LiDAR De-Snow Score (DSS): Combining Quality and Perception Metrics for Optimized De-Noising

Valentina Donzella, Pak Hung Chan, Daniel Gummadi, Abu Mohammed Raisuddin, Eren Erdal Aksoy

Abstract: The testing and safety cases of assisted and automated driving (AAD) functions require considerations for nonideal environmental conditions, such as adverse and extreme weather. In these extreme conditions, perception sensors (e.g., camera, LiDAR, and RADAR), which build the situational awareness of the vehicle, might produce noisy and degraded data. Therefore, it is key to consider: 1) how to reliably and robustly measure data degradation; 2) how to evaluate de-noising techniques; and 3) linking perception performance to de-noising quality. This article focuses on de-snowing of LiDAR data, as falling snow is one of the most variable and dangerous conditions encountered while driving, and LiDAR can provide essential 3-D information to still enable safe vehicle navigation. Using the Winter adverse driving dataset (WADS) dataset, which contains segmented point clouds including falling and deposited snow points, state-of-the-art de-snowing techniques are compared using an array of adapted point cloud quality metrics, combined with downstream segmentation (perception) performance after de-snowing evaluated using perception-based metrics. The different metrics can capture different aspects/effects of the data degradation, and hereby the novel De-Snow score (DSS) is proposed and applied to have a holistic evaluation of the de-noising techniques considering both data quality and expected perception performance. Based on DSS, the most promising de-noising algorithms are identified. The proposed methodology and DSS can pave the way for a standardized approach when evaluating perception sensor data degradation and de-noising techniques.

External IDs:doi:10.1109/jsen.2025.3570478