Earticle

초록

영어

In strawberry cultivation, simultaneous detection of fruit ripeness and leaf diseases is very important to maximize yield and ensure crop health. In this study, we propose an ensemble technique combining an improved YOLOv8l and a ResNet50-based Faster R-CNN model. The YOLOv8l model reduces the number of parameters by 29% by replacing the standard backbone with ConvNeXtV2, introducing the BiFPN neck structure, and integrating the GRN layer, but the accuracy is lowered. However, the accuracy is improved by integrating the model ensemble technique and the Weighted Box Fusion algorithm. The experiment is conducted using 5,000 strawberry images collected from a smart farm in Cheonan-si, South Korea from January to April 2025. The dataset has a total of nine classes, including five maturity stages (Flower, Green, White, Turning Red, and Red) and four leaf disease states (Chlorosis, Tip Dieback, wilt, and Plauge). Experimental results show that the proposed model architecture achieves 0.8623 at mAP@0.5, which improves the performance by about 8% compared to the single-model approach. In particular, the system demonstrates excellent performance in detecting visually distinct classes such as flowers, while achieving good results even for classes with subtle features such as early-stage diseases. Visual analysis confirms the robustness of the model even in complex agricultural environments with various lighting conditions and overlapping objects. Therefore, it contributes to the development of an automated monitoring system for strawberry cultivation in greenhouse environments and has significant potential for application in various smart agricultural environments.

목차

Abstract
1. Introduction
2. Background Theory
2.1 YOLOv8 Architecture
2.2 Related Work
3. Proposed Method
3.1 Improved YOLOv8 Architecture
3.2 Ensemble Model Techniques for Multi-Task Detection
4. Experimental environment and Results
4.1 Data Acquisition
4.2 Training Configuration and Hyperparameters
4.3 Evaluation Metrics
4.4 Performance Comparison
5. Conclusion
Acknowledgement
References

키워드

저자

• Taewook Kim [ M.S, Department of Information Convergence System, Kwangwoon University, Seoul, South Korea ]
• Heejun Youn [ 2Doctoral Course, Department of Plasma Bio Display, Kwangwoon University, Seoul, South Korea ]
• Yuseong Lee [ M.S, Department of Information Convergence System, Kwangwoon University, Seoul, South Korea ]
• Jinyoung Park [ CEO, Needs Robot Inc., Seoul, South Korea ]
• Soonchul Kwon [ Associate professor, Graduate School of Smart Convergence, Kwangwoon University, Seoul, South Korea ] Corresponding Author

참고문헌

발행기관

간행물

표지보기 관심저널 등록

• 간행물명

  The International Journal of Advanced Smart Convergence

• 간기

  계간

• pISSN

  2288-2847

• eISSN

  2288-2855

• 수록기간

  2012~2025

• 십진분류

  KDC 326 DDC 380