Surface defect detection of aluminum profiles based on deep learning

With the rapid advancement of industrial automation, detecting surface defects in aluminum profiles has become crucial for enhancing product quality and production efficiency. This study enhances the YOLOv8 object detection algorithm by integrating SKconv convolution to improve the CNN model’s feature representation across various scales, thereby boosting its perceptual capabilities and recognition performance.SKconv dynamically selects convolution kernel sizes to better adapt to targets of different scales, enhancing both generalization capability and robustness. A comprehensive system for detecting surface defects in aluminum profiles has been developed. By training convolutional neural networks on extensive datasets of defect samples, this system enables real-time detection and classification of various surface defects.Compared to traditional methods, our enhanced model significantly improves accuracy and efficiency in defect identification while reducing human intervention. Moreover, it holds great potential for application beyond the aluminum profile industry, extending to other sectors that require surface defect detection in metallic or non-metallic materials.