TR2015-139
Extended Target Localization with Total-Variation Denoising in Through-the-Wall-Imaging
-
- "Extended Target Localization with Total-Variation Denoising in Through-the-Wall-Imaging", IEEE International Workshop on Computational Advances in Multi-Sensor Adaptive Processing (CAMSAP), DOI: 10.1109/CAMSAP.2015.7383832, December 2015, pp. 445-448.BibTeX TR2015-139 PDF
- @inproceedings{Handa2015dec,
- author = {Handa, H. and Mansour, H. and Liu, D. and Kamilov, U.},
- title = {Extended Target Localization with Total-Variation Denoising in Through-the-Wall-Imaging},
- booktitle = {IEEE International Workshop on Computational Advances in Multi-Sensor Adaptive Processing (CAMSAP)},
- year = 2015,
- pages = {445--448},
- month = dec,
- publisher = {IEEE},
- doi = {10.1109/CAMSAP.2015.7383832},
- url = {https://www.merl.com/publications/TR2015-139}
- }
,
- "Extended Target Localization with Total-Variation Denoising in Through-the-Wall-Imaging", IEEE International Workshop on Computational Advances in Multi-Sensor Adaptive Processing (CAMSAP), DOI: 10.1109/CAMSAP.2015.7383832, December 2015, pp. 445-448.
-
MERL Contacts:
-
Research Area:
Abstract:
We propose a target detection and clutter removal technique for through the wall radar imaging that captures the extended reflections of targets behind the wall and determines target consistency using total variation denoising. Our approach is based on the multipath elimination by sparse inversion (MESI) algorithm which models the clutter removal problem as a structured blind deconvolution problem with sparsity constraints on the scene and the multipath reflections. In this work, we extend the MESI algorithm by incorporating the spatial correlation of extended target reflections into the target detection stage. This in turn improves the clutter mitigation performance by ensuring that a separate convolution kernel is computed for each detected target to match the corresponding multipath reflections. When MIMO measurements are available, we apply total variation denoising on the clutter-suppressed SIMO images followed by incoherent summation to generate a single noise free target image. We present numerical experiments that demonstrate the improved performance of our approach compared to standard MESI and MIMO imaging.