On the implementation of land cover classification systems for SAR images

The Synthetic Aperture Radar (SAR) sensor is widely used to record ground data under all atmospheric conditions. The high resolution of the SAR images contains useful information about the different ground regions. In this work, Khoros, a data visual language provided the tools to design and develop several land cover classification systems. The supervised classification tries to improve and automate the processes of recognition and classification of regions such as mountains and lakes. There are two supervised approaches: the segmentation and the multidimensional ones. Khoros is used as the main tool to provide a texture analysis of the images. The texture features under analysis includes: Statistics, Invariant Moments, Fractal Dimension, and Haralick's and Laws' texture features. The L5L5 and L5E5 (80.26%) features prove to be the best set of features to classify SAR images. The work proves the efficiency of Khoros in the visualization and manipulation of image processing techniques and shows a collection of several supervised land cover classification. Some of these systems prove the poor performance of several features for the efficient classification of SAR images.