基于改进YOLOv7算法的连铸坯表面缺陷检测

Surface Defect Detection of Continuous Casting Billet Based on Improved YOLOv7

针对连铸坯表面缺陷检测存在的参数量大、小目标检测效果较差、人工标注数据集质量的不稳定性等问题,提出了一种基于改进YOLOv7算法的连铸坯表面缺陷检测模型。首先,引入MobileNetv3 block模块轻量化主干网络,降低计算量,减少参数与复杂度;其次,融合双层路由注意力机制,实现更灵活的计算分配和内容感知;最后,使用WIoU损失函数替换原损失函数,使模型获得更好的泛化能力。改进后的模型检测精度达到了89.3%,角部横裂的检测精度提升了6.6%;权重文件大小减少为原来的56.5%,检测速度达到了98 fp/s。结果表明,所提出的算法参数量小、检测速度快、精度高,能够满足连铸坯表面缺陷在线检测工作需求。

Aiming at the problems of large number of parameters,poor detection effect of small objects,and instability of the quality of artificial annotation datasets in the detection of surface defects of continuous casting billet,this paper proposes a surface defect detection model of continuous casting billet based on the improved YOLOv7 algorithm.Firstly,the MobileNetv3 block module is introduced to reduce the amount of calculation,parameters and complexity.Integrate the two-layer routing attention mechanism to achieve more flexible computing distribution and content awareness.Finally,the WIoU loss function is used to replace the original loss function to make the model obtain better generalization ability.The detection accuracy of the improved model reached 89.3%,and the detection accuracy of corner transverse cracking was increased by 6.6%.The weight file size has been reduced to 56.5%,and the detection speed has reached 98 fp/s.The results show that the proposed algorithm has small parameters,fast detection speed and high precision,which can meet the requirements of online detection of surface defects of continuous casting billet.