Detection of Defects on Aluminum Profile Surface Based on Improved YOLO

Aluminum material is widely used in production and life, and it is a material with high requirements on surface treatment. Detecting its surface defects is the key to improving its utilization efficiency. To improve the accuracy and reliability of surface defect detection of aluminum material, this paper uses YOLO X with flexibility, lightness, and accuracy to build a training network, and proposes a defect detection model based on YOLO X, which replaces the original CSP-DarkNet with CSP-ResNeXt and integrates the Attention Mechanism. The network's ability to classify defects is strengthened, so the detection accuracy of multiple defects is improved. The Transfer Learning method is used in training, which shortens the training cycle and improves the detection performance of the short-term training network. The experimental results show that the Average Precision (AP) and mean Average Precision (mAP) of the model have been significantly improved, and the detection speed Frame Per Second (FPS) has not decreased significantly.