Online citations, reference lists, and bibliographies.

Land Use Classification Via Multispectral Information

Cem Unsalan, Kim L. Boyer
Published 2011 · Computer Science
Cite This
Download PDF
Analyze on Scholarcy
Share
In the previous chapter, our land use classification approach was based on the organization of straight lines (structure) in panchromatic images. It is well-known that multispectral information also offers a great deal of information for land use classification. This chapter describes an approach to combining structural information, obtained from 1m panchromatic Ikonos images with spectral information, obtained from the corresponding 4 m multispectral images with application to identifying areas of significant land development. There are several contributions in the literature in which spatial and spectral features have been combined in land use classification and related problems. However, none to date use the line support region structural feature, as we do. Finally, we introduce additional spatial information, over a broader area than the structural information captured in the line support regions, by means of probabilistic relaxation. Although relaxation improves classification slightly, the improvement comes at substantial computational cost. Therefore, we recommend that this approach be used only in applications where the improvement is absolutely necessary.



Semantic Scholar Logo Some data provided by SemanticScholar