Technical Content

Introduction: To support local industries in Georgia, including agriculture and manufacturing, the state updated the permissible weights of trucks on its roads. This led to a rise in heavy vehicle usage, putting added stress on Georgia’s rural road infrastructure, particularly with the growth of the agricultural sector and the new electric vehicle industry. With sparse populations in these rural areas, there’s a smaller tax base, making road maintenance challenging. This study introduces a cutting-edge approach using remote sensing and machine learning (ML) algorithms to monitor and assess the condition of these rural roads. The method promises to be cost-effective, eliminating the need for extensive field inspections and aiding governments in prioritizing maintenance projects. Literature Review: Previous studies, notably GDOT RP 20-27, have explored the use of satellite images and remote sensing for highway management. This study achieved a classification accuracy of 74% using the XGBoost classifier to predict road conditions, drawing from a massive dataset of 5.8 million data points spanning 2010-2022. The importance of local road systems to communities cannot be overstated, yet many face challenges due to budget constraints and the absence of reliable road condition data. This current study takes a data-centric approach to address these issues, emphasizing the unique characteristics of chip-sealed pavements, which are common in local roads. It also seeks to refine and enhance ML techniques for more accurate road condition monitoring.

Objectives:

• Develop and evaluate ML models to predict the surface condition of rural roads using remote sensing data.

• Examine the potential of multispectral satellite images to assess the quality of chip-sealed asphalt pavements.