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.