Pansharpening and Semantic Segmentation of Satellite Imagery

Abstract

The fusion of high spatial resolution panchromatic (PAN) data with simultaneously acquired lower spatial resolution multispectral (MS) data is called pansharpening. The conventional techniques like Brovey and IHS to perform the same are lengthy and time consuming. Hence, this paper proposes a Convolutional Neural Network for pansharpening which gives better results than the conventional methods. This paper also looks at binary and multiclass semantic segmentation using U-Net and its variations. Finally, a model based on 3D convolutions is introduced which performs semantic segmentation on Hyperspectral Imagery with high accuracy and computational efficiency.