Deep Manifold Network-Based Radar Target Detector With False Alarm Control via NP Criterion
IEEE Transactions on Aerospace and Electronic Systems · 2026 · 62 · 17031-17043
DOI: 10.1109/taes.2026.3731335 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Aerospace and Electronic Systems, 2026. Verfasst von Yangyang Su und 6 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 46 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Hunan Provincial Natural Science Foundation of China.
Abstract (Originalsprache)
The detection performance of radar systems is constrained by the complex and variable nature of real-world clutter backgrounds. Traditional energy-based detectors face limitations due to the difficulty of accurately modeling clutter statistics. Convolutional neural networks typically learn discriminative features between targets and clutter in Euclidean space and may therefore not fully exploit the non-Euclidean geometry inherent in matrix representations of second-order statistical features, such as frequency correlations. This article designs a network-based radar target detection method on manifolds. The core detection network is a symmetric positive definite matrix network (SPDNet). Radar echo data are transformed into the time-frequency domain to construct time-frequency feature SPD matrices, which serve as the input to SPDNet. SPDNet distinguishes targets from clutter by learning discriminative features such as frequency correlation. To control the probability of false alarm during SPDNet detection, a differentiable loss function is designed based on the Neyman–Pearson (NP) criterion to replace the traditional cross-entropy loss, and a Neyman–Pearson criterion-constrained SPDNet (NPC-SPDNet) is constructed. Finally, experiments on NPC-SPDNet are carried out using measured data, which verify the rationality and effectiveness of the designed differentiable NP loss function, as well as the feasibility and superiority of network-based radar target detection on manifolds.
Förderung
- National Natural Science Foundation of China
- Hunan Provincial Natural Science Foundation of China
Zitieren
APA
Su, Y., Hua, X., Wan, Z., Lin, H., Cheng, Y., Wu, H., & Wang, H. (2026). Deep Manifold Network-Based Radar Target Detector With False Alarm Control via NP Criterion. IEEE Transactions on Aerospace and Electronic Systems, 62, 17031-17043. https://doi.org/10.1109/taes.2026.3731335
Vancouver
Su Y, Hua X, Wan Z, Lin H, Cheng Y, Wu H, et al. Deep Manifold Network-Based Radar Target Detector With False Alarm Control via NP Criterion. IEEE Transactions on Aerospace and Electronic Systems. 2026;62:17031-17043. doi:10.1109/taes.2026.3731335
BibTeX
@article{Su2026,
author = {Su, Yangyang and Hua, Xiaoqiang and Wan, Zhonghao and Lin, Huai and Cheng, Yongqiang and Wu, Hao and Wang, Hongqiang},
title = {Deep Manifold Network-Based Radar Target Detector With False Alarm Control via NP Criterion},
journal = {IEEE Transactions on Aerospace and Electronic Systems},
year = {2026},
volume = {62},
pages = {17031--17043},
doi = {10.1109/taes.2026.3731335}
}
EUVIMED