Scholarlink

<p>As a critical flexible component in industrial applications, the detection of surface defects in steel wire ropes is essential for ensuring production safety. To address the limitations of existing detection methods in terms of accuracy and efficiency, this paper proposes a lightweight detection model, YOLOv7-GPSS, based on improvements to YOLOv7. First, the backbone network is restructured by introducing the lightweight GCBS module to reduce the model’s complexity. Second, the SPAConv, a separated local attention convolution, is designed to form the ELAN-SPA composite structure, which enhances fine-grained feature representation and improves small target detection sensitivity. The SimAM, a parameterless attention mechanism, is then integrated into the pooling layer to strengthen spatial information interaction during feature learning. Finally, SPD-Conv is applied to optimize the detection head, improving small defect detection and significantly enhancing small defect feature extraction. Experimental results demonstrate that the proposed model increases detection accuracy by 6.2%, while reducing the number of parameters by 18.1% and FLOPs by 18.9%, showing better engineering applicability in wire rope damage detection scenarios.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

YOLOv7-GPSS: a YOLOv7-based wire rope surface defect detection algorithm

  • 期刊论文

  • 发表日期：

    2025年7月30日

查看全文
  • Rui Shi,
  • Hongwang Li,
  • Tianbing Ma,
  • Changpeng Li,
  • Cuiwei Tian,
  • Qinghui Shi

摘要

As a critical flexible component in industrial applications, the detection of surface defects in steel wire ropes is essential for ensuring production safety. To address the limitations of existing detection methods in terms of accuracy and efficiency, this paper proposes a lightweight detection model, YOLOv7-GPSS, based on improvements to YOLOv7. First, the backbone network is restructured by introducing the lightweight GCBS module to reduce the model’s complexity. Second, the SPAConv, a separated local attention convolution, is designed to form the ELAN-SPA composite structure, which enhances fine-grained feature representation and improves small target detection sensitivity. The SimAM, a parameterless attention mechanism, is then integrated into the pooling layer to strengthen spatial information interaction during feature learning. Finally, SPD-Conv is applied to optimize the detection head, improving small defect detection and significantly enhancing small defect feature extraction. Experimental results demonstrate that the proposed model increases detection accuracy by 6.2%, while reducing the number of parameters by 18.1% and FLOPs by 18.9%, showing better engineering applicability in wire rope damage detection scenarios.