基于多维交互感知的输电线覆冰厚度检测

打开文本图片集

中图分类号：TP391.4 文献标识码：A文章编号：1673-9868（2026）02-0246-10

DOI： 10.13718/j. cnki. xdzk. 2026. 02.020

Abstract：Aiming at the problem of low accuracy of existing ice thickness detection methods for transmission lines，an ice thickness detection method based on multi-dimensional interactive perception was proposed.This method used an encoder-decoder architecture and proposed a multi-dimensional interactive perception module（MIPM） at cross layer connections. This module consisted of three parts：multi-dimensional channel attention，multi-dimensional spatial attention，and interactive perception module. Within the multi-dimensional channel atention，the multi-scale convolution kernels were utilized to extract channelinformation for the detailed features of transmission lines.Within the multi-dimensional spatial attention，the multi-scale strip convolution was utilized to extract spatial information based on the shape characteristics of transmission lines. The interactive perception module cross-assigned attention weights to encoding and decoding features to guide edge generation，and the ice thickness was calculated by determining the pixel diference between the upper and lower edges.Experiments on ice-covered datasets showed that the detection accuracy of this method was superior to other methods. The optimal dataset scale （ODS） and optimal image scale（OIS）reached O.930 and O.933，respectively，with the thickness detection eror controlled within 8.8%

Key Words： transmission lines； edge detection； ice thickness； attention mechanism; encoder-decoder

输电线路的稳定运行对保障人们的日常生活与生产活动至关重要[1]。（剩余13050字）