Airport Runway Foreign Object Detection Method Based on Improved YOLOv8

A modified YOLOv8 FOD object detection algorithm is proposed to address the issue of foreign object debris (FOD) target size being too small on airport runways. Add space to depth convolution to eliminate the negative impact of convolution and pooling operations on small targets; Introducing lightweight and efficient convolutional attention modules to focus on key features of the network from both spatial and depth dimensions; Improve multi-scale feature fusion to effectively improve the detection accuracy of small targets; Use weighted bounding box fusion algorithm to improve the accuracy of bounding boxes and the overall detection performance of the model. The experimental results show that the improved YOLOv8 object detection algorithm achieves 94% mAP50 and 93% mAP50-90 while meeting real-time requirements, effectively solving the problems of false detection, missed detection, and low positioning accuracy in FOD detection tasks.