🤖 AI Summary
To address high false-negative and false-positive rates, as well as low inference efficiency in foreign object detection on railway transmission lines, this paper proposes MRS-YOLO—a YOLO11-based detector. It introduces the C3k2_MAKDF module for multi-scale adaptive kernel deep feature fusion, an RCFPN (re-calibrated feature pyramid network) architecture, and an SC_Detect head enabling spatial-channel collaborative localization. Additionally, channel pruning is integrated for model lightweighting. Experiments on a custom railway foreign object dataset show that MRS-YOLO achieves mAP₅₀ = 94.8% and mAP₅₀:₉₅ = 86.4%, outperforming the baseline by 0.7 and 2.3 percentage points, respectively. Model parameters and computational cost (GFLOPs) are reduced by 44.2% and 17.5%, significantly enhancing small-object detection accuracy and edge-deployment efficiency.
📝 Abstract
Aiming at the problems of missed detection, false detection and low detection efficiency in transmission line foreign object detection under railway environment, we proposed an improved algorithm MRS-YOLO based on YOLO11. Firstly, a multi-scale Adaptive Kernel Depth Feature Fusion (MAKDF) module is proposed and fused with the C3k2 module to form C3k2_MAKDF, which enhances the model's feature extraction capability for foreign objects of different sizes and shapes. Secondly, a novel Re-calibration Feature Fusion Pyramid Network (RCFPN) is designed as a neck structure to enhance the model's ability to integrate and utilize multi-level features effectively. Then, Spatial and Channel Reconstruction Detect Head (SC_Detect) based on spatial and channel preprocessing is designed to enhance the model's overall detection performance. Finally, the channel pruning technique is used to reduce the redundancy of the improved model, drastically reduce Parameters and Giga Floating Point Operations Per Second (GFLOPs), and improve the detection efficiency. The experimental results show that the mAP50 and mAP50:95 of the MRS-YOLO algorithm proposed in this paper are improved to 94.8% and 86.4%, respectively, which are 0.7 and 2.3 percentage points higher compared to the baseline, while Parameters and GFLOPs are reduced by 44.2% and 17.5%, respectively. It is demonstrated that the improved algorithm can be better applied to the task of foreign object detection in railroad transmission lines.