Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Road Surface Friction Estimation for Winter Conditions Utilising General Visual Features

Risto Ojala, Eerik Alamikkotervo

TL;DR

This work tackles estimating road surface friction under winter conditions from roadside camera imagery by framing it as a regression problem to predict the friction factor $f$ from images. It introduces WCamNet, a hybrid architecture that fuses general features from the visual foundation model DINOv2 with a CNN processing path to predict $\hat{f}$. The authors train and evaluate on a Finnish dataset of 48,791 image/friction pairs collected from 31 station sites and 58 cameras, showing that WCamNet achieves lower MAE and RMSE than CNN- and ViT-based baselines. They discuss limitations due to absolute error and spatial ambiguity in single-image labels and suggest future work combining weather context data to further boost performance. Overall, the work demonstrates that foundation-model features can improve roadside-camera friction estimation for potential ITS applications.

Abstract

In below freezing winter conditions, road surface friction can greatly vary based on the mixture of snow, ice, and water on the road. Friction between the road and vehicle tyres is a critical parameter defining vehicle dynamics, and therefore road surface friction information is essential to acquire for several intelligent transportation applications, such as safe control of automated vehicles or alerting drivers of slippery road conditions. This paper explores computer vision-based evaluation of road surface friction from roadside cameras. Previous studies have extensively investigated the application of convolutional neural networks for the task of evaluating the road surface condition from images. Here, we propose a hybrid deep learning architecture, WCamNet, consisting of a pretrained visual transformer model and convolutional blocks. The motivation of the architecture is to combine general visual features provided by the transformer model, as well as finetuned feature extraction properties of the convolutional blocks. To benchmark the approach, an extensive dataset was gathered from national Finnish road infrastructure network of roadside cameras and optical road surface friction sensors. Acquired results highlight that the proposed WCamNet outperforms previous approaches in the task of predicting the road surface friction from the roadside camera images.

Road Surface Friction Estimation for Winter Conditions Utilising General Visual Features

TL;DR

This work tackles estimating road surface friction under winter conditions from roadside camera imagery by framing it as a regression problem to predict the friction factor from images. It introduces WCamNet, a hybrid architecture that fuses general features from the visual foundation model DINOv2 with a CNN processing path to predict . The authors train and evaluate on a Finnish dataset of 48,791 image/friction pairs collected from 31 station sites and 58 cameras, showing that WCamNet achieves lower MAE and RMSE than CNN- and ViT-based baselines. They discuss limitations due to absolute error and spatial ambiguity in single-image labels and suggest future work combining weather context data to further boost performance. Overall, the work demonstrates that foundation-model features can improve roadside-camera friction estimation for potential ITS applications.

Abstract

In below freezing winter conditions, road surface friction can greatly vary based on the mixture of snow, ice, and water on the road. Friction between the road and vehicle tyres is a critical parameter defining vehicle dynamics, and therefore road surface friction information is essential to acquire for several intelligent transportation applications, such as safe control of automated vehicles or alerting drivers of slippery road conditions. This paper explores computer vision-based evaluation of road surface friction from roadside cameras. Previous studies have extensively investigated the application of convolutional neural networks for the task of evaluating the road surface condition from images. Here, we propose a hybrid deep learning architecture, WCamNet, consisting of a pretrained visual transformer model and convolutional blocks. The motivation of the architecture is to combine general visual features provided by the transformer model, as well as finetuned feature extraction properties of the convolutional blocks. To benchmark the approach, an extensive dataset was gathered from national Finnish road infrastructure network of roadside cameras and optical road surface friction sensors. Acquired results highlight that the proposed WCamNet outperforms previous approaches in the task of predicting the road surface friction from the roadside camera images.
Paper Structure (13 sections, 5 figures, 2 tables)

This paper contains 13 sections, 5 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Computer Vision-based Road Surface Condition Monitoring
  4. General Visual Features
  5. Research Gap
  6. Methods
  7. Problem Formulation and Dataset
  8. Model Architecture
  9. Training and Testing
  10. Results
  11. Accuracy Results
  12. Ablation Study
  13. Discussion

Figures (5)

  • Figure 1: Samples of image data with corresponding friction factor values.
  • Figure 2: Distribution of distances between the corresponding camera and weather stations.
  • Figure 3: Distribution of friction factor values in the dataset.
  • Figure 4: Image pairs demonstrating resized image samples and their respective DINOv2 patch tokens reduced to three feature dimensions with principal component analysis.
  • Figure 5: Proposed WCamNet architecture. Tensor sizes are reported for processing a single input image (batch size of one).