Multiple data-dependent kernel for classification of hyperspectral images

作者：

Highlights：

• Propose a MDK for classification of hyperspectral image.

• Adopt centered kernel alignment (CKA) to optimize the weights of multiple basic kernels.

• Transform the CKA problem into quadratic programming (QP) problem.

• Utilize Fisher’s discriminant analysis (FDA) to optimize the coefficients of data dependent kernel.

• Transform the Fisher scalar maximization problem into generalized eigen-decomposition problem.

摘要

•Propose a MDK for classification of hyperspectral image.•Adopt centered kernel alignment (CKA) to optimize the weights of multiple basic kernels.•Transform the CKA problem into quadratic programming (QP) problem.•Utilize Fisher’s discriminant analysis (FDA) to optimize the coefficients of data dependent kernel.•Transform the Fisher scalar maximization problem into generalized eigen-decomposition problem.

论文关键词：Classification,Hyperspectral image (HSI),Data-dependent kernel (DK),Support vector machine (SVM),Sparse representation classifier (SRC)

论文评审过程：Available online 16 September 2014.

论文官网地址：https://doi.org/10.1016/j.eswa.2014.09.004