SimData-NRTI: A Novel Radio Tomographic Imaging Based on the Simulation-Fingerprint Database for Multitarget Localization

Device-free localization (DFL) is a promising positioning technique that can localize targets without their awareness and any device. Radio tomographic imaging (RTI) is a favorable DFL method that can achieve real-time localization without prior data collection. However, current RTI systems are incapable of achieving accurate indoor multitarget localization due to the ill-posed nature of its positioning model and environment complexity. To eliminate the negative influence, we propose a novel RTI based on the simulation-fingerprint database for multitarget localization (SimData-NRTI). Here, the database, utilizing RTI reconstructed image as the fingerprint, records all images corresponding to the target appearing at all potential locations. SimData-NRTI extracts target-edge samples from the measurement-reconstructed image and database fingerprint according to the gray gradient and the distance between the image’s edge pixels, respectively. Then, we develop a modified Hungarian method to match these sample histograms with measurement-reconstructed image and the fingerprint, resulting in original matching pairs of these samples with them. Since the interference from actual indoor environment can induce distortions in the locations and shape of targets within the measurement-reconstructed image, we adopt the thin-plate spline (TPS) to deform the image’s edge for further seeking higher matching pairs from the original pairs. Finally, we design a similarity evaluation method to assess the similarities between the measurement-reconstructed image and database fingerprints based on the histogram difference and TPS’s characteristics. The coordinates of targets corresponding to the fingerprint with the highest similarity are the estimated locations of targets to be localized. The simulation and experiment show that SimData-NRTI outperforms the other methods in terms of the positioning accuracy and target identification accuracy rate (TIAR).