Integration of C band SAR and optical temporal data for identification of paddy fields

The problem of identification of single crop fields is a challenge when single date optical remote sensing image is used. The use of temporal images solves this problem. However, issues like cloud cover in optical images influence accuracy of results. Microwave data, which penetrate through the atmosphere, solve this problem. The existence of mixed pixels in satellite images and nonlinearity in image classification is also overlooked. These issues were considered and worked on by integrating C band RISAT-1 with Formosat-2 temporal images and using possibilistic c-means classifier with similarity and dissimilarity norms to identify late transplanted paddy (Oryza sativa) fields in Haridwar District of India. Three datasets in different temporal combinations of microwave and optical images were classified for various similarity and dissimilarity norms for different values of weighted constant. Favorable results were achieved for Manhattan and mean absolute difference norm at weighted constant m = 1.3. Classification of late transplanted paddy for datasets containing multiple RISAT-1 and single Formosat-2 images with transplanting, growth stages was found to yield best results as compared to other combination of temporal images.