Synthetic Aperture Radar (SAR) images (Microwave data) were classified using Multi-Layer Feed Forward, Cascade Forward Neural Networks and Random Forest (RF) algorithms. For the Random Forest, a general model for classification of Remotely Sensed Radar dual-polarization data based on RF is implemented and classified of SAR image (microwave data) classifications. The RF model exploits spatial context between neighbouring pixels in an image, and temporal class dependencies between different images of the same region, in the case of multi-temporal data. Based on the well-founded experimental on basis of random forest techniques for classification tasks and the encouraging experimental results in RF algorithm , the authors conclude that the proposed RF algorithm is useful for classification of SAR (Sentinel 1A) imagery and evaluate its accuracy and kappa coefficient.

No abstract