Improved ML Models on Feature Extraction in HSI

Hyper Spectral Image (HSI) is widely used in remote sensing applications that take into account thousands of spectral channel compositions over a single scene. HSI requires accurate models of learning to extract the features of the HSI. Due to the presence of its spatial and spectral resolution, the image learning model presents a core challenge due to its complicated nature of image frames. In order to assist it during the learning process, several attempts have been made to address its complicated nature. These methods, however, failed to provide HSI understanding. Using ML techniques addresses the issues due to the presence of mixed pixels, large amounts of data and limited samples of training. The process of ML addresses the complex relationship between image data. We present various methods of ML used to learn HSI in this paper. Initially, for different image processing techniques, we present an overview of different ML methods. A system review is then performed on the basis of deep learning based on various HSI learning models. Finally, for future study, we are discussing the challenges and possible directions.