Abstract:

Roads, as the country’s vital infrastructure, play a very important role in the country’s economic and social development. Therefore, proper road maintenance is of utmost importance. Timely road maintenance can prevent more serious damage and reduce future repair costs. With the advancement of technology, automated data acquisition methods have replaced traditional methods, and the use of machine vision and deep learning techniques has enabled the automatic processing and analysis of road pavement images.

In recent years, self-driving vehicles have emerged as an effective tool in data acquisition and detection of asphalt pavement damage. These vehicles, equipped with advanced sensors and cameras, can continuously collect data from the road surface without human intervention. In this study, using data collected by autonomous vehicles and employing the YOLO algorithm version 8, which is considered one of the most powerful machine vision algorithms, eight types of pavement damage, including lizard skin cracks, longitudinal and transverse cracks, patches, and potholes, each with low and high intensities, were identified, and a model was developed that achieved an average accuracy of 0.815 for all damages.