree of heterogeneity in targeted QoI, the uni?ed framework outperforms its existing counterparts and results in up to 72% energy savings with as low as 94% reliability.