Urban-landuse/land-cover mapping with high-resolution SAR imagery by integrating support vector machines into object-based analysis
2008 (English)In: SPIE Europe Remote Sensing Conference, 2008, 2008, Vol. 7110Conference paper (Refereed)
Thispaper investigates the capability of high-resolution SAR data for urbanlanduse/land-cover mapping by integrating support vector machines (SVMs) into object-basedanalysis. Five-date RADARSAT fine-beam C-HH SAR images with a pixelspacing of 6.25 meter were acquired over the rural-urban fringeof the Great Toronto Area (GTA) during May to Augustin 2002. First, the SAR images were segmented using multi-resolutionsegmentation algorithm and two segmentation levels were created. Next, arange of spectral, shape and texture features were selected andcalculated for all image objects on both levels. The objectson the lower level then inherited features of their superobjects. In this way, the objects on the lower levelreceived detailed descriptions about their neighbours and contexts. Finally, SVMclassifiers were used to classify the image objects on thelower level based on the selected features. For training theSVM, sample image objects on the lower level were used.One-against-one approach was chosen to apply SVM to multiclass classificationof SAR images in this research. The results show thatthe proposed method can achieve a high accuracy for theclassification of high-resolution SAR images over urban areas.
Place, publisher, year, edition, pages
2008. Vol. 7110
high resolution, SAR, landuse/land-cover, classification, SVM, object-based analysis
IdentifiersURN: urn:nbn:se:kth:diva-27205DOI: 10.1117/12.800298ScopusID: 2-s2.0-62449238125OAI: oai:DiVA.org:kth-27205DiVA: diva2:375854
2008 SPIE Europe Remote Sensing Conference, Cardiff, UK.
QC 201012092010-12-092010-12-092016-06-08Bibliographically approved