Fusion Segmentation Network Guided by Adaptive Sampling Radius and Channel Attention Mechanism Module for MLS Point Clouds

Road high-precision mobile LiDAR measurement point clouds are the digital infrastructures for maps, autonomous driving, twins, etc. High-precision automated semantic segmentation of road is a crucial research direction. Aiming at problem low accuracy existing deep learning networks inhomogeneous sparse system measurements (MLS), method that adaptively adjusts sampling radius region groups according to density proposed. We construct dataset based on self-developed train and test segmentation. The overall 98.08%, with an mIOU 0.73 mIOUs 0.99, 0.983, 0.66, 0.51 roads, guardrails, signs, streetlights, lane lines, respectively. experimental result shows can achieve more accurate systems. Compared methods, significantly improved.