基于成对约束半监督降维的高光谱遥感影像特征提取

半监督降维(Semi|Supervised Dimensionality Reduction,SSDR)框架下,基于成对约束提出一种半监督降维算法SCSSDR。利用成对样本进行构图,在保持局部结构的同时顾及数据的全局结构。通过最优化目标函数,使得同类样本更加紧凑,异类样本更加离散。采用UCI数据集对算法进行定量分析,发现该方法优于PCA及传统流形学习算法,进一步的UCI数据集和高光谱数据集分类实验表明:该方法适合于进行分类目的特征提取。

Feature Extraction from Hyperspectral Remote Sensing Imagery based on Semisupervised Dimensionality Reduction with Pairwise Constraints

This paper proposed a new Semi|supervised dimensionality reduction method called Side information Constraint Semi|Supervised Dimensionality Reduction(SCSSDR) based on Semi|supervised dimensionality reduction framework.,which can preserve the intrinsic structure of the unlabeled data as well as global structure.For classification purpose,the SCSSDR take into account the discriminant information of labeled data,and finally get the projection matrix by solving an optimization problem.UCI data sets is used for quantitative analysis of the algorithm,and find that the method is superior to the traditional PCA and manifold learning algorithm,further experimental results on UCI data sets and hyperspectral data sets demonstrate the effectiveness of the method.