LiDARDustX: A LiDAR Dataset for Dusty Unstructured Road Environments

TL;DR

LiDARDustX tackles the scarcity of perception data for dusty, unstructured roads by introducing a large, multi-sensor LiDAR dataset collected in mining-like environments. The dataset provides 30,000 frames with $7$-DOF bounding boxes $(c_x,c_y,c_z)$, $(l,w,h)$, and heading $ heta$, across 14 object categories and 16 semantic classes, with more than 80% of scenes dust-affected. A six-step annotation pipeline enables high-quality ground-truth labels, including point cloud stitching and ground-plane estimation, yielding precise annotations for detection and segmentation. Benchmark results across six detectors, four segmentation methods, and a LiDAR multi-task baseline reveal substantial degradation of perception performance under dust, while models with robust feature processing and multi-task objectives show improved resilience; the dataset thus provides a valuable resource for advancing perception in dusty unstructured environments.

Abstract

Autonomous driving datasets are essential for validating the progress of intelligent vehicle algorithms, which include localization, perception, and prediction. However, existing datasets are predominantly focused on structured urban environments, which limits the exploration of unstructured and specialized scenarios, particularly those characterized by significant dust levels. This paper introduces the LiDARDustX dataset, which is specifically designed for perception tasks under high-dust conditions, such as those encountered in mining areas. The LiDARDustX dataset consists of 30,000 LiDAR frames captured by six different LiDAR sensors, each accompanied by 3D bounding box annotations and point cloud semantic segmentation. Notably, over 80% of the dataset comprises dust-affected scenes. By utilizing this dataset, we have established a benchmark for evaluating the performance of state-of-the-art 3D detection and segmentation algorithms. Additionally, we have analyzed the impact of dust on perception accuracy and delved into the causes of these effects. The data and further information can be accessed at: https://github.com/vincentweikey/LiDARDustX.

Figures (7)

• Figure 1: Overview of dust point clouds from 6 types of LiDAR in different perspectives, where pink represents dust, red represents the ground, light blue represents walls, and green/blue indicates different types of vehicles.
• Figure 2: Collection Platform LiDAR Position Schematic.
• Figure 3: The LiDARDustX dataset showcases point clouds from various LiDAR sensors. The points in the cycle are zoomed in and shown in the white boxes for a better view.
• Figure 4: The methodologies for point cloud segmentation annotation.
• Figure 5: The number of 3D box annotations in different detection categories.