Pixel, object and hybrid classification comparisons
- 1 June 2010
- journal article
- research article
- Published by Taylor & Francis in Journal of Spatial Science
- Vol. 55 (1) , 43-54
Abstract
The choice of the best classification approach for thematic map generation relies on many factors, such as image resolution and minimum mapping unit. The generalized GIS-ready products derived from the results of pixel-based approaches and the availability of higher-resolution imagery have directed research towards object-based classification approaches. In this paper we present the superior performance of a hybrid methodology that combines the results of automatic segmentation with the land cover information derived from a pixel classification by means of the Winner Takes All (WTA) algorithm. Land use and land cover results obtained through this hybrid classification approach are compared with those of a One Against All (OAA) object-oriented classification approach.Keywords
This publication has 20 references indexed in Scilit:
- Using texture analysis to improve per-pixel classification of very high resolution images for mapping plastic greenhousesISPRS Journal of Photogrammetry and Remote Sensing, 2008
- Feature group matching for appearance-based localization2008
- Spatial Techniques for Image Classification2006
- Image Segmentation Methods for Object-based Analysis and Classification2004
- An experimental assessment of minimum mapping unit sizeIEEE Transactions on Geoscience and Remote Sensing, 2003
- A Competitive Winner-Takes-All Architecture for Classification and Pattern Recognition of Structures2003
- A comparative study of filter based texture operators using Mahalanobis distance2002
- 10.1162/15324430152733133Applied Physics Letters, 2000
- Wrappers for feature subset selectionArtificial Intelligence, 1997
- Statistical and structural approaches to textureProceedings of the IEEE, 1979