This is a page that curates AI-related papers published worldwide. All content here is summarized using Google Gemini and operated on a non-profit basis. Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.
This paper proposes an ensemble of YOLO-based deep learning models that leverage both visible and thermal images for defect detection in key components such as blades and towers of wind power plants. This approach combines the common YOLOv8 model with a thermal-specific model and integrates the prediction results through a sophisticated bounding box fusion algorithm. Experimental results demonstrate that the proposed method achieves a mean accuracy (mAP@.5) of 0.93 and an F1-score of 0.90, significantly improving performance compared to a single YOLOv8 model (mAP@.5 of 0.91). This suggests that leveraging multiple YOLO architectures and fused multispectral data can enhance the reliability of visual and thermal defect detection.
Takeaways, Limitations
•
Takeaways:
◦
We demonstrate the potential for improving the accuracy of wind power plant fault detection by leveraging multispectral (visible and thermal) data and YOLO-based model ensembles.
◦
Confirming the feasibility of developing a practical defect detection system based on the YOLOv8 model.
◦
Demonstrating the effectiveness of integrating multi-model prediction results using a bounding box fusion algorithm.
•
Limitations:
◦
Only experimental results on a limited dataset are presented, requiring further verification of generalization performance.
◦
Lack of comparative analysis of detection performance for different types of defects.
◦
Lack of real-time performance evaluation and applicability review in actual wind power plant environments.
