文献类型：期刊文献

中文题名：稀疏表示模型及高光谱遥感应用研究

英文题名：Research of Sparse Models and Applications in Hyperspectral Images

作者：张敬尊[1];张睿哲[1];徐光美[1];王金华[1];何宁[1]

第一作者：张敬尊

机构：[1]北京联合大学智慧城市学院,北京100101

第一机构：北京联合大学智慧城市学院

年份：2020

卷号：30

期号：10

起止页码：173-178

中文期刊名：计算机技术与发展

外文期刊名：Computer Technology and Development

收录：CSTPCD

基金：国家自然科学基金(61872042);北京市教委科技计划一般项目(KM201811417004);北京联合大学2019科研项目(ZK50201903)。

语种：中文

中文关键词：稀疏表示;稀疏编码;字典学习;高光谱;降维;分类

外文关键词：sparse representation;sparse coding;dictionary learning;hyperspectral;dimension reduction;classification

摘要：稀疏表示是一种新型的数据挖掘技术,与传统算法相比,稀疏表示类算法更善于发现隐藏在数据背后的知识,具有优秀的特征发现和保持能力,近年来成为多领域的研究热点。然而,各领域对此技术的表征和描述不尽相同,不利于遥感高光谱图像处理领域的扩展,应用潜力有待深挖。该文在对生物视觉、统计学以及机器学习等领域中稀疏表示的理论基础、研究进展进行总结的基础上,提出了遥感适用的稀疏表示框架,就稀疏表示的模型进行了系统而详尽的描述,重点介绍了稀疏编码及字典学习两个关键问题。基于稀疏表示遥感应用适用性及应用潜力分析的需求,梳理了稀疏表示模型遥感领域的应用,重点分析并统计了高光谱各分支的应用热点与难点。最后,对稀疏表示框架的优势以及高光谱遥感图像处理应用面临的问题进行了总结。
As a new technology of data mining,sparse representation is better at discovering the knowledge hidden behind the data and has excellent feature discovery and retention ability compared with traditional algorithms.In recent years,it has become a research hotspot in many fields.However,the characterization and description of this technique vary from field to field,which is not conducive to the expansion of remote sensing hyperspectral image processing field,and the application potential remains to be deeply explored.On the basis of summarizing the theoretical basis,research progress and application of sparse representation from three fields of biological vision,statistics and machine learning,we comb out the remote sensing oriented framework,systematically and detailedly describe the model and terminology system of sparse representation,and focus on two key issues of sparse coding and dictionary learning.Based on the requirements of application applicability and application potential analysis of sparse representation remote sensing,we review the application of sparse representation model in the field of remote sensing,and focus on the analysis and statistics of application hotspots and difficulties of each branch of hyperspectral.Finally,the advantages of sparse representation framework and the problems in remote sensing image processing are summarized.