中文摘要：

当前基于压缩感知的传感器网络数据融合方案中,不论数据字段有何特征,均假设网络具有固定而均匀的压缩阈值,从而导致数据通信量过高,能耗浪费较大。提出一种基于多分辨率和压缩感知的数据融合方案。首先,对传感器网络进行配置,以生成多个层次类型不同的簇结构,用于过渡式数据收集,在该结构上,最低层的叶节点只传输原始数据,其他层的数据收集簇进行压缩采样;然后将其测量值向上发送,当母数据收集簇收到测量值时,利用基于反向DCT和DCT模型的CoSaMP算法恢复原始数据;最后,在SIDnet-SWANS平台上部署了该方案,并在不同的二维随机部署传感器网络规模下进行了测试。实验结果表明,随着分层位置的变化,大部分节点的能耗均显著降低,与NCS方案相比,能耗下降50％～77％,与HCS方案相比,能耗下降37％～70％。

英文摘要：

A data aggregation scheme based on multi-resolution with compressed sensing was proposed. Firstly, the network was configured to achieve the muhiple-level and the different types of cluster structure for intermediate data collection, on this structure, the leaf nodes in the lowest level only transmit the raw data. The collecting clusters in other levels perform the compressed sampling and then transmit them to their parent cluster heads. When parent collecting clusters receive random measurements, they use inverse DCT and DCT model based CoSaMP algorithm to recover the original data. The proposed scheme was implemented on a SIDnet-SWANS simulation platform and test different sizes of two-dimensional randomly deployed sensor network. The experiment results show that the substantial energy savings are reported for a large portion of sensors on the different hierarchical positions, ranging from 50% to 77% when compared with NCS, and from 37% to 70% when compared with HCS.