Video-rate Radio Tomographic Imaging with Virtual Fingerprint

Radio tomographic Imaging (RTI) provides highprecision localization of targets in a monitoring area by exploiting the relationship between target location and radio link attenuation. Most recent RTI approaches leverage pre-processing and data-driven strategies to improve the accuracy of moving targets. However, these approaches suffer from limitations such as training time and low frame rate, which constrain RTI to quickly and accurately localize the moving target. To address these limitations, we develop a coarse-to-fine RTI (CF-RTI) system that achieves a maximum frame rate of 29 frames per second (fps), surpassing the minimum requirement of 16 fps for a typical RGB camera while maintaining remarkable accuracy and resolution. First, CF-RTI enables high frame rate imaging by constructing a visual fingerprint database that establishes connections between target locations and potentially affected links, thereby enabling multi-step acquisition, reconstruction, and estimation. Second, the multi-step processing exploits the knowledge gained from localization efforts during acquisition processing, enabling high-quality imaging by rapidly identifying and collecting measurements from the target's affected links. Finally, the experimental results show that our CF-RTI can achieve centimeter-level positioning accuracy and a maximum frame rate of 29 fps while using only 9% of the traditional RTI measurements. The maximum frame rate of our CF-RTI is comparable to that of standard RGB cameras. These results suggest that our CF-RTI has excellent potential for implementing RTI techniques in real-world scenarios