B2G-YOLO11-S: An efficient intelligent grading model for strawberry maturity with integrated causal analysis
Smart Agricultural Technology · 2025-12 · 12 · 101607
DOI: 10.1016/j.atech.2025.101607 ↗Kurz gefasst
Fachartikel in der Zeitschrift Smart Agricultural Technology, 2025-12. Verfasst von Qian Zhao und 7 weiteren Autor:innen. Die Arbeit wurde bisher 7-mal zitiert (Crossref) und nennt 28 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0). Gefördert von: Shanxi Provincial Key Research and Development Project, Shanxi Agricultural University, Shanxi Provincial Education Department.
Abstract (Originalsprache)
• Dual-stream B2-Net: Learns causal features, solves data shift interpretability. • Improved C3kGFPN: Suppresses non-target interference, aids fine-grained grading. • ACE metric: Quantifies feature-ripeness causal links, fuses causal theory. • Strawberry multi-source dataset: 2 cultivars, UAV/smartphone multi-scene images. ABSTRACT Strawberry (Fragaria × ananassa Duch.), pose notable constraints on the advancement of modern precision agriculture, primarily owing to their short postharvest storage life and the requirement for labor-intensive harvesting operations. As a core enabling technology for automated harvesting, strawberry maturity detection exhibits critical limitations: insufficient robustness under the complex conditions of high-rise greenhouse cultivation—where the environment includes sensors, drip irrigation, and other equipment that introduce complexities like occlusion and reflective interference—and a lack of causal interpretability in its decision-making procedures. To mitigate these challenges, this study presents an innovative integrated framework aimed at delivering a systematic solution for a precision-based strawberry maturity detection system. Based on an improved YOLO11 architecture, the framework innovatively integrates a dual-stream B2-Net backbone, an efficient HGNetv2-C feature extraction module, and a novel causal analysis metric (Average Causal Effect, ACE). Meanwhile, by fusing aerial and ground multi-source image data, it achieves comprehensive and dynamic evaluation of strawberry maturity in real fields, breaking through the limitations of single-view data. Experiments show that the model's performance is leading: its mAP50:95 reaches 82.9%, with mAP50 reaches 95.6%; precision and recall are 89.6% and 92.2% respectively, outperforming benchmarks. Its strong adaptability in complex scenarios verifies its practical value. ACE causal analysis reveals that under light intensity perturbations, the model's average absolute percentage change in ACE is strictly controlled within ±0.5%Δ, outperforming YOLOv8 and YOLO11, demonstrating excellent balance and stability and providing reliable support for strawberry maturity detection.
Förderung
- Shanxi Provincial Key Research and Development Project
- Shanxi Agricultural University
- Shanxi Provincial Education Department
Zitieren
APA
Zhao, Q., Hao, C., Cui, J., Zan, J., Han, X., Chen, Q., Zhang, X., & Li, F. (2025). B2G-YOLO11-S: An efficient intelligent grading model for strawberry maturity with integrated causal analysis. Smart Agricultural Technology, 12, 101607. https://doi.org/10.1016/j.atech.2025.101607
Vancouver
Zhao Q, Hao C, Cui J, Zan J, Han X, Chen Q, et al. B2G-YOLO11-S: An efficient intelligent grading model for strawberry maturity with integrated causal analysis. Smart Agricultural Technology. 2025;12:101607. doi:10.1016/j.atech.2025.101607
BibTeX
@article{Zhao2025,
author = {Zhao, Qian and Hao, Chunxu and Cui, Jianhua and Zan, Jiangchen and Han, Xiongwei and Chen, Qingqiang and Zhang, Xiaoying and Li, Fuzhong},
title = {B2G-YOLO11-S: An efficient intelligent grading model for strawberry maturity with integrated causal analysis},
journal = {Smart Agricultural Technology},
year = {2025},
volume = {12},
pages = {101607},
doi = {10.1016/j.atech.2025.101607}
}
EUVIMED