基于形态学属性剖面的高光谱影像集成分类

发布时间：2018-03-02 13:38

本文选题：形态学属性剖面　切入点：集成学习　出处：《遥感技术与应用》2016年04期 　论文类型：期刊论文

【摘要】：传统高光谱遥感影像逐像素分类方法未考虑像元之间的空间关联性且泛化性能较低。形态学属性剖面是表征影像空间结构的有效方法,同时集成学习可显著提升分类算法的泛化能力。为了在高光谱影像分类中充分利用影像的空间信息并提高分类的稳定性,提出一种基于形态学属性剖面高光谱遥感影像集成学习分类方法。首先,用主成分分析和最小噪声变换进行特征提取,并借助形态学属性剖面获取影像的多重空间特征;然后用极限学习和支持向量机的方法进行分类;最后将多个分类结果以多数投票的方式集成。区别于已有集成学习方法,综合考虑了不同特征提取和不同分类方法的联合集成,并将形态学属性剖面引入其中以充分利用影像的空间信息。采用AVIRIS和ROSIS两组高光谱数据检验该方法的分类性能,实验结果表明该方法可获得高精度和高稳定性的分类结果,总体精度分别达到83.41%和95.14%。
[Abstract]:Traditional hyperspectral remote sensing image pixel by pixel classification method does not take into account the spatial correlation among pixels and its generalization performance is low. Morphological attribute profile is an effective method to characterize the spatial structure of the image. In order to make full use of spatial information in hyperspectral image classification and improve the stability of classification, ensemble learning can significantly improve the generalization ability of classification algorithm. An integrated learning method based on morphological attribute profile hyperspectral remote sensing image is proposed. Firstly, feature extraction is performed by principal component analysis (PCA) and minimum noise transform (MNT), and multiple spatial features of the image are obtained by morphological attribute profile. Then we use the method of limit learning and support vector machine to classify. Finally, we integrate the classification results by majority vote. Different from the existing integrated learning methods, we consider the different feature extraction and the joint integration of different classification methods. The morphological attribute profile is introduced to make full use of the spatial information of the image. Two groups of hyperspectral data, AVIRIS and ROSIS, are used to test the classification performance of the method. The experimental results show that the proposed method can obtain high accuracy and high stability classification results. The overall accuracy reached 83.41% and 95.14 respectively.
【作者单位】：天津市地质调查研究院;卫星测绘技术与应用国家测绘地理信息局重点实验室南京大学;江苏省地理信息技术重点实验室南京大学;
【基金】：江苏省杰出青年基金项目(BK2012018) 国家重大科学仪器设备开发专项(012YQ050250)资助
【分类号】：TP751

【相似文献】