<![CDATA[The rapid improvement of remote sensing technology has given rise to three paradigms of remote sensing image classification methods, namely pixel-based, object-based, and scene-based. This article aims to explain or reveal the development of remote sensing image classification methods and their relationship with Thomas Kuhn's scientific revolution process (pre-paradigm, normal science, anomaly, crisis, and scientific revolution) that occurs in the development of these classification methods. The preparation of this article uses a descriptive qualitative method. Reference sources are journal articles collected from the Scopus database with topics related to classification and remote sensing. Other reference sources are data extracted from review articles. From all the references collected, a literature study is then carried out by analyzing the article's title, abstract, and overall content. After that, the stages of the scientific revolution related to the development of classification methods in remote sensing images were described. Based on the review of the articles, it can be explained that the development of classification methods for remote sensing imagery began in the 1970s when the Landsat satellite was first launched. In this early period, the classification method used was based on pixels or sub-pixels, because the spatial resolution of remote sensing imagery was shallow. As remote sensing technology developed, in the 2000s a new approach was discovered that was more efficient than the pixel-based approach for classifying high-resolution imagery, namely object-based classification methods. Then, with the release of the land use dataset (UC-Merced) in the 2010s, scene-based remote sensing image interpretation began to be used, as pixel- and object-based methods were insufficient to classify correctly.]]>
Copyrights © 2023
