Effective Attributed Scattering Center Extraction via Gaussian Representation and Deep Learning
IEEE Transactions on Antennas and Propagation · 2026-04 · 74 (4) · 3559-3574
DOI: 10.1109/tap.2026.3654265 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Antennas and Propagation, 2026-04. Verfasst von Yanjin Zhang und 3 weiteren Autor:innen. Die Arbeit wurde bisher 5-mal zitiert (Crossref) und nennt 42 Literaturangaben. Gefördert von: National Natural Science Foundation of China, National Key Research and Development Program of China.
Abstract (Originalsprache)
The attributed scattering center (ASC) model characterizes target scattering behavior using a compact set of physically meaningful parameters and plays a crucial role in synthetic aperture radar (SAR) image interpretation and target recognition. However, the conventional SAR images are typically reconstructed via inverse Fourier transforms. As a result, ASCs appear as sinc-shaped responses with wide mainlobes and high sidelobes, which limits parameter estimation accuracy. To address this, we propose a novel ASC extraction framework based on Gaussian representation and deep learning. Specifically, the sinc kernel is replaced with a Gaussian kernel to construct a super-resolved, energy-concentrated Gaussian scattering map (GSM). A deep neural network, Gaussian transform network (G-TransNet), is designed to learn a nonlinear, end-to-end mapping from 2-D radar echoes to the GSM. Coarse ASC parameters are then estimated via peak detection and local Gaussian fitting on the GSM, followed by refinement using the orthogonal matching pursuit (OMP) algorithm. Notably, the network is trained entirely on synthetic data generated from parametric models, which substantially reduces data preparation costs while maintaining consistent performance across both simulated and real-world measurements. Experimental results demonstrate that the proposed method attains high estimation accuracy and robustness with significantly reduced computational time, making it well-suited for practical applications.
Förderung
- National Natural Science Foundation of China
- National Key Research and Development Program of China
Zitieren
APA
Zhang, Y., Yan, H., Feng, X., & Yin, H. (2026). Effective Attributed Scattering Center Extraction via Gaussian Representation and Deep Learning. IEEE Transactions on Antennas and Propagation, 74(4), 3559-3574. https://doi.org/10.1109/tap.2026.3654265
Vancouver
Zhang Y, Yan H, Feng X, Yin H. Effective Attributed Scattering Center Extraction via Gaussian Representation and Deep Learning. IEEE Transactions on Antennas and Propagation. 2026;74(4):3559-3574. doi:10.1109/tap.2026.3654265
BibTeX
@article{Zhang2026,
author = {Zhang, Yanjin and Yan, Hua and Feng, Xuejian and Yin, Hongcheng},
title = {Effective Attributed Scattering Center Extraction via Gaussian Representation and Deep Learning},
journal = {IEEE Transactions on Antennas and Propagation},
year = {2026},
volume = {74},
number = {4},
pages = {3559--3574},
doi = {10.1109/tap.2026.3654265}
}
EUVIMED