Hyperspectral Data Classification Algorithm considering Spatial Texture Features

Algorithm 1

STF-CT algorithm.

Data: a labeled training sample set ;

   Sample set containing unlabeled samples;

1. Initialization: randomly select samples from as the “connected” dataset .

2. Repeat the following steps:

(a) A probabilistic SVM is used to train a classifier using spectral features on the set £

(b) Probabilistic SVM is used to train a classifier using texture features on set 3

(c) Use to mark most trusted samples from

(d) Use to mark most reliable samples from

(e) Add the most trusted samples of these new annotations to the third

(f) Randomly select enough samples from to add to

3. Until iterations are performed.