José Oniram de Aquino Limaverde Filho,  Díbio Leandro Borges,  Osmar Abílio de Carvalho Junior. 

Automatic object detection in satellite images is still an unsolved problem that many researchers are trying to address. In this context road detection is of particular importance for transportation and environmental applications. In this paper we propose a novel algorithm for detection roads in high resolution pancromatic imagery. It first detects and labels five levels of discontinuities in the images, and using the same original images fractal dimension and homogeneity measures of texture are computed. Those information are then fused in order for having a set of hypotheses of the pixels being actually roads. Results are presented showing roads detected in ALOS images of a region in Brazil.

http://www.lbd.dcc.ufmg.br/colecoes/wvc/2009/001.pdf

Processando consulta.