EUVIMEDEUVIMED entdecken
Systematische ÜbersichtOpen Access · CC BY 4.0

From bench to field: A systematic review of computer vision for tomato detection in precision agriculture (2018–2025)

Philippe Lyonel Mbouembe Touko, Guoxu Liu

Computing and Artificial Intelligence · 15.12.2025 · 3 (4)

DOI: 10.59400/cai4296 ↗

Kurz gefasst

Systematische Übersicht in der Zeitschrift Computing and Artificial Intelligence, online erschienen am 15.12.2025. Verfasst von Philippe Lyonel Mbouembe Touko und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Accurate tomato detection enables robotic harvesting, crop yield estimation, and tomato quality control, among other agricultural tasks. Despite remarkable advances in computer vision, particularly YOLO models, a significant gap persists between laboratory research and field deployment. This PRISMA-guided review of 110 publications (2018-2025) analyzes the disparity between lab-tested model performance and reliable real-world performance in this context. To quantify current capabilities, we construct a taxonomy based on the sensing platform, task, and environment. Across the reviewed literature, the mean average precision has increased from 78.3% for YOLOv3 to 94.7% for YOLOv11, while a controlled benchmarking study on an identical dataset (LaboroTomato) reveals smaller differences. However, tomato detection performance significantly drops in deployment, with a mean cross-domain performance loss of 8.24% due to occlusion, illumination changes, and weather conditions. Our reproducibility audit shows that most research lacks protocols for model development and that 12% releases make their public code available. Finally, 73% of high-accuracy models have requirements above the popular edge-device sizes commonly used in agricultural robotics. To bridge this implementation gap, we outline: 1) reporting guidelines to promote reproducibility, 2) decision frameworks to translate pragmatic agricultural considerations into concrete technical specifications, and 3) open research directions centered on reliability, cross-domain validation, and real-world deployment. This survey will support practitioners in agriculture, robotics, and machine learning design, deployable computer vision systems for tomatoes and other crops.

Zitieren

APA

Touko, P. L. M., & Liu, G. (2025). From bench to field: A systematic review of computer vision for tomato detection in precision agriculture (2018–2025). Computing and Artificial Intelligence, 3(4). https://doi.org/10.59400/cai4296

Vancouver

Touko PLM, Liu G. From bench to field: A systematic review of computer vision for tomato detection in precision agriculture (2018–2025). Computing and Artificial Intelligence. 2025;3(4). doi:10.59400/cai4296

BibTeX
@article{Touko2025,
  author = {Touko, Philippe Lyonel Mbouembe and Liu, Guoxu},
  title = {From bench to field: A systematic review of computer vision for tomato detection in precision agriculture (2018–2025)},
  journal = {Computing and Artificial Intelligence},
  year = {2025},
  volume = {3},
  number = {4},
  doi = {10.59400/cai4296}
}
Kostenlos starten →