| 基于邻域距离ISOMAP算法的高光谱遥感降维算法 |
| |
| 引用本文： | 周颂洋,谭琨,吴立新.基于邻域距离ISOMAP算法的高光谱遥感降维算法[J].遥感技术与应用,2014,29(4):695-700. |
| |
| 作者姓名： | 周颂洋 谭琨 吴立新 |
| |
| 作者单位： | (中国矿业大学江苏省资源环境信息工程重点实验室,江苏 徐州221116) |
| |
| 基金项目： | 国家自然基金项目(41101423);中央高校基本科研业务费专项资金(2010QNA18);中国博士后科学基金资助项目(2011M500128);江苏高校优势学科建设工程资助项目(SZBF2011-6-B35) |
| |
| 摘 要： | 提出一种以邻域距离改进ISOMAP的算法(Neighborhood Distance ISOMAP,ND\|ISOMAP),该方法采用邻域距离逐步逼近流形距离来表达高维数据的流形结构。同时针对ISOMAP算法的计算复杂度高、运算时间长的特点,提出了一种基于矩阵分块和自动调图的ISOMAP算法(Block\|matrix and Auto\|color ISOMAP,BA\|ISOMAP)以提高运算速率。通过对高光谱遥感影像进行分类比较算法优劣性,基于邻域距离的ISOMAP算法较原始的ISOMAP算法降维效果有了较大的提升,最高分类精度达到97.36%,而原始的ISOMAP算法仅能达到75.01%的分类精度,而基于矩阵分块与自动调图ISOMAP与邻域距离相结合降维后精度达到89.61%,但是其计算速率得到了较大提升,为原始ISOMAP算法的近40倍。
|
| 关 键 词： | 高光谱 遥感分类 邻域距离 ISOMAP 支持向量机 |
Hyperspectral Image Classification based on ISOMAP Algorithm using Neighborhood Distance |
| |
| Zhou Songyang,Tan Kun,Wu Lixin.Hyperspectral Image Classification based on ISOMAP Algorithm using Neighborhood Distance[J].Remote Sensing Technology and Application,2014,29(4):695-700. |
| |
| Authors: | Zhou Songyang Tan Kun Wu Lixin |
| |
| Affiliation: | (Jiangsu Key Laboratory of Resources and Environment Information Engineering,China University;
of Mining and Technology,Xuzhou 221116,China) |
| |
| Abstract: | In this paper,we proposed an improved ISOMAP algorithm using neighborhood distance(ND\|ISOMAP).Neighborhood distance is used to successively approximate manifold distance indicating the manifold structure of high-dimensional data.In order to improve the computer rate of traditional ISOMAP,block\|matrix and auto\|color methods are applied in the ISOMAP(BA-ISOMAP).The dimension reduction of hyperspectral data is classified by Support Vector Machine(SVM),and the results show that the ND-ISOMAP performs the best overall accuracy (97.31% )than the traditional ISOMAP (75.01%).The combination method of ND\|ISOMAP and BA\|ISOMAP is with 89.61% overall classification accuracy,while its operation rate is almost 40 times than the traditional ISOMAP.
|
| |
| Keywords: | Hyperspectral Remote sensing classification Neighborhood distance ISOMAP Support Vector Machine |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《遥感技术与应用》浏览原始摘要信息 |
|
点击此处可从《遥感技术与应用》下载全文 |
