Skip to main navigation Skip to search Skip to main content

Assessing the Efficiency of One-shot Visual Object Trackers for Underwater Robot Position Locking

  • Hamad bin Khalifa University
  • Vrije Universiteit Brussel

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

Abstract

This study presents, to our knowledge, the first comprehensive evaluation of seven Machine Learning (ML)based one-shot object tracking algorithms for the vision-based position stabilization of remotely-operated underwater vehicles (ROVs). We introduce a position-locking framework that analyzes images of a target object, in front of which the ROV must maintain stability. The system leverages the outputs of various object-tracking algorithms to autonomously adjust the ROV's position in response to external disturbances. Extensive realworld experiments were conducted using a BlueROV2 platform in an indoor pool, highlighting the advantages and limitations of each tracking method. Additionally, to address the lack of publicly available underwater ROV datasets, we are releasing our collected data as open-source, aiming to support and advance future research in this field.

Original languageEnglish
Title of host publication2025 Ieee/acs 22nd International Conference On Computer Systems And Applications, Aiccsa
PublisherIEEE Computer Society
Number of pages7
ISBN (Electronic)9798331556938
ISBN (Print)979-8-3315-5694-5
DOIs
Publication statusPublished - 22 Oct 2025
Event22nd ACS/IEEE International Conference on Computer Systems and Applications, AICCSA 2025 - Doha, Qatar
Duration: 19 Oct 202522 Oct 2025

Publication series

NameInternational Conference On Computer Systems And Applications

Conference

Conference22nd ACS/IEEE International Conference on Computer Systems and Applications, AICCSA 2025
Country/TerritoryQatar
CityDoha
Period19/10/2522/10/25

Keywords

  • One-shot object tracking
  • Robot position control
  • Underwater robot navigation

Fingerprint

Dive into the research topics of 'Assessing the Efficiency of One-shot Visual Object Trackers for Underwater Robot Position Locking'. Together they form a unique fingerprint.

Cite this