兵工自动化2025，Vol.44Issue(6)：23-27,5.DOI:10.7690/bgzdh.2025.06.006
基于轻量级Fast-Unet网络的绝缘子图像分割方法
Insulator Image Segmentation Based on Lightweight Fast-Unet Network
袁新平 1王忠军 2路辉1
作者信息
- 1. 云南电网有限责任公司信息中心,昆明 650500
- 2. 南方电网数字平台科技(广东)有限公司 广东 深圳 518000
- 折叠
摘要
Abstract
In order to quickly find the insulator defects in the image and reduce the probability of power accidents,an insulator image segmentation algorithm is designed based on the lightweight Fast-Unet network.The method comprises the following steps of preprocessing an aerial image of an insulator to realize reconstruction and normalization;calculating the element symbiosis probability by converting image elements so as to obtain the color,texture and shape characteristics of the image;refining the channel space information characteristic value to form an initial network structure and generate a segmentation result dendrogram.By calculating the modularity and relative modularity of the image,a lightweight Fast-Unet network segmentation model is established.The experimental results show that the average comprehensive quality of the proposed algorithm is 0.72 in noise-free images,and 0.57 and 0.46 in simple and complex background images,respectively.Therefore,denoising has a positive effect on the quality of image segmentation.关键词
轻量级Fast-Unet网络/绝缘子串/绝缘子图像/图像分割算法/图像预处理Key words
lightweight Fast-Unet network/insulator string/insulator image/image segmentation algorithm/image preprocessing分类
信息技术与安全科学引用本文复制引用
袁新平,王忠军,路辉..基于轻量级Fast-Unet网络的绝缘子图像分割方法[J].兵工自动化,2025,44(6):23-27,5.
