3D Road extraction using a stereo pair of aerial images

• Autor(a) principal: Dal Poz, A. P. [UNESP]
• Data de Publicação: 2010
• Outros Autores: Silva, J. F.C. [UNESP], Gallis, R. A.B. [UNESP]
• Tipo de documento: Artigo de conferência
• Idioma: eng
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://hdl.handle.net/11449/227937
• topic: Aerial Images; Dynamic Programming; Feature Extraction; Road Extraction; Road Model

Road extraction from aerial and satellite imagery is of fundamental importance in the context of spatial data capturing and updating for GIS (Geographic Information System) applications. This article proposes a semiautomatic methodology for 3D road extraction from digital images. The dynamic programming (DP) algorithm is used to carry out the optimisation process in the object-space, instead of doing it in the image-space such as the DP traditional methodologies. This means that road centrelines are directly traced in the object-space, implying that a mathematical relationship is necessary to connect road points in object- and image-space. This allows the integration of radiometric information from images into the associate energy function. Although the methodology can operate in different modes (mono-plotting or stereo-plotting), and with several image types, including multisensor images, this paper presents details of our stereo-based approach involving a stereo pair of aerial images. In order to initialise the extraction procedure, a few seed points are necessary to be supplied in the object-space to sparsely and coarsely describe the roads. We visually identify these points on the stereo images and project them onto the object-space. Each road is extracted separately by an iterative scheme based on the DP optimisation algorithm. Each iteration cycle is carried out by means of a polyline densification followed by a DP optimisation. The polyline densification is carried out by linearly interpolating middle points between every adjacent vertices of the current polyline. Preliminary results have shown that the proposed methodology works properly.