基于DL1图和KNN图叠加图的高光谱图像半监督分类算法  

Semi-supervised classification algorithm of hyperspectral image based on DL1 graph and KNN superposition graph

作　　者：王君言 张春梅[1] 张云斌 刘瑶瑶 王晓峰[1] 李丹 张少敏[1] 白静[1] 

机构地区：[1]北方民族大学计算机科学与工程学院,银川750021 [2]中软国际有限公司欧美线speech组,北京100070

出　　处：《中国科学：信息科学》2017年第12期1662-1673,共12页Scientia Sinica(Informationis)

基　　金：国家自然科学基金(批准号:61461002);宁夏自然科学基金(批准号:NZ15105);北方民族大学校级科研项目(批准号:JSKY06);北方民族大学研究生创新项目(批准号:YCX1657)资助

摘　　要：基于少数已标记样本的高光谱图像分类是一个具有挑战的任务.本文将概率矩阵与L1图的权值矩阵叠加,形成了强鉴别力的DL1图.将空间的局部信息与光谱的全局信息通过KNN图和DL1图结合在一起,构建了空谱信息联合的图框架结构,使用该框架构建的图,能更精细地反映高光谱图像数据的图谱结构.利用图的标记传播达到半监督分类的目的,以此提高小样本高光谱图像自动分类的精度,实验表明,在标记样本比例为5%时,分类精度提升亦非常显著.The classification of hyperspectral images with a paucity of labeled samples is a challenging task. This paper describes the use of a superpose probability matrix and weight matrix of an L1 graph, thereby forming a strong discriminating DL1 graph. Combining the local information of the space with the global information of the spectrum through the superposition of a KNN graph and a DL1 graph, a graph-based framework is built that combines the spatial and spectral information. This framework of a DL1 KNN graph can reflect the more sophisticated structure of hyperspectral image data. Experimental results show that the improvement in classification accuracy is significant when the percentage of labeled samples is 5% through the use of the label propagation of the graph to achieve semi-supervised classification for improving the automatic classification accuracy of hyperspectral data with a small number of samples.

关 键 词：高光谱图像 半监督分类 稀疏图 KNN图 标记传播 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] TP751[自动化与计算机技术—控制科学与工程]