Exploring the spatial correlation in radio tomographic imaging by block-structured sparse Bayesian learning

Abstract

Radio Tomographic Imaging (RTI) is a low-cost computational imaging method realised by the Radio Frequency (RF) signal sensing. The target-induced shadowing effect in the RF sensing network is reconstructed as a probability image to estimate the target's position. Then, the RTI-based Device-free Localization (DFL) is becoming a promising research topic in the Location-based Services applications by the Internet of Things (IoT). However, the multipath interference in the RF sensing network often induces the imaging degradation and decreases the DFL accuracy. To deal with the multipath-induced imaging degradation, considering that the target's shadowing occupies a small spatial range in the RF network and expresses some spatial structure, this article explores the spatial correlation in the target's shadowing. Then, a new RTI reconstruction method based on the Structured Sparse Bayesian Learning is proposed to model the spatial correlation implied in the sparse target's shadowing image. Further, the localisation experiments in actual scenes are conducted to validate the utilisation of the spatial correlation in target's shadowing is able to improve the imaging quality of the RTI system by enhancing the robustness towards the multipath-induced imaging degradation.