Gradient Iteration Regularization to Solve Radio Tomographic Imaging Model in UHF RFID Scenarios

Abstract

Radio tomographic imaging (RTI) is one of potential device-free localization (DFL) technologies. Solving RTI model is an ill-posed problem that the number of measurements is less than pixels in a reconstruction image. Currently, the Tikhonov regularization based least square method (TRLS) is used to handle the problem, but its artificial regularization parameter disturbs initial RTI model and results in low localization accuracy. Gradient iteration regularization method (GIRM) is proposed to solve RTI model and has the advantages of high accuracy, low storage cost, good convergence and high stability. Unlike the TRLS, it uses mathematical deviation to obtain its parameter required for the solution and adopts several iterations to further weaken the influence of the parameter on the model. The image processing eliminates false targets and artifacts in a reconstruction image to obtain the number and the locations of real targets. Simulation shows that the localization performance of proposed method is higher than TRLS.