High-resolution imaging using virtual sensors from 2-D autoregressive vector extrapolation

Virtual sensors are used to attain a robust high-resolution imaging capability that detects weak signals in the presence of strong signals, when the sensors are limited in number due to space, weight, power, and cost constraints. Such conditions are becoming commonplace with the influx of smart systems, wireless networks, remote sensing, and autonomous vehicles/systems. The virtual sensor data is created autonomously in real time from the original data using a novel two-dimensional (2-D) Autoregressive Vector Prediction algorithm. A 2-D transform is then applied to the new virtual data set, which includes the original data, to give a robust high resolution imaging capability. Simulations are used to compare this super-resolution capability with a high-resolution technique and the truth, to resolve previously obscured low-level signals in the presence of a dominant source. The virtual sensor data is also compared to the truth data. We also summarize the computational cost and extrapolation stability to achieve this high-resolution capability.