Detection of road pavement deteriorations by using unmanned aerial vehicle (UAV)

Assessing pavement condition is crucial for reducing future economic losses and improving structural reliability and durability. Data from assessing pavement condition is used to evaluate pavement performance and reliability. Assessing the condition of pavement is challenging because it is expensive, poses safety concerns, and has accessibility limitations, especially once it starts to deteriorate. The objective of this research is to create an advanced classification model that can swiftly categorize four significant types of pavement distresses: transverse cracks, longitudinal cracks, alligator cracks, and patch deteriorations. High resolution images representing transverse cracks, longitudinal cracks, crocodile cracks and patch deteriorations for flexible pavements were collected using an Unmanned Aerial Vehicle (UAV) within Afyon Kocatepe University Ahmet Necdet Sezer campus. The image classification model utilizes YOLOv5, a cutting-edge CNN architecture designed specifically for object detection. It divides the image into multiple regions to detect and classify objects.When the complexity matrix of the YOLOv5 model used was examined for 70 epoks and 16 batch size, it was seen that a success rate of 88,25 was achieved. Based on these results, the YOLOv5m model emerged as the optimal choice, delivering superior performance in both test results and detection times. The methodology behind the developed model will help decision-making institutions and organizations to reduce the need for on-site presence, increase safety, and respond urgently to decide the safest route to take after meteorological events. Furthermore, implementing this model will enable transportation engineers to swiftly assess pavement degradation, expedite decision-making regarding road repairs and recovery, and determine the most suitable maintenance and repair approach.