Deep Learning-based Detection of Bubble Defects on the Surface of Precast Beams
An algorithm for detecting bubble defects on the surface of precast beams based on YOLOv5s was proposed.Based on the original model,the algorithm introduced the CBAM attention module to enhance the relevance of information between channels and the attention of interest features.In the neck network,BiFPN weighted bidirectional pyramid structure was used to improve the network fea-ture fusion module and realize fast multi-scale feature fusion.After detecting bubble defects,two e-valuation indexes based on area and diameter were proposed to classify bubbles.The results show that the improved model has stronger feature extraction ability,the average detection accuracy(mAP)is 95.8％,which is 2.3％higher than the original model,the accuracy is 6.5％higher,and the recall is 3.5％higher.In the task of bubble defect detection,the missed detection and false detection are effec-tively reduced,and the detection performance is better.
万方数据
维普
