利用偏最小二乘方法修复高光谱影像等距映射降维中遗失点的坐标  

Manifold Coordinate Repairing of Lost Points with PLS for Isomap Nonlinear Dimensionality Reduction of Hyperspectral Image

作　　者：孙伟伟[1] 刘春[1,2] 施蓓琦[3] 李巍岳[1] 

机构地区：[1]同济大学测量与国土信息工程系,上海市四平路1239号200092 [2]现代工程测量国家测绘地理信息局重点实验室,上海市四平路1239号200092 [3]上海师范大学城市信息研究中心,上海市桂林路100号200234

出　　处：《武汉大学学报（信息科学版）》2012年第5期550-554,共5页Geomatics and Information Science of Wuhan University

基　　金：上海市教育委员会科研创新基金资助项目(10ZZ25;11YZ290);地理空间信息工程国家测绘局重点实验室经费资助项目(200914);现代工程测量国家测绘局重点实验室开放基金资助项目(TJES1010)

摘　　要：将Isomap流形学习方法应用于高光谱影像非线性降维时,在构建最短路径过程中,其边界点往往被忽略而没有低维流形坐标。对此,引入偏最小二乘方法来模拟修复遗失点的流形坐标,并从两个方面进行了综合评价。实验结果表明,模拟流形坐标与实际坐标吻合很好。As a manifold learning method,Isomap has been widely used for making nonlinearly reduction for hyperspectral image.However,during the construction process of the shortest path graph,the boundary points,which are not noise points,have always been omitted for the consideration of the stability of the graph.Therefore,the PLS method is introduced to repair and simulate the manifold coordinates of the lost points in the shortest path graph.And the simulated manifold coordinates have been evaluated from two different aspects to verify our method.The results show that the simulated manifold coordinates agree well with the real one.It will be quite useful for further classification or visualization with low dimensional manifold image.

关 键 词：等距映射 偏最小二乘 流形坐标修复 最短路径 

分 类 号：P237.3[天文地球—摄影测量与遥感]