Filter Methods for Removing Falling Snow from Light Detection and Ranging Point Clouds in Snowy Weather

For autonomous driving systems to effectively replace human drivers, they must be able to adapt to harsh weather conditions. Rain and snow can cause noise to be introduced into light detection and ranging (LiDAR) point cloud data, which can interfere with the work of the perception module of autonomous driving systems. In this work, we collected LiDAR point cloud data of snowy weather in Beijing, China, applied current state-of-the-art point cloud filtering methods such as dynamic statistical outlier removal (DSOR) and dynamic radius outlier removal (DROR) filters, verified the effectiveness of filtering and real-time performance of these methods under the snowy weather environment in Beijing, and proposed possible improvements to the methods. Experiments showed that the DSOR filter has better performance than the DROR filter in snowfall scenarios and is better suited for use in automated driving systems.