An optimal dimension reduction-based feature selection and classification strategy for geospatial imagery

Driven by the explosive growth on the available data nowadays and advancement of technologies, the strong need arises for utilising and maintaining the available data. However, while building an expert prediction system, the inconsistency present in the information system, incompleteness of available knowledge base, continuous natured attribute values and noise present in the system (especially in case of spatial image data handling), are prime factors which may degrade the process of classification with available traditional methods. Our proposed construction adopts an efficient strategy for classification. Here we explore the problem of classifying remote sensing satellite images. Image data pre-processing and its categorisation refers to the labelling of individual pixel object instances into one of a number of predefined categories. Although this is usually not a much intractable task for humans, it has proved to be an extremely difficult problem for machines. We performed experimental analysis for classification using NWPU-RESISC45 dataset. Experiment result shows the improvement in classification by adopting our proposed strategy over other significant state of the art.