Link Selection in Radio Tomographic Imaging with Backprojection Transformation

Abstract

Multi-path interference in Radio Tomographic Imaging(RTI), often brings unpredictable degeneration to the reconstructed image and degrades the accuracy of Device-Free Localization(DFL). By analyzing the reconstruction process of RTI, this paper certifies that the shadow fading can be transformed as a linear combination of the contribution of RF links. This transformation named backprojection indicates that the selection of informative RF links is helpful to resist the multi-path noise. Then a method based on Bayesian Compressive Sensing(BCS) and backprojection is proposed to figure out the contributive RF links and reconstruct the image. Besides, by transforming the reconstruction issue of high-dimensional image into the analysis problem of low-dimensional measured data, the proposed method also decreases the time complexity of BCS without reducing the accuracy. The experimental results show the effectiveness and practicability of the method in RTI and DFL.