EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

EFA-RadNet: Efficient Feature Aggregation with Balanced Attention for Raw Radar Multi-Task Learning

Chengliang Zhong iD, Xiuping Li iD, Jingjing Li iD, Juan Liu iD, Xiyan Sun iD

Sensors · 25.03.2026 · 26 (7) · 2050

DOI: 10.3390/s26072050 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sensors, online erschienen am 25.03.2026. Verfasst von Chengliang Zhong und 4 weiteren Autor:innen. Beteiligte Einrichtung: Guangxi Key Laboratory of Precision Navigation Technology and Application, Guilin University of Electronic Technology, Guilin 541004, China. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 25 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: the Guangxi Science and Technology Project, the National Natural Science Foundation of China, Natural Science Foundation of Guangxi.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Original high-definition radar data contains rich environmental information, including distance, Doppler velocity, and azimuth. However, extracting robust features from such sparse and noisy frequency-domain data remains a challenge. To address this issue, this paper proposes an improved multi-task network, the Efficient Feature Aggregation with Balanced Attention Radar Network (EFA-RadNet). This network introduces the VoVNetV2 architecture into the field of raw radar perception and effectively preserves feature diversity across different receptive fields through a One-Shot Aggregation (OSA) module, avoiding signal aliasing. In addition, we propose an attention mechanism module, Balanced effective Squeeze–Excitation (B-eSE), which is better suited for sparse radar processing and effectively addresses the problem of weak target loss in the radar spectrum. Experiments on the RADIal dataset show that our EFA-RadNet achieves excellent target detection performance while also attaining optimal accuracy in free space segmentation.

Schlagwörter

Stichwörter der Autor:innen

attention mechanismautonomous drivingdeep learningfeature aggregationmillimeter-wave radarobject detection

Autor:innen und Institutionen

  • Guangxi Key Laboratory of Precision Navigation Technology and Application, Guilin University of Electronic Technology, Guilin 541004, China.
  • Information and Communication School, Guilin University of Electronic Technology, Guilin 541004, China.
  • National & Local Joint Engineering Research Center of Satellite Navigation Positioning and Location Service, Guilin 541004, China.
  • GUET-Nanning E-Tech Research Institute Co., Ltd., Nanning 530031, China.

ORCID-Profile: Chengliang Zhong, Xiuping Li, Jingjing Li, Juan Liu, Xiyan Sun

Förderung

  • the Guangxi Science and Technology Project
  • the National Natural Science Foundation of China
  • Natural Science Foundation of Guangxi
  • Guangxi Science and Technology Base and Talent Special Project: Research and Application of Key Technologies for Precise Navigation
  • Nanning Scientific Research and Technology Development Project
  • Basic Research Capabilities Enhancement Project for Young and Middle-aged Teachers in Guangxi's Universities
  • Guangxi Science and Technology Project
  • National Natural Science Foundation of China https://ror.org/01h0zpd94

Zitieren

APA

Zhong, C., Li, X., Li, J., Liu, J., & Sun, X. (2026). EFA-RadNet: Efficient Feature Aggregation with Balanced Attention for Raw Radar Multi-Task Learning. Sensors, 26(7), 2050. https://doi.org/10.3390/s26072050

Vancouver

Zhong C, Li X, Li J, Liu J, Sun X. EFA-RadNet: Efficient Feature Aggregation with Balanced Attention for Raw Radar Multi-Task Learning. Sensors. 2026;26(7):2050. doi:10.3390/s26072050

BibTeX
@article{Zhong2026,
  author = {Zhong, Chengliang and Li, Xiuping and Li, Jingjing and Liu, Juan and Sun, Xiyan},
  title = {EFA-RadNet: Efficient Feature Aggregation with Balanced Attention for Raw Radar Multi-Task Learning},
  journal = {Sensors},
  year = {2026},
  volume = {26},
  number = {7},
  pages = {2050},
  doi = {10.3390/s26072050}
}
Kostenlos starten →