Gini Index Based Sparse Reconstruction for Radio Tomographic Imaging

Radio tomographic imaging (RTI) is a computational imaging technique that estimates the distribution of shadowing loss caused by obstructions in a monitored area using received signal strength (RSS) measurements between radio frequency (RF) network nodes. However, finding a solution to RTI under noise interference is generally an ill-posed inverse problem, and additional measures are required to reduce noise and achieve the desired image properties. The sparse distribution of targets in the monitored areas is intuitively an important prior for accurate image reconstruction. In this article, we propose a Gini index (GI)-based sparse reconstruction approach for RTI that fully leverages sparsity for high-quality imaging. Based on the linear cumulative shadowing model, RTI is formulated using the classical GI from economics as the sparsity measure to construct the optimal objective function. The GI-based objective function is quasi-convex, and RTI is solved by the simultaneous perturbation stochastic approximation (SPSA) algorithm. Experimental studies in three different environments are conducted to validate the superior performance of the proposed approach with respect to commonly used regularization and sparsity measures.