Towards an analysis of damage progression in concrete pavements using machine learning and computer vision

Abstract

In the field of pavement damage analysis, long‐term damage monitoring and assessment remain a formidable challenge due to the dynamic nature of damage development and environmental changes. This study presents preliminary progress towards a system for tracking and analyzing the progression of pavement deterioration over time using image datasets of real concrete pavement. The research focuses on specific road sections, with annual data collection over three years, resulting in a time‐series dataset with detailed spatiotemporal information. The proposed methodology integrates GPS data for precise location mapping and employs advanced image analysis techniques to consistently identify and monitor pavement damages across different time frames. Image rectification and alignment processes are applied to ensure comparability in detail and orientation. The current status of the project is presented, including initial implementations of the methodology, limitations encountered, and suggestions for improvement in data collection. The goal is to establish a robust and replicable methodology that facilitates understanding the evolution of pavement damage. This spatiotemporal matching approach aims to lay the foundation for agencies to monitor the progression of issues and enable timely interventions throughout the pavement life cycle.