Skip to main navigation Skip to search Skip to main content

SAR Ship OOD Detection Combining Deep Learning with Guidance and Constraint Mechanisms

  • Qiansheng Ma
  • , Zhe Chen
  • , Yun Zhang*
  • , Zhiquan Ding
  • , Qing Hua
  • *Corresponding author for this work
  • Harbin Institute of Technology
  • System Engineering Institute of Sichuan Aerospace

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Synthetic Aperture Radar (SAR) [1] is an active ground detection system capable of performing all-weather, all-time surveillance of the Earth's surface [2]. However, the Out-Of-Distribution (OOD) [3] problem causes a significant degradation in the performance of traditional deep learning algorithms. This paper designs a SAR ship detection method combining deep learning with Guidance and Constraint mechanisms. Based on the original deep learning model, CFAR preprocessing results are introduced as guidance to assist the model in more accurately localizing and extracting more precise features. Additionally, the salient features of SAR are employed as a constraint mechanism to filter out irrelevant features. Experimental results show that the proposed method outperforms a pure deep learning model, especially when there are significant distribution differences between the training and validation sets. It demonstrates higher robustness and stronger generalization ability for out-ofdistribution (OOD) samples.

Original languageEnglish
Title of host publicationProceedings of the 2025 IEEE Radar Conference, RadarConf 2025
EditorsMarek Rupniewski, Shannon Blunt, Jacek Misiurewicz, Maria Sabrina Greco, Braham Himed
PublisherInstitute of Electrical and Electronics Engineers
Pages367-372
Number of pages6
ISBN (Electronic)9798331544331
DOIs
StatePublished - 2025
Event2025 IEEE Radar Conference, RadarConf 2025 - Krakow, Poland
Duration: 4 Oct 20259 Oct 2025

Publication series

NameProceedings of the IEEE Radar Conference
ISSN (Print)1097-5764
ISSN (Electronic)2375-5318

Conference

Conference2025 IEEE Radar Conference, RadarConf 2025
Country/TerritoryPoland
CityKrakow
Period4/10/259/10/25

Keywords

  • CFAR
  • OOD
  • deep learning
  • ship target detection

Fingerprint

Dive into the research topics of 'SAR Ship OOD Detection Combining Deep Learning with Guidance and Constraint Mechanisms'. Together they form a unique fingerprint.

Cite this