Autofocusing for RF tomography using particle swarm optimization

RF tomography employs geometric diversity to obtain high resolution images of targets, potentially with narrow band waveforms. In order to properly capitalize on this spatial diversity, precise knowledge of the sensor positions is required to sub-wavelength accuracy. While GPS can provide good estimates for these locations, remaining uncertainties distort the images produced and limit their utility. In this paper, an autofocusing approach is proposed to compensate for these uncertainties. A cost function is established for the image by evaluating the sharpness of image sub-regions in the neighborhood of known point-like targets. These targets can be identified either from a knowledge base or by examining preliminary imaging results. The cost function is then minimized by adjusting the estimated sensor positions using particle swarm optimization. The algorithm is shown to significantly improve the quality of RF tomographic images, allowing the separation of closely spaced targets that were severely distorted in uncorrected images. Results are provided for both simulated and measured data. A discussion of potential enhancements and wider applications of the algorithm is also included.