Strip Steel Surface Quality Detection System Based on YOLOv5
Xiaoli Zhang,Li Wu,Yali Guo,Shizhong Guo,Sisi He,Na Hao
DOI: https://doi.org/10.1080/10584587.2024.2325876
2024-10-29
Integrated Ferroelectrics
Abstract:For identifying and classifying surface defects of strip steel in industrial production and improving the efficiency of defect detection, a detection and recognition system for surface quality defects of strip steel in industrial production environments is designed in this article. Firstly, industrial raw material strip steel was used as the research object, and an experimental operating environment was established to label the dataset. Then, entered into the image, image scaled by YOLOv5s, and then the image is scaled into S × S using convolutional neural networks. Feature maps are extracted from each small grid. Secondly, the extracted feature maps were pooled and subjected to a series of sharding, stitching, and multi-scale feature fusion. Finally, under the GPU acceleration of the PyTorch learning framework, the processing speed was improved by identifying categories through a fully connected layer network, all indicators have been detected, as well as information on the category and confidence level of defects identified. The experimental results show that the system can identify four types of defects on the surface of the strip steel and achieve an accuracy of 90% in identifying defect types.
engineering, electrical & electronic,physics, condensed matter, applied