Application of the Deep Learning Methodology for the Detection of Cracks in Asphalt Roads

Insufficient data availability and suboptimal monitoring systems notably reduced the lifespan of flexible pavements. This study addressed these challenges by introducing an innovative tool to enhance control over pavement conditions. Initial field observations identified various types of cracking, forming the basis for a comprehensive photogrammetric data survey. This dataset was then employed to train a Deep Learning model for object detection. The results showcased the model’s exceptional reliability in identifying pavement cracks, achieving an impressive accuracy rate of 83.33%. The study emphasizes the practical viability of the proposed tool as an effective means of monitoring roadway conditions. By overcoming data limitations and monitoring deficiencies, this research not only contributes to the progression of pavement maintenance practices but also establishes a solid foundation for creating a maintenance and repair priority map. This serves as a valuable tool for targeting interventions, enhancing the longevity and overall performance of flexible pavements, and represents a significant advancement in sustainable infrastructure management.