Services on Demand
- Cited by Google
- Similars in SciELO
- Similars in Google
Sociedade & Natureza
On-line version ISSN 1982-4513
FRANCISCO, Cristiane Nunes and ALMEIDA, Cláudia Maria de. Orbital images interpretation by means of an expert system for land cover mapping in highlands. Soc. nat. [online]. 2012, vol.24, n.2, pp.283-302. ISSN 1982-4513. http://dx.doi.org/10.1590/S1982-45132012000200009.
Land cover maps are an important data source for land planning and management, and hence, are crucial in zoning projects, environmental impact assessment, risky areas mapping, among other applications. They are usually derived from the interpretation of airborne or orbital images and/or the analysis of cartographic products, associated with field work. The traditional methods of remote sensing images classification consist either in pixel-per-pixel or region-based analyses, focusing on spectral differences for information extraction. The object-based image analysis (OBIA), although also based on the use of regions, represents an advance in relation to the traditional region-based classification approaches, for it relies on a knowledge model (semantic network) appended to the scene interpretation process, which renders the interpreter´s knowledge explicit, in a way to resemble the human cognitive processes. This paper aims to analyze land cover mapping resulting from the interpretation of remote sensing images using OBIA. Statistical and textural attributes extracted from ALOS/AVNIR images pan-sharpened with ALOS/PRISM image as well as relief data from the TOPODATA geomorphometric database were used as input data. The study area is Nova Friburgo County, with an area of 933 km2, located in the mountainous region of Rio de Janeiro State. The land cover map was validated by the Kappa index, which relates classified samples with field data. The Kappa value obtained in this paper was 0.85, which showed to be greater than the ones found in similar works that used traditional classification techniques.
Keywords : remote sensing; images classification; object-based image analysis; land cover; ALOS.