期刊
REMOTE SENSING OF ENVIRONMENT
卷 119, 期 -, 页码 105-117出版社
ELSEVIER SCIENCE INC
DOI: 10.1016/j.rse.2011.12.010
关键词
RGB-clustering; Change-detection; Landsat-TM, dune-fields; Dune dynamics
资金
- Interfaculty Council for Development Co-operation (IRO)
- Katholieke Universiteit Leuven
- STEREO II-program of the Belgian Science Policy [SR/00/132]
Studying dune dynamics at the dune-field scale requires intensive fieldwork and a relatively huge dataset. However, this is not always possible with the high cost and the limited availability of data. Here, we present a technique based on the analysis of multi-temporal Landsat-TM images for studying the dynamics of different dune morphologies from five dune-field sites around the world. A pair of Landsat-TM images for each site has been used after performing several steps of image pre-processing and enhancement (i.e. image co-registration, radiometric calibration and histogram equalization). RGB-clustering is a commonly used algorithm for data compression and iso-clustering. In this study, we implemented this technique as an unconventional change-detection algorithm with 32 clustering partitions per band. A bi-temporal layerstack of the near-infrared bands was the input for the RGB-clustering process. RGB-clustering produced thematic maps that gave an immediate insight into the dune dynamics at the dune-field scale given the availability of Landsat images. The technique gave the best results with dunes developed under uni-directional wind pattern. In addition, it was also easier to detect migration of simple dunes than in compound or complex dunes. These maps can be used as a bi-temporal digital atlas of large dune-systems and help in defining areas for detailed field measurements on dune dynamics and migration. Moreover, it can be implemented in analyzing aeolian systems where field measurements are practically impossible (i.e. large and extra-terrestrial aeolian systems). (C) 2011 Elsevier Inc. All rights reserved.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据