Adaptive Two-Stage Filter for De-snowing LiDAR Point Clouds

The Light Detection And Ranging (LiDAR) is integrated into self-driving cars to enhance system reliability. Unfortunately, its effectiveness is severely limited under the extreme weather conditions such as rain, fog, and cloud. This paper proposes an adaptive filter that can remove snow particles (de-snowing) from raw LiDAR point clouds. The proposed method adopts the intensity threshold with a two-stage filter that provides outstanding performance on both accuracy (98%) and processing speed (1.42 frames per second) under extreme environment for the autonomous driving system.