Insulator Location and Segmentation Method Based on Improved S2ANet

Insulator is one of the important components of power system, which greatly affects the safe operation of transmission line. Aiming at the problem that insulator in aerial images has arbitrary rotation angle which leads to inaccurate location of insulator region and contains redundant background information, S2ANet model based on PAN is proposed to locate insulator region in aerial images. Firstly, rotation, cutting and noise were used to enhance the insulator data set and expand the number of samples. Secondly, S2ANet model based on PAN is used to locate the insulator and enhance the feature extraction ability of insulator edge. Finally, the insulator region is segmented accurately based on the image outspread clipping method. Experiments show that AP of improved S2ANEet model on the test data set is 92.6%, and outspread clipping method can completely segment the insulator region. The method proposed for insulator location and segmentation has good engineering application value.