EBS-YOLO: Foreign object detection algorithm for transmission lines based on improved Yolov10

When a foreign body touches a transmission line it may have serious consequences. If it is not handled in time it may lead to accidents such as short circuits and blackouts, affecting the normal operation of the power system and the stability of social life. In order to detect foreign objects on transmission lines, this paper proposes an EBS-YOLO method based on Yolov10. Firstly, in the structure of the backbone network, we adopt C2f-Efficient Multi-Scale-Conv plus (C2f-EMSCP) as the convolutional layer for feature extraction, replacing part of the C2f standard convolutional layer, and obtaining a richer feature representation by combining different scales of feature mapping. Secondly, a Bidirectional Feature Pyramid Network (BiFPN) is used, which enables the model to fuse features of different scales better. Then, the SEAMHead module with multi-head attention is utilized to augment the original features, enhance head detection, and reduce the effect of object occlusion. Finally, in order to solve the consistency problem between the predicted and real bounding boxes, the SIoU loss is used to replace the original CIoU loss in Yolov10. The experimental results show that EBS-YOLO achieves an average detection accuracy of 90.4 %, which is 4.1 % better than Yolov10, and Recall and Precision are improved by 5.3 % and 1.8 %, respectively. Compared with other methods, our EBS-YOLO has higher accuracy in detecting foreign objects on transmission lines.