Online Rail Fastener Detection Based on YOLO Network

Traveling by high-speed rail and railway transportation have become an important part of people’s life social production. Track is the basic equipment transportation, its performance directly affects service lifetime lines vehicles. The anomaly detection fasteners in a priority, while traditional manual method extremely inefficient dangerous to workers. Therefore, this paper introduces efficient computer vision into system not only locate normal fasteners, but also recognize states. To be more specific, mainly studies fastener based on improved You can Only Look Once version 5 (YOLOv5) network, completes real-time classification YOLOv5 network proposed contains five sections, which are Input, Backbone, Neck, Head Detector read-only Few-shot Example Learning module. main purpose project improve precision shorten time. Ultimately, confirmed superior other advanced algorithms. This model achieves on-line completing “sampling-detection-recognition-warning” cycle single sample before next image sampled. Specifically, mean average reaches 94.6%. And speed 12 ms per deployment environment NVIDIA GTX1080Ti GPU.