Bionote
Kaiwen Wang is a Ph.D. candidate in Agricultural Robotics at Wageningen University & Research. His research focuses on UAV-based localization and mapping, 3D reconstruction, and semantic understanding for precision agriculture. He holds an M.S. in Robotics from Cranfield University and a B.S. in Computer Science from Sichuan University. In 2024–2025, he conducted a research visit at Bonn University. Kaiwen has published in leading journals such as Computers and Electronics in Agriculture and Journal of Field Robotics, contributing to agricultural automation through computer vision and SLAM technologies.
Presentation Abstract
Accurate fruit shape reconstruction under real-world field conditions is essential for high-throughput phenotyping, yield estimation, and orchard management. However, existing approaches based on 2D imaging, or explicit 3D reconstruction often suffer from occlusions, sparse views, and complex scene dynamics. This paper presents a novel UAV-based monocular 3D panoptic mapping framework for robust and scalable fruit shape completion in orchards. The proposed method integrates (1) Grounded-SAM2 for multi-object tracking and segmentation (MOTS), (2) photogrammetric Structure-from-Motion (SfM) for 3D scene reconstruction, and (3) DeepSDF, an implicit neural representation, for completing occluded fruit geometries. A new MOTS evaluation protocol is introduced to assess tracking performance without requiring ground truth annotations. Experiments conducted in both controlled laboratory conditions and an operational apple orchard demonstrate the effectiveness of the proposed method. DeepSDF consistently outperforms traditional shape approximation methods in reconstruction accuracy, while Grounded-SAM2 enables robust fruit tracking across challenging viewpoints. The approach is highly scalable, user-friendly, and applicable to real-world agricultural scenarios, offering a promising solution for precise 3D fruit phenotyping at large scale.
Co-Researcher (Supervisor): Lammert Kooistra, João Valente, Wensheng Wang