Abstract

PAZ PELLAT, Fernando  and  MEDRANO RUEDAFLORES, Edgardo. Discrimination of land cover classes using multi-angular spectral data with polder-1 sensor: scopes and limitations. Terra Latinoam [online]. 2016, vol.34, n.2, pp.187-200. ISSN 2395-8030.

Multi-angular spectral information (vision of a pixel from different viewing angles and different solar illumination angles) obtained from remote sensing has the potential to discriminate adequately land cover classes. According to several efforts to analyze the capacity of discrimination of land cover classes, a conceptual theoretical framework for the analysis of the spectral, angular and temporal information (fixed pixel size) is introduced. In this paper the use of the POLDER-1 sensor is explored. The database was analyzed by fitting a model function of bidirectional reflectance distribution (BRDF) over bandwidths available for different classes of land cover GLC2000 system. Experimental results show appropriate adjustments on pixel-level daily data. With the set parameters of the BRDF model the potential for discrimination using spectral band spaces, red and near infrared bands, with different temporal and spatial resolution (pixel group) were analyzed. The results were highly confusing (position overlaps in spectral spaces), pointing to limitations of such approaches because of confusion caused by mixtures of classes or their temporal dynamics. At the end, a scheme is defined to approximate the vegetation classification by coupling the information available on optical sensors and classes that can be discriminated.

Keywords : BRDF; BUPM; GLC2000; normalized reflectances; g and G parameter; discrimination.

        · abstract in Spanish     · text in Spanish     · Spanish ( pdf )