Infrared Image Segmentation Detection of Surge Arrester Bushing Based on U-Net with Attention Mechanisms

The accurate infrared thermal image detection and segmentation of surge arrester bushing is a prerequisite for intelligent temperature analysis. In response to the low accuracy and high false detection issues in existing surge arrester bushing instance segmentation methods, this paper proposes a surge arrester bushing segmentation detection method based on the Casing-Unet model. The Casing-Unet model is constructed by improving the Unet with the CBAM attention mechanism. Through qualitative and quantitative experiments, this paper compares and verifies the performance of the Casing-Unet model in surge arrester bushing infrared image segmentation detection. Compared with the traditional Unet model, the accuracy is improved by 9.3%. The results indicate that the model has a good detection and segmentation effect on surge arrester bushing, which is of great significance for subsequent temperature detection of surge arrester bushing, reducing labor costs, improving equipment intelligence, and ensuring the safe operation of substation equipment.