| 彭艳斌,郑志军,潘志刚,李晓勇,金诚.基于流形波段选择的高光谱图像分类[J].光电子激光,2016,27(6):670~674 |
| 基于流形波段选择的高光谱图像分类 |
| Hyperspectral image classification based on manifold band selection |
| 投稿时间：2015-10-22 |
| DOI： |
| 中文关键词: 高光谱图像 分类 波段选择 流形学习 无监督 |
| 英文关键词:hyperspectral image classification band selection manifold learning unsuperv ised |
| 基金项目:国家自然科学基金(61379074)和浙江省自然科学基金(LQ13F020015)资助项目 (1.浙江科技学院 信息学院,浙江 杭州 310023； 2.浙江大学 计算机学院,浙江 杭州 310027) |
|
| 摘要点击次数: 1056 |
| 全文下载次数: 0 |
| 中文摘要: |
| 为解决高光谱图像中高维数据和有标记训练样本不 足的矛盾导致“维度灾难”问题,提出一种无监督的基于流形学习的波段选择(MLBS)方法。 首先通过流形学习方法,得到原始数据的流形嵌入映射；然后通过LASSO优化过程,运用顺 向坐标下降算法,得到原始波段对每个流形结构 维度的贡献度；最后统计每个波段的贡献度,选取贡献度大的波段形成波段子集。用 真实的AVIRIS高光谱图像对算法进行仿真实验的结果表明,本文方法在小样本下的高光谱 地物分类识别问题上具有良好的效果。 |
| 英文摘要: |
| For solving the “curse of dimensionality” problem.Caused by the contradiction between high dimensional da ta and insufficient labeled training samples in huperspectral image, this paper presentes an unsupervised manifold learning based band selection (MLBS) method.Firstly,the embedding projection of each pixel is calculated through manifold learning algorithm;secondly,the contribution degree of each d imensional of manifold structure is calculated through a lasso optimization process and coordi nate-wise descent algorithm;at last,the contribution degree of each band is calculated,while the b ands with large contribution degree form band subset.The proposed band selection approach is ex perimentally evaluated using real AVIRIS hyperspectral data set.Experimental results in AVIR IS data demonstrate that the proposed algorithm can yield good performance in hyperspect ral land-cover classification with small samples. |
| 查看全文 下载PDF阅读器 |
| 关闭 |
|
|
|
|
|