数据内蕴结构驱动的深度学习高光谱图像分类  

Data Intrinsic Structure Driven Deep Learning Method for Hyperspectral Image Classification

作　　者：龚智强 胡卫东[2] 杜小勇[2] 钟平 姚雯 GONG Zhiqiang;HU Weidong;DU Xiaoyong;ZHONG Ping;YAO Wen(Defense Innovation Institute,Chinese Academy of Military Science,Beijing 100000,China;College of Electronic Science and Technology,National University of Defense Technology,Changsha 410073,China)

机构地区：[1]军事科学院国防科技创新研究院,北京100000 [2]国防科技大学电子科学学院,长沙410073

出　　处：《遥感信息》2023年第2期104-111,共8页Remote Sensing Information

基　　金：国家自然科学基金项目(62001502)。

摘　　要：针对深度学习性能受限于高光谱遥感图像有限样本的问题,提出一种描述类别样本先验分布模型的数据内蕴结构,并利用特定样本点与类别先验分布模型以及样本子集与类别先验分布模型间数据内蕴结构关系作为先验信息构造数据内蕴结构驱动的新型深度学习方法,提高了有限样本下深度学习高光谱图像分类性能。此外,文章介绍了新型学习方法的具体表现形式并展示了传统深度学习方法以及数据内蕴结构驱动新型深度学习方法在3个实际高光谱遥感图像上的相关分类性能。实验结果表明,基于数据内蕴结构驱动的新型深度学习方法能显著提高深度模型的分类性能。Faced with the limited labelled samples of hyperspectral images in deep learning process,this work proposes a novel intra-class relationship model which is called data intrinsic structure as class prior information,and further develops the data intrinsic structure driven deep learning method based on the class model using the point-to-set correlation and set-to-set correlation,which significantly improves the classification performance.Furthermore,this work introduces several special representations of the novel deep learning methods and the corresponding experiments over three real-world hyperspectral image datasets,and the experimental results demonstrate that the data intrinsic structure driven deep learning methods can significantly improve the classification performance of the deep model.

关 键 词：深度学习方法 数据内蕴结构 类别模型 FISHER判别准则 交叉熵 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]