Abstract
To address the limitations of current transmission line point cloud segmentation algorithms in accurately segmenting fine-grained structures, this paper proposes a model named TLSNet. The network addresses the issue of uneven point cloud distribution through a dynamic density adaptation mechanism (DALNF-Layer), achieves cross-scale context modeling with a hierarchical-offset transformer (DLCTransformer), and optimizes gradient propagation using the inverted residual module (InvResMLP). Together, these components form an efficient point cloud segmentation framework tailored for the complex scenarios of transmission lines. Experimental results demonstrate that TLSNet achieves precise segmentation of key regions in transmission line point clouds, outperforming existing algorithms in segmentation accuracy. This advancement provides a solid technical foundation for the digital operation and maintenance of transmission lines, as well as for autonomous unmanned aerial vehicle (UAV) inspection path planning.