Unsupervised classification of satellite images

The work had an ambition to do the parallel between the results of the principal methods, and unsupervised classifications applied to remote sensing data. To achieve this goal, it is essential to point out certain characteristics of the remote sensing thus the fundamental principles of the unsupervised methods of classification. 

Three methods of classification were tested on remote sensing image :

• Hierarchical ascending classification
• K-means clustering
• C-means clustering (Fuzzy)

Context : collaboration between University of sciences of Tunis and the National Mapping and Remote Sensing Center (Tunisia).

Keywords: Remote sensing, Image processing, classification.

Download: report

• ---

• Test area

 Landsat TM satellite image- Mareth area (south of Tunisia)

Classification results samples 

K-mean classification (5 classes)