Segmentation of Power Tower Point Clouds With Color-Guided Perception and Self-Supervised Pretraining
TL;DR
This paper introduces a method combining Color-Guided Hierarchical Feature Perception and Structure-Aware Self-Supervised Pretraining to improve point cloud segmentation for power towers, addressing challenges like weak category differentiation and limited labeled data.
Segmentation of Power Tower Point Clouds With Color-Guided Perception and Self-Supervised Pretraining
Wei Liu; Yang Jin; Meng Jiang
https://doi.org/10.1109/ACCESS.2025.3649252
Volume 14
Point cloud semantic segmentation for power towers is essential in smart grid inspections, yet it faces challenges like weak category differentiation, complex structural composition, and scarcity of labeled data. To overcome these challenges, a method that combines Color-Guided Hierarchical Feature Perception (CGFHP) and Structure-Aware Self-Supervised Pretraining (SASSP) for point cloud segmentat...