IDSOR: Intensity- and Distance-Aware Statistical Outlier Removal for Weather-Robust LiDAR Point Clouds

LiDAR point clouds captured in rain or snow are often corrupted by weather-induced returns, which can degrade perception and safety-critical scene understanding. This paper proposes Intensity- and Distance-Aware Statistical Outlier Removal (IDSOR), a range-adaptive filtering method that jointly exploits intensity cues and neighborhood sparsity. By incorporating an empirical, range-dependent distribution of weather returns into the threshold design, IDSOR suppresses weather-induced points while preserving fine structural details without cumbersome manual parameter tuning. We also propose a variant that uses a previously proposed method to estimate the weather return distribution from data, and integrates it into IDSOR. Experiments on simulation-augmented level-crossing measurements and on the Winter Adverse Driving dataset (WADS) demonstrate that IDSOR achieves a favorable precision-recall trade-off, maintaining both precision and recall above 90% on WADS.

💡 Research Summary

The paper addresses a critical problem in autonomous‑driving perception: LiDAR point clouds become heavily polluted by weather‑induced returns (rain, snow) that appear as dense, low‑intensity points near the sensor and as sparse points at longer ranges. Existing geometric outlier‑removal methods—Distance Outlier Removal (DOR), Statistical Outlier Removal (SOR), Dynamic Radius Outlier Removal (DROR), and Distance‑Statistical Outlier Removal (DSOR)—rely solely on spatial density and ignore the intensity channel, which is a strong cue for distinguishing true surface returns from weather particles. A recent extension, DDIOR, adds intensity but requires manual, distance‑wise weighting, making it impractical for real‑world deployment.

To overcome these limitations, the authors propose IDSOR (Intensity‑ and Distance‑aware Statistical Outlier Removal). The core idea is to embed an empirically derived, range‑dependent probability density function (PDF) of weather returns into the outlier‑detection threshold and to modulate this threshold with the normalized intensity of each point. The PDF is obtained by aggregating simulated rain/snow point clouds (using the augmentation method of