A Comparative Study of Vision-based Road Surface Classification Methods for Dataset From Different Cities

In order to classify road surface for intelligent vehicles, we develop a data set including day and night driving from different cities. Meanwhile, the dataset contains various scenarios such as dry, wet, and icy conditions. Based on the developed data set, we employ both classical machine learning...

Full description

Saved in:
Bibliographic Details
Published in2022 IEEE 5th International Conference on Industrial Cyber-Physical Systems (ICPS) pp. 01 - 06
Main Authors Liang, Huan, Zhang, Hui, Sun, Zhifeng
Format Conference Proceeding
LanguageEnglish
Published IEEE 24.05.2022
Subjects
Cyber-physical systems
Decision making
Deep learning
Intelligent vehicles
Machine learning
Road conditions
Road surface classification
Roads
Support vector machines
Urban areas
Online AccessGet full text

Cover

More Information
Summary:In order to classify road surface for intelligent vehicles, we develop a data set including day and night driving from different cities. Meanwhile, the dataset contains various scenarios such as dry, wet, and icy conditions. Based on the developed data set, we employ both classical machine learning and deep learning techniques for road surface classification. In the machine learning approach, we utilize support vector machine, k nearest neighbor, and k means to classify road texture features extracted from images. In the deep learning approach, ResNet and MobileNet are used as a backbone to extract features in images to be classified by a softmax layer. Experimental results show that deep learning techniques can achieve 98.09% accuracy to classify road surfaces between dry, wet, and icy.
DOI:10.1109/ICPS51978.2022.9816956