Bionote
Zhen Cao is a PhD candidate at Wageningen University & Research and at the Agricultural Information Institute of the Chinese Academy of Agricultural Sciences (CAAS). His research focuses on AI-driven solutions for smart agriculture, including real-time object detection, edge computing, and UAV-based sensing systems. His current work emphasizes lightweight vision systems for real-time deployment, particularly UAV-based real-time object detection for field monitoring. He is dedicated to developing practical, resource-efficient algorithms that can be deployed in real-world agricultural environments, supporting quick decision-making.
Presentation Abstract
This presentation introduces an edge offloading strategy to enable real-time grape bunch segmentation and tracking with UAVs. The proposed system integrates a TensorRT-accelerated YOLOv8n detector, an IoU-based keyframe selection algorithm, and a parallel processing pipeline on the edge device. By selectively transmitting only informative frames to an edge server, the method reduces data load by up to 69% and significantly improves segmentation accuracy and tracking stability. Evaluation on real vineyard datasets demonstrates enhanced performance in sMOTSA, MOTSA, and MOTSP metrics, supporting efficient and scalable UAV-based grape monitoring in precision viticulture.
Co-Authors: Lammert Kooistra, Hilmy Baja, Wensheng Wang, João Valente