A Study on the Advanced Method for Image Classification of Remote Sensing Datasets Using Deep Learning Algorithms

Abstract

Geospatial technology and advanced AI methodologies can improve the processing of enormous spatial datasets, provide accurate forecasts, rapid user-defined models, and more. Machine-Deep-Learning algorithms, a branch of artificial intelligence, are supported by powerful computing platforms and can be used with geospatial science to visualise, analyse, and predict real-time COVID-19 issues.  The main aim of this study is to emphasise geospatial-analytical methods, advanced machine-deep learning algorithms in big data mining, spatial visualisation, and web-based spatial analysis that can give decision-makers new predictive models and more intuitive information. Multi-dimensional sensors simplify data collection and enable global research. This study analyses complex remote sensing datasets using free optical and microwave data. Landsat -7,8, Sentinel-2, and MODIS are optical datasets, while Sentinel-1 SAR is microwave. Machine learning and the most common deep learning models of convolutional neural network (CNN) for large-scale mapping can automatically and independently extract information without human interaction. It depicts numerous steps and the complete procedure. Data is injected and transmitted across layers to extract key features by removing picture dimensions. Google Earth Engine's cloud platform helps do the tasks. These cutting-edge approaches can use heterogeneous and complex huge data in remote sensing applications. This laid the groundwork for the new information age's geographical database.