Deep transformer and few‐shot learning for hyperspectral image classification
作者机构：College of Information Science and TechnologyBeijing University of Chemical TechnologyBeijingChina Key Laboratory of Computational Optical Imaging TechnologyAerospace Information Research InstituteChinese Academy of SciencesBeijingChina China Greatwall Technology Group Co.LtdShenzhenChina Department of Computer ScienceNational University of Computer and Emerging SciencesChiniotPakistan
出 版 物：《CAAI Transactions on Intelligence Technology》 (智能技术学报（英文）)
年 卷 期：2023年第8卷第4期
页 面：1323-1336页
核心收录：
学科分类：0502[文学-外国语言文学] 050201[文学-英语语言文学] 05[文学]
主 题：deep learning feature extraction hyperspectral image classification
摘 要：Recently,deep learning has achieved considerable results in the hyperspectral image(HSI)***,most available deep networks require ample and authentic samples to better train the models,which is expensive and inefficient in practical *** few‐shot learning(FSL)methods generally ignore the potential relationships between non‐local spatial samples that would better represent the underlying features of *** solve the above issues,a novel deep transformer and few‐shot learning(DTFSL)classification framework is proposed,attempting to realize fine‐grained classification of HSI with only a few‐shot ***,the spatial attention and spectral query modules are introduced to overcome the constraint of the convolution kernel and consider the information between long‐distance location(non‐local)samples to reduce the uncertainty of ***,the network is trained with episodes and task‐based learning strategies to learn a metric space,which can continuously enhance its modelling ***,the developed approach combines the advantages of domain adaptation to reduce the variation in inter‐domain distribution and realize distribution *** three publicly available HSI data,extensive experiments have indicated that the proposed DT‐FSL yields better results concerning state‐of‐the‐art algorithms.
维普期刊数据库