YoPNet:A YOLO Based Detection Algorithm for Railway Fasteners

This study proposes an innovative algorithm YoPNet for automatic fastener detection in heavy-load railroad maintenance. To address challenges including fastener overlap, complex spatial arrangements, and environmental variations, we constructed a specialized dataset containing 2,000+ annotated images enhanced through multi-source acquisition and hybrid data augmentation strategies. The proposed method synergizes YOLOv11's efficient detection capability with PSPNet's pyramid pooling module for multi-scale feature extraction, achieving a 99.0% mAP@ 0.5 on our dataset. Experimental results demonstrate significant improvements over baseline YOLOv11, with a 2% accuracy increase and 28% reduction in false detection rates under challenging conditions (low-light, rain/fog). The framework's hardware compatibility (tested on NVIDIA RTX 4090) and computational efficiency highlight its potential for realworld deployment. This research advances automated railroad maintenance by addressing critical technical bottlenecks in visual detection tasks.