AI-Driven Pavement Condition Assessment for Municipalities Using Computer Vision

Abstract: Road infrastructure is a cornerstone of economic growth, providing essential connections between cities and states and facilitating the efficient movement of users and commerce. This directly impacts a country's GDP and improves the quality of life for its citizens. However, many regional municipalities in the U.S. face significant challenges in maintaining their road networks due to limited funding and a shortage of skilled personnel. As a result, the costs of maintaining and improving critical infrastructure continue to rise.
To maintain or increase the funding, the municipalities often rely on state and national grants for which they have to justify the need for funding by providing accurate pavement condition information. Traditional manual methods for assessing road conditions are labor-intensive, error-prone, and pose safety risks. While automated solutions offer greater accuracy and safety, they are often cost-prohibitive, requiring specialized vehicles equipped with sensors, 3D scanners, LIDARs, and vibration-based technologies, which are beyond the budget of many municipalities. This research aims to provide a cost-effective solution that enables all municipalities to assess pavement conditions without budgetary constraints. Current studies on pavement distress detection predominantly rely on deep learning techniques and computer vision algorithms to estimate potholes and cracks, but they often fall short in quantifying the damage severity. This paper introduces a novel approach leveraging Artificial Intelligence to address this gap. A Computer Vision Algorithm, You Only Look Once (YOLO), was trained to accurately detect pavement deformations in real-time using a mobile phone application. The phone, mounted on the rear of a vehicle, captures images as it moves. The YOLO model detects pavement distresses, assigns a score to quantify the severity, and records the geolocation of each deformation. This solution eliminates the need for expensive specialized vehicles, making it a cost-effective option for municipalities. Additionally, the AI model continuously learns from new data, ensuring it adapts to various conditions. This approach has the potential to revolutionize how municipalities manage road infrastructure, positively impacting local economic growth and community development.