Aluminum Defect Detection Based on Weighted Feature Fusion Mechanism

In the detection of defects on the surface of aluminum materials, such as jet stream and scratch. These defects have the problem of background similarity of different sizes, which brings difficulties and challenges to the detection. Our method introduces a cross-layer linking network to further fuse shallow and deep features. Taking into account more location, semantics, and detailed information, the detection accuracy of the network for aluminum surface defects is improved. Moreover, A weighted feature fusion mechanism is introduced to solve the negative impact of features from different layers and enhance the feature extraction ability of the model. Experimental results show that the improved YOLOv5 network model has good defect detection performance. The mAP on the Tianchi dataset reached 78.4%, which is 2.3% higher than the original YOLOv5 network. The model in this paper can quickly and accurately detect aluminum surface defects while keeping the original detection speed basically unchanged.