Info

Title: Semi-supervised Hpyerspectral Image Classification via Discriminant Analysis and Robust Regression
Authors: Cheng et al.
Year: 2016
Source: here
Journal: IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing

What is about?

Developed Discriminant Analysis and Robust Regression (DARR) algorithm, a regression based semi-supervised technique to establish landuse classification by using hpyerspectral image.

What are amazing points comapred to the previous study?

  • Applied a graph based semi-supervaised methodology which allows to build a classification model with less labeled data
  • Effectively and algorithmically processing number of hyperspectral images to create features into a model
  • Achieved higher accuracy of landuse classification compared to previous methodologies

What are the “keys” of technology and methodology?

  • Graph based manifold structures
  • Pairwise constraints based discriminant analysis
  • Robust regression with adaptive loss function
  • Robust hyperspectral image classification via discriminant analysis and aaptive loss function

How did they varified effectiveness of this study?

Evaluate different algorithms (SVM-SS, SSL, MLR-AL-SS, EPF, MLRsubMLL) and proposed methodology by three metrics (overall accuracy, average accuracy and kappa). They used four different hyperspectral image datasets to process those algorithms.

Is there any discussion?

[TBD]

Is there any paper need to read?

[TBD]