高技术通讯2025，Vol.35Issue(11)：1153-1162,10.DOI:10.3772/j.issn.1002-0470.2025.11.001
基于昇腾处理器的小样本声呐图像分割方法
Few-shot sonar image segmentation method based on Ascend
摘要
Abstract
Sonar equipment is currently commonly used underwater detection equipment.However,sonar image data in-volves sensitive information,is difficult to obtain,and has limited samples,which leads to the issue that traditional neural networks require large-scale dataset for training but perform inadequately.To this end,this paper proposes a sonar image few shot segmentation model called Meta-attention-UNet(MAUNet)based on meta-learning that in-tegrates local parameter freezing mechanism and attention mechanism.This model enhances the model's ability to extract key information by introducing an attention mechanism;it introduces a meta-learning module to quickly adapt to segmentation tasks,and achieves adaptability to target domain tasks through a local parameter freezing mechanism in the meta-learning module.Experimental results show that on the public sonar dataset,mean intersec-tion over union(mIoU)reaches 67.2%,which is better than the existing mainstream few shot segmentation mod-el.After the algorithm was transplanted to Ascend,the lake test in Qiandao Lake in Hangzhou achieved quasi-real-time of 6.2 Hz performance.Comparative experiments were conducted on the performance of the algorithm on dif-ferent hardware platforms,forming a complete data collection and real-time segmentation system.关键词
昇腾处理器/元学习/声呐图像/图像分割/小样本学习Key words
Ascend/meta-learning/sonar images/image segmentation/few-shot learning引用本文复制引用
ZHAO Dongdong,WANG Yuhang,CHEN Peng,LI Yiran,GUO Xinxin..基于昇腾处理器的小样本声呐图像分割方法[J].高技术通讯,2025,35(11):1153-1162,10.基金项目
国家自然科学基金(62371421,62001418,62005245,62076220,62036009),国家自然科学基金联合基金(U1909203),浙江省领军型创新创业团队(2021R01002)和浙江省自然科学基金(LD24F020005)资助项目. （62371421,62001418,62005245,62076220,62036009）
