UAV obstacle perception based on transmission line detection from 3D point cloud

To ensure the safe operation of low-altitude UAVs, this paper presents an obstacle perception method for transmission towers utilizing 3D point cloud cylindrical segmentation. To address the challenges posed by the sparsity and low resolution of point cloud data, we propose a short-time neighborhood point cloud registration method based on LiDAR SLAM to enhance point cloud density. For the obstacle avoidance strategy within the cylindrical area surrounding the UAV, we employ a 3D convolutional network that utilizes a cylindrical representation of the segmented point cloud for obstacle identification. Experimental results demonstrate that the method proposed in this paper effectively provides real-time visual perception for obstacle avoidance in low-altitude UAVs.