Share Email Print

Journal of Applied Remote Sensing

Impervious surface extraction using coupled spectral–spatial features
Author(s): Xinju Yu; Zhanfeng Shen; Xi Cheng; Liegang Xia; Jiancheng Luo
Format Member Price Non-Member Price
PDF $20.00 $25.00
cover GOOD NEWS! Your organization subscribes to the SPIE Digital Library. You may be able to download this paper for free. Check Access

Paper Abstract

Accurate extraction of urban impervious surface data from high-resolution imagery remains a challenging task because of the spectral heterogeneity of complex urban land-cover types. Since the high-resolution imagery simultaneously provides plentiful spectral and spatial features, the accurate extraction of impervious surfaces depends on effective extraction and integration of spectral–spatial multifeatures. Different features have different importance for determining a certain class; traditional multifeature fusion methods that treat all features equally during classification cannot utilize the joint effect of multifeatures fully. A fusion method of distance metric learning (DML) and support vector machines is proposed to find the impervious and pervious subclasses from Chinese ZiYuan-3 (ZY-3) imagery. In the procedure of finding appropriate spectral and spatial feature combinations with DML, optimized distance metric was obtained adaptively by learning from the similarity side-information generated from labeled samples. Compared with the traditional vector stacking method that used each feature equally for multifeatures fusion, the approach achieves an overall accuracy of 91.6% (4.1% higher than the prior one) for a suburban dataset, and an accuracy of 92.7% (3.4% higher) for a downtown dataset, indicating the effectiveness of the method for accurately extracting urban impervious surface data from ZY-3 imagery.

Paper Details

Date Published: 10 August 2016
PDF: 14 pages
J. Appl. Rem. Sens. 10(3) 035013 doi: 10.1117/1.JRS.10.035013
Published in: Journal of Applied Remote Sensing Volume 10, Issue 3
Show Author Affiliations
Xinju Yu, Institute of Remote Sensing and Digital Earth (China)
Zhanfeng Shen, Institute of Remote Sensing and Digital Earth (China)
Xi Cheng, Chengdu Univ. of Technology (China)
Liegang Xia, Zhejiang Univ. of Technology (China)
Jiancheng Luo, Institute of Remote Sensing and Digital Earth (China)

© SPIE. Terms of Use
Back to Top