Intelligent Scratch Detection Based on YOLOv5-CBAM Model
As a core product of modern steel industry,strip steel plays a crucial role in surface quality control.Scratch detection occupies a vital position in surface quality control and is extensively applied in fields such as automobile manufacturing,metal processing,and electronic product production.However,due to the diverse shapes and sizes of scratches and the impact of factors such as illumination,background,and noise,the detection task is extremely challenging.In recent years,target detection algorithms based on deep learning(such as Faster R-CNN,SSD,and YOLO)have performed outstandingly in detection tasks.Through automatic learning,feature detecting,and precise target localization,these algorithms enable accurate detection even with complex backgrounds.Based on the YOLOv5 model,the Spatial Pyramid Pooling-Fast(SPPF)module is used instead of the Spatial Pyramid Pooling(SPP)module,and an attention mechanism is incorporated to improve the existing target detection algorithm to enhance the accuracy and robustness of scratch detection.Based on the YOLOv5 model,a YOLOv5-CBAM model is constructed by integrating the YOLOv5 with the Convolutional Block Attention Module(CBAM).The CBAM makes the model focus more accurately on the scratch area by paying atten-tion to the information in the channel and spatial dimensions,improving the detection effect.Experimental results show that,compared to YOLOv5,the YOLOv5-CBAM model achieves great performance in precision,recall,and mAP@0.5 across various Intersection over Union(IoU)thresholds,with increasement of 5.6％,9.1％and 5.9％respectively.Meanwhile,with the continuous progress of spatial data intelligence technology,it is expected to provide more innovative ideas and solutions for scratch detection in the future.
scratch detectionYOLOv5CBAMmodel construction and training
万方数据
维普
