Railway Intrusion Detection Based on YOLOV8

The issue of foreign object intrusion on railways has become a significant hidden threat to the safety of train operations and the efficiency of traffic scheduling. To enhance the intelligence level of object detection in railway scenarios, this study conducts a target detection investigation based on the YOLOv8 model. A dataset consisting of 802 images covering seven categories of typical foreign objects is utilized. Through image preprocessing, model training, and performance evaluation, the effectiveness of YOLOv8 in complex railway environments is experimentally verified. The model achieves high detection accuracy across multiple object categories, with a mean Average Precision (mAP@0.5) reaching 83.2%. In addition, two sets of ablation experiments are designed to systematically analyze the impact of data augmentation strategies and feature fusion methods on detection performance. The results demonstrate that YOLOv8 exhibits strong robustness and practicality, offering an efficient and reliable technical foundation for real-time monitoring and intelligent early warning of railway foreign object intrusions.