基于YOLOv3的航拍输电通道航拍绝缘子检测方法研究

针对已有绝缘子检测算法在航拍图像检测中绝缘子检测精确度低、泛化能力差的问题,提出一种基于YOLOv3改进的输电线路航拍绝缘子检测方法。在YOLOv3模型的基础上使用ResNet50-vd替换主干网络,并组合使用CoordConv、DropBlock及SPP等技巧提取复杂背景中的绝缘子特征,然后利用该文收集的绝缘子数据集对改进后的网络进行训练。实验结果表明,改进后的YOLOv3网络模型相较于改进前精确度提升5.5%,有效提升航拍图像绝缘子检测的准确率。

Aiming at the problems of low detection accuracy and poor generalization ability of existing insulator detection algorithms for aerial images,an improved method based on YOLOv3 for aerial image insulator detection of transmission lines is proposed.Based on the YOLOv3 model,ResNet50-vd is used to replace the backbone network,and CoordConv,DropBlock,SPP and other techniques are combined to extract insulator features from complex background,and then this paper collects insulator data sets to train the improved network.The experimental results show that the accuracy of the improved YOLOv3 network model is 5.5%higher than that before the improvement,and effectively improves the accuracy of insulator detection in aerial images.