摘要
为了更准确地描述高光谱数据的内在几何流形结构,提高高光谱数据的分类准确度,提出了一种基于双图结构的标签传递算法。首先,提出了一种基于属类概率选择图近邻结构的方法;然后,利用高斯核函数计算近邻点的边权值,构造属类概率图;最后,将这种属类概率图与K-NN图线性组合并嵌入标签传递算法框架中,得到一种基于双图结构的标签传递算法。将算法应用到两种高光谱数据分类实验中,实验结果表明于双图结构的标签传递算法能有效提高高光谱数据的分类准确度。
In order to have a better description about intrinsic geometric manifold structure of hyperspectral data, and improve the classification accuracy, a novel method named couple graph based on label propagation method is proposed. First of all, the class-probability between data is proposed to select the graph adjacency. Then, the weights of class-probability graph are calculated by using Gauss kernel function. Finally, the couple graph based on label propagation algorithm is proposed in this paper, which combines the class-probability graph with the K-NN graph into the framework of label propagation algorithm. In the experiments, the couple graph based label propagation algorithm is applied to classify two kinds of hyperspectral data, and the experimental results show that the novel algorithm can effectively improve the classification accuracy of hyperspectral data.
作者
王小攀
胡艳
WANG Xiaopan;HU Yan(Geographic Information Center of Chongqing,Chongqing 401121)
出处
《计算机与数字工程》
2018年第10期2117-2122,共6页
Computer & Digital Engineering
关键词
图
标签传递
半监督分类
高光谱遥感
graph
label propagation
senti-supervised classification
hyperspectral remote sensing
