Publications

Li, Wenliang; Wu, Changshan (2015). Incorporating land use land cover probability information into endmember class selections for temporal mixture analysis. ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING, 101, 163-173.

Abstract
As a promising method for estimating fractional land covers within a remote sensing pixel, spectral mixture analysis (SMA) has been successfully applied in numerous fields, including urban analysis, forest mapping, etc. When implementing SMA, an important step is to select the number, type, and spectra of pure land covers (also termed endmember classes). While extensive studies have been conducted in addressing endmember variability (e.g. spectral variability of endmember classes), little research has paid attention to the selection of an appropriate number and types of endmember classes. To address this problem, in this study, we proposed to automatically select endmember classes for temporal mixture analysis (TMA), a variant of SMA, through incorporating land use land cover probability information derived from socio-economic and environmental drivers. This proposed model includes three consecutive steps, including (1) quantifying the distribution probability of each endmember class using a logistic regression analysis, (2) identifying whether each endmember class exists or not in a particular pixel using a classification tree method, and (3) estimating fractional land covers using TMA. Results indicate that the proposed TMA model achieves a significantly better performance than the simple TMA and a comparable performance with the METMA with an SE of 2.25% and an MAE of 3.18%. In addition, significantly better accuracy was achieved in less developed areas when compared to that of developed areas. This may indicate that an appropriate endmember class set might be more essential in less developed areas, while other factors like endmember variability is more important in developed areas. (C) 2014 International Society for Photogrammetry and Remote Sensing, Inc. (ISPRS). Published by Elsevier B.V. All rights reserved.

DOI:
10.1016/j.isprsjprs.2014.12.007

ISSN:
0924-2716