Rational feature selection from the varieties of spectral channels in the optical wavelengths of electromagnetic spectrum (VIS and NIR) is very important for effective analysis and information extraction of remote sensing data. Feature selection is one of the most important steps in recognition and classification of remote sensing images. Therefore, it is necessary to select features before classification. Three factors-the information quantity of bands, the correlation between bands and the spectral characteristic (e.g. absorption specialty) of classified objects in test area Romanian North Western Black Sea coastal area have been considered in our study being suggested a method of multi-level feature selection. Spectral signatures of different terrain features have been used to extract structural patterns aiming to separate surface units and to classify the general land cover categories. The synergetic analysis and interpretation of the different satellite images (LANDSAT: TM, ETM; IKONOS) acquired over a period of 20 years reveals significant aspects regarding impacts of climate and anthropogenic changes on coastal area as well as in Constanta town urban/periurban environment. Information on the spatial pattern and temporal dynamics of marine coastal areas land cover is critical to address regarding sustainability and rational planning policy.

Date Published: 25 October 2010
PDF: 7 pages
Proc. SPIE 7831, Earth Resources and Environmental Remote Sensing/GIS Applications, 78311P (25 October 2010); doi: 10.1117/12.864793

Published in SPIE Proceedings Vol. 7831:
Earth Resources and Environmental Remote Sensing/GIS Applications
Ulrich Michel; Daniel L. Civco, Editor(s)