Revolutionizing Road Maintenance: A Data Fusion and AI-Based Approach

Abstract

The deterioration of pavement conditions over time can lead to significant costs for rehabilitation once the pavement conditions fall below a certain level of deterioration. To mitigate these costs, it is possible to extend the service life of pavement through periodic maintenance at a relatively low cost. To increase the frequency of inspections and enhance the effectiveness of maintenance activities, this study proposes a crowdsourcing-based inspection system for autonomous road condition assessment that can be mounted on multiple vehicles. The proposed inspection system can evaluate the pavement condition based on the Pavement Surface and Evaluation Rating (PASER) system. Vehicles equipped with this data acquisition system are driven over the road through the cities twice per week, thus comprehensive RGB-D data of the road surface are collected timely and widely. The system can detect various defects, including transverse cracks, longitudinal cracks, alligator cracking, and 3D defects such as rutting and potholes through a trained deep learning-based classification model. Furthermore, the proposed autonomous road inspection system can monitor the evolutionary changes of defects such as cracks and potholes to enable damage prognosis.