Detection of Rice Fields in Rasht Township Using Multi-Temporal Landsat-8 Images

Rice has become one of the most important food security items in many countries, especially Iran. In this study, a model was proposed to select Landsat-8 satellite time-series images in order to prepare a map of paddy lands. The method is based on the phenological characteristics of rice plants and annual surface temperature data from the MODIS sensor. After preprocessing satellite images, they were classified using an object-based approach and fuzzy functions. Various data such as a digital elevation model, land surface temperature, and spectral indices including NDVI, EVI, NDBI and LSWI are used to improve the classification process. In addition, information about the segmentation of the image is employed during the process of classification. Because of the different traits of paddy fields, a digital elevation model with a resolution of 12.5 meters was used to help differentiate paddy lands from other vegetation. In addition, a comparison was made between the results of classification based on object-based and pixel-based methods. The results showed that the object-based classification yields better results than the pixel-based method with special considerations. The classification result following validation using ENVI software pixel-based classification indicated an overall accuracy of 92 percent and a kappa value of 0.89. This is in contrast to the object-based classification technique in the eCognition software, which yielded an overall accuracy of 94 percent and a kappa coefficient of 0.92.