High Precision Steel Strand Defect Detection Based on RepHELAN-EYolo

A steel strand damage detection model based on RepHELAN-EYolo is proposed to address safety issues caused by steel strand breakage. Due to the existence of strand damage of different sizes and uneven distribution, adding the RepHELAN feature extraction module to YOLOv10 to improve the extraction of strand details, the introduction of Channel Attention and Spatial mechanism to help increase the degree of recognition of helpful features. Also, to improve the prediction framework's robustness, the loss function is enhanced by substituting CIOU with EIOU. Experimental results demonstrate that the RepHELAN-EYolo algorithm outperforms the original YOLOv10 model, significantly enhancing steel strand damage detection accuracy with a 20.7% increase in mean average precision (mAP) while delivering superior overall performance.