Systematic Analysis of Effective Segmentation and Classification for Land Use Land Cover in Hyperspectral Image using Deep Learning Methods: A Review of the State of the Art : Reviewing Deep Learning Techniques for Land Use and Cover in Hyperspectral Images

This review paper delves into the intricate realm of segmenting and classifying Hyperspectral Images (HSI), complex visuals spanning numerous electromagnetic spectrum bands. HSI, garnering significant research attention, holds pivotal roles in geospatial applications, notably Land Use/Land Cover (LULC) mapping, demanding precise object identification. Addressing challenges posed by imbalanced data and limited labelled examples, the survey scrutinizes how researchers navigate HSI segmentation and classification. Briefly touching upon the foundational background of these techniques, the study navigates through diverse processing methods—thresholding, clustering, watershed, Deep Learning (DL) and others. Systematically exploring literature trends, DL advancements, attention mechanisms, data types, achieved accuracies and existing weaknesses, it strives to guide future research. Critically evaluating current knowledge, the paper illuminates’ gaps in HSI segmentation and classification, culminating in discussions on pertinent issues and prospective projects in this domain. Ultimately, this work aims to propel advancements by addressing knowledge lacunae and charting potential pathways for upcoming research initiatives.