摘要
局部线性嵌入(LLE)等流形学习算法中需要通过欧氏距离来度量数据点之间的近邻关系,但欧氏距离只表示两点间的直线距离,在高维空间中不一定能真实反映出图像数据点之间的空间分布情况。针对此问题,本文提出了融合数据间夹角和欧氏距离度量LLE近邻和分类的方法。该方法通过融合图像数据间的夹角和欧氏距离来度量图像数据点之间的近邻关系,寻找k个近邻点,实现更有效的局部重构,提取鉴别特征,然后用融合了数据间夹角的最近邻分类器对数据进行分类。在KSC和Indian Pine高光谱遥感影像数据集上的实验结果表明:在总体分类精度上,本文算法比LLE提升了1.54%～6.91%。
Locally Linear Embedding （LLE） manifold learning algorithm needs to calculate the neighbor points of each image based on Euclidean distance.But this method represents only the straight line distance between two points and does not necessarily reflect the actual distribution relationship of the image data sets in the high dimensional space.In order to solve this problem,an approach based on the fusion data between angle and Euclidean distance of images is proposed to calculate the neighbor points of LLE and to classify data.This method uses the fusion data between angle and Euclidean distance of images to measure the adjacent relations of image data points and find k neighbor points,which can achieve more effective local reconstruction to extract the distinguishing features.Finally,the nearest neighbor classifier with angle of images is used to classify the image data.Experiments on KSC and Indian Pine database show that the overall accuracy of this proposed algorithm is improved by 1.54%～6.91% compared with LLE algorithm.
出处
《光电工程》
CAS
CSCD
北大核心
2013年第6期97-105,共9页
Opto-Electronic Engineering
基金
国家自然科学基金项目(61101168)
中国博士后科学基金资助项目(2012M511906)
重庆市博士后科研项目特别资助(XM2012001)
关键词
高光谱影像
流形学习
局部线性嵌入(LLE)
近邻点
hyperspectral images
manifold learning
locally linear embedding （LLE）
neighbor points
