Deep Learning for GPS-Denied SAR Image Focusing and Vehicle Trajectory Estimation


Christopher Beam (University of North Carolina at Charlotte), Andrew R. Willis (University of North Carolina, Charlotte), Kevin M Brink (Air Force Research Laboratory)
The 35th British Machine Vision Conference
PDFPosterVideo (Right click to download)

Abstract

This article integrates AI algorithms into a GPS-denied Synthetic Aperture Radar (SAR) processing framework to transform coherent electromagnetic (EM) radiofrequency (RF) backscatter into 2D scene images through rain, fog, smoke, clouds, and other atmospheric phenomena that visible light imagery cannot penetrate. The work extends prior efforts on the NP-hard problem that seeks to solve the trajectory of the radar platform during the SAR measurement from highly inaccurate knowledge of the required measurement geometry where the error significantly exceeds the nominal wavelength of the radar. Our approach uses massively parallel GPU acceleration to implement search and focusing algorithms to quickly find candidate solutions. The approach is demonstrated to converge and yield accurate trajectory estimates and focused SAR images for an experiment consisting of 1,200 different trajectories. These results significantly improve over previously published results. Gains are made possible by two new components to the algorithm: (1) a new performance functional that is better at identifying candidate solutions from many thousands of possible solutions and (2) a new AI-based method to reliably detect correct solutions from the list of candidate solutions.

Citation

@inproceedings{Beam_2024_BMVC,
author    = {Christopher Beam and Andrew R. Willis and Kevin M Brink},
title     = {Deep Learning for GPS-Denied SAR Image Focusing and Vehicle Trajectory Estimation},
booktitle = {35th British Machine Vision Conference 2024, {BMVC} 2024, Glasgow, UK, November 25-28, 2024},
publisher = {BMVA},
year      = {2024},
url       = {https://papers.bmvc2024.org/0939.pdf}
}


Copyright © 2024 The British Machine Vision Association and Society for Pattern Recognition
The British Machine Vision Conference is organised by The British Machine Vision Association and Society for Pattern Recognition. The Association is a Company limited by guarantee, No.2543446, and a non-profit-making body, registered in England and Wales as Charity No.1002307 (Registered Office: Dept. of Computer Science, Durham University, South Road, Durham, DH1 3LE, UK).

Imprint | Data Protection