特异小样本工业产品表面缺陷检测方法研究
- 分享到:
摘要：
基于机器视觉的工业产品表面缺陷检测设备和系统大量应用在工业制造领域，目前其难点在于工业检测数据的采集，由于训练样本缺失导致深度学习网络模型无法有效训练。为解决上述问题，首先，提出一 种基于不规则掩码的伤痕样本生成算法，改善了钢板表面缺陷检测任务中特异小样本数据集正负样本不均衡的情况；然后，在YOLOv8主干网络引入MHSA多头自注意力，提高对钢板表面缺陷的关注度；最后，使用SIoU替换原损失函数，增强网络模型的定位能力 ，提高检测的准确性 。基于热轧钢板表面缺陷检测问题的实验结果表明，该方法能够有效解决特异小样本工业探伤的具体问题。
Machine vision-based industrial product surface defeet detection equipment and systems are widely used in the industrial manufacturing field, Currently, the main diffeulty lies in the collection of industrial inspection data and the inabilitvy of deep learning network models to be efectively trained due to the lack of training samples. To solve these problems, firstly,this paper proposes a scatr sample generation algorithm based on irregular masks to improve the imbalance of positive and nega-ive samples in the special small sample dataset for steel plate surface deleet detection task: then, the MHSA multi-head self-attention is introduced into the YOl.Ov8 backbone network to enhanee the attention to steel plate surface defects; finally, the SloU loss function is used to replace the original loss funetion to enhanee the network model's localization ability and improve detection accuracy. The experimental results on the hot rolled steel plate surlace defeet detection problem based on this method show that can be effectively solved.
作者：
郑李明，许天赐，高浩然，李庆华，胡晨光，窦智
Zheng Liming， Xu Tianci，Gao Haoran，LiQingHua，Hu Chenguang，Dou Zhi
机构地区：
金陵科技学院机电工程学院；betway官方app 计算机与信息工程学院；莱芜钢铁集团银山型钢铁有限公司
引用本文：
郑李明，许天赐，高浩然等。特异小样本工业产品表面缺陷检测方法研究[J].betway官方app 学报(自然科学版) ,2024, 52(6):88-96. (Zheng Liming, Xu Tianci, Gao Haran, et al. Research on the detection method for specialsmall-sample defects in industrial products[J] . Journal of Henan Normal University(Natural Science Edition) ,2024,52(6) :88-96. DOI:10. 16366/j. cnki.1000-2367. 2023. 06. 26. 0001. )
基金：
国家自然科学基金；山东钢铁股份有限公司科技创新项目
关键词：
深度学习；目标检测；YOLOv8；注意力机制；数据增强；特异小样
deep learning;object detection; YOLOv8; attention mechanism;data enhancement;special small samples
分类号：
TP391. 41
