Electronics, Journal Year: 2025, Volume and Issue: 14(5), P. 831 - 831

Published: Feb. 20, 2025

The accurate and efficient detection of steel surface defects remains challenging due to complex backgrounds, diverse defect types, varying scales. existing CNN-based methods often struggle with capturing long-range dependencies handling background noise, resulting in suboptimal performance. Meanwhile, although Transformer-based approaches are effective modeling global context, they typically require large-scale datasets computationally expensive, limiting their practicality for industrial applications. To address these challenges, we introduce a novel attention-based salient object detector, called the ASOD, enhance effectiveness detectors strip defects. In particular, first design channel-attention-based block including max/average pooling focus on relevant channel-wise features while suppressing irrelevant channel responses, where maximizing extracts main local regions, removing average obtain overall details. Then, new based spatial attention is designed emphasize area areas. addition, cross-spatial-attention-based fuse feature maps multiple scales filtered through proposed produce better semantic information such that detector adapts sizes. experiments show ASOD achieves superior performance across evaluation metrics, weighted F-measure 0.9559, an structure measure 0.9230, Pratt’s figure meri 0.0113, mean absolute error 0.0144. demonstrates strong robustness noise interference, maintaining consistently high even 10–20% dataset which confirms its stability reliability.

Language: Английский