Based on Multi-Module Synergistic Optimization: MSDA-YOLO Method for Foreign Object Detection on Power Transmission Lines

This paper presents a novel approach to foreign object detection on power transmission lines using MSDA-YOLO (YOLOv26), RT-DETR, Grad-CAM, and Soft-NMS under a multi-module synergistic optimization framework. The system aims to identify foreign objects — specifically balloons, kites, nests, and trash — by processing image datasets through advanced object detection techniques. The MSDA-YOLO model (122 layers, 9,466,728 parameters, 20.5 GFLOPs) and RT-DETR model (310 layers, 31,991,960 parameters, 103.4 GFLOPs) are employed for detecting objects in images, with Soft-NMS applied to enhance the accuracy of object localization by reducing redundant predictions. Grad-CAM is utilized for visual interpretability, providing insights into the decision-making process of the models by generating class activation maps. The detection process focuses on four classes — balloon, kite, nest, and trash — and optimizes the performance by fine-tuning hyperparameters and leveraging the strengths of these advanced models. MSDA-YOLO achieves an overall mAP50 of 0.942 and mAP50-95 of 0.642, while RT-DETR achieves mAP50 of 0.911 and mAP50-95 of 0.643, demonstrating robust and precise foreign object detection. The methodology improves the robustness and precision of foreign object detection in real-time applications, offering a scalable solution for automated power line inspection and safety systems.

Keywords: MSDA-YOLO, YOLOv26, RT-DETR, Grad-CAM, Soft-NMS, foreign object detection, power transmission lines, multi-module synergistic optimization, object detection, hyperparameter optimization, machine learning, deep learning.