My country’s economy has developed rapidly in recent years, the country has built more airports to satisfy people’s travel. The airport runway carries hundreds of planes taking off and landing every day. Which has caused certain damage to the airport runway and brought potential safety hazards. The maintenance of the airport runway has become an important issue. The government invests a lot of money to maintain airport runway every year. Therefore, detecting cracks on the surface of the runway is very important to reduce maintenance costs and ensure safety. In view of the poor real-time performance of traditional runway crack detection and the long time-consuming manual inspection. We proposed a runway crack detection method based on YOLOv5, and annotated 3281 collected data sets. Finally, we use the YOLOv5 model with different parameters for training to obtain the optimal weight file and test it. Compare and analyze with several other classic machine learning models. The experimental results show that the yolov5 algorithm can effectively detect and identify airport runway cracks, and compared with traditional identification methods, the crack detection accuracy is also improved.