Performance assessment of phased array type L-band Synthetic Aperture Radar and Landsat-8 used in image classification

Owing to its large spatial and periodic temporal coverage, satellite remote sensing has emerged for formulating and implementing strategies for natural resources management. This study focuses on an appraisal of satellite sensors and artificial intelligence techniques such as kernels-based support vector machines (SVMs) and artificial neural networks (ANNs). These methods are used for land cover classification on multispectral and microwave satellite images acquired from Landsat-8 and Advanced Land Observing Satellite (ALOS-2) Phased Array type L-band Synthetic Aperture Radar (PALSAR) over Varanasi, India. The analysis shows comparable the performance of the microwave-classified image compared with the multispectral Landsat-8 image. ANNs and SVMs performed best with an overall accuracy of 97.3% and kappa coefficient of 0.97 for the Landsat-8 image, whereas SVM radial basis function was the best classifier for the ALOS PALSAR image with 94% overall accuracy. Other statistical indices such as kappa total disagreement and allocation disagreement scores revealed similar performances. The analysis demonstrated the ability of microwave data in land cover classification studies with satisfactory performance. These data can be used in nearly all weather and environmental conditions for broad image classification purposes when optical and infrared imagery such as Landsat are unavailable.