<abstract xmlns="http://www.w3.org/1999/xhtml">

<p>Potholes detection is an essential aspect of road safety and road infrastructure maintenance. Potholes, which are typically caused by a combination of heavy traffic and weather, are depressions or holes in the road surface that can cause damage to specific parts of a vehicle. Autonomous vehicles, in particular, must be capable of detecting and avoiding them. Hitting a deep or sharp-edged pothole at high speed can lead to loss of control or even an accident. This makes pothole detection all the more important. The accuracy of pothole detection systems installed in autonomous vehicles may be significantly impaired by adverse weather and bad light conditions. Therefore, the classification accuracy of selected well-known computer vision models for pothole detection under these specific conditions has been investigated. The results were then compared with state-of-the-art methods. Our findings showed that we outperformed many of them when used under adverse weather and low light situations. This paper presents valuable insights into the precision of various computer vision models for potholes detection. It may aid in selecting the optimal model for a specific application.</p>
</abstract>

Potholes detection is an essential aspect of road safety and road infrastructure maintenance. Potholes, which are typically caused by a combination of heavy traffic and weather, are depressions or holes in the road surface that can cause damage to specific parts of a vehicle. Autonomous vehicles, in particular, must be capable of detecting and avoiding them. Hitting a deep or sharp-edged pothole at high speed can lead to loss of control or even an accident. This makes pothole detection all the more important. The accuracy of pothole detection systems installed in autonomous vehicles may be significantly impaired by adverse weather and bad light conditions. Therefore, the classification accuracy of selected well-known computer vision models for pothole detection under these specific conditions has been investigated. The results were then compared with state-of-the-art methods. Our findings showed that we outperformed many of them when used under adverse weather and low light situations. This paper presents valuable insights into the precision of various computer vision models for potholes detection. It may aid in selecting the optimal model for a specific application.

Analysis of Pothole Detection Accuracy of Selected Object Detection Models Under Adverse Conditions

Department of Quantitative Methods and Economic Informatics Faculty of Operation and Economics of Transport and Communications

Department of Electronics and Communication Engineering University of Engineering & Management

Transport and Telecommunication Journal

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Potholes detection is an essential aspect of road safety and road infrastructure maintenance. Potholes, which are typically caused by a combination of heavy...

Analysis of Pothole Detection Accuracy of Selected Object Detection Models Under Adverse Conditions

Publié en ligne: 23 avr. 2024

Pages: 209 - 217

DOI: https://doi.org/10.2478/ttj-2024-0016

Mots clés
Computer vision, object detection, potholes, YOLOv7, FRCNN

© 2024 Jaroslav Frnda et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Analysis of Pothole Detection Accuracy of Selected Object Detection Models Under Adverse Conditions

Publié en ligne: 23 avr. 2024

Pages: 209 - 217

DOI: https://doi.org/10.2478/ttj-2024-0016

Mots clésComputer vision, object detection, potholes, YOLOv7, FRCNN

© 2024 Jaroslav Frnda et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Mots clés
Computer vision, object detection, potholes, YOLOv7, FRCNN