TiHAN Adverse Weather Dataset

The increasing use of LiDAR sensors in autonomousdriving, drones, and intelligent transportation system highlightstheir ability to provide accurate 360-degree environmentalperception. However, unfavorable weather conditions such asfog, snow, rain, dust, low temperature introduce noise intoLiDAR point cloud data, complicating the perception anddecision-making process. Since labeling every point in pointcloud data is very hard, this paper addresses this challenge usingan energy-based model (EBM). We have proposed a LiDARadverse weather detection score (LAWDS) metric to classify theLiDAR frame as normal or adverse, an energy-based denoisingalgorithm to filter the noise by calculating the energy valueof each point. The TiHAN Adverse Weather Dataset (TiHANAWD) captured in various adverse weather conditions: rain,fog, snow, and high altitude in low temperature, is capturedoffering a valuable resource for research. Additionally, thepaper provides an introduction of a ring formed aroundthe LiDAR sensor during adverse weather. This approachimproves the robustness and reliability of autonomoussystems by removing noise from LiDAR data, leading tosafer and more efficient operations.