Geometric and Radiometric Enhancements of Vehicles in LiDAR Point Clouds with Snow Accumulation

Document Type

Conference Proceeding

Publication Date

1-1-2026

Abstract

LiDAR perception is critical for autonomous driving, yet research primarily focuses on falling snow as atmospheric noise that degrades performance. The impact of accumulated snow on target objects, however, remains largely unexplored. Following our preliminary evaluations on the WADS-3D and CADC datasets, which revealed a counter-intuitive improvement in vehicle detection accuracy during heavier snowfall due to increased object point counts, we investigate the underlying physical mechanisms. Through a controlled experiment, we validate two hypotheses: (1) geometric inflation and (2) a radiometric shift from specular to diffuse reflectance. We analyze LiDAR point counts and intensity distributions across four different LiDARs (Ouster, Velodyne, Ruby, and Iris). Our results confirm that accumulated snow not only increases the effective surface area of vehicles but also alters surface physics by converting vehicles from specular-like to diffuse-like reflectors. This radiometric shift effectively increases point density and return intensity, significantly improving the visibility of distant vehicles.

Publication Title

2026 IEEE International Workshop on Metrology for Automotive Metroautomotive 2026 Proceedings

ISBN

[9798331551285]

Link to Full Text

Share

COinS