Electrical equipment identification method in infrared images based on VGG

Abstract Temperature is widely used to detect the state of electric equipment. Almost all the data is collected and sorted. And image analyzed by manual, which cause some trouble. In order to improve the intelligence level of power equipment detection, this paper proposes an infrared scene identification method for 220 kV current transformers, including target identification, temperature extraction, and image fusion. Firstly, a semantic segmentation model is built on VGG, which can effectively extract the target temperature information in complex environments. Then combined with the method of data set expansion, 3224 images of 220 kV current transformers were used for training. Finally, an image fusion method is used to extract equipment temperature. The experimental show that the method proposed in this paper for semantic segmentation and identification of 220 kV current transformers accuracy reaches 99.68%. In contrast, the traditional CNN identification method is 44%, effectively improving the target identification efficiency of 220 kV current transformers and has a good reference for electric equipment intelligence detection.