Airport runway FOD detection model based on improved YOLOv10m

Abstract Foreign Object Debris (FOD) poses a significant safety risk to aircraft operations, making efficient and accurate detection methods crucial. Existing detection techniques often struggle to meet the demands of busy airports due to limitations in real-time performance and detection accuracy. To address these challenges, we proposed a lightweight FOD detection model based on YOLOv10. By incorporating PConv and EMA attention mechanisms, our model enhances detection accuracy, achieving a 0.4% increase in mAP@50 and a 0.7% improvement in mAP@50-95 compared to the original YOLOv10m. Additionally, our model reduces the number of parameters by 19.8% and decreases computational complexity by 20.9%, thanks to the integration of Feature Pyramid Shared Convolution (FPSConv) and a lightweight shared detection head (Detect-LSCD). These improvements make our model more suitable for real-time FOD detection on airport runways, offering a more efficient and accurate solution than the original YOLOv10m.