Abstract
With the extensive development in infrastructures, many airports are built in order to satisfy travelling needs of people. The frequent arrival and departure of numerous plans lead to substantial runway damage and related safety concerns. So, the regular maintenance of runway has become an essential task specially for detection and classification of cracks in terms of owing to the intensity heterogeneity of cracks such as low real-time performance and the long time-consuming manual inspection. This paper introduces a new dataset named as ARID with 8 different crack classes. A runway crack detection model based on YOLOv5 and Faster RCNN has been proposed which is annotated on 8,228 collected datasets. Then the model is trained with different parameters for training to obtain the optimal result. Finally, based on experimental result, the crack detection precision has improved from 83% to 92%, while the recall has increased from 62.8% to 76%.
Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
