JOURNAL ARTICLE

Object detection of steel surface defect based on multi-scale enhanced feature fusion

Abstract

针对轻量级目标检测算法在钢表面缺陷检测任务中识别精度低的问题,提出一种多尺度增强特征融合的钢表面缺陷目标检测算法。该算法采用提出的自适应加权融合模块为不同层级特征自适应计算融合权重,将深层语义与浅层细节进行加权融合,使得浅层特征在不丢失细节信息的同时获得丰富的深层语义。利用提出的空间特征增强模块从3个独立方向强化融合特征,通过引出残差旁路增强网络结构的稳定性,使卷积过程能够挖掘到更多的关键信息。根据先验框与真实框的整体交并程度为模型选择更为合适的训练样本。实验结果表明,该算法的检测精度达到80.47%,相比原始算法提升6.81%。该算法的参数量为2.36 M,计算量为952.67 MFLOPs,能快速且高精度检测钢材表面的缺陷信息,具有较高的应用价值。

Keywords:
Feature (linguistics) Fusion Artificial intelligence Pattern recognition (psychology) Scale (ratio) Computer science Surface (topology) Object (grammar) Image fusion Semantics (computer science) Computer vision Image (mathematics) Mathematics Geography Geometry

Metrics

0
Cited By
0.00
FWCI (Field Weighted Citation Impact)
0
Refs
0.06
Citation Normalized Percentile
Is in top 1%
Is in top 10%

Topics

Industrial Vision Systems and Defect Detection
Physical Sciences →  Engineering →  Industrial and Manufacturing Engineering
Welding Techniques and Residual Stresses
Physical Sciences →  Engineering →  Mechanical Engineering
Infrastructure Maintenance and Monitoring
Physical Sciences →  Engineering →  Civil and Structural Engineering
© 2026 ScienceGate Book Chapters — All rights reserved.