基于改进YOLOv10n模型的高压输电线路异物检测方法OA
Foreign object detection method for high-voltage transmission lines based on improved YOLOv10n model
针对实际环境中高压输电线路异物检测存在背景复杂、检测精度低等问题,本文提出一种改进YOLOv10n的检测模型——YOLOv10n-OMW.首先,在主干网络中引入了全维动态卷积(ODConv),结合通道注意力、空间注意力和动态卷积核选择机制,动态提取异物特征;其次,设计了多尺度增强空间注意力机制(MES Attention),通过多尺度自适应池化提取空间特征,抑制复杂背景干扰,突出异物关键区域;最后,引入 Wise-IoU(WIoU)损失函数,替代原有的Complete Intersection over Union(CIoU)损失函数,提升了模型的总体性能.实验结果表明,改进后的 YOLOv10n-OMW 模型在检测精度上达到98.0%,相较YOLOv10n的提高了3.9%,在复杂背景下表现出更优性能,为智能电力巡检提供了技术参考.
Addressing the issues of complex backgrounds and low detection accuracy in foreign object detec-tion for high-voltage transmission lines in real-world environments,this paper proposed an improved YOLOv10n detection model,YOLOv10n-OMW.Firstly,an all-dimensional dynamic convolution(ODConv)was introduced into the backbone network,combining channel attention,spatial attention,and dynamic con-volution kernel selection mechanism to dynamically extract foreign object features.Secondly,a multi-scale enhanced spatial attention mechanism(MES Attention)was designed to extract spatial features through multi-scale adaptive pooling,suppress complex background interference,and highlight key areas of foreign objects.Finally,the Wise-IoU(WIoU)loss function was introduced to replace the original Complete Inter-section over Union(CIoU)loss function,enhancing the overall performance of the model.Experimental re-sults show that the improved YOLOv10n-OMW model achieves a detection accuracy of 98.0%,an im-provement of 3.9%compared to YOLOv10n,and demonstrates superior performance in complex back-grounds,providing a technical reference for intelligent power inspection.
李鑫;赵佳
长春工业大学 计算机科学与工程学院,吉林 长春 130102长春工程学院 计算机技术与工程学院,吉林 长春 130021
信息技术与安全科学
异物检测YOLOv10注意力机制损失函数卷积模块
foreign object detectionYOLOv10attention mechanismloss functionconvolution module
《长春工程学院学报(自然科学版)》 2026 (1)
75-83,9
吉林省科技发展计划杰出青年人才项目(20240602004RC)
评论