| 空谱协同多尺度顶点成分分析的高光谱影像端元提取EI北大核心CSCD |
| |
| 引用本文： | 孙伟伟,常明会,孟祥超,杨刚,任凯.空谱协同多尺度顶点成分分析的高光谱影像端元提取EI北大核心CSCD[J].测绘学报,2022,51(4):587-598. |
| |
| 作者姓名： | 孙伟伟 常明会 孟祥超 杨刚 任凯 |
| |
| 作者单位： | 1. 宁波大学地理与空间信息技术系, 浙江 宁波 315211;2. 宁波大学信息科学与工程学院, 浙江 宁波 315211 |
| |
| 基金项目： | 国家自然科学基金(42122009;41971296;41671342;41801256;41801252)；;浙江省自然科学基金(LR19D010001;LQ18D010001)； |
| |
| 摘 要： | 针对顶点成分分析方法无法实现复杂地表环境下的高光谱影像端元精确提取问题,提出了一种基于空谱协同多尺度顶点成分分析的端元提取方法,通过影像空谱特征融合和聚类分割,对不同分辨率空间尺度下的分割影像进行端元协同提取,并考虑噪声对影像端元提取精度的影响,提升端元提取的精度。首先,对影像进行预处理,采用低秩矩阵分解去除噪声。其次,对高光谱影像进行空谱多特征提取,利用多特征融合和K-means算法进行聚类分割,获取地物分布的空间异质性信息,提升后续端元提取的精度。然后,对高分辨率影像空间降采样,利用顶点成分分析方法对降采样后的低分辨率分割图像进行端元提取,并利用坐标映射寻找高分辨率影像中的相应端元,利用光谱角来判定是否为纯端元。最后,遍历上述方法至所有分割影像以获取最终的端元集合。使用模拟数据和真实的高分五号高光谱数据对提出的方法进行实验验证。实验结果表明,空谱协同多尺度顶点成分分析方法可提取高精度的纯净端元,且计算效率较高。
|
| 关 键 词： | 光谱端元 高光谱遥感 空谱协同 多尺度顶点成分分析 |
| 收稿时间： | 2021-10-26 |
| 修稿时间： | 2022-03-21 |
Spatial-spectral collaborative multi-scale vertex component analysis for hyperspectral image endmember extraction |
| |
| SUN Weiwei,CHANG Minghui,MENG Xiangchao,YANG Gang,REN Kai.Spatial-spectral collaborative multi-scale vertex component analysis for hyperspectral image endmember extraction[J].Acta Geodaetica et Cartographica Sinica,2022,51(4):587-598. |
| |
| Authors: | SUN Weiwei CHANG Minghui MENG Xiangchao YANG Gang REN Kai |
| |
| Institution: | 1. Department of Geography and Spatial Information Techniques, Ningbo University, Ningbo 315211, China;2. Faculty of Electrical Engineering and Computer Science, Ningbo University, Ningbo 315211, China |
| |
| Abstract: | Current endmember extraction methods cannot accurately extract the endmembers of complicated ground features, and therefore this paper proposed a spatial-spectral collaborative multi-scale vertex component analysis (VCA) method. Hyperspectral images are firstly jointly clustered and segmented based on multi-feature fusion using spectral features, texture features, and shape features, which makes full use of the spatial heterogeneity information of ground features. Then, multi-scale low-rank matrix decomposition is used to decompose the segmented images and reduce the influence of noise on endmember extraction. Meanwhile, VCA is used to extract endmembers from low-resolution images, coordinate mapping is implemented to search these endmembers of high-resolution images, and vertex component analysis is used to extract endmember from low resolution image. After that, coordinate mapping is used to ferret about the corresponding endmembers in the high-resolution image, and the spectral angle between them is calculated to help accurately decide the pure endmembers. Finally, the proposed method is traversed into all segmented images to obtain the final pure endmembers. The proposed method is verified experimentally by using simulated and real GF-5 hyperspectral data. Experimental results show that the CVCA method can extract high-precision pure endmembers and has high calculation efficiency. |
| |
| Keywords: | |
| 本文献已被 维普 万方数据 等数据库收录！ |
| 点击此处可从《测绘学报》浏览原始摘要信息 |
| 点击此处可从《测绘学报》下载免费的PDF全文 |
|
