Classification of Road Surface and Weather-Related Condition Using Deep Convolutional Neural Networks

verfasst von

Alexander Busch, Daniel Fink, Max Heinrich Laves, Zygimantas Ziaukas, Mark Wielitzka, Tobias Ortmaier

Abstract

In order to achieve the goal of autonomous driving, a precise perception of the vehicle’s environment is required. In particular, the weather-related road condition has a major influence on vehicle dynamics and thus on driving safety. In this paper, we compare Deep Convolutional Neural Networks of different computational effort, namely Inception-v3, GoogLeNet and the much smaller SqueezeNet, for classification of road surface and its weather-related condition. Previously, different regions of interest were compared in order to provide the networks with optimal input data.

Organisationseinheit(en)

Institut für Mechatronische Systeme

Typ

Aufsatz in Konferenzband

Seiten

1042-1051

Anzahl der Seiten

10

Publikationsdatum

13.02.2020

Publikationsstatus

Veröffentlicht

Peer-reviewed

Ja

ASJC Scopus Sachgebiete

Fahrzeugbau, Luft- und Raumfahrttechnik, Maschinenbau, Fließ- und Transferprozesse von Flüssigkeiten

Elektronische Version(en)

https://doi.org/10.1007/978-3-030-38077-9_121 (Zugang: Geschlossen)