Robust Intensity-Based Localization Method for Autonomous Driving on Snow–Wet Road Surface

Aldibaja, Mohammad; Suganuma, Naoki; Yoneda, Keisuke · 2017 · Crossref

DOI: 10.1109/tii.2017.2713836

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of robust autonomous vehicle localization in adverse weather conditions, specifically on snow-covered and wet road surfaces. Standard LIDAR-based localization systems suffer from two primary issues in these environments: reduced image quality due to weak laser reflectivity on wet roads, and lateral localization errors caused by snow lines that mimic lane markings. The authors propose a method to enhance LIDAR image quality and improve lateral positioning accuracy to ensure stable autonomous driving under such conditions. The proposed solution involves two main technical improvements. First, to address low LIDAR image density and intensity variations, the authors implement an online accumulation strategy that stacks multiple LIDAR frames to increase data density. They then apply Principal Component Analysis (PCA) to reconstruct the accumulated LIDAR images based on a predefined high-definition map. This process filters noise, recovers missing areas, and aligns the intensity levels of the online LIDAR data with the map, thereby improving the reliability of intensity-based template matching. Second, to mitigate lateral drift caused by snow lines, the authors introduce an edge matching strategy. They extract edge profiles from both the map and LIDAR images using Sobel filters, focusing on gradients parallel to the vehicle’s heading. By matching these edge profiles, the system encodes static road structures like lane lines and roadside edges, reducing the influence of transient snow patterns. The method was evaluated using real-world data collected during the winter of 2016/2017 in Suzu and Kanazawa, Japan. The experimental platform utilized a Velodyne HDL-64E LIDAR, GNSS/IMU, and other sensors, operating at speeds up to 60 km/h. The results demonstrated that the proposed method significantly improved localization robustness. Specifically, the PCA-based reconstruction enhanced the texture and context of LIDAR images on wet surfaces, while the edge matching strategy provided stable lateral localization despite the presence of snow lines. The combined approach significantly reduced overall localization error compared to standard intensity-based methods. The significance of this work lies in its contribution to the commercialization of autonomous vehicles by addressing critical environmental challenges often neglected in research. By improving the reliability of LIDAR-based localization in snow and rain, the method supports the transition from controlled testing to real-world deployment. The findings suggest that integrating structural edge information with enhanced intensity matching can effectively overcome the limitations of sparse sensor data in adverse weather, ensuring safer and more accurate autonomous navigation.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• situational awareness