Improved Multi-Object Tracking System Using 3D–2D Image Data Fusion for Beach Cleaning Robots
-
- Yatigul Rut
- Kyushu Institute of Technology
-
- Jie Tan Chi
- Kyushu Institute of Technology
-
- Munjer M.A
- Kyushu Institute of Technology
-
- Jitviriya Wisanu
- King Mongkut's University of Technology North Bangkok
-
- Sittiwanchai Teppakorn
- King Mongkut's University of Technology North Bangkok
-
- Tangsuksant Watcharin
- King Mongkut's University of Technology North Bangkok
-
- Hayashi Eiji
- Kyushu Institute of Technology
書誌事項
- 公開日
- 2026-01-29
- バージョン
- 01
- DOI
-
- 10.5954/icarob.2026.os1-2
- 公開者
- 株式会社ALife Robotics
説明
This research introduces a 3D-2D image data fusion mechanism to enhance Multi-Object Tracking for marine debris detection. Traditional systems relying solely on 2D data cannot reidentify long-occluded objects that leave and reenter the frame. The proposed system combines RGB and depth data from an RGBD camera on a Beach Cleaning Robot, registering each tracked object with 3D coordinates referenced to the robot's starting point for accurate ID reassignment. The system achieved 80.02% reidentification accuracy in simulation and 78.33% in real-world tests despite odometry drift. Results demonstrate that 3D-2D fusion provides robust tracking for long-occluded objects, offering practical value for Multi-Object Tracking across computer vision applications.
収録刊行物
-
- 人工生命とロボットに関する国際会議予稿集
-
人工生命とロボットに関する国際会議予稿集 31 18-21, 2026-01-29
株式会社ALife Robotics
- Tweet
キーワード
詳細情報 詳細情報について
-
- CRID
- 1390308043759221376
-
- ISSN
- 21887829
-
- 本文言語コード
- en
-
- データソース種別
-
- JaLC
- Crossref
-
- 抄録ライセンスフラグ
- 使用不可