基于改进 YOLOv5 输电线路异物检测算法研究
- 分享到:
摘要：
输电线路异物检测对于电力系统的安全可靠运行具有重要意义。为提高输电线路异物检测的精度，提出了一种基于改进YOLOv5的输电线路异物检测算法。该算法以YOLOv5为基础网络，在YOLOv5的C3模块嵌入Swin Transformer模块，形成C3STR，将其作为一个整体模块嵌入网络，借助其窗口自注意模块将自注意计算限制在偏移后的局部窗口，减少计算量，还允许跨窗口连接来提高效率，增强目标的深层语义信息和特征表示能力。并增加CBAM注意力机制，将空间注意力机制和通道注意力机制相结合，使网络能够关注到图片中的重要信息，提升网络特征提取能力。最后将回归函数的损失函数CIoU_loss替换为 SIoU_loss以提升网络的收敛速度。实验结果表明，模型改进后的平均精度均值(mAP)为98.8%，较原模型提高了3.3%。
Transmission lines foreign opject detection is of great significance to the safe and stable operation of electric power system. In order to improve the accuracy of foreign object detection on transmission lines, this paper proposes a foreign object detection algorithm of transmission lines based on improved YOLOv5. Based on the YOLOv5 network, Swin Transformer module is embedded in C3 module of yOLOy5 to form C3STR, which is embedded into the network as a whole, With the help of its window self-attention module, self-attention calculation is limited to the local window after migration, which reduces the calculation amount and allows cross-window connection to improve efficiency, enhance the deep semantic information and feature representing ability of the target. Moreover, CBAM attention mechanism is added, combining spatial attention mechanism and channe attention mechanism, so that the network can pay attention to the important information in the picture and improve the feature extracting ability of the convergence speed of the network. The experimental results show that the mAP of the improved model is 98.8%, which is 3.3% higher than that of the original model.
作者：
刘聪，李丽，许婷婷，胡胜，孔祥斌
Liu Cong，LiLi，Xu Tingting，Hu Sheng，Kong Xiangbin
机构地区：
湖北工业大学电气与电子工程学院；太阳能高效利用湖北省协同创新中心；武汉华安科技股份有限公司博士后科研工作站
引用本文：
刘聪，李丽，许婷婷等。基于改进YOLOv5输电线路异物检测算法研究[J] . 学报(自然科学版) , 2025,53(2) : 115-123. (Liu Cong, LiLi, Xu Tingting, et al. Research on foreign object detection algorithm of transmission lines based on improved YOLOv5[J] . Journal of Henan Normal University(Natural Science Edi- tion) ,2025,53(2) :115-123. DOI:10. 16366/j. cnki.1000-2367. 2023. 09. 02. 0001. )
基金：
湖北省自然科学基金
关键词：
YOLOv5；异物检测；输电线；Swin Transformer；CBAM；SIoU
YOLOv5; foreign object detection; transmission lines; Swin Transformer; CBAM; SloU
分类号：
TP391
