PROJECT TITLE :
Weighted Sparse Graph Based Dimensionality Reduction for Hyperspectral Images
Dimensionality reduction (DR) is an important and useful preprocessing step for hyperspectral image (HSI) classification. Recently, sparse graph embedding (SGE) has been widely used in the DR of HSIs. SGE explores the sparsity of the HSI data and can achieve sensible results. But, in most cases, locality is additional important than sparsity when learning the options of the data. In this letter, we have a tendency to propose an extended SGE technique: the weighted sparse graph based mostly DR (WSGDR) methodology for HSIs. WSGDR explicitly encourages the sparse coding to be native and pays additional attention to those coaching pixels that are a lot of just like the check pixel in representing the take a look at pixel. Furthermore, WSGDR can offer knowledge-adaptive neighborhoods, which ends up in the proposed technique being more strong to noise. The proposed method was tested on 2 widely used HSI knowledge sets, and the results recommend that WSGDR obtains sparser illustration results. Furthermore, the experimental results conjointly make sure the superiority of the proposed WSGDR technique over the other state-of-the-art DR ways.
Did you like this research project?
To get this research project Guidelines, Training and Code... Click Here