Abstract:
This paper describes a method to identify different types of vegetation (under the domain limits of the powerlines in the forest) by a vegetation index and the maximum li...Show MoreMetadata
Abstract:
This paper describes a method to identify different types of vegetation (under the domain limits of the powerlines in the forest) by a vegetation index and the maximum likelihood classification techniques in multispectral QuickBird images. These images were chosen due to its spatial characteristics, since the domain limits width established by Brazilian standards is 100m (in that study, a 50m band width was used) and the average distance of two electric towers is 420m. The initial identification of the areas with a higher concentration of biomass was obtained by the Atmospheric Resistance Vegetation Index (ARVI), calculated from the 1, 3 and 4 channels. The final classification process was developed using (as the accepted threshold and the change threshold respectively of 99% and 5%) the Maximum Likelihood (Interacted Conditional Modes - ICM) classifiers. Training samples were collected in the monitored area covered by a 40km of the powerline providing an overall accuracy of 85.90% and the worst performance was observed in the pasture category (74.8% correctly classified). The areas with the highest vegetation density were identified by ARVI, it discriminated bare soil areas from the category: water, pasture and dense vegetation. However details of that last class (water, pasture and dense vegetation) were not available since their spectral responses were very close in the domain of QuickBird channels.
Date of Conference: 23-28 July 2007
Date Added to IEEE Xplore: 07 January 2008
ISBN Information: