半监督条件随机场的高光谱遥感图像分类

发布时间：2018-03-18 03:01

本文选题：高光谱　切入点：遥感　出处：《遥感学报》2017年04期 　论文类型：期刊论文

【摘要】：善于捕捉空间信息的条件随机场模型虽然已被应用于高光谱遥感图像分类,但条件随机场的性能受到了标注训练样本数量的制约。为解决上述问题,本文提出了一种半监督条件随机场模型用于高光谱遥感图像分类。在该模型中,首先,利用空间-光谱拉普拉斯支持向量机定义关联势函数,以利用未标注样本中包含的信息获取样本类别概率;然后,在交互势函数中嵌入未标注的空间邻域样本,以充分利用空间信息实现对样本类别概率的修正;最后,采用分布式学习策略和平均场完成半监督条件随机场的训练和推断。本文在两个公开的高光谱数据集(Indian Pines数据集,Pavia University数据集)上进行了实验。实验结果表明Kappa系数提升3.94%。
[Abstract]:The conditional random field model, which is good at capturing spatial information, has been applied to the classification of hyperspectral remote sensing images, but the performance of conditional random field is restricted by the number of training samples. In this paper, a semi-supervised conditional random field model is proposed for hyperspectral remote sensing image classification. In this model, the spatial spectral Laplace support vector machine is used to define the correlation potential function. The probability of sample category is obtained by using the information contained in the unlabeled sample. Then, the unlabeled spatial neighborhood sample is embedded in the interactive potential function to make full use of the spatial information to correct the probability of the sample category. Finally, Using distributed learning strategy and mean field to complete the training and inference of semi-supervised conditional random fields, the experiments were carried out on two open hyperspectral data sets, Indian Pines datasets and Pavia University datasets. The experimental results show that the Kappa coefficient is increased by 3.94.
【作者单位】：北京航空航天大学宇航学院;数字媒体北京市重点实验室;
【基金】：国家自然科学基金(编号:61501009,61371134和61071137)~~
【分类号】：TP751

【参考文献】