Share Email Print

Journal of Applied Remote Sensing

Classification of semiurban landscapes from very high-resolution satellite images using a regionalized multiscale segmentation approach
Author(s): Taskin Kavzoglu; Merve Yildiz Erdemir; Hasan Tonbul
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

In object-based image analysis, obtaining representative image objects is an important prerequisite for a successful image classification. The major threat is the issue of scale selection due to the complex spatial structure of landscapes portrayed as an image. This study proposes a two-stage approach to conduct regionalized multiscale segmentation. In the first stage, an initial high-level segmentation is applied through a “broadscale,” and a set of image objects characterizing natural borders of the landscape features are extracted. Contiguous objects are then merged to create regions by considering their normalized difference vegetation index resemblance. In the second stage, optimal scale values are estimated for the extracted regions, and multiresolution segmentation is applied with these settings. Two satellite images with different spatial and spectral resolutions were utilized to test the effectiveness of the proposed approach and its transferability to different geographical sites. Results were compared to those of image-based single-scale segmentation and it was found that the proposed approach outperformed the single-scale segmentations. Using the proposed methodology, significant improvement in terms of segmentation quality and classification accuracy (up to 5%) was achieved. In addition, the highest classification accuracies were produced using fine-scale values.

Paper Details

Date Published: 15 September 2017
PDF: 20 pages
J. Appl. Rem. Sens. 11(3) 035016 doi: 10.1117/1.JRS.11.035016
Published in: Journal of Applied Remote Sensing Volume 11, Issue 3
Show Author Affiliations
Taskin Kavzoglu, Gebze Technical Univ. (Turkey)
Merve Yildiz Erdemir, Gebze Technical Univ. (Turkey)
Hasan Tonbul, Gebze Technical Univ. (Turkey)

© SPIE. Terms of Use
Back to Top