多级多尺神经网络自搜索的焊缝缺陷语义分割  

Semantic Segmentation Method of Weld Defects Using Multilevel and Multi-Scale Neural Network Self-Search

作　　者：张睿 李吉 Zhang Rui;Li Ji(College of Computer Science and Technology,Taiyuan University of Science and Technology,Taiyuan 030024;Shanxi Electromechanical Design and Research Institute Co.,Ltd,Taiyuan 030009)

机构地区：[1]太原科技大学计算机科学与技术学院,太原030024 [2]山西省机电设计研究院有限公司,太原030009

出　　处：《计算机辅助设计与图形学学报》2024年第11期1750-1760,共11页Journal of Computer-Aided Design & Computer Graphics

基　　金：山西省基础研究计划(20210302123216);山西省机械产品质量司法鉴定中心企业委托项目(2021168);山西省研究生教育改革研究课题(2021YJJG244);太原科技大学研究生联合培养示范基地项目(JD2022004),太原科技大学研究生教育创新项目(SY2022064)。

摘　　要：为进一步提高焊缝缺陷X-ray底片低质影像语义分割精度,降低人工设计网络主观影响及耗时等问题,提出一种多级多尺神经网络自搜索的焊缝缺陷语义分割方法.通过对多尺度轻量化候选操作、通道注意力机制和多层级动态神经网络架构的设计,从不同维度提升网络对低质影像缺陷特征提取的表达能力;同时对网络训练早期与最终识别性能之间的潜在关联探索,提出使用固定采样逐步确定最优候选操作的渐进式快速神经架构搜索方法.在架构搜索阶段使用自行采集、标注的483幅X-ray焊缝缺陷图像,经过随机裁剪、旋转、平移等数据增强操作进行架构寻优,最终以较低的搜索成本自动构建出焊缝缺陷语义分割网络.实验表明,所提方法对X-ray焊缝缺陷进行语义分割最终mIoU指数达到了49.23%,高于人工设计网络的45.41%和直接使用模型迁移的28.86%,网络自搜索速度和分割效果提升明显.To further improve the accuracy of semantic segmentation of low-quality X-ray images of weld de-fects and reduce the subjective impact and time-consuming of artificial design networks,a semantic segmenta-tion method of weld defects using multilevel and multi-scale neural network self-search is proposed.Through the design of multi-scale lightweight candidate operations,channel attention mechanism and multi-level dy-namic network,the expression ability of the network to extract defect features of low-quality images is im-proved from different dimensions;at the same time,through the exploration of the correlation between the recognition performance of the early and final stages of network training,a gradually fast neural architecture search method using fixed sampling to gradually determine the optimal candidate operation is proposed.In the architecture search phase,483 X-ray weld defect images collected and annotated by oneself were used for ar-chitecture optimization through random cropping,rotation,translation,and other data augmentation operations.Finally,a semantic segmentation network for weld defects was automatically constructed at a lower search cost.Experiments show that using the above ideas for semantic segmentation of X-ray weld defects,the final mIoU index reaches 49.23%,which is higher than 45.41%of the artificial design network and 28.86%of the direct use of model transfer.The self-search speed and segmentation effect of the network are significantly im-proved.

关 键 词：语义分割 多分辨率多尺度特征融合 神经架构搜索 焊缝缺陷 无损检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]