Research on Long-Distance Railway Train Detection Technology Based on Multi-modal Sensor Fusion

To address the issues of low efficiency, poor safety, limited coverage, and reduced reliability under adverse weather conditions inherent in traditional contact-based train detection methods, a railway train detection system based on multi-modal sensing fusion of LiDAR and industrial-grade vision modules is designed. The system centers on high-precision LiDAR and cameras, integrating power supply through a switch-mode power supply and junction boxes. A mechanical pan-tilt unit is controlled by an STM32 microcontroller to enable multi-track, long-range train detection. Field data is collected at railway yards to accomplish temporal synchronization and spatial alignment, and a multi-modal fusion dataset is constructed. To improve fusion efficiency and real-time performance, a decision-level fusion algorithm is introduced to achieve accurate real-time detection and position tracking of trains across multiple tracks.