In:
Advanced Materials Research, Trans Tech Publications, Ltd., Vol. 605-607 ( 2012-12), p. 2274-2278
Abstract:
Mangroves are among the richest ecosystems in terms of natural resources. Because of the similar spectral characteristics and complex spatial structure among different mangrove communities, it is still difficult to extract accurate mangrove communities covers from the very high resolution (VHR) satellite images, based on the conventional pixel-based classification methods. Object-oriented classification methods are proposed to process VHR images because they can incorporate as much information on spatial neighbourhood properties as possible into the classification process. On the basis of object-oriented classification method, the paper extracted the Zhangjiangkou mangrove communities using Quickbird image, through image segmentation, merge segmentation, computing and selecting attributes, K Nearest Neighbor classification, et al. Mangrove communities were eventually divided into six types, i.e. Aegiceras corniculatum, Phragmites karka, Kandelia candel, Cyperus malaccensis, Avicennia marin, Aegiceras corniculatum & Kandelia candel. The extraction accuracy is 85.7%. In this paper, through many experiments, the scale level was determined to 50, and the merge threshold was chosen as 90. In this case, the resulting segments of mangrove communities had a obvious spatial characteristics, and the boundary can be delineated well, which would be help for later information extraction. Considering mangrove communities characters, the following object attributes are generated and added : NDVI, texture, and AVGBAND. This research shows very encouraging results for the use of image segmentation and Quickbird data for mapping mangrove communities.
Type of Medium:
Online Resource
ISSN:
1662-8985
DOI:
10.4028/www.scientific.net/AMR.605-607
DOI:
10.4028/www.scientific.net/AMR.605-607.2274
Language:
Unknown
Publisher:
Trans Tech Publications, Ltd.
Publication Date:
2012
detail.hit.zdb_id:
2265002-7
Bookmarklink