3D Reconstruction of Cables for Live-Working Robots in Distribution Networks

Good environmental perception and target recognition capabilities are prerequisites for the autonomous operation of live-working robots. This paper focuses on the identification of flexible cables in 10kV distribution networks and proposes an improved 3D model reconstruction method. A point cloud semantic segmentation network based on PointNet++ is designed to achieve precise separation of cables from complex scenes. Building on this, a piecewise fitting reconstruction strategy is proposed, taking into account the geometric characteristics of the cables, incorporating an improved initial centroid selection method in the K-Means algorithm, significantly enhancing the segmentation results for cables with varying curvature levels. Experimental results validate the effectiveness of the proposed method, providing a feasible solution for the recognition of flexible cables.