Application of unsupervised learning of finite mixture models in ASTER VNIR data-driven land use classification. Issue 1 (2nd January 2021)

• Record Type: Journal Article
• Authors: Zhao, Bo; Yang, Fan; Zhang, Rongzhen; Shen, Junping; Pilz, Jürgen; Zhang, Dehui
• Is Part Of: Journal of spatial science. Volume 66:Issue 1(2021)
• Page Start: 89
• Page End: 112
• Publication Date: 2021-01-02
• Subjects: Mixture Gaussian distribution -- land use -- topographical analysis -- remote sensing; Geographic information systems -- Periodicals; Cartography -- Periodicals; Surveying -- Periodicals; Geodesy -- Periodicals; Photogrammetry -- Periodicals; Cartography; Geodesy; Geographic information systems; Photogrammetry; Surveying; Periodicals; 526.05
• DOI: 10.1080/14498596.2019.1570478
• Languages: English
• ISSNs: 1449-8596

Abstract

ABSTRACT: Based on an ASTER VNIR image, we studied the applicability of the MML-EM (Minimum Message Length Criterion-Expectation Maximization) algorithm for land-use classification in southern Austria. Firstly, the RVI (ratio vegetation index) and PC1 (first principal component) bands have been utilized to enhance the targeted information; secondly, the MML-EM algorithm and the terrain analysis-based imagery clipping were jointly used for surface type discrimination. Findings showed that the MML-EM method can provide refined imagery classification results and this is the first time it has been applied in this realm.