@article{ART003399323},
author={Wan Q and Min, Byung-Won},
title={Adaptive Cropped-SGBM Stereo Ranging with Edge-Deployed Real-Time Perception for Floating-Debris Retrieval by Inland-Waterway USVs},
journal={Journal of Internet of Things and Convergence},
issn={2466-0078},
year={2026},
volume={12},
number={5},
pages={7}
TY - JOUR
AU - Wan Q
AU - Min, Byung-Won
TI - Adaptive Cropped-SGBM Stereo Ranging with Edge-Deployed Real-Time Perception for Floating-Debris Retrieval by Inland-Waterway USVs
JO - Journal of Internet of Things and Convergence
PY - 2026
VL - 12
IS - 5
PB - The Korea Internet of Things Society
SP - 7
EP -
SN - 2466-0078
AB - Reliable three-dimensional localization of floating debris is essential for autonomous retrieval by unmanned surface vehicles (USVs) on inland waterways. However, stereo matching is degraded by specular reflection, ripple noise, and the texture-less surfaces of floating debris, and dense matching is too costly for edge platforms. This paper proposes a semantically guided binocular localization method that confines Semi-Global Block Matching (SGBM) and disparity refinement to a local region defined by a 2D detection box under epipolar constraints, and applies adaptive local weighted-least-squares (AL-WLS) filtering, a 𝛾-mask, and median disparity extraction. On a 400-frame real-waterway sequence, the method reduces the mean absolute percentage error (MAPE) from 13.85% to 2.73% and the SGBM latency by 98.07%, with a mean deviation of 14.84 mm from full-frame matching. Deployed with TensorRT on an NVIDIA Jetson Orin Nano 8GB, and after correcting a training –inference geometric inconsistency (detection rate from 80.1% to 96.7%) and optimizing local-region rectification (rectification latency from 27.32 to 0.84 ms), the system runs at 12.67 FPS with a root mean square error (RMSE) of 128.91 mm at a mean power of 6.84 W, and its perception-to-control-state chain was implemented as a distributed ROS2 pipeline. These results show that guided local stereo matching enables real-time, low-power 3D ranging on embedded hardware within the tested conditions (a single sequence, ground-truth range about 3.3–3.7 m).
KW - Unmanned Surface Vehicle;Binocular Stereo Vision;SGBM;Edge Computing;Jetson Orin Nano;Real-Time Ranging;Floating Debris
DO -
UR -
ER -
Wan Q and Min, Byung-Won. (2026). Adaptive Cropped-SGBM Stereo Ranging with Edge-Deployed Real-Time Perception for Floating-Debris Retrieval by Inland-Waterway USVs. Journal of Internet of Things and Convergence, 12(5), 7.
Wan Q and Min, Byung-Won. 2026, "Adaptive Cropped-SGBM Stereo Ranging with Edge-Deployed Real-Time Perception for Floating-Debris Retrieval by Inland-Waterway USVs", Journal of Internet of Things and Convergence, vol.12, no.5 7.
Wan Q, Min, Byung-Won "Adaptive Cropped-SGBM Stereo Ranging with Edge-Deployed Real-Time Perception for Floating-Debris Retrieval by Inland-Waterway USVs" Journal of Internet of Things and Convergence 12.5 7 (2026) : 7.
Wan Q, Min, Byung-Won. Adaptive Cropped-SGBM Stereo Ranging with Edge-Deployed Real-Time Perception for Floating-Debris Retrieval by Inland-Waterway USVs. 2026; 12(5), 7.
Wan Q and Min, Byung-Won. "Adaptive Cropped-SGBM Stereo Ranging with Edge-Deployed Real-Time Perception for Floating-Debris Retrieval by Inland-Waterway USVs" Journal of Internet of Things and Convergence 12, no.5 (2026) : 7.
Wan Q; Min, Byung-Won. Adaptive Cropped-SGBM Stereo Ranging with Edge-Deployed Real-Time Perception for Floating-Debris Retrieval by Inland-Waterway USVs. Journal of Internet of Things and Convergence, 12(5), 7.
Wan Q; Min, Byung-Won. Adaptive Cropped-SGBM Stereo Ranging with Edge-Deployed Real-Time Perception for Floating-Debris Retrieval by Inland-Waterway USVs. Journal of Internet of Things and Convergence. 2026; 12(5) 7.
Wan Q, Min, Byung-Won. Adaptive Cropped-SGBM Stereo Ranging with Edge-Deployed Real-Time Perception for Floating-Debris Retrieval by Inland-Waterway USVs. 2026; 12(5), 7.
Wan Q and Min, Byung-Won. "Adaptive Cropped-SGBM Stereo Ranging with Edge-Deployed Real-Time Perception for Floating-Debris Retrieval by Inland-Waterway USVs" Journal of Internet of Things and Convergence 12, no.5 (2026) : 7.
