Snakes-based approach for extraction of building roof contours from digital aerial images

Snakes-based approach for extraction of building roof contours from digital aerial images

Aluir P. Dal Poz and Antonio J. Fazan

São Paulo State University – Dept. of Cartography, R. Roberto Simonsen 305 – 19060-900 Presidente Prudente, SP – Brazil (aluir@fct.unesp.br, ajfazan@gmail.com.cr)

Abstract: This paper presents a method for building roof contour extraction from digital images taken over complex urban scenes. The proposed method is based on a snakes-based energy function that represents building roof contours in digital images, which is optimized by using the dynamic programming (DP) algorithm. As most of the building roof contours are characterized by rectilinear sides intercepting at right angles, appropriate geometrics constraints are enforced into the original snakes energy function. The main advantage of using the DP algorithm for optimizing the proposed snakes-based energy function is its better radius of convergence, when compared to the one that is usually obtained in the original solution based on variational approaches. Experimental evaluation, including visual inspection and numeric analysis, was performed by using real data and the obtained results showed the potentiality of the proposed method for extracting building roof contours from digital images. Keywords: Snakes, dynamic programming, building extraction, image analysis.

AttachmentSize
PozAccuracy2012.pdf117.67 KB