基于改进YOLOv5的输电线路绝缘子 识别方法

针对输电线路绝缘子识别准确率低、识别花费时间长的问题，提出一种改进的YOLOv5绝缘子识别方法。首先，通过引入超分辨率卷积网络提升数据集中图像样本质量；其次，通过引入k3-Ghost结构替换原始网络BCSP模块中的普通卷积，减少模型主干网络参数量，在主干网络尾部引入SENet注意力模块，加强模型对于通道信息的关注提升目标检测性能；在颈部网络引入DC-BiFPN结构替换原始结构，对不同尺度特征赋予不同权重以使多尺度特征进行更好的融合，提升绝缘子识别效果。最后，使用CIOU作为回归损失函数，加快网络收敛速度。实验结果表明：本文提出的方法在保证绝缘子识别准确率的同时拥有更高的识别速度，检测准确率达到89.5%，检测速度达到35.7FPS，验证了改进方法的有效性。

Insulator identification method of transmission line based on improved YOLOv5

To solve the problems of low accuracy and long identification time of insulators in transmission lines, An improved method for identification of YOLOv5 insulators is proposed. Firstly, the quality of image samples in the dataset is improved by introducing super-resolution convolutional network. Secondly, by introducing k3-Ghost structure to replace common convolution in BCSP module of original network, the number of parameters in main network of model is reduced, the SE attention module is introduced in the tail of the trunk network to strengthen the model''s attention to channel information and improve the performance of target detection; In the neck network, DC-BiFPN structure was introduced to replace the original structure, and different weights were assigned to different scale features to make better fusion of multi-scale features, so as to improve the insulator recognition effect. Finally, CIOU is used as regression loss function to speed up network convergence. The experimental results show that the proposed method has a higher recognition speed while ensuring the accuracy of insulator recognition, with detection accuracy up to 89.5% and detection speed up to 35.7FPS, which verifies the effectiveness of the improved method.