MTF-YOLO: a steel surface defect detector with multiscale texture fusion and variance-guided convolution
Year: 2026
Type: article
Abstract: In industrial surface inspection, defects are often small, weakly textured, and bounded by uncertain edges—conditions under which state‑of‑the‑art detectors struggle, especially in low‑contrast scenes and for thin cracks. To address these challenges, this paper proposes MTF‑YOLO, a YOLO‑based detector that integrates three modules: Defect Variance‑Guided Convolution (DVGC), Defect Edge Enhancement (DEE), and Defect‑Oriented Texture Fusion (DOTF). Firstly, DVGC explicitly models the statistical disparity between foreground and background, improving separability for small, low‑contrast targets. Secondly, DEE sharpens contours and fine details at the feature‑level, mitigating localization drift caused by blurred boundaries. Thirdly, DOTF aggregates texture cues across multiple scales and orientations. In addition, this paper employs a Path Aggregation Network–FPN (PAN–FPN) neck together with Spatial Pyramid Pooling–Fast (SPPF) to efficiently fuse semantic and low‑level information across pyramid levels. Experiments show that MTF‑YOLO achieves [email protected] scores of 84.1% and 82.7% on NEU‑DET and GC10‑DET, respectively; it strikes a favorable balance among accuracy, efficiency, and deployability, and demonstrates robust performance for steel surface defect detection.
Institution: Shanghai Polytechnic University
Language: English
Subfield: Industrial and Manufacturing Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7164353840
