Journal Article

Detection model of foreign object on airport runway based on improved RT-DETR

Zhang, Yi, Sun, Mingkai, Meng, Ruifeng

2026

Overview

Various foreign objects on the runway will pose major hidden dangers to the safety of the aircraft. It is of great significance to be able to detect foreign objects on airport runways quickly and in real time. The existing detection technology has dual limitations in real-time performance and accuracy, which makes it difficult to be competent in busy airport scenarios with high flight take-off and landing frequency and a complex operating environment. To solve this kind of problem, we improve the model based on RT-DETR. First, the size and parameter amount of the model are reduced by replacing MobileNetV3 _ Small in the trunk; Secondly, adding the MSMHSA attention mechanism to AIFI has richer feature representation and better performance. Finally, improving the RepC3 module and adding the RetBlock module significantly improves the training stability and convergence efficiency of the extremely deep model. Experiments show that the improved model improves the AP 0. 5 index by 0.2% and the AP 0.5-0.95 index by 0.6%. In addition, the number of model parameters is reduced by 58.5%, and the computational complexity is reduced by 70%. This optimized model is more appropriate for real-time FOD detection of airport runways, as it surpasses the original RT-DETR in terms of both efficiency and accuracy.

• Publisher: IOP Publishing
• Subject: Accuracy / Aircraft landing / Airports / Complexity / Modules / Parameters / Real time / Runways