Volume 10, Issue 2 (2006)                   MJSP 2006, 10(2): 1-31 | Back to browse issues page

XML Persian Abstract Print


Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:

samadi Z, Alimohammadi A. Using Edge Density Information in Classification of Rural-Urban Region from Remotely-Sensed Data. MJSP 2006; 10 (2) :1-31
URL: http://hsmsp.modares.ac.ir/article-21-3943-en.html
1- Ph.D. Student of Climatology and M.Sc. in Remore sensing , Tehran University , zsamadi@ut.ac.ir
2- Assistant Professor of Remote Sensing and GIS , Department of GIS , Faculty of Geomatic Engineering , Khage nasir University, alimoh-abb@ yahoo.com
Abstract:   (5677 Views)
Operation manner in most of the conventional classification algorithms in remote sensing is based on pixels spectral information.Classification of these data ignore information obtained from adjacent pixels. In additional to with increasing of spatial resolution in satellites , increase harmful information( noise) and spectral similarity between classes , consequently increase internal variance of classes and finally decrease classification accuracy. To remove or decrease this problems , the proper incorporation and use of spectral and contextual information can efficiently help distinguish land-uses which are similar spectrally. In this study, effectiveness of incorporating structural information with classification procedures have been investigated. The technique is based on the use of edge-density information generated from the classified data. “ Maximum Likelihood ”(ML) , “ Minimum Distance to Means ”(MD) and “ Mahalanobis ” classification procedures have been used to classify data together with the edge-density information as an additional band. The performance of using edge-density data has been evaluated using the data of SPOT-XS and aerial photographs of the Anzali Wetlands ( Anzali Talab ) located in Gilan province north of Iran. This region is very heterogeneous. Results show that use of the structural information leads to increases in accuracy of some classes particularly those with low spectral separabilities. Mahalanobis classifier using spatial and spectral information in rural-urban ( 74.60) and river and channel ( 66.87) classes show 14.06 and 6.57 percent increases respectively in accuracies as compared to the spectral classification of satellite data. Application of this approach also in aerial photographs for patches of trees , river , agricultural and residential classes show 11.78 , 36.61 , 28.09 and 53.29 percent increases in accuracies respectively. Result show that considering the complex environmental conditions of the study site, the proper incorporation and use of spectral and spatial information can result in more efficient discrimination of some spectrally similar classes. The information of edge-density seems to be more promising in high resolution imagery and heterogeneous classes such as urban features.
Full-Text [PDF 410 kb]   (4399 Downloads)    

Received: 2010/05/25 | Accepted: 2010/05/25 | Published: 2010/05/25

Add your comments about this article : Your username or Email:
CAPTCHA

Rights and permissions
Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.