Large Model Based Interactive Segmentation of Infrared Image for Power Equipment
Intelligent monitoring of power systems has become an important guarantee for modern industrial production,and the segmen-tation of power equipment in infrared images collected during daily inspections is an important part of the power fault diagnosis process.However,current deep learning-based infrared image segmentation methods require a large number of manually annotated samples,which is time-consuming and labor-intensive.Therefore,this paper takes advantage of large models pre-trained from visible light image datasets and applies the image segmentation model,Segment Anything Model(SAM),to interactive segmentation of power e-quipment in infrared images.With a small amount of user input,accurate power equipment segmentation results can be quickly ob-tained.In the self-constructed infrared image dataset,the model's performance was evaluated using simulated user clicks.The ex-periments show that the proposed method can accurately achieve interactive segmentation of power equipment without the need for addi-tional training.This method can effectively help workers improve the efficiency of annotating power equipment,quickly build large-scale infrared image datasets,promote the training of infrared image segmentation and fault diagnosis models,and help improve the lev-el of intelligent monitoring of power systems.
Power equipmentInfrared imageInteractive segmentationLarge ModelImage annotation
NETL
NSTL
万方数据
