Development of a Framework for Identifying Asphalt Pavement Cracking Distresses Using Machine Learning

Asphalt pavement cracking is one of the most critical distresses affecting pavement performance and service life. When pavement deteriorates, it can lead to safety hazards, higher vehicle maintenance costs, and expensive repairs for cities and states—making early detection essential for everyone who relies on the roadway system. To address this challenge, the research team developed a prototype cracking identification system that integrates a customized machine learning model with computer vision algorithms. High-resolution images collected from drones or ground-based cameras are processed within the system to automatically detect and classify major cracking types. The core of the framework utilizes the You Only Look Once (YOLO) architecture, which enables fast detection and accurate localization of cracking distresses. Experimental results demonstrate that the model achieves over 80% accuracy across multiple crack categories. In addition to detection efficiency, the system emphasizes usability, providing an effective tool for transportation agencies to support pavement evaluation and management. This innovative approach represents a step toward more automated, reliable, and scalable pavement condition assessment, which supports safer, more cost-effective transportation systems.