New classification techniques for analysis of remote sensing integrated data

The aim of this work is the application of techniques based on fuzzy logic for integrating and classifying spectral data and spatial information. The analysis was performed on a Landsat-TM image and on a set of map field data. Firstly fuzzy membership values of Landsat image pixels were calculated and a buffer area was defined around map field data for each considered class. The membership values were then modulated using spatial information derived from known targets. Finally the accuracy of the classification was assessed on a set of test data. The obtained results shown a significant improvement compared to more traditional methods of classification.