RoboAI

Opinion

Vision-Based Pose Detection for Dynamic Docking of Unmanned Surface Vehicles

Daily briefingElena PetrovaMay 12, 2026· 3,343 views

A stereo marker and robust recognition method enable reliable pose estimation for USV dynamic docking under large angles, motion blur, and occlusion.

Dynamic docking of unmanned surface vehicles (USVs) requires precise target pose estimation, yet motion-induced challenges such as large observation angles, motion blur, and occlusion can degrade features of conventional planar markers, hindering reliable omnidirectional detection. To address this, researchers propose a vision-based method using a novel 3D fiducial marker. The marker is a four-faced cube with each face featuring a multi-ring nested structure, leveraging circular projection invariance and stereo distribution to enable full-attitude detection. A recognition algorithm based on ring projection and concentricity constraints identifies the marker even when features are degraded or partially missing, using dynamic thresholds and arc segment matching to complete incomplete elliptical rings.

Key takeaways

  • Marker design: A regular quadrangular prism with four orthogonal faces ensures at least one face is visible from any viewpoint, avoiding blind spots and enabling mutual verification when two faces are visible.
  • Robust recognition: The algorithm filters elliptical regions, checks center coincidence with dynamic thresholds, and uses arc segment matching to recover missing concentric ellipses, maintaining stability under motion blur and partial occlusion.
  • Pose estimation: Affine transformation corrects circular features, and the geometric properties of the quadrangular prism optimize localization. Temporal trend integration enables continuous attitude estimation.
  • Performance: Simulations within 30 m show max position error below 0.8 m and heading error under 5°. Real-ship static tests yield position error of 1.2 m and heading error under 6.2°; dynamic docking final-stage position error is below 0.16 m with 85% of heading errors within 5°.

This work provides a practical solution for USV autonomous docking, enhancing reliability in challenging marine environments. The method's robustness to feature degradation and its omnidirectional coverage make it suitable for integration into USV control systems, supporting autonomous reconfiguration and cooperative missions.

Source: 《机器人》期刊 (robot.sia.cn) · Published 2026-05-12 · “水面机器人面向动态对接的目标位姿视觉检测”