Road condition estimation using deep learning with hyperspectral images: detection of water and snow

Road surface condition monitoring is one of the most crucial tasks for vehicle perception systems. The presence of water, snow, ice, or any other substance covering the road surface directly affects the rolling resistance and controllability of the vehicle, which is directly related to the safety of...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Proceedings of the Estonian Academy of Sciences 2024-01, Vol.73 (1), p.77-91
Hauptverfasser: Valme, Daniil, Galindos, Javier, Liyanage, Dhanushka Chamara
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Road surface condition monitoring is one of the most crucial tasks for vehicle perception systems. The presence of water, snow, ice, or any other substance covering the road surface directly affects the rolling resistance and controllability of the vehicle, which is directly related to the safety of the traffic participants. Many sensors, such as RGB cameras, infrared sensors, and mmWave sensors, are used to monitor and inspect road surfaces. The research aims to provide a tool to segment an input image into correct classes. The DeepHyperX toolbox was used for the rapid prototyping of deep learning (DL) classification models for hyperspectral images. The effectiveness of the developed algorithm in several case studies is presented, and it is verified that the low number of iterations is enough to detect the water, snow, and ice on the road surface. Keywords: road surface condition monitoring, water, snow, advanced driver assistance system.
ISSN:1736-6046
1736-7530
DOI:10.3176/proc.2024.1.09