高光谱遥感影像多特征优化模型与协同表示分类

There are many open issues in hyperspectral remote sensing image classification, such as small sample, feature selection, classification accuracy reliability etc. As the new machine learning algorithms, swarm intelligence optimization and collaborative representation provide new tools for hyperspectral imagery classification. In this project, the efficient multiple features optimization model and high-reliable classification approaches for hyperspectral remotely sensed imagery based on swarm intelligence optimization and collaborative representation technology are proposed. The feature quality evaluation algorithm is designed based on feature extraction methods and specific applications for hyperspectral remote sensing imagery; the firefly algorithm-supported multiple features optimization model with improved optimization objective functions and levy flights for hyperspectral imagery will be designed. For high-reliable classification of hyperspectral images in the case of small number of samples, the global / local sample selection mehtods based on collaborative representation, semisupervised learning and active learing are proposed with respect to the distribution characteristics of the ground truth samples in high dimensional space. With the support of multi-features and sample selection, the dynamic multiple classifier ensenmble system based on classifier diversity and fusion rules can be built. The experiment results of this project are expected to provide a new method and technique for hyperspectral image analysis, and will expand the development of hyperspectral remote sensing technology, and promote the applications of hyperspectral remote sensing.

高光谱遥感影像分类目前在特征选择、小样本、分类精度可靠性等方面仍存在许多需要解决的问题；群体智能优化和协同表示作为新型机器学习算法，可为高光谱影像的分类研究提供新的工具。本项目拟在群体智能优化和协同表示理论的基础上，研究高光谱影像多特征优化模型与协同表示高可信分类方法。针对高光谱特征提取和应用目标特点，基于信息测度研究高光谱影像特征质量的评价方法，利用levy飞行和随机惯性权重改进萤火虫优化算法，研究高光谱影像的多特征优化模型；分析小样本的分布特征，基于半监督学习提升小样本的数量，基于主动学习提升小样本的质量，研究全局/局域协同表示的样本优选方法；通过构建多分类器集合，基于分类器多样性和融合规则研究协同表示引导的多分类器动态集成方法；实现多特征、小样本情况下高光谱影像的高可信分类。本项目成果有望拓展高光谱遥感影像分类的理论方法，推动高光谱遥感理论与技术的发展，促进高光谱遥感的深入应用。

高光谱遥感影像分类目前在特征选择、小样本、分类精度可靠性等方面仍存在许多需要解决的问题，为解决以上问题，课题组在群体智能优化和协同表示理论的基础上，研究了高光谱影像多特征优化模型与协同表示高可信分类方法。针对高光谱特征提取和应用目标特点，基于信息测度研究了高光谱遥感特征质量的评价方法，利用levy飞行和随机惯性权重改进萤火虫优化算法，提出了高光谱影像的多特征优化模型；基于小样本的分布特征，提出了全局/局域协同表示的样本优选方法；基于协同表示分类器构建多分类器集合，提出了协同表示引导的多分类器动态集成方法；实现多特征、小样本情况下高光谱影像的高可信分类。本课题提出的方法具有较好的性能，进一步拓展了高光谱遥感影像分类的理论方法，提升了高光谱遥感分类精度的可靠性，为后续高光谱遥感的深入应用奠定了坚实理论基础。围绕项目研究，共发表学术论文15篇，其中在IEEE Transactions on Geoscience and Remote Sensing IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing、IEEE Geoscience and Remote Sensing Letters等国际权威期刊发表SCI论文8篇，在《武汉大学学报信息科学版》发表研究论文1篇，在国际会议发表论文3篇，协助培养博士生2人，独立培养硕士生9人，获得国家发明专利4项，申请国家发明专利2项。