Regularization techniques for floor plan estimation in Radio Tomographic Imaging

Abstract

An important potential application of Radio Tomographic Imaging is the estimation of building floor plans and interior features in a wireless network. This paper proposes a novel technique for the regularization of solutions to RTI problems with the goal of enhancing building floor plan images. This is done by exploiting a-priori information about the structure, and therefore spatial covariance, of typical buildings. An elliptically-shaped covariance function is used to better model the spatial covariances of pixels in floor plan images. Our approach is verified through simulation, showing improvements over the widely applied Tikhonov regularization method. Our results show reductions in mean-squared error, while also increasing SSIM, a popular measure of subjective image quality.