Steel surface defect detection method based on improved YOLOv9
Abstract With the development of industrial automation and intelligent manufacturing, steel surface defect detection has become a critical step in ensuring product quality and production efficiency. However, the diverse types and significant size variations of defects on steel surfaces pose great ch...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2025-07-01
|
| Series: | Scientific Reports |
| Subjects: | |
| Online Access: | https://doi.org/10.1038/s41598-025-10647-1 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract With the development of industrial automation and intelligent manufacturing, steel surface defect detection has become a critical step in ensuring product quality and production efficiency. However, the diverse types and significant size variations of defects on steel surfaces pose great challenges. Among these defects, small-sized defects are characterized by their subtle appearance on the surface, making them difficult to distinguish from the background. This often results in high false detection and missed detection rates during the inspection process. To address this issue, this paper proposes an improved steel surface defect detection algorithm based on YOLOv9. First, introducing Depthwise Separable Convolution (DSConv) can effectively reduce the computational complexity of the model, thereby enhancing its operational efficiency. Second, the C3 module is incorporated to effectively fuse feature maps from different levels, enhancing the model’s ability to detect multi-scale targets. To improve the recognition accuracy of small targets, a bidirectional feature pyramid network (BiFPN) is integrated, enabling the model to capture small target features more precisely. Additionally, the DySample upsampling operator is employed to address the issue of detail loss in traditional upsampling methods, enhancing the model’s sensitivity and localization accuracy for small targets. Experimental results demonstrate that the improved model achieves a mean average precision (mAP) of 78.2% and an accuracy of 82.5%, representing increases of 1.8% and 7.4%, respectively, compared to the baseline model. Moreover, the number of model parameters is reduced by 8.9%. The findings of this study hold significant practical value for improving the quality and efficiency of industrial products in the field of steel surface defect detection. |
|---|---|
| ISSN: | 2045-2322 |