A full-process closed-loop UAV inspection system for citrus disease detection in communication-constrained orchards based on SPN-YOLO
Smart Agricultural Technology · 2026-12 · 15 · 102490
DOI: 10.1016/j.atech.2026.102490 ↗Kurz gefasst
Fachartikel in der Zeitschrift Smart Agricultural Technology, 2026-12. Verfasst von Wanmei Feng und 5 weiteren Autor:innen. Die Arbeit nennt 47 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Basic and Applied Basic Research Foundation of Guangdong Province, Guangzhou Municipal Science and Technology Program key projects, Fund for National Key Laboratories.
Abstract (Originalsprache)
Timely and accurate detection of citrus diseases is crucial for preventing disease spread, requiring efficient and reliable orchard inspection methods. However, existing UAV-based deep learning methods often remain loosely integrated at the system level and are difficult to apply in communication-constrained orchards. To address these issues, this study proposes a full-process closed-loop UAV inspection system for citrus disease detection based on SPN-YOLO. First, a custom-built quadrotor platform with an industrial camera, an onboard computer, and a LoRa module was developed. Based on a preliminary ground sampling distance (GSD) experiment, the optimal UAV working distance was determined to be approximately 2.5 m, and a dataset of 7,419 images across six citrus disease and pest categories was constructed. Second, a lightweight YOLO11-based SPN-YOLO model was proposed for tiny-lesion detection. Specifically, SmartADown preserves fine-grained lesion information during downsampling, PKI-EMA enhances tiny-lesion feature representation while reducing computational redundancy, and NWD-Wise IoU stabilizes bounding-box regression. Finally, the system integrates grid-based path planning, autonomous patrolling, onboard inference, real-time result feedback via LoRa, and offline review. Experimental results show that SPN-YOLO achieves an mAP@0.5 of 89.2% with only 14.2 M parameters. Field verification shows that edge inference for each grid image takes approximately 3.8 s and can be executed concurrently with UAV’s flight to the next waypoint. The LoRa communication latency remains within 110 ms without packet loss. These outcomes confirm the feasibility and stability of the integrated “platform–algorithm–system” solution in communication-constrained orchards, providing technical support for intelligent orchard management.
Förderung
- Basic and Applied Basic Research Foundation of Guangdong Province
- Guangzhou Municipal Science and Technology Program key projects
- Fund for National Key Laboratories
- National Natural Science Foundation of China
- National Key Research and Development Program of China
Zitieren
APA
Feng, W., Wu, Z., Luo, J., Lyu, S., Li, Z., & Wen, W. (2026). A full-process closed-loop UAV inspection system for citrus disease detection in communication-constrained orchards based on SPN-YOLO. Smart Agricultural Technology, 15, 102490. https://doi.org/10.1016/j.atech.2026.102490
Vancouver
Feng W, Wu Z, Luo J, Lyu S, Li Z, Wen W. A full-process closed-loop UAV inspection system for citrus disease detection in communication-constrained orchards based on SPN-YOLO. Smart Agricultural Technology. 2026;15:102490. doi:10.1016/j.atech.2026.102490
BibTeX
@article{Feng2026,
author = {Feng, Wanmei and Wu, Ziliang and Luo, Jiayuan and Lyu, Shilei and Li, Zhen and Wen, Wei},
title = {A full-process closed-loop UAV inspection system for citrus disease detection in communication-constrained orchards based on SPN-YOLO},
journal = {Smart Agricultural Technology},
year = {2026},
volume = {15},
pages = {102490},
doi = {10.1016/j.atech.2026.102490}
}
EUVIMED