UNSUPERVISED LEARNING FOR HYPERSPECTRAL IMAGE CLASSIFICATION
Out of all the data sources accessible to geographic information systems (GIS), remote sensing is one of the most crucial ones. Remote sensing is the process of gathering data about the surface of the Earth without physically being there. It detects radiations that are emitted and reflected and are normally captured by sensors that are installed on an aircraft or a satellite. Modeling and monitoring activities on the Earth's surface as well as identifying elements in the land cover by analysing spectral characteristics collected by sensors are the two main goals of remote sensing. When a specific sensor device collects and processes information from the electromagnetic spectrum, it is known as Hyperspectral Imaging (HSI). The data it generates is a goldmine of information. Use this data to solve a variety of problems in a variety of applications. A digital image's pixels are divided up into groups using hyperspectral imaging classification. These techniques were used to classify hyperspectral images using unsupervised hyperspectral image classification algorithms. KMeans and ISODATA algorithms are employed. ENVI is used to apply two algorithms to a hyperspectral image of Washington DC, USA. The accuracy of the procedure was assessed using Principle Component Analysis (PCA) and K-Means or ISODATA algorithm in this paper. The ISO-DATA algorithm outperforms the K-Means algorithm in terms of precision. Since The K-Means algorithm has a classification accuracy of 78.3398 percent, whereas the ISODATA algorithm has a classification accuracy of 81.7696 percent. When the number of classification iterations increased, so did the processing time.