| Abstract |
Remote sensing technology is indispensable for comprehending and monitoring the
Earth’s surface through the acquisition of data from satellite imagery. This literature review delves
into the realm of satellite image processing across various study objectives, employing diverse
learning paradigms. The scope of study encompasses a broad array of applications, including
land cover classification, change detection, object detection, segmentation, image fusion, and
retrieval systems. The methodologies explored in extracting meaningful insights from satellite
data span supervised, unsupervised, semi-supervised, and self-supervised learning techniques.
Supervised learning entails training models with labeled data to categorize and identify specific
features, while unsupervised learning facilitates pattern and structure extraction from unlabeled
data. Bridging the gap between supervised and unsupervised methods, semi-supervised learning
amalgamates labeled and unlabeled data. In contrast, self-supervised learning exploits inherent
data properties for representation learning without manual labeling. By scrutinizing the application
of these learning paradigms across various study objectives in remote sensing, this literature
review offers valuable insights into the progress and challenges in satellite image processing for
comprehending Earth’s surface dynamics. |
