Research on a Deep Learning-Based Foreign Object Intrusion Detection Model for Railway Tracks Using Mamba-YOLO

Railway safety is vital for national economic development, with track foreign object intrusion posing significant risks. This study explores the feasibility of the Mamba-YOLO model for intelligent rail intrusion detection by leveraging multi-scale feature extraction and adaptive attention mechanisms to balance accuracy and resource efficiency, comparing network architectures under special scenarios to propose a lightweight deployment strategy for intelligent maintenance systems. Aiming to address traditional models’ limitations such as accuracy drops in complex environments, lack of multimodal fusion, and insufficient training integration, it develops a multimodal detection framework combining novel architectural features and multi-source data fusion to enhance intrusion detection precision and intelligence for railway safety advancement. Covering theoretical principles, algorithmic foundations, model evolution, and comparative analysis, the research offers practical insights for optimizing rail safety systems via advanced deep learning techniques.