Skip to main navigation Skip to search Skip to main content

SPTF-YOLO: A Sonar-Parameter-Embedded and Time-Frequency-Feature-Guided YOLO for Object Detection in Sonar Images

  • Xiaoguang Chen
  • , Wei Li*
  • , Fanyang Meng
  • , Lin Mei
  • *Corresponding author for this work
  • Harbin Institute of Technology Shenzhen
  • Peng Cheng Laboratory

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

Abstract

Sonar image detection faces challenges from noise, low resolution, and limited annotated data. We propose a Sonar-Parameter-Embedded and Time–Frequency-Feature-Guided YOLO (SPTF-YOLO), which integrates sonar-specific priors and time–frequency features via a prior knowledge branch. Experiments on the underwater acoustic target detection (UATD) sub-dataset show that SPTF-YOLO outperforms YOLOX and YOLOv8, achieving higher mean average precision (mAP).

Original languageEnglish
Title of host publicationWUWNet 2025 - The 19th International Conference on Underwater Networks and Systems
PublisherAssociation for Computing Machinery, Inc
ISBN (Electronic)9798400723162
DOIs
StatePublished - 22 May 2026
Event19th International Conference on Underwater Networks and Systems, WUWNet 2025 - Shenzhen, China
Duration: 29 Oct 202531 Oct 2025

Publication series

NameWUWNet 2025 - The 19th International Conference on Underwater Networks and Systems

Conference

Conference19th International Conference on Underwater Networks and Systems, WUWNet 2025
Country/TerritoryChina
CityShenzhen
Period29/10/2531/10/25

Keywords

  • Object detection
  • Sonar images
  • YOLO
  • underwater target detection

Fingerprint

Dive into the research topics of 'SPTF-YOLO: A Sonar-Parameter-Embedded and Time-Frequency-Feature-Guided YOLO for Object Detection in Sonar Images'. Together they form a unique fingerprint.

Cite this